Questions, answered plainly

• Can I keep a calendar without a Google or Microsoft account?

  Yes. A calendar is dates and text, and a browser can store both locally. What you give up is syncing between devices and invitations arriving by email, both of which need a server that holds your schedule.

• What is an .ics file?

  The standard calendar exchange format, and what every invitation actually is. It is plain text, so it can be read, edited and moved between applications. It also means an invitation carries more than the event: attendee addresses, the organiser, and often a join link and passcode.

• Does a calendar reveal much about me?

  A great deal. Who you meet, how often, where, and what you call it. A shared calendar or an exported calendar file is one of the more revealing documents most people own, which is worth remembering before granting an application access to it.

• Can I move events to Apple, Google or Outlook Calendar?

  Yes. Export an .ics file and import it there, and import their .ics exports here the same way. Everything is kept in this browser; there is no account and nothing is uploaded.

• Can I keep more than one calendar?

  Yes. Each one has a name and a colour, and every event belongs to exactly one of them. Unticking a calendar takes its events off the grid without deleting anything, and removing a calendar moves its events to another one rather than throwing them away.

• Will a reminder reach me if the tab is closed?

  No, and the page says so rather than letting you find out by missing a meeting. A reminder fires from a timer in an open tab of this site. One that came due while the site was shut is still told when you come back, but as a missed one, because a toast saying “Standup” three days late is a lie about the time.

• Does the desktop notification need an account?

  No. It is the browser’s own notification, asked for only when you press the button, and it never leaves the machine. Without it, reminders still appear in the page.

• Does the .ics file carry the reminder?

  It does, as a VALARM with a TRIGGER counted back from the start, which is what Apple Calendar, Google Calendar and Outlook read. The calendar’s name and colour go on the file too, so a file that has been out and back comes home to the same calendar rather than making a second one.

• Where do the events in a file I import end up?

  You are asked before anything is written. You can put them all in one calendar, or let each event go to the calendar it names in the file, which makes any calendar that is not here yet.

• Can I move an event to another day without opening it?

  Drag it to another day or time. For a repeating event it asks whether to move this event, this and the following ones, or all of them.

• What are the chips on some days?

  Tasks from the board that are due that day, shown but not owned: the chip is a link to Tasks and this page never writes a task. A finished task does not appear.

• What is the week view for?

  Seven columns of clock time, with each event drawn where it falls and a grip at the foot of it for changing when it ends. On a phone there is no room for seven columns, so the week goes back to seven day cells.

Open Calendar

• What is a vCard file?

  The standard format for exchanging contact details, and what the .vcf attachment your phone produces actually is. It is plain text, so it is portable between address books and readable in any editor, which also means an exported address book is a plain-text list of everyone you know.

• Can I keep an address book without handing it to an app?

  Yes, locally in the browser, with vCard import and export to move it around. The reason to bother is that address book access is one of the most requested and most abused permissions on a phone: an application granted it receives everyone in the book, not just you.

• How do I move contacts between phones or services?

  Export to vCard from the old one and import to the new one. Check a few entries afterwards, because photographs, custom fields and notes are the parts that most often fail to survive the trip, and delete the exported file once you are done.

• Where are my contacts kept?

  In this browser's local storage on this device, saved on every change. Nothing is sent anywhere: no account, no sync, no address book harvesting. Clear the browser's site data and they are gone, so export a copy of anything that matters. To carry contacts to a phone, export a .vcf and open it there.

• Which vCard fields does it keep?

  Everything an address book uses: every part of the name, with the phonetic name and the nickname, the company, job title and department, every email, phone and address with the kind each one is, the birthday with or without a year, anniversaries and other dates, related people, websites, custom fields, the labels, the notes and the picture. A label somebody made up, such as a phone called Studio, travels the way iPhone and Google write one. The two it leaves are a picture held as a link to a web address rather than carried in the card, which is left alone because fetching it would tell that server who is reading the card, and chat addresses, which have nowhere to go here. Both are said out loud when they happen rather than passed over.

• What happens when I import somebody who is already here?

  The two are made one. An import used to skip a card whose name and email matched, which is right when the two are identical and wrong the rest of the time: the copy on your phone usually has the mobile number and the copy here has the work address, and skipping means the two halves never meet. Merging adds what is new (another number, another address, a group, a birthday, a picture where there was none) and changes nothing that is already filled in, so the same file can be imported twice with no harm done. Two cards count as the same person when they share an email address, or when they have the same full name and at most one of them has an email at all.

• What are the labels?

  What a phone calls groups and a vCard calls categories. Make one with the + beside Labels, then tick contacts in the list and put them under it with Manage labels; each label is listed down the left with how many people are in it. A label made here travels to a phone in the .vcf, and a group made on a phone arrives here as a label. Deleting a label asks whether to keep its contacts or to move them to the trash with it.

• Is the picture kept at full size?

  No, and it says so when you choose one. A photograph off a phone is several megabytes and this browser keeps about five in all, so the picture is squared off and drawn down to 192 pixels, which is about twelve kilobytes and is as much as a thumbnail beside a name can show. The file you chose is not touched; this is the copy the address book keeps, and it is the one written into the .vcf.

Open Contacts

• Can I write a document in the browser without an account?

  Yes. A word processor is ordinary browser capability, and nothing about editing text requires a server. The document stays in this browser's storage on this device, so there is nothing to sign into and nothing uploaded, and equally nothing syncing to your other devices.

• Where is my document stored?

  In this browser's storage on this device. Clearing site data for this site deletes it, a private window will not see it, and another browser or another machine is a separate world entirely. Export anything you want to keep.

• Can I open and save Word files?

  Yes. File > Open (Ctrl+O) reads a .docx, and File > Save (Ctrl+S) or Save As (F12) writes one. The Word file carries the headings, lists, tables, links and pictures, the comments and their replies, tracked changes, footnotes and endnotes, and the page size, margins, header, footer and page numbers. The watermark is printed but not written into the Word file. Both happen in this tab; the file is never uploaded.

  Opening a Word file brings back its words, formatting, pictures, comments, footnotes and endnotes. Its tracked changes arrive already accepted, and its own header, footer and page setup are not read.

• Can I set the paper size and margins?

  Yes. Layout > Size offers Letter, Legal, Executive, A4, A5, B5 and Tabloid; Layout > Orientation, portrait or landscape; and Layout > Margins, Normal, Narrow, Moderate or Wide, or Custom Margins... for any four figures. The sheets on screen take the paper's size and margins, the print takes all of it, and the Word file gets the paper and the margins. The settings are kept with the document.

• Can I put a header or footer on every page?

  Yes. Insert > Header > Edit Header, or double-click in the top margin, and type; the footer is the same, from Insert > Footer > Edit Footer or the bottom margin. Each is one line of text, the same on every page, shown on every sheet on screen, printed, and written into the Word file. Page numbers go at the foot from Insert > Page Number.

• How do I start a new page?

  Ctrl+Enter, or Insert > Page Break (also Layout > Breaks > Page). It shows as a dotted line marked Page Break, the text after it starts on the next sheet on screen and on paper, and it is a page break in the Word file. Insert > Blank Page puts in a whole empty page.

• Can it make a table of contents?

  References > Table of Contents > Automatic Table lists every Heading 1, 2 and 3 with the page it is on and a link to it. After the headings change, References > Update Table brings it up to date in place.

• Can I comment on a document?

  Select the words and press Ctrl+Alt+M, or use Review > New Comment or Insert > Comment. The comment is anchored to the words, so it cannot drift off them, and replies thread under it. Your name is typed once and kept in this browser; it never leaves the device.

• Do comments survive being saved as Word?

  Yes. They are written as a real word/comments.xml with the range markers around the right words and the replies threaded through commentsExtended, which is how Word itself stores them, so Word, LibreOffice and Google Docs all show them. Opening that file here brings them back onto the same words.

• Are comments printed?

  Not unless you ask. On paper the highlights are hidden, because a highlight nobody asked for is not part of the document. Ticking “print with comments” keeps them and adds the threads as a list at the end.

• Can I get an earlier version of a document back?

  File > Info > Version History keeps the document as it stood at each autosave that changed the words, with the time, the word count and what changed since the version before it. You can preview one read only, name one to keep it for good, or restore it. A restore keeps the state it replaced first, so it can itself be undone.

• How many versions are kept?

  The last twenty-five, or about two megabytes a document, oldest out first. A version you have named is never dropped. Browser storage is small and one pasted picture fills it, so history is never allowed to be the reason a document will not save.

• Can I number the printed pages?

  Insert > Page Number puts the number at the bottom of every page: Plain Number 1 on the left, 2 in the centre, 3 on the right; Remove Page Numbers takes it off. In Print Layout each sheet on screen shows its number, the print counts the sheets itself, and the Word file gets a page-number field, so Word numbers its pages however much you edit.

• How do I type a character my keyboard does not have?

  Insert > Symbol shows the ones used most recently; More Symbols... offers about three hundred, searchable by name: type euro, dash, alpha or acute. Clicking one types it where the caret was, and the window stays open for the next.

• Can I resize a picture?

  Click it and the Picture Format tab appears. There you can set its width in pixels, put it in line with the text or on the left, centre or right with Position, give it Alt Text for a reader who cannot see it, or reset it to its own size. The Word file gets the same size and the alt text, but not the position.

Open Docs

• Where are files stored in a browser-based file manager?

  In this browser's storage on this device, using the storage the browser makes available to a page. Nothing is uploaded. The consequence is that another browser, another device, and a private window all see nothing at all.

• Can browser storage be lost?

  Yes. Clearing site data deletes it, some browsers evict storage belonging to sites you have not visited recently when disk space runs short, and reinstalling the browser or the operating system takes it with them. Treat it as working storage rather than backup, and export what matters.

• How much can I store in a browser?

  Browsers grant a page a share of free disk, often a substantial fraction of it, but the exact quota is the browser's decision and is not guaranteed. Large files work until they do not, which is another reason to keep an export somewhere else.

• Can I back it up?

  Yes, three ways, and they are for different things. The .json backup holds what you have made in Docs, Sheets, Slides, Forms and Notes. A folder of files goes out as a zip, built here in the tab, which is the thing to keep somewhere safe. And on Chrome, Edge or Brave on a computer, every file here can be copied straight out into a folder you choose, with each folder here as a folder in there. There is no account and nothing to pay.

• Does the drive keep itself in step with a folder on my disk?

  No, and it says so rather than pretending. Files you add live in this browser's own storage (IndexedDB) on this device. You can copy them all out to a folder, and you can add every file from a folder, but each is a copy made when you press the button: nothing is watched afterwards, and changing a file on one side does not change it on the other. Keeping two places in step means guessing which one wins when both changed, and a file manager that guesses is worse than one that copies when asked.

• Can I see what is in a file without downloading it?

  Yes. Open a file, with a double click or P, and it is shown here: a picture at the size it really is, a PDF drawn by the same reader the rest of the site uses, with Page Up and Page Down to turn its pages, or the first stretch of anything that is text. The arrows either side, and the Left and Right keys, go to the file before and after it. Sound and video play in the page too. Nothing is sent anywhere to make that happen: the bytes come out of this browser's storage and the reader is this site's own copy. A kind of file that cannot be shown on a page says so plainly instead of drawing nothing.

• Is a deleted file gone for good?

  Not straight away. Move to bin puts a file or a folder in the bin, where it keeps its bytes and the folder it came from, and Restore puts it back there. It is still taking up space while it sits there, which the Storage view says out loud. Emptying the bin, or deleting one thing forever, is the step that cannot be undone, and that is the one you are asked about.

• Can I get a whole folder out at once?

  Yes. Download on a folder, or on several things selected at once, puts them into one .zip, made here in the tab, with each folder a folder in it and two files of the same name kept apart rather than one landing on top of the other. The zip is stored rather than squeezed, because what people keep in a drive is mostly already compressed and squeezing it again would only cost you a frozen tab.

Open Drive

• Where are the notes kept?

  In this browser, on this device, until you connect a folder. Connect one and every note becomes a plain .md file inside it, with your folders as real subfolders, and that copy is the one that outlives the browser. Clearing site data cannot touch it.

• Can I use my existing folder of markdown files?

  Yes. Press the folder button and choose it. Every .md file in it and in its subfolders is read, and from then on edits are written straight back to those files. Nothing is converted and nothing is moved.

• What are wikilinks and backlinks?

  A wikilink is a note's name in double brackets, and it links to that note. A backlink is the same thing seen from the other end: the side panel lists every note that links to the one you are reading, along with mentions of its name that are not yet links, so you can turn them into links if you want.

• Is anything uploaded?

  No. There is no account and no server holding notes. Turn the network off and writing, linking, searching and the folder on your disk all still work, which is the simplest way to check it rather than believe it. The one thing that asks this site for anything is the zip export, which fetches a zip writer from obscuraos.com the first time you use it and says so plainly if it cannot; everything else, saving a single note as .md included, is built here in the tab.

• Does it work on a phone?

  Yes, with the panels sliding in over the editor rather than sitting beside it. Connecting a folder is a desktop feature, because phone browsers do not offer that permission; on a phone the notes stay in the browser and can be exported or carried across in the encrypted bundle in Settings.

• Can the site read my notes?

  The page's own code reads them in order to show them, and nothing leaves the browser. Anything running on the device can potentially reach browser storage, so a shared or compromised machine is still a shared or compromised machine, and genuinely secret material belongs behind encryption rather than in plain notes.

• What are the panels either side of the editor?

  Folders on the left, tabs across the middle, and on the right the outline of the note you are reading, every note that links to it, and its tags.

• What can I write in a note?

  Headings, lists, tables, quotes, code blocks with a language label, task boxes you can tick, footnotes, highlights, callouts and front matter. The source stays exactly as you typed it while it reads as formatted text.

• Can I link to a section, or pull one note into another?

  Yes. [[Note name]] links to a note, [[Note#Heading]] to a section of it, and ![[Note]] pulls one note into another. The names you already have are offered as soon as you type the brackets. Renaming a note rewrites the links that point at it.

• How do I find a note?

  Search across every note, with tag:, path: and file: to narrow it. There is also a tag panel, starred notes, a jump-to-note box on Ctrl+O, a command list on Ctrl+P, and a graph of how the notes connect.

• Does it have tabs, templates and daily notes?

  Yes. Tabs, a split view, daily notes, templates with {{date}} and {{title}}, and you can drag a note between folders.

• Can I get everything out again?

  Yes. Export the whole notebook as a zip of ordinary .md files, folders and all. It imports straight back with nothing lost, and because every note is a plain file nothing is kept in a format only this page can open.

• Does it sync between machines?

  No. There is no sync because there is no server, which is the trade for nothing being uploaded. To have the same notes on two machines, put the folder inside whatever already syncs for you. The Sync page on this site explains how.

• Can I put a picture in a note?

  Yes. Drop one into the editor, paste it, or use Picture on the toolbar. The note gets the ordinary markdown line every other markdown app writes, ![alt](attachments/name.png), and the picture itself is saved under attachments beside your notes: in the connected folder if you have one, in this browser if you have not. The note stays a text file you could read in any editor, rather than a megabyte of base64 in the middle of a sentence.

• What happens to the pictures when I export?

  They come with it. The zip export puts each picture under the same attachments path the markdown points at, so the export opens in any markdown app with its pictures rather than with holes in it. If a note points at a picture that is not here, the reading view says so in words instead of showing a broken icon.

• Can I print a note, or send one to somebody who does not want a text file?

  Yes, both. Print (Ctrl+Shift+P, or Cmd+Shift+P) puts the note on paper as the reading view draws it, pictures and all, with the app and the site around it left off. Save as Word writes a real .docx with the pictures in it at the size they are, which is the thing to send to somebody who would rather not be handed a markdown file. Both are built here in the tab.

Open Notes

• Is my photo uploaded?

  No. The picture is decoded in this tab and every edit happens in memory on your machine. The page works with the internet switched off once it has loaded.

• Does it keep layers?

  Yes. Save the project and you get one file that holds every layer, its position, opacity and blend mode. Open it again later and carry on. Export gives you a flat PNG, JPEG or WebP.

• Can it open HEIC from an iPhone?

  Yes. A decoder is fetched from this site the first time one is opened, about 1.4 MB, and kept on your device.

• Does it have layer masks and adjustment layers?

  Yes. Add a mask to any layer and paint it white to show and black to hide, as in Photoshop, or make it from a selection. An adjustment layer applies levels, curves, colour balance or any of the other adjustments to everything under it without touching a pixel, and you can double-click it later to change it. Both are saved in the project file.

• Can it remove an object?

  Select it and choose Remove what is selected, or use the patch tool to borrow a clean part of the picture and blend it over, or content-aware move to slide the thing elsewhere and fill the gap. The area is filled from its surroundings, first by spreading the colours inward and then by fitting small patches from just outside it. It works on wires, marks and small things against fairly even backgrounds. It does not understand what the thing was, and it does not invent anything.

• Can it open and save Photoshop files?

  Yes. A PSD opens with its layers, groups, masks, clipping masks, opacity, blend modes, guides and the common adjustment layers, and the editor saves a PSD back with the same. Text goes over as pixels and layer styles are not written, and the status line says so when it happens.

• Is it a Photoshop replacement?

  For most retouching, compositing, colour work and annotation, yes: layers in groups, masks, clipping masks, adjustment layers, selections, brushes, text and shapes that stay editable, pen paths, effects, guides and thirty adjustments and filters, with undo through all of it, a project file that keeps the layers and PSD files in and out. Its plugins are filters written in JavaScript, each run shut off from the page and the network; Photoshop’s .8bf plugins cannot run in a browser, and it says so rather than pretending. Camera raw is there, like the rest of the editor, as part of OBSCURA Pro.

• Does it have a histogram and channels?

  Yes. A histogram panel for the picture or the active layer, an info panel with the colour under the pointer, a navigator for panning and zooming, and a channels panel that shows red, green and blue on their own and keeps selections as alpha channels. A channel mixer, replace colour, match colour and .cube lookup tables are under the Image menu.

• Can it warp and liquify?

  Yes. Free transform has skew, distort, perspective and a sixteen-point warp as well as scale and rotate. Liquify pushes, twirls, puckers and bloats with a brush and can reconstruct what it moved. Content-aware scale takes the least noticeable pixels out so the picture gets narrower without the people in it getting thinner, and a perspective crop pulls a sign or a page photographed at an angle square.

• Can it merge exposures or a focus stack?

  Yes. Open the exposures or the focus slices as layers and choose exposure fusion or focus stack from the Layer menu; the best-exposed or sharpest parts of each are blended through Laplacian pyramids into a new layer, with no seams and no model.

• Can it make a GIF?

  Yes. Visible layers become frames, at a frame time you set, saved as an animated GIF or an animated PNG. Both are encoded in the tab; nothing is uploaded.

• What if the tab crashes?

  A few seconds after each change the open picture is kept in the browser's own database on this device, layers and all, and offered back on the next visit. Nothing is sent anywhere; closing a picture on purpose forgets its copy.

• Does it have smart objects and actions?

  Yes. A smart object keeps its source and is redrawn from it, so scaling loses nothing, and filters on it stay editable as smart filters. Actions record menu steps with their numbers and play them on a picture or on a batch of files. Several pictures can be open at once as tabs.

• Does it have one-click filters or looks?

  Yes, thirteen, under Image, Looks. Each is shown on your picture before you choose. A look is made of the editor's own adjustments with numbers already in them, so it goes on as a group of adjustment layers you can open, change, switch off or delete; or onto the pixels if you prefer.

• What is missing?

  More than eight bits a channel, Photoshop’s .8bf plugins, a photo library, editable text and layer styles in a saved PSD, and anything that needs a server, such as filling a picture from a prompt. For print it proofs on screen against a printer’s ICC profile, opens CMYK TIFF and PSD files and writes a CMYK TIFF, but it edits in RGB. The tools that are here are real: they work on the pixels the way their names say.

• When I remove something, what fills the hole?

  The picture itself. The spot heal brush, Remove what is selected, the patch tool and content-aware move all use PatchMatch with the iterative vote, which fills the hole with pieces of texture taken from elsewhere in the same picture and then matches the colour and the light at the seam. Spreading the colours inward from the edge, which is what most browser editors do, works on a plain wall and smears everything else.

• Is the healing quick enough to use?

  It runs coarse to fine in a Worker, so the page stays answerable and a long fill can be stopped with Escape. A hole eighty pixels across in a small photograph takes about a quarter of a second.

• Can it save a TIFF?

  Yes, plain or LZW, eight bits a channel, written here byte by byte because no browser can encode one. There is a sixteen bit PNG too, and the dialog is honest that every layer in this editor is eight bits a channel, so it is eight bit data in a sixteen bit container for a program that insists on one.

• What happens to the Exif in a photo I open?

  It is read from the bytes before the pixels go on the stage and shown in the Info panel, the position called out plainly, because that is the field people are surprised to find in a picture they are about to send. Exporting offers to strip it all, to keep the camera fields and drop the position, or to keep everything.

• Can I choose how a picture is resized?

  Image size offers nearest, bilinear, bicubic and Lanczos 3 as well as the browser’s own. Every one but nearest widens its window by the factor the picture shrinks by, which is what stops a shrunk photograph shimmering: a fine pattern is averaged rather than skipped.

• Can I see what a Levels move is throwing away?

  Hold Alt over the dialog. The picture becomes a map of what is being clipped at that end, in the colours of the channels that survive, which is the one view that shows clipping before it is permanent.

• Is it free?

  No. The whole Photo Editor, from the first layer to the last export, is US$4 a month or US$35 a year on its own, or part of OBSCURA Pro. Without Pro the page shows what the editor looks like and asks for a plan. Like the rest of the site it runs on your device and uploads nothing.

Open OBSCURA Photo Editor

• Can I edit a PDF without uploading it?

  Yes. The PDF is drawn and rewritten in this browser tab. Pages can be turned, moved, removed and added, the form filled, text and marks written on, a signature placed and areas blacked out, and the file saved to your disk. The page works with the network switched off.

• Are the marks real, or a layer on top?

  Both kinds are real, and each is what Acrobat would write. Text, ticks, pictures and a signature are drawn into the page's content when you save, and the form's answers are written into its fields. Comments - sticky notes, highlights, underlines, text boxes, drawings - are saved as annotations, the PDF's own comments, with your name, the time and any replies, so every PDF reader shows and lists them and they can be answered or taken away later.

• Does blacking something out really remove it?

  Yes. A page with a black box on it is written back as a picture, with the box burned in, so the words under it are not in the file. The whole page becomes a picture, which means its text can no longer be selected or searched. For a redaction with a certificate of what was removed, SEAL does the same job and keeps a record.

• Is the signature a legal digital signature?

  It is a picture of your signature, drawn or typed, placed on the page: a mark that you made, as a pen on paper is. It is not a certificate-based digital signature and nothing verifies it later. PACT signs a PDF with a record that can be checked.

• Can it open a PDF with a password?

  Yes, if you type the password; it is read here but cannot be changed, because the file is decrypted only for reading. It can also put a password on a PDF: File > Protect Using Password... locks the saved file with 256-bit AES, with a password to open it, a password to change the restrictions on printing, editing and copying, or both. Edit > Protection > Remove Security takes that off again before you save.

• What can it not do?

  It does not edit the text that is already on the page, only adds text, pictures and marks over it. It does not compress a PDF, convert one to Word, or read the words in a scan; OPTIC does the last. Text you add uses a standard font, so characters outside Western European alphabets come out as question marks.

Open PDF

• What is a critical path, and why does it matter?

  It is the longest chain of dependent tasks through the plan, and it sets the finish date. Every task on it has no float: lose a day on any of them and the project finishes a day later. Every task off it has slack, which you can spend without the end date moving. Knowing which is which is the difference between a plan and a list of tasks with dates on it.

• What is float?

  How long a task can be delayed before it starts pushing the finish date. A task with five days of float can start five working days late and cost nothing. The number is worked out for every task here, and the ones with none are drawn in red.

• What do FS, SS, FF and SF mean?

  The four ways one task can depend on another. Finish to start is the usual one: this starts when that finishes. Start to start means they begin together, or a set number of days apart. Finish to finish means they have to end together. Start to finish is rare and used mostly for handovers. Write them as 3SS+2, meaning two days after row three starts.

• Does it understand weekends?

  Yes. Durations are working days, so a five-day task starting on a Thursday finishes the following Wednesday rather than promising work on Saturday. Change Working Time on the Project tab sets which days of the week are worked, for a project that genuinely runs seven days a week, and takes holidays and shutdowns as exceptions.

• What is a baseline for?

  Taking a snapshot of the plan as agreed, so you can see slippage later. Without one, a plan that is edited as it slips always looks on time, because the dates move and nothing records that they moved. The Tracking Gantt draws the baseline under each bar, and the Variance table and Project Statistics say how many days behind or ahead each task and the finish now are.

• Can I open a plan from Microsoft Project or Asana?

  From Microsoft Project, through its XML format: save the project as XML there and open that file here, and Save As XML here takes it back. Its own .mpp files cannot be read. From Asana or anything else, export a CSV with a task name column and a duration column; dependencies come across if they are written as row numbers, and resource names come too. The plan also saves as its own JSON, and the chart as an SVG.

• Does it level resources?

  When you ask it to. Until then it only flags a person booked for more than they have, in the indicator column and on the Resource Sheet, because levelling changes the plan on your behalf. Level All on the Resource tab delays tasks, never shortens them, until nobody is overbooked; each delay is kept on its task, and Clear Levelling takes them all out again.

Open Plan

• Can I open an Excel file without uploading it?

  Yes. File, Open takes an .xlsx, a CSV or a tab-separated file, and an .xlsx opens as a workbook of its own: every sheet becomes a tab, with its values and formulas, fonts, fills, borders, alignment, number formats, merged cells, column widths and row heights, hidden rows, columns and sheets, frozen panes, notes, data validation, the filter, conditional formatting, defined names and the page setup. The engine here has more than a hundred and sixty of Excel's functions; a formula that uses one it does not have keeps the value Excel saved. The file is unzipped and read in this tab, and nothing is uploaded.

• Can it save as Excel?

  Yes. File, Save writes the workbook as an .xlsx on this computer, and where the browser allows it the next Save goes back into the same file. Save As and Export write it too, or the sheet in view as CSV or tab-separated text. The .xlsx keeps every sheet with its formulas and their values, the formatting, widths and heights, merged cells, frozen panes, notes, validation, the filter, conditional formatting, defined names and page setup, tables, threaded comments, sparklines, pictures and shapes, charts as Excel's own charts and PivotTables as Excel's own PivotTables, and opens in Excel, Numbers and LibreOffice.

• Can a spreadsheet have several sheets?

  Yes. The tabs along the bottom switch between them, as do Ctrl+Page Up and Ctrl+Page Down. The + beside them adds one, a double-click on a tab renames it, and a right-click on a tab inserts, deletes, renames, moves or copies, colours, hides and unhides sheets. A formula can name another sheet, as Excel writes it: =Costs!B4, or ='Notes & dates'!C1 when the name has a space, and renaming the sheet changes the formulas that name it.

• Can it make a chart?

  Yes. Select a block of cells and choose Column, Bar, Line or Pie on the Insert tab, or press Alt+F1. A first row of words names the series and a first column names the categories. The chart sits on the sheet, can be moved and resized, follows the cells as they change and is kept with the sheet; Save as PNG in its dialog gives a picture to put in a document. In an Excel file it goes as Excel's own chart over the same cells, not a picture, and comes back out as one.

• Is there a spreadsheet that works offline and without an account?

  A browser can run a spreadsheet engine locally, which is what this is. Formulas, ranges and CSV import and export all work with no network at all. What you do not get is collaboration or sync, because both of those require a server holding your data.

• Where does the data live?

  In this browser's storage on this device. It is not on a server, which is the point, and that also means it is not backed up by anyone. Export anything you would be sorry to lose.

• Can I sort the rows?

  Sort A to Z and Sort Z to A, on the Data tab and under Sort & Filter on the Home tab, sort the table around the active cell by its column, numbers as numbers and blanks last, keeping the first row as a header when it looks like one; select a block first to sort only that. Sort opens a dialog to choose the column and the order. Whole rows move with their formats, and a formula moves with its row as a copy of it would, so it still points at its own row. Undo puts the rows back.

• Is there find and replace?

  Ctrl+F opens Find, and Ctrl+H opens it with the cursor in the replace box; both are under Find & Select on the Home tab, beside Go To. Find next lands on each cell in turn, Replace changes that one, and Replace all changes every one and says how many. Match case and Whole cell narrow it.

• Can I freeze the header row?

  Freeze Panes on the View tab keeps the rows above the active cell and the columns to its left in place while the rest scrolls; Freeze Top Row and Freeze First Column hold the first row or column alone. They are kept with the sheet and go into an Excel file as a frozen pane, and a workbook that arrives with one opens with it.

• Can I colour cells by their value?

  Conditional Formatting on the Home tab colours cells by what they hold: greater than, less than, between or equal to a value, text that contains a word, duplicates, the top or bottom items or per cent, above or below the average, and data bars and colour scales. The rules follow the numbers as they change, are kept with the sheet, and go into an Excel file and come back out of one.

• Can I change a column's width?

  Drag the right edge of a column's letter, or double-click it to fit the widest entry in the column; Format on the Home tab sets a width in characters, as Excel counts them. Rows work the same way at the bottom edge of a row's number. Widths and heights are kept with the sheet and written into an Excel file.

• Can I print a sheet, or save it as a PDF?

  File, Print (Ctrl+P) sets the orientation, paper size and margins, whether gridlines and headings print, and whether it is the sheet, or its print area when Page Layout has set one, or only the selection. Print sends the cells and nothing else of the window to the printer; choose Save as PDF as the printer for a file.

• Does Sheets understand VLOOKUP?

  Yes, along with HLOOKUP, XLOOKUP, MATCH and INDEX. An exact lookup that finds nothing gives #N/A rather than the nearest row, and an approximate one takes the band at or below the value, which is what a rate table needs.

• Can I add up only the rows that match two conditions?

  SUMIFS, COUNTIFS, AVERAGEIFS, MAXIFS and MINIFS all take as many condition pairs as you give them. A condition can be a value, a comparison written as text such as ">1000", or a pattern with * and ?. Matching ignores case.

• What can Sheets do with dates?

  DATE, DATEVALUE, YEAR, MONTH, DAY, DAYS, EDATE, EOMONTH, WEEKDAY, WEEKNUM and DATEDIF, plus NETWORKDAYS and WORKDAY, which leave the weekends out and take a list of holidays if you have one. Dates are the serial numbers Excel uses, counted from 30 December 1899.

• Does TEXT understand format codes?

  It does. Date codes such as "dddd d mmmm yyyy" and "mmm yyyy", and number codes such as "#,##0.00", "0.0%" and a comma after the last digit place to scale by a thousand. An m between an hour and seconds is read as minutes, not months, the way Excel reads it.

• Are the financial functions there?

  PMT, FV, PV, NPER, RATE, IPMT, PPMT, NPV and IRR. They follow Excel’s sign convention, so money you pay out is negative and IPMT and PPMT add up to PMT. RATE and IRR are solved numerically and give #NUM! when no rate fits the numbers.

• How do I know the answers are right?

  A test drives the grid with about a hundred formulas whose answers were worked out by hand against Excel, then writes the workbook as .xlsx and opens it again to check the formulas survived. It runs before anything ships.

Open Sheets

• Can I build a presentation in the browser without signing in?

  Yes. Slides opens as a presentation program of its own, laid out the way PowerPoint is: the ribbon, the slide thumbnails, the notes and the status bar. The deck stays in this browser's storage on this device, so there is no account and nothing is uploaded, and it can be installed as an app from the browser's menu.

• How do I present from a browser?

  F5 starts the slide show from the beginning and Shift+F5 from the slide you are on, full screen, with transitions and animations as you set them. The arrow keys, Space and a click move it on; Esc ends it. Before presenting on someone else's machine, save the deck to a .pptx or export it as a web page rather than relying on their browser having your local storage, because it will not.

• Can I open and save PowerPoint files?

  Both ways. File, Open takes a .pptx and opens it as a presentation of its own: the words with their formatting, lists, pictures, shapes, tables, groups, backgrounds, transitions, hidden slides and speaker notes come in. File, Save writes a .pptx with all of those, editable in PowerPoint, Keynote and LibreOffice. PowerPoint's own layouts and charts do not come in, and animations stay in this browser's copy. Print gives a PDF that looks the same everywhere.

• Is there a speaker's view, with notes and a timer?

  Presenter View (Alt+F5, or Use Presenter View on the Slide Show tab) shows the slide with the next one beside it, the speaker notes in a size you can change, the time since the show began (click it to start from zero) and the clock. B blacks the screen for a pause and W whites it; any key brings the slide back. Home and End go to the ends, and a number then Enter goes to that slide.

• Can the audience see the slide on another screen while I see my notes?

  Yes. Second screen, in the show's controls, opens the same deck in a window of its own that draws only the slide. Drag it to the projector and click it for full screen. It follows this window slide for slide, the arrow keys in it steer this one back, and ending the show closes it. The deck is read from this browser's storage; nothing is sent anywhere.

• Can I put something anywhere on a slide?

  Yes, on any slide. The Insert tab puts a text box, a picture, a table or one of the shapes on it; drag it with the mouse, resize it by a handle, turn it by the round handle above it, and nudge it with the arrow keys, and its edges snap to the other things on the slide and to the slide's middle. Shape Format has the fill, outline, shadow and glow, and the order, alignment, grouping and size.

• Does a free canvas slide survive going to PowerPoint?

  Yes. Everything on a slide is held as a percentage of the slide, which is the one unit that means the same thing in the editor, in the show, on paper and in the EMU coordinates PowerPoint counts in, so it goes out exactly where you put it. Shapes are written as PowerPoint's own preset shapes rather than as pictures of them, with their outline, rotation and effects, and opening that file here brings them back.

• What happens to a PowerPoint slide that was laid out by hand?

  It comes in as a Blank slide with the boxes, shapes and pictures where they were, read against the file's own slide size, so a four by three deck is not stretched. A slide that uses PowerPoint's placeholders comes in with its title and text in the matching layout, and anything else on it stays where it was.

• Can I put a footer or page numbers on every slide?

  Insert, Header & Footer has all three that PowerPoint has: the date, fixed or today's, the slide number and a footer, with the choice to leave them off the title slide. They are drawn outside the layout, so nothing on a slide can sit on top of them, and they go into the printed PDF and the PowerPoint file as well.

Open Slides

• Is there a task board that does not need an account?

  Yes, for a personal board. Boards, lists, cards and dragging between them are ordinary browser work, and everything stays in this browser's storage on this device. Shared boards need a server, which is the line between this and a team tool.

• Where is the board stored?

  In this browser's storage on this device, with attached files in the browser's own file store beside it. Nothing is uploaded, nothing syncs, and clearing site data deletes it. Export a board if you want a copy that outlives the browser.

• Can I bring a board over from Trello?

  Yes. Export the board from Trello as JSON and import the file from this board's menu. Lists, cards, labels, dates, checklists, comments and covers come across; Trello's attachments come across as links, because the files themselves are on Trello's servers.

• Can I move my boards to another tool?

  Export a board as JSON or CSV from its menu and import it there. Keeping an export is worth doing regardless, because a browser's storage is not a backup.

• How do I use the same board on two machines?

  Export it as JSON on one and import it on the other. There is no account and no sync, and nothing is sent anywhere. The JSON names attached files but does not carry them, so attach those again on the other machine.

• When does a card count as overdue?

  When its due date has gone by and it is not marked complete. That is judged against this machine's clock and the calendar day where you are, not UTC, so a card due today stays due today until your own midnight.

• Can I load a CSV into the board?

  Yes, and it becomes a board of its own. The first row names the columns and only title is required; a list column puts each card on the list of that name. JSON is the exact copy, so use that when you want a board back precisely as it was.

• Can I drag cards with a finger?

  Yes. Hold a card for a moment and then drag it, between lists or up and down inside one; a quick swipe still scrolls the board. Lists drag by their titles the same way.

• Can a card remind me?

  Yes, a chosen number of days before it is due, at nine in the morning. It fires from an open tab of this site, never from a server, and the board says so where the reminder is set.

• What happens when I finish a repeating card?

  When its due date is marked complete, the next one is made in the board's first list, due one period later, with its checklists unticked and its reminder still on it. The finished one stops repeating, so it cannot breed copies of itself.

Open Tasks

• What is a good Cronbach's alpha?

  There is no line that makes a scale good. 0.70 is the figure most often quoted, from Nunnally, who also wrote that it depends on the purpose. Alpha rises with the number of items, so twenty mediocre items can beat five good ones, and a very high alpha (above about 0.95) often means items that ask the same question twice. Report the figure and the number of items, and look at the item statistics.

• What does a negative item-total correlation mean?

  That the item goes the opposite way to the rest of the scale. Almost always it is a reversed item ('I rarely feel calm' in a calm scale) that was not reverse-scored before alpha was worked out. Name it in the Reverse-score box and the page flips it, using the lowest and highest possible scores you give.

• What is alpha if item deleted?

  What alpha would be with that one item left out. If it is higher than the alpha for the whole scale, the item is pulling consistency down. That is a reason to look at the item's wording, not an automatic reason to drop it: a scale built by deleting items until alpha is highest tends to measure something narrower than intended.

• How are missing answers handled?

  Anybody with any item unanswered is left out of the whole calculation, which is what SPSS does by default (listwise deletion). The page says how many people that removed. If a lot are removed, the answer may not describe the people who were.

• What is the difference between alpha and standardised alpha?

  Alpha uses the items' variances and covariances as they are. Standardised alpha is worked out from the average correlation between items, as if every item had been put on the same scale first. They are close when the items have similar spreads; when they differ, report alpha unless the items are going to be standardised before they are added up.

Open ALPHA

• Why does my GIF look grainy or banded?

  A GIF can hold at most 256 colours in each frame, and a photograph has many thousands. The colours are chosen well for each frame, but smooth skies and skin can still show bands. Choose Animated PNG to keep every colour; it plays in every current browser, though some chat and email apps show only its first frame.

• How do I make the GIF smaller?

  Make it narrower, use fewer frames, or pick fewer colours. Width matters most: halving it makes the file about a quarter of the size. Many chat apps refuse animations over about 8 MB, and the page warns you when a file is over that.

• Can one frame stay on screen longer?

  Yes. Each frame has its own time under its picture. A common pattern is a short time for every frame and two or three seconds on the last, so it rests on the result before it loops.

• What is the shortest time a frame can show?

  20 milliseconds, 50 frames a second. Browsers slow down anything faster than that, and many show a frame given less than 20 ms for a tenth of a second, so the page does not go below it.

Open ANIMATE

• Why is the name so large?

  Because that is the entire job of a badge. It is read at two or three metres by somebody who has met you once and cannot remember your name, and a name set in eleven point under a logo is unreadable at that distance. The organisation already knows who it is; the badge is for everybody else.

• What is the colour band for?

  Marking a group (staff, visitor, speaker, first aider) in a way that is read from much further away than any word. It is the one piece of information that works across a room.

• Does the name shrink if it is long?

  Only as far as it must. It steps down a point at a time until it fits the badge width, so a long name is smaller than a short one but a sheet of badges still looks consistent.

• Does the list go anywhere?

  No. It is a list of people attending something, and there is no network call on this page.

Open BADGE

• Why does it change my check digit?

  It does not; it tells you. EAN, UPC and similar symbologies end in a digit computed from the ones before it, and a scanner refuses a symbol where it does not agree. If the digit you typed is wrong, the barcode drawn carries the correct one and the page says so, because the alternative is a label that looks perfect and scans nowhere.

• What is the module width?

  The width of the narrowest bar, and the dimension that decides whether the symbol reads. Below roughly a quarter of a millimetre, ordinary handheld scanners start refusing and a laser printer stops holding the edge cleanly. This is why a barcode scaled to fit a box on a label often stops working: the scaling changed the module.

• Why should I use the SVG rather than the PNG?

  Because it is resolution-independent, so it prints at whatever your printer can do rather than at whatever the bitmap was made at. The PNG here is rendered at 600 dots per inch for the places that insist on one, but a barcode from a screen-resolution bitmap loses the edge of every bar.

• Why is Code 128 sometimes much narrower?

  Because it has a character set that encodes two digits in one symbol. A run of digits long enough to be worth the switch is packed that way automatically, so a sixteen-digit number is close to half the width it would otherwise be. The page says which sets it used.

• Does my data go anywhere?

  No. Asset numbers, batch numbers and patient identifiers all end up in barcodes, and there is no network call on this page.

Open BARCODE

• How is this different from SHEETS?

  A spreadsheet is a grid of cells where anything can be anything and the meaning lives in the formulas. A database is a set of records with a stated shape, where the meaning lives in the columns. If you are modelling a calculation, you want the grid. If you are keeping a list of things of the same kind, and you keep wishing the grid would stop letting you type nonsense into it, you want this.

• What types can a column be?

  Text, long text, number with a set number of decimal places, date, checkbox, single select with defined options and colours, multi select, URL, and a link to a row in another table. You can change a column's type at any time under Columns, sorting and filters; the values you already have are re-read as the new type and anything that cannot be read stays as it was typed.

• Can I edit in the grid or do I have to open a row?

  Both. Every cell in the grid is editable where it sits, including the selects and the checkboxes, and the tags in a multi select are added and removed from the cell itself. The side panel gives the same row one field at a time with room to write, which is where the long text fields are actually usable.

• What does the export contain?

  Exactly what the view in front of you shows: those columns, those rows, in that order, after the filters and the search. The line under the export buttons says how many of each, so there is never a question of whether a filter was still on. The JSON also carries the type of every column and the options of every select.

• What happens if I delete something?

  You are asked first, and told what will go: how many rows and columns a table holds, how many filled values a column holds, and how many links from other tables point at the row or the table you are about to remove. Those links are then emptied rather than left pointing at nothing. There is no undo, so the export is the safety net.

• A row I have just edited still shows, though it no longer matches the filter. Why?

  Because it would be worse the other way. The browser tells a page that a cell has been finished with at the moment you click the next one, so re-running the filter there would delete the row, and with it the cell you were aiming at, half a keystroke before your typing arrived in it. Rows therefore stay where they are while you work and the view re-asks its question when you next change the search, the sort, the filters or the view. The figures at the foot of each number column follow every keystroke regardless, so the totals are never behind what you can see.

• Why can it not open my Excel file?

  Because reading the xlsx format properly means a library, and every library this site serves has to be vendored and pinned by hand. Save the sheet as CSV, which every spreadsheet does from its own save menu, and this reads it and works out the types. CLEANSHEET is the tool for looking inside an xlsx before you share it.

• Where is the data, and how much can it hold?

  In this browser's local storage, under one key, on this device, and nowhere else. Browsers give a site a few megabytes of that, which in practice is thousands of rows rather than millions. Export the JSON if you want a copy that outlives the browser.

• Can two people share a table?

  No, and that is deliberate: sharing needs a server, a server needs accounts, and the whole point of this site is that it asks nobody for anything. Export the CSV or the JSON and send it the way you would send any other file.

• How does the CSV import decide what each column is?

  Every column is read and its type guessed, and the guess is shown with the reason for it under "What each column was taken to be". You can change any guess before the table is created. Nothing is committed until you have looked and pressed Create the table.

• Why does a column need a type?

  A grid will let you type anything anywhere, which is why so much of the world's data is a date column with the word soon in it. Here a table is a set of records with the same shape: a date column holds dates, a number column holds numbers to a stated number of decimal places, and the grid can say no to anything else.

Open BASE

• Can it convert a book I bought from Kindle, Kobo or Apple Books?

  Only if the book has no DRM. A book sold with DRM is encrypted so that only the seller's app can open it, and this refuses it and says so rather than trying. Many publishers and shops sell DRM-free EPUBs, and those convert.

• Does it keep the book's own design?

  It keeps the order, the chapters, the headings, emphasis, lists, tables and pictures, and sets the text in a plain book style in DejaVu Sans, the way an e-reader shows a book with the publisher's styles turned off. The book's own fonts and style sheets are not used.

• Why start every chapter on a new page?

  Because that is how a printed book reads, and because an EPUB often splits a chapter into several files. Untick the box to run the chapters on without a break.

• What about a fixed-layout book, like a picture book or a comic?

  It is reflowed onto the paper you choose, which changes how it looks, and the page says so when it opens one. The pictures are all there, one after another.

• Is the book uploaded?

  No. It is unzipped and read in your browser, and the PDF is written there too, with every library loaded from this site.

Open BINDERY

• Will the form work in Acrobat, Preview and a browser?

  Yes. The fields are written as the standard kind of PDF form field (AcroForm) with their appearance drawn, which is what Acrobat, Reader, macOS Preview, Chrome, Edge, Firefox and the PDF editor on this site fill in.

• How do radio buttons work?

  Radio buttons with the same set name are one question: choosing one clears the others. Give each button in the set its own name, which is the answer recorded when it is chosen.

• Can I place fields without a mouse?

  Yes. Add a field with the button and it appears in the middle of the page; then type its position and size in millimetres, measured from the top left of the page. Every field is also listed with buttons to edit or remove it.

• What about the form fields the PDF already has?

  They are kept as they are. A new field cannot share a name with one already there; the page says so if you try.

• Is the PDF uploaded?

  No. It is drawn by pdf.js and the fields are written by pdf-lib, both loaded from this site, in your browser tab.

Open BLANKS

• Can I make an EPUB from a Word document without uploading it?

  Yes. Drop the .docx here: its headings become chapter breaks, its bold, italic, lists, tables and pictures come across, and the EPUB is written in the tab and saved to your disk. Nothing leaves the browser.

• How are the chapters decided?

  At the headings. If the document has several Heading 1s, each starts a chapter; if not, Heading 2s; failing that, horizontal rules; failing that, it is one chapter. The choice is shown and can be changed, and every chapter title can be edited before the book is written.

• Does it open on a Kindle, a Kobo or in Apple Books?

  It is a standard EPUB 3 with the EPUB 2 contents table as well, which Kobo, Apple Books, Google Play Books, calibre and every reader app open. A Kindle takes an EPUB by Send to Kindle, or calibre converts it. READER on this site opens it too.

• Can it make a cover?

  Yes. Choose a picture, which is resized and stored as a JPEG, or have one drawn from the title and author. A book without a cover is still a valid book; readers show the title.

• What about pictures in the document?

  Pictures inside a Word file come across and are stored in the book. A picture that points at a web address is left out and counted, because a reader offline would show a broken box for it.

• Can I go the other way, from an EPUB to text?

  READER opens an EPUB and can save its text as markdown, and PACK opens the zip itself to take any file out.

• Do the pictures in my Word document end up in the EPUB?

  Yes, each one as a file of its own inside the book, named in the package the way a reader expects, byte for byte what was in the .docx rather than re-encoded on the way through. The words that describe a picture come with it, so somebody reading with their ears is told what is there. A picture that points at a web address rather than being carried in the document is left out and said so, because fetching it would tell that server you are making a book.

Open BOOK

• What is the rule?

  On sheet i of a booklet of n pages, the front carries page n−1−2i on the left and page 2i on the right; the back carries 2i+1 on the left and n−2−2i on the right. For eight pages that gives 8 and 1, then 2 and 7, then 6 and 3, then 4 and 5. The first two sheets are written out on the page so you can see that it is right.

• Why does it add blank pages?

  Because a folded sheet carries four pages, so a booklet is always a multiple of four. A nine-page document becomes twelve, with three blanks placed where they belong rather than at the end.

• What is a signature?

  A group of sheets folded together. A hundred-page document folded as a single booklet has a spine that will not close and pages that creep badly at the fore edge. Splitting it into signatures of four or eight sheets, folded separately and then stacked, is how books are actually made.

• Short edge or long edge?

  Short. Flipping on the long edge turns every second sheet upside down, and that is the other half of why home-printed booklets come out wrong.

• Does the file leave my machine?

  No. It is read into memory, rearranged and written back out in this tab. A document about to be made into a booklet for a meeting is usually one you would rather not upload.

Open BOOKLET

• What is operating leverage, in plain terms?

  Contribution divided by profit. If it is four, then every one per cent that volume moves, profit moves four per cent, up or down. A business with high fixed costs and a good margin has high leverage and is wonderful right up until the month sales dip. It is an identity, not an estimate, and it is the number people's intuition gets furthest wrong.

• Why does the sales mix change the answer?

  Because break-even is fixed costs divided by what a sale contributes, and different products contribute different amounts. Averaging the margins across products answers a question about an imaginary average product nobody buys. Treating the mix as a bundle and dividing by what a bundle contributes answers the real one, and it moves when you sell more of the good stuff.

• Why is price nearly always the biggest lever?

  Because a price rise carries no cost with it. Sell one per cent more and you pay to make one per cent more; charge one per cent more and the whole of it lands in contribution. The table moves all four levers by the same amount so you can see the size of the difference rather than assume it.

• What is the discount figure?

  The across-the-board price cut that would take the entire profit at current volume. It is usually much smaller than people expect, because a discount comes out of contribution rather than out of price: ten per cent off a product with a forty per cent margin is a quarter of the profit on it.

• Do my costs go anywhere?

  No. There is no network call on this page. A cost base is not something to type into somebody else's server.

Open BREAKEVEN

• Why not just write it in a document?

  Because a document has no opinion about what is missing, and what is missing from a briefing is always the same five things. A blank page will happily let you give a briefing with no time in it.

• What is the card for?

  Nobody reads eleven headings standing up. The card is the intention and the method and nothing else, and it falls out of the full brief automatically so the two cannot drift apart, which is what happens the moment somebody maintains them separately.

• Is IIMARCH better than SMEAC?

  No. They are the same discipline for different work. IIMARCH ends with risk, communications and proportionality, which suits a police operation that will be reviewed. SMEAC front-loads the situation and the mission, which suits a group about to move. Pick the one your organisation already knows.

• Does it check whether the briefing is any good?

  It cannot, and it does not pretend to. It checks for absence, which is mechanical, and leaves quality to you. Read it out loud once before you give it: that finds more than any tool.

Open BRIEF

• Why does 0.1 + 0.2 come out wrong everywhere else?

  Because computers hold fractions in binary, and 0.1 in binary is a recurring fraction in the same way a third is in decimal. It gets cut off at 53 bits, so the stored number is very slightly more than a tenth, and adding two of those shows the error at the seventeenth digit. Money arithmetic done that way drifts by a penny over enough lines and nobody can see where the penny went. This page never converts your figures to binary at all.

• How do I use one line's answer in another?

  Name it. Write subtotal = 120 * 3 and then use subtotal anywhere below that line. Names are one word of letters, digits and underscores. There is also prev, which is whatever the line directly above worked out, for when naming it would be more ceremony than it is worth. A name is allowed to be the same word as a unit, because a line called days or hours is a far more ordinary thing to want than the word days spelled out as a unit; the name wins from that line down, and any line that then tries to use it as a unit says so.

• What goes into the running total?

  Every line that is a plain number and has not been given a name. That is the rule that makes a tape usable: your working lines get names and stay out of the total, and the lines you actually want added up are the bare ones. A line with a unit on it is left out too, since adding kilograms to gigabytes would be a nonsense.

• What does 120 + 15% actually do?

  It gives 138. The percentage is taken of the number on the left, which is what a person means when they say add fifteen percent, and what a pocket calculator with a % key does. 120 - 15% is 102 for the same reason. On its own, 15% is 0.15.

• Why will it not add 20 C to 30 C?

  Because a temperature in Celsius or Fahrenheit is a point on a scale rather than an amount of something, so adding two of them is meaningless in a way that adding two lengths is not. Twenty degrees plus thirty degrees is not fifty degrees of anything. Conversion between the three scales works normally.

• Is anything sent anywhere?

  No. The page loads its own script and nothing else, works with the network switched off, and keeps the tape in this browser's local storage under one key. Clear the tape, or clear the site's data in your browser, and it is gone.

• Can I keep a tape for each job?

  Not in one page: there is one tape, kept where you left it. Copy a finished tape out with the copy button and paste it wherever the working belongs, which is usually the thing you were going to have to do anyway.

• How do I move between lines on the tape?

  Enter starts a new line. The up and down arrows move between lines, and backspace on an empty line takes it away. Every line is an expression you can go back to: change one and everything below it works itself out again.

• Which units does it know?

  Length: mm cm m km in ft yd mi nmi um nm. Mass: mg g kg t oz lb st ton longton. Volume: ml cl dl l m3 cm3 tsp tbsp floz cup pt qt gal impfloz imppt impgal. Temperature: C F K. Data: bit B kB MB GB TB PB KiB MiB GiB TiB kbit Mbit Gbit. Time: ms s min h d wk mo yr.

  Long spellings work too: inches, kilograms, minutes. The word to converts whatever the line has worked out so far, as in 2 m + 30 cm to mm.

• How can I write a number?

  As 1,234.50 or as 1.2345e3. A minus sign, a multiplication sign or a space pasted out of a document is read as the character you meant. A comma has to group in threes: 1,5 is refused rather than read as fifteen.

• Can I put a note on the tape?

  Yes. A line beginning with # is a note. It stays on the tape and is not worked out.

Open CALC

• Is there a whiteboard that works without an account?

  Yes, for working alone. Drawing, arranging and panning an infinite canvas are all ordinary browser capabilities. What needs a server is other people seeing it at the same time, which is why every collaborative whiteboard asks you to sign in.

• Can I share a drawing with someone?

  Export it as an image or a file and send that. There is no live link, because a live link means a server holding the board. For a diagram that ends up inside a document, the export was what you wanted anyway.

• Will my board still be there tomorrow?

  In the same browser on the same device, yes. Clearing site data removes it, and it does not follow you to another machine. Export anything you would be sorry to lose.

• How do I move around the board?

  Scroll or pinch to zoom. To pan, hold Space and drag, use two fingers on a touch screen, or pick the Pan tool. Recenter brings you back to the middle.

• How do I select, move and resize several things at once?

  With the Select tool, drag a box round the things you want, and shift-click to add one more. Pull the blue squares on the selection to resize it. Draw with the pen, double click anywhere for a sticky note, and connect ideas with arrows.

• Can I use the whiteboard without a mouse?

  Yes. Click the board once, then Tab steps through what is on it. The arrow keys move the selection, and holding Shift moves it further. + and - resize it. Enter opens a note for typing or drops a new one in the middle, and Delete removes it.

• How do I keep a copy of the board?

  Everything stays on this device and autosaves in your browser. Use Save to keep a copy as a file, Open to load it again later, or PNG to share a picture of the board.

Open Canvas

• Why take percentages of strength?

  Because a vacant post does not take annual leave, go sick or attend a course. Applying an abstraction rate to establishment counts absence for people who do not exist, and the answer comes out higher than the truth, which is exactly the direction nobody notices.

• Why is filling one post not worth one person?

  Because the new post is abstracted at the same rate as every other one. At a thirty per cent abstraction rate, putting one more body on every shift takes roughly one and a half posts per shift, not one. That is the figure that makes recruitment arguments go differently.

• Is this a workforce model?

  No. It is one page of arithmetic that anybody could do and almost nobody does, because the five numbers it needs live in five different places. Putting them in one column is the whole trick.

• Can I use rates and counts together?

  Yes, and you usually have to. Leave and sickness are naturally rates; three people on restricted duties is a count. Mixing them in one column is fine as long as it is visible which is which, and it is.

Open CAPACITY

• How many cards should I review a day?

  Whatever is due, which the scheduler keeps small: a card you know well comes back in weeks or months. New cards are the cost; twenty a day is the default and a good ceiling, because every new card is a review you owe later. Each deck keeps its own number, next to the Review button, so a deck you are in a hurry with and a deck you are keeping ticking over do not have to be paced the same. Setting it to nothing stops new cards without stopping the reviews.

• What do Hard, Good and Easy mean?

  How it went. Good is the normal answer: you recalled it. Hard: you got there slowly, so the card's ease drops and every interval after this one is shorter. Easy: instant, so the ease rises and this interval is stretched as well. Again: you did not, so the card starts over and comes back in this session. Hard and Good set the same next interval, which is what SM-2 does; the difference between them shows from the following review on.

• Can I move a deck to another device?

  Export it as JSON, which carries the schedule, and import it there. The tab-separated export carries only the cards, for a spreadsheet or for Anki.

• Where is the deck stored?

  In this browser on this device, and in your folder as a file if you have connected one. Clearing this site's data removes it; the export is the backup.

• Can I take a piece out of a sentence rather than write a question?

  Yes. Write the sentence in the front box with the piece in double braces, The capital of {{c1::Peru}} is {{c2::Lima}}., and you get one card for each number: each asks about its own piece with the rest of the sentence left to read, which is how a fact learnt in context is learnt. A second pair of colons puts a hint in the blank, {{c1::Ecuador::which country}}. Editing one of those cards edits the sentence, and the blank is worked out from it again.

• Can I be asked the other way round as well?

  Tick the box under the front and back, and adding the card makes the pair both ways round. Recognising a word and producing it are two different things to know, and the two cards are scheduled apart, because one of them is usually harder than the other. The box applies to the paste box too; a file you import is taken as it is, so importing five hundred cards cannot quietly become a thousand.

Open Cards

• Why will it not let me cut the axis on a bar chart?

  Because the length of a bar is the quantity. A bar chart with a cut axis is not a chart with a caveat, it is a picture of a different number, and it is the single most common way a true table becomes a misleading graphic. To show a small difference between large values honestly, use a line; this will draw one, and its axis does not have to include zero.

• Why SVG rather than PNG?

  Because an SVG scales to any size, has real text in it that can be searched and corrected, and opens in almost anything for editing. A PNG is offered too, rendered from the same SVG on your device, for the places that still insist on one.

• Are the colours arbitrary?

  No. They are a set chosen to stay distinguishable under the commonest kinds of colour blindness, ordered so the first two are the pair that survives best. About one man in twelve cannot tell a certain red from a certain green, and a chart is a thing other people have to read.

• Does my data go anywhere?

  No. The picture is drawn by this page, in this tab, and the file is made in your browser. There is no network call, which is the whole reason this exists, because a column of figures is very often about people or money.

Open CHART

• How do I check whether a card number is valid?

  Its last digit is a Luhn check digit: double every second digit from the right, add the digits of the results and the rest, and the total ends in 0. 4111 1111 1111 1111 passes; change any one digit and it fails. Passing says the number was not mistyped, not that a card with that number exists.

• How is an IBAN checked?

  Move the first four characters to the end, turn letters into numbers (A is 10, B is 11 and so on to Z at 35), and the whole number divided by 97 leaves 1. The length must also be the one the SWIFT registry gives for the country: 22 for the UK and Germany, 27 for France. GB82 WEST 1234 5698 7654 32 is the usual example, and it passes.

• Does a valid check digit mean the number is real?

  No. It means the number is internally consistent, which catches almost every single-digit typing mistake. Whether a card was issued, an account is open, a VAT number is registered or a book exists needs the issuer's own service, and nothing on this page asks anyone.

• Why does a European VIN fail?

  The check digit in position 9 is required in North America, and vehicles built for other markets often use that position for something else. A VIN that fails the check may still be genuine if the vehicle was not built for the US or Canada; the page says so beside the answer.

• Can I check a whole column of numbers?

  Yes. Paste them one a line into the second box and press Check them all. Each line is read as whatever it looks like, or as the kind chosen above, and the results can be saved as a CSV.

Open CHECKDIGIT

• What does the digest prove?

  That this record has not been altered since you sealed it, and nothing else. It does not prove the run happened, and a digest computed and kept in the same place as the record proves very little on its own. Write it down somewhere outside this tool, and then it starts being worth something.

• Why does skipping need a reason?

  Because a checklist where not-applicable and done look identical is a checklist that hides the one item somebody decided not to do, which is the item that matters. A skipped item stays visibly skipped in the record, with the reason next to it.

• Can I keep several checklists?

  One at a time on this page. Starting a fresh run keeps the items and clears every tick, time and reason, which is the ordinary case. For a different list, paste it in.

• Does it work offline?

  Entirely. There is no network call on this page at all, and the digest is computed by the browser itself.

Open CHECKLIST

• What does the warning about expected counts below 5 mean?

  The chi-square test's p-value comes from an approximation that is good when every cell's expected count is reasonably large. The usual rule, from Cochran, is that no expected count should be below 1 and no more than a fifth below 5. When a table breaks it, the page says how many cells do; on a two-by-two table use Fisher's exact test, which needs no approximation and is shown beside it.

• Should I use Yates's correction?

  On a two-by-two table it makes the test more conservative, often more than it needs to. Many statisticians now prefer the uncorrected Pearson test with large counts and Fisher's exact test with small ones. Both corrected and uncorrected figures are shown so you can report the one your field expects.

• What is Cramér's V?

  A measure of how strong the association is, from 0 (none) to 1 (each row predicts the column exactly), that does not grow with the sample size the way chi-square does. On a two-by-two table it equals the absolute value of phi. A significant chi-square with a small V means a real but weak association, which is common with large samples.

• Can I paste raw data instead of a table?

  Yes. Choose 'Two columns of categories' and paste two columns, one row per person or case. The page counts them into a table and runs the test on that. A row with either value blank is left out, and the count says how many were.

• What is the goodness-of-fit test for?

  Whether counts in one set of categories match the shares you expected: a die that should be fair, survey responses against census proportions. Type the counts and, if the expected shares are not equal, a second column of proportions or counts. The shares are scaled to the total, so 25, 25, 50 and 0.25, 0.25, 0.5 mean the same.

Open CHI-SQUARE

• How do I stop faces being cut off?

  Each picture is cropped from the middle when it fills its space. If that cuts something off, choose Shown whole, which fits every picture in its space on the background colour, or pick a layout whose spaces are closer to your pictures' shape: side by side for tall pictures, one above another for wide ones.

• What size should I make it?

  Square or portrait 4:5 for a social post, wide 16:9 for a slide or a screen, tall 9:16 for a story, and A4 at 300 dpi for printing on a page.

• Can the background be see-through?

  Yes: tick the box and save as PNG. A JPEG cannot be see-through, so the gaps come out white in a JPEG.

Open COLLAGE

• Why does a PDF keep its properties twice?

  The document information dictionary is the original place, from the first versions of PDF. The XMP packet is a block of XML added later so that the same details could be read by any program that reads XMP, including photo and asset managers. Most PDFs from Word, Acrobat or a scanner carry both. A property editor that changes only one leaves the old author sitting in the other, so this replaces the XMP packet with one that says exactly what the list says.

• Does clearing the properties remove everything that identifies the file?

  It removes the properties and the XMP packet. It does not touch what is on the pages, the fonts' names, the file's ID, comments, form fields or hidden layers. For a wider clean-up before sending a document, use SCRUB for Office files, FLATTEN for forms and SEAL to redact the pages themselves.

• What happens to earlier versions saved inside the file?

  They go. A PDF that was edited and saved by appending, which is how Acrobat saves by default, still holds every earlier version after the last one; STRATA lists them. This writes the file out whole, so only the current version, with its new properties, is in it.

• Can it change a password-protected PDF?

  Not while the protection is on. Take the password off with PADLOCK first if you know it, then change the properties here.

• Is the PDF uploaded?

  No. It is read and rewritten by pdf-lib, loaded from this site, in your browser tab.

Open COLOPHON

• How does it find the table?

  The way you would. Words that sit on the same line are a row. Where the page draws the table's vertical lines, those are the column edges; otherwise the columns are the bands of the page that the words of every tabular row share and nothing crosses. The preview shows the result before you save it.

• Will the numbers add up in Excel?

  Yes. 1,250.00, £45.00, (320.00) and 12.5% are stored as the numbers 1250, 45, -320 and 0.125, with a number format that shows them as they were printed. Tick the box to keep everything as text instead, for account numbers and codes that must keep their leading zeros.

• What will not come out right?

  Cells that span several columns, text wrapped onto a second line inside a cell (it becomes a row of its own), and pages laid out in several columns of prose. It also needs the PDF to have words in it: a scanned page has none, and OPTIC can give it a text layer first.

• Can I have everything on one sheet?

  Yes, tick the box and every page's rows are put one after another on a single sheet, with a blank row between pages. Otherwise each page gets its own sheet, named for the page.

• Is the PDF uploaded?

  No. It is read by pdf.js, loaded from this site, and the spreadsheet is written in your browser tab.

Open COLUMNS

• How do I combine several Excel files into one sheet?

  Drop them all here, keep One sheet, every row stacked, and save. The columns are lined up by their headings, so a file whose columns are in a different order, or that has an extra column, still lands in the right place, and a Source column says which file each row came from.

• How do I merge several workbooks into one with a tab for each?

  Choose One workbook, each sheet as its own tab. Every sheet of every file becomes a tab, named after the file (and the sheet, where a file has more than one), with its formulas and their last results.

• Does it keep the formatting?

  No. It carries the data: text, numbers, true and false, dates as dates, and formulas. Fonts, colours, borders, column widths, charts, pictures, comments and macros are not carried, and the page says so before you combine.

• What happens to formulas when rows are stacked?

  Each formula's result, as Excel last saved it, is used. The formula itself would refer to cells that are somewhere else in the combined sheet. A formula that was never calculated in its own file has no saved result, and is written as text starting with =, counted in the message.

• Can it read .xls files?

  No, only .xlsx and .csv. Open the .xls in Excel or LibreOffice and save it as .xlsx first.

Open COMBINE

• Why are gallons listed twice?

  Because they are two different volumes. An imperial gallon is 4.54609 litres and a US gallon is 3.785, a seventeen per cent difference, and the word on both sides of the Atlantic is “gallon”. A converter that picks one for you is guessing about the thing that matters.

• Is the fluid ounce the same way round?

  No, and that is why it catches people. The imperial fluid ounce is smaller than the US one, which is the opposite direction to the gallon. An imperial pint is twenty imperial fluid ounces and a US pint is sixteen US ones.

• Which factors are exact?

  The ones with exact definitions: the inch is exactly 25.4 mm by international agreement, the pound is exactly 0.45359237 kg, the nautical mile is exactly 1852 m, the imperial gallon is exactly 4.54609 litres. Where a factor is approximate (the speed of sound, for instance), the page says so rather than offering twelve digits of false precision.

• Why two kinds of mil?

  Because the NATO mil divides a circle into 6400 and the Soviet one into 6000. They are both called a mil, they are used for the same job, and they differ by about seven per cent, which over a few kilometres is a long way.

• What about kilobytes?

  A kilobyte is 1000 bytes and a kibibyte is 1024. Storage is sold in the first and reported by most operating systems in the second, which is the entire explanation for the space that appears to be missing from a new drive. Both are here, labelled.

Open CONVERT

• What is a CSR?

  A certificate signing request: your public key, the names the certificate is for, and a signature made with the matching private key, which proves the request came from whoever holds it. You paste it into the certificate authority, and it sends back a certificate. The private key never goes to the authority.

• Why put the host name in the alternative names too?

  Browsers have ignored the common name since 2017 and look only at the subject alternative names, so a certificate whose host name is only in the common name is refused. The page adds the common name to the alternative names if you did not, and lists every name the certificate will cover.

• ECDSA or RSA?

  ECDSA P-256 is smaller and faster and every current browser and authority accepts it. Use RSA 2048 or 3072 where an older system or an authority's form asks for RSA.

• Should the private key have a passphrase?

  For a key that is kept or moved around, yes. A web server has to be given the passphrase every time it starts, so server keys are usually left without one and protected by who can read the file: chmod 600, readable by the server's own account only.

• What is a self-signed certificate for?

  A test server, a device on your own network, or a development machine. Nobody vouches for it but itself, so browsers warn about it until you trust it on each device. For a public website use a certificate authority; many, such as Let's Encrypt, are free.

Open CSR

• Is this CV builder free, and does it need an account?

  No account, and the PDF, through the browser's own print, is free. Saving the Word file, the plain text or the CV's own file is part of OBSCURA Pro. The form, the page and the files are all made in this browser tab. There is no sign-up, no watermark, no paid layout and no upload; the CV is kept in this browser and written to your disk when you ask.

• How do I get the CV as a PDF?

  Print / PDF opens the browser's print dialogue; choose Save as PDF as the printer. The page on the right is laid out at paper size, so the PDF is what you see, with the layout's colours kept. A4 and Letter are both there in the tool row.

• Can I save it as a Word file?

  Yes. Save to computer, or Export and Word file, writes a .docx in one column: the name, the contact line, each section as a heading, entries with their dates, and lists as lists. It opens and is editable in Word, LibreOffice and Pages. The two-column layout is for paper; Word gets the words and lays them out its own way.

• Can I open my old CV from Word?

  Open takes a .docx. The first heading becomes the name, the lines above the first section heading become the contact details and the summary, each heading after that becomes a section, and under Experience and Education a paragraph that opens in bold starts an entry. It is a best reading of a document that was never a form, so check the entries and dates against the original. A .cv.json written here opens exactly as it was.

• Which layout should I use?

  Five to choose from, with the same words in all of them. Classic for most jobs: centred name, serif headings, one column. Modern puts a coloured band across the top and the short things in a side column. Compact is smaller type and tighter spacing for a CV that has to fit one page. Sidebar runs a column the full height of the page with your name, contact details, skills and languages in it, and the work beside it. Plain is for the machine that reads a CV before a person does: no colour, no rules, one column, no photograph, nothing that turns into gibberish when it is scraped out of the PDF. Switch between them and see.

• Will it fit on one page?

  The line under the tool row says how many pages the CV runs to on the paper chosen, and the page itself shows a faint line where each sheet ends. Compact is the layout for squeezing; after that, shorter bullets.

• Where is the CV kept?

  In this browser's storage on this device, saved as you type. Another browser or device does not have it; take it there as a .cv.json (Export), which opens here again exactly as it was, or as a Word file. Clearing the site's data removes it, so keep a file of a finished CV.

• Should I put my address on it?

  A town is enough; the form asks for a location, not an address. A CV goes through recruiters, agencies and whoever they forward it to, and a street address gives none of them anything they need. For a CV that has to be judged blind, BLIND takes the name and the rest out.

• Can I put a photograph on it?

  Yes, and the page says where that helps and where it does not: a portrait is usual on a CV in much of Europe and is a liability in Britain, Ireland, the United States, Canada and Australia, where employers often ask for CVs without one to keep hiring blind. Choose a PNG or a JPEG and it is squared off from the middle and drawn down to 320 pixels, which is enough for the 28 millimetres a printed CV gives it and small enough to live in this browser and inside a Word file. It goes on the page and into the .docx, and the plain layout leaves it off on purpose.

• Can I move the sections around?

  Yes. Every section has an up and a down beside its title, and so does every entry inside it, so experience can come before education or the other way round, and the job you want to be read first can be first. The order is the order the PDF, the Word file and the plain text all come out in.

Open CV

• When is a data URI a good idea?

  For small things used once: an icon in a stylesheet, a logo in an email signature, a picture in a single-file HTML report. Base64 makes a file about a third larger and it cannot be cached on its own, so above roughly 10 KB a linked file is usually better.

• What does it accept when decoding?

  Plain Base64, a full data URI, Base64 split over several lines, the URL-safe alphabet with - and _, and Base64 with its = padding missing. The kind of picture is read from its first bytes, so it is named correctly even when the text says otherwise.

• Is a decoded SVG safe to look at here?

  Yes. It is shown as an image, and a browser never runs script inside an SVG shown that way. Saving it gives you the file exactly as decoded, script and all, so check it with SVGCLEAN before putting it on a website.

Open DATAURI

• Why two answers for days between?

  Because the question has two meanings. From Monday to Wednesday is two days that pass and three days involved. Notice periods, interest and ages count the first; bookings, custody clocks and inclusive statutory periods count the second. Giving one number and hoping is how a deadline is missed by exactly one day.

• What happens adding a month to the 31st?

  It clamps to the last day of the target month (31 January plus one month gives 28 February), and the page says that it clamped. The other common behaviour is to roll forward into March. Both are defensible, they disagree, and the only unacceptable thing is not saying which you did.

• Why no built-in holiday list?

  Because there is no such thing as the holiday list. They differ by country, by region within a country, by year, and by whether your organisation observes them. A list shipped in software would be quietly wrong for most people, which is worse than no list at all.

• What is an ISO week?

  The week numbering where week 1 is the one containing the year's first Thursday, and weeks run Monday to Sunday. It is why the 1st of January is frequently in week 52 or 53 of the previous year, which surprises people every January and is not a bug.

Open DATES

• What is the difference between the two lengths?

  A hot debrief happens while everybody is still standing there, takes ten minutes and asks three questions. A full review happens later, with the timeline in front of you, and separates what was supposed to happen from what did and why. Both are real and neither replaces the other.

• Why does it insist on an owner and a date?

  Because "communications need to be looked at" is not an action and everybody in the room knows it. It has no owner, no date, and no chance. The page does not refuse it; it shows it as something said rather than agreed, which is usually enough.

• Does it stop the debrief turning into blame?

  No. Nothing can do that except whoever is running it. What it can do is keep the questions apart, so that what happened is written before why, and ask for what went right, because a debrief that only lists failures teaches people not to volunteer.

• Where does it go?

  Nowhere. It is held in this browser, and can be printed, copied or downloaded. A debrief is often the frankest document an organisation produces, which is a good reason for it not to be on somebody else's server.

Open DEBRIEF

• Does changing the DPI make my picture better quality?

  No. DPI is a label that says how many pixels go to an inch of paper. Changing it changes how large the picture prints, never how much detail it has. A 1200 pixel wide photo at 300 dpi prints 4 inches wide; at 72 dpi it is labelled to print 16.7 inches wide, with exactly the same pixels.

• Someone wants 300 dpi. What do I actually need?

  Enough pixels for the printed size at 300 per inch: 2480 × 3508 for a full A4 page, 1800 × 1200 for a 6 × 4 inch print. The page shows what each of your pictures prints at the DPI you choose, so you can see whether it is enough before you send it.

• Why not just re-save the picture at 300 dpi?

  Re-saving a JPEG compresses it again and loses a little detail every time. The DPI is only a number in the header, so this changes that number and copies the rest of the file across exactly.

• Why can I not set the DPI of a WebP, GIF or HEIC?

  Those formats have no standard place to store it. Convert the picture to PNG or JPEG with SHRINK first if a DPI is required.

Open DPI

• How is it different from a quiz?

  A quiz blocks the wrong answer and moves on. DRILL lets it play out: choose to power off the phone to “preserve” it and you watch it lock and encrypt, then read why. The consequence is the lesson. Nothing is scored as a percentage; the debrief counts sound decisions, minor slips and serious errors and shows the path you took beside the one that holds up.

• What scenarios are built in?

  Seizing a suspect's phone found unlocked and live, handling a USB stick found during a search, and being recognised by name in the street while under cover. Each branches: what you choose changes what happens next, and there is more than one way to get it wrong and more than one to get it right.

• Can I write my own?

  Yes. A scenario is a small JSON file: a name, a starting node, and nodes that each have some text and either choices - each with a grade of good, ok or bad, a line of feedback, and which node it leads to - or an ending with an outcome line. Use the template button to download one to fill in, then load it back. It is checked whole before it runs, so a broken link is caught before a trainee hits it.

• Where do my scenarios and runs go?

  Nowhere but this browser. Loaded scenarios stay on this device until you remove them, runs are not saved between sessions, and nothing is uploaded. Export a debrief as text if you want to keep it.

• Is this a substitute for training?

  No. It rehearses judgement between courses and lets people fail safely, which a real deployment does not. The built-in answers reflect widely-taught good practice, not any one force's policy; where yours differs, write a scenario that teaches yours.

Open DRILL

• What is the test print for?

  It prints the envelope outline and the window on ordinary A4 so you can hold it against a real envelope and see whether the address lands in the window. One sheet of paper instead of four ruined envelopes, which is the entire reason this tool exists.

• Which corner are the measurements from?

  The top left of the envelope as it is fed into the printer, which is usually long edge first with the flap away from you. If everything comes out mirrored or rotated, the envelope is going in a different way round: turn it and the same numbers work.

• Why does the window check say roughly?

  Because the exact width of a line of text depends on the font the printer uses, and the page measures it approximately. It is accurate enough to catch an address that is plainly outside the window, which is the failure that matters; the test print settles anything marginal.

• Does the address go anywhere?

  No. There is no network call on this page.

Open ENVELOPE

• When should I not use the mean?

  When the distribution is skewed, when a few values account for much of the total, or when the shape has two humps. All three are checked here and reported in words. The rule of thumb is simple: if the mean and the median disagree by much, quoting only the mean is a choice and somebody will eventually ask why you made it.

• What is an outlier here?

  A value more than one and a half interquartile ranges beyond the quartiles, which is the usual convention. It is worth saying plainly that outside the fences is not the same as wrong: the 412,000 in a column of salaries may be the chief executive, which is a fact rather than an error. Check them; do not delete them by reflex.

• Why does the quartile definition matter?

  Because on a sample of nine values the nine published definitions can give visibly different quartiles, and two people using different software will disagree about the same data. Naming the method is what makes the number checkable.

• What does two humps mean?

  That you are almost certainly looking at two populations added together: two teams, two products, two shifts. A mean sits in the valley between them and describes nothing at all. The fix is to separate them, not to summarise harder.

• Does my data go anywhere?

  No. There is no network call on this page.

Open FIGURES

• Are my photos uploaded?

  No. The camera is your browser's own, and its picture, like any photo you choose, is drawn and flattened in this tab. The page asks for no network while it works, and the PDF is written there too.

• How does it find the page?

  The photo is made small and grey, the brightness that best separates the paper from what is behind it is worked out, and the largest bright shape is the page. Its corners are the four points furthest along each diagonal. On a pale table or with a folded page it can miss, so the corners can always be moved by hand.

• Which look should I choose?

  Colour, cleaned, for most pages: the paper comes out white and shadows go. Black and white makes text crisp and files small, like a photocopy, but greys out photographs. Grey is in between. Colour as photographed keeps the photo's own light.

• Is the text searchable?

  Not yet: a scan is pictures of pages, from a phone or from a scanner. OPTIC reads the words in a scan, in your browser, and can make it searchable.

• What makes a good photo?

  The whole page in the frame, against something darker than the paper, in even light, with the phone held over the page rather than at a steep angle. The page does not need to be straight; that is what the corners are for.

Open FLATBED

• Why layers by longest path?

  Because a step that can be reached both early and late belongs where it actually happens, not where it first could. Laying out by shortest path pulls those steps up the page and the flow stops reading downwards, which is the only reason to draw one.

• What happens to a loop?

  It is drawn, routed round the outside rather than through whatever is in the way, and the layering stops rather than spinning. A flow with a loop in it is perfectly ordinary; a layout that hangs on one is not.

• Why flag a decision with one branch?

  Because the branch nobody writes is the one the chart existed for. A question with only a yes leaves the reader to invent the no, and they will each invent a different one.

• What happens to a line I have half typed?

  Nothing. A line that is not an arrow is skipped and listed below, rather than blanking the picture while you are still writing it.

Open FLOW

• Can I use this font on my website?

  This page cannot tell you that and neither can any page. It shows you the two things that bear on it: the fsType bits, which say what the foundry recorded in the file, and the licence description and licence URL from the name table, which is where the actual terms live. Read the licence. A font marked installable in fsType can still be under a licence that forbids web use.

• Why does my WOFF2 show no tables?

  Because its tables are Brotli-compressed, and a web page has no way to decompress Brotli: the browser has a decoder, but it does not hand it to JavaScript. The table directory inside a WOFF2 is not compressed, so the tags and their original lengths are listed, and the specimen still works because the browser decompresses the font internally to render it. For everything else, convert to TTF or WOFF first.

• What does a 78% on a Unicode block mean?

  That the cmap maps 78% of the assigned code points in that block to a glyph. Unassigned code points are not counted against the font. It does not mean the glyphs are good, and it does not mean the font will compose them correctly: a complex script needs GSUB rules as much as it needs glyphs, and coverage says nothing about those.

• Why are the metrics three different numbers?

  Because there are three sets and they disagree by design. hhea ascender and descender are what most desktop applications use. OS/2 sTypoAscender and sTypoDescender are what the specification says should be used for line layout. OS/2 usWinAscent and usWinDescent are the clipping box Windows uses. A font whose three sets differ will set at different line heights on different systems, and seeing all three is usually how you find out why.

• Can it tell a real font from a pirated copy?

  No. It reads what the file says about itself, and every one of those fields is text somebody typed. A copy with the name records rewritten looks identical here to the original. What it can do is show you the fields, so you can see when the copyright, the designer and the vendor ID do not agree with each other.

• Does the glyph count include blanks?

  Yes. numGlyphs in maxp counts every glyph slot in the font, including .notdef, the space glyphs, and any glyph that is a component of another. It is the number the font declares, and it is usually larger than the number of characters the font can show.

• I switched a feature on and nothing changed. Why?

  Switching a feature on writes it into font-feature-settings on the specimen. If nothing changes, the font has no rule for that feature that applies to the characters in the sample. Change the sample text and try again.

• What does FONTLENS read out of the file?

  A font file is a database with outlines in it. FONTLENS parses the table directory itself, reads the name records the foundry wrote, decodes the embedding permission bits into English, works out which parts of Unicode the font actually covers, and loads it into the page so you can see a specimen, switch each OpenType feature on and off, and drag every variable axis. The file never leaves this tab.

Open FONTLENS

• How do I convert TTF to WOFF2?

  Drop the .ttf here, choose WOFF2 and save. The tables are compressed together with Brotli into a WOFF2 file every current browser reads. It comes out around a tenth larger than Google's woff2_compress, which also rewrites the outline table before compressing it.

• How do I turn a WOFF2 web font back into a TTF I can install?

  Drop the .woff2 here and choose TTF (or OTF, for a font with PostScript outlines). WOFF2's rewritten outlines and metrics are put back as the specification sets out, every table is checksummed, and the result is an ordinary font file of the kind operating systems install. Check that the font's licence lets you install it.

• Can it convert OTF to TTF?

  Not the outlines. An .otf with PostScript (CFF) outlines stays PostScript, and is saved as .otf, because turning its curves into TrueType's is a different job that changes the shapes slightly. Renaming an .otf file to .ttf changes nothing inside it.

• Will the converted font look the same?

  Yes. Only the wrapper changes: the outlines, hinting, kerning, character map and names are the same tables, byte for byte where the format allows, and the tests compare every glyph with the original using fontTools.

Open FONTSWAP

• Can I build a form without a service collecting the responses?

  You can build and fill one locally, which suits a checklist, an intake sheet, or a questionnaire completed on the same device. What a local tool cannot do is collect responses from other people over the internet, because that needs a server, and a server means somebody holds the answers.

• Where do the responses go?

  Into this browser's storage on this device. Nothing is transmitted, which is the point when the questions are sensitive, and it also means you have to export the results to get them anywhere else.

• Is it safe to use an online form for sensitive questions?

  It depends who runs it and where the answers land. Health, financial and identity questions put into a free form service mean the answers sit in that company's database under their terms. Read where the data is stored, for how long, and who can see it, before you ask anyone anything.

• How do I get the answers out?

  Read them as a table on the page, or export them as CSV. The answers are kept in this browser and nothing is uploaded, so the export is how they get anywhere else. There is no account and nothing to pay.

• Can a question appear only when an earlier answer says so?

  Yes. Each question can carry rules against the answers above it: show it when a choice is one thing, is not another, is answered, or contains some text, matching all the rules or any of them. A question hidden by its rules is not asked and its answer is not kept.

• Can I break a long form into pages?

  A page break splits it, and the person filling it in walks the pages with Next and Back. What they have typed stays as they move.

• What do I get back when people answer?

  A summary that counts each answer, averages a scale, and draws the counts as a bar, a pie or a line, with a PNG of each chart. The chart is drawn by the same code Sheets uses, so a bar here and a bar in a spreadsheet are the same picture in the same colours.

• Where are the responses kept?

  In this browser, on this device. There is no server to collect them, so a form you share is filled in and the answer comes back only when the person filling it in is on the same device, or hands you the file. Send them to Sheets to work with them as a table.

Open Forms

• How do I make a GIF smaller?

  In order of effect: a smaller width, fewer colours, and dropping every second frame. Halving the width makes the file about a quarter of the size. Dropping every second frame keeps the speed the same by showing each remaining frame for twice as long.

• Why is my edited GIF larger than the original?

  Some GIFs are optimised so that each frame stores only the part that changed. This page puts every frame together into a whole picture before editing, and writes whole frames back. The new size is shown next to the old before you save, so you can pick fewer colours or a smaller width if it has grown.

• Can I get the frames out as pictures?

  Yes. Save the frames as PNGs gives a zip with one picture per frame, numbered in order, each as the GIF shows it at that moment, with any crop and resize applied.

• What does there and back do?

  It plays the frames forward, then backward, then loops, so the animation never jumps from its last frame to its first. It is sometimes called a boomerang.

Open FRAMES

• How do I open a GEDCOM file without family tree software?

  Drop the .ged file on this page. It is read in the browser, and lists everybody in it with their dates and places; choose a name to see their parents, partners and children. Nothing is installed and the file is not uploaded.

• How do I get a GEDCOM file out of a family tree website?

  Most of them have an export: look in the tree's settings or its menu for Export, Download or Share, and choose GEDCOM. RootsMagic, Family Tree Maker, Legacy and Gramps write one from their File menu. The file that arrives ends in .ged.

• Why do the names in my GEDCOM file show strange characters?

  Older programs wrote GEDCOM in ANSEL, a character set in which an accent comes before its letter, and a reader that takes the file for something else shows the wrong characters. This page reads the character set the file says it uses (ANSEL, UTF-8, UTF-16 or Windows-1252) and, when a file says UTF-8 and is not, reads it as Windows-1252 and says so.

• What counts as an impossible date?

  A death before a birth, a burial before a death, a child born before a parent or after the mother died, a parent under twelve at a birth, a marriage before the person was born or after they died, and a life of more than 122 years. Dates marked about, estimated or calculated are left out, since a guess set against a guess proves nothing.

• Can I edit the tree here?

  No. This reads a GEDCOM file and changes nothing in it. OBSCURA Roots, the family tree application on this site, opens the same file for editing, sources, charts and books, and keeps the tree on your computer.

Open GEDCOM

• What does naming them in the order to post them do?

  A profile shows the newest post first, in the top left. So to see the picture whole, the bottom-right piece has to be posted first and the top-left piece last. With this chosen, the pieces are numbered post-01, post-02 and so on in exactly that order.

• Why are some pieces a pixel bigger than others?

  Because the picture's width or height does not divide evenly by the number of pieces, and a piece cannot hold part of a pixel. The extra pixels are spread out one at a time rather than stretched. The page says when this happens.

• Will the pieces line up when printed as a poster?

  Yes, edge to edge, with nothing lost between them. Print each at the same scale; if your printer cannot print to the edge, leave a margin and trim it off each sheet before taping them together.

Open GRID

• Will the slides look exactly as they do in PowerPoint?

  Close, not exact. Each slide's layout, backgrounds, text boxes, bullets, sizes and colours are read from the deck, its layout and its master the way PowerPoint reads them. The text is set in DejaVu Sans rather than the deck's fonts, so a line can wrap in a different place, and the PDF has one page for each slide.

• What is not drawn?

  Charts, SmartArt, video and sound, pictures in Windows' own EMF and WMF formats, and shapes other than rectangles, rounded rectangles, ellipses and lines (their words are still drawn). Turned boxes are drawn upright, and animations have nothing to show on paper. The result counts each of these for your deck, so you know what to check.

• Is the text in the PDF real text?

  Yes. Every word is written into the PDF in the font it was laid out in, where it was laid out, so it can be selected, searched and copied. It is not a picture of the slide.

• What happens to hidden slides?

  They are left out, as PowerPoint leaves them out when it saves a PDF. Tick the box to include them.

• Is the deck uploaded?

  No. It is unzipped and read in your browser, laid out by the browser, and written into a PDF by pdf-lib, all loaded from this site.

Open HANDOUT

• What does the carry count actually do?

  It makes visible the thing everybody has stopped hearing. An item on its fifth handover is not a live issue any more; it is a decision the organisation has not taken, and putting the number on the screen is generally enough to get it taken.

• Why close an item instead of deleting it?

  Because how something stopped being carried is worth as much as that it was. Closing records the time and the reason; deleting leaves a gap that nobody can explain three weeks later.

• Why does it insist on naming who is receiving?

  A handover with nobody at the receiving end did not happen. It is the single most common way an item is lost: passed to the room rather than to a person.

• Is this a log book?

  No. It is the state of play at the moment of handover. If you need a running record of events with times, that is a different job and THREAD does it.

Open HANDOVER

• Does it have opinions of its own?

  None. It ships with an example sheet you can delete in one go, and after that everything it says came from rules you wrote. That is deliberate: a checker that argues with your house style is worse than no checker, because people learn to ignore it and then ignore the real findings too.

• Which rules can it apply automatically?

  The ones with a single right answer: prefer-this-word and capitalisation. A banned phrase has no replacement (the right wording depends on the sentence), so those are marked and left for you. It also keeps a capital letter at the start of a sentence when it swaps a word in.

• Why does it list rules that found nothing?

  Because a sheet that has grown to eighty rules, sixty of which never match anything, is a sheet nobody maintains. Knowing which ones are earning their place is how a style sheet stays short enough to be read.

• Can a rule be a phrase?

  Yes. Spaces in a rule match any amount of whitespace, so a rule about “at this moment in time” still matches across a line break.

• Does the document go anywhere?

  No. There is no network call on this page, which is the other reason the checkers that already exist are unusable for a draft.

Open HOUSESTYLE

• What playbooks are built in?

  Six of the situations that need the first hour handled well: a device seized or found, a device or account that may be compromised, a cover identity that may be blown, a lost or stolen device, a phishing or social-engineering attempt, and a personal-data exposure. Each has its steps grouped into the phases they belong to, in a sensible order.

• Are the steps official procedure?

  No. They are a sound default order to work from, not law or your organisation's policy, and they say so. Where your own procedure differs, follow it; the value here is that the record of what was actually done is kept as you do it, which is the thing that is otherwise never captured at the time.

• How does hashing evidence work?

  A step that preserves something offers to hash a file. The SHA-256 is computed here in the browser and recorded against the step with the file's name and the time, so the exhibit can later be tied to the moment it was taken. The file itself is never stored in the tool and never uploaded.

• What is the timeline?

  Every timestamped thing in one place, in order: the declaration, each step as it was ticked or skipped, and each action as it was closed. It is assembled for you, because reconstructing the order of events afterwards from memory is exactly what the review cannot rely on.

• Where is it kept?

  In this browser only, and nothing is uploaded. Seal the record with a digest, export it, and clear it from the browser once it is filed. Related: BREACH runs the notification clock for a data breach, CHAIN keeps the continuity record for an exhibit, and RENDEZVOUS holds the extraction plan a blown cover needs.

Open INCIDENT

• How do I open a .pages file on Windows?

  Drop it here to see the preview of its first page and to save every picture in it. To get the text as a Word document, sign in to icloud.com with any Apple Account, open Pages there, upload the file and download a copy as Word or PDF. iCloud's web apps are free and run in any browser, but the file is uploaded to Apple.

• Why can it not show the whole document?

  Since 2013 the document itself is stored in Apple's IWA format: compressed protocol buffers whose layout Apple has never published and changes between versions. A reader built by guessing would show a document that is subtly different from the one in the file, which is worse than not showing it. The preview and the media are stored as ordinary files, so those are exact.

• Can I get the photos out of a Keynote or Pages file?

  Yes. Every picture, film and sound put into the document is inside it as the original file, and IWORK lists them with a Save button each, or saves them all as one zip.

• My Pages file is a folder. What do I do?

  Older versions of Pages on the Mac could save a document as a package, a folder that looks like one file. On the Mac, right-click it and choose Compress, then open the zip here.

Open IWORK

• What is a second cousin once removed?

  Either your second cousin's child, or your parent's second cousin. Second cousins share great-grandparents; once removed means one of you is a generation further from those ancestors than the other.

• What is the difference between a second cousin and a first cousin once removed?

  Second cousins are the same generation and share great-grandparents. A first cousin once removed is your first cousin's child, or your parent's first cousin: you share grandparents, with one of you a generation lower.

• Is a great-aunt the same as a grandaunt?

  Yes. Your grandparent's sister is called a great-aunt by most people and a grandaunt by many genealogists, by analogy with grandparent. Her daughter is your parent's first cousin, which makes her your first cousin once removed.

• How much DNA do second cousins share?

  On average a thirty-second, about 3.1%, because the generations between them through the shared couple each halve what is passed on. Real pairs share more or less than the average, and this page does not give centimorgan figures.

• What is a half first cousin?

  The child of your parent's half-sibling. You share one grandparent rather than a couple of them, and on average half the DNA full first cousins share: a sixteenth rather than an eighth.

Open KINSHIP

• Why not a list of brand codes?

  Because the codes change, the same code means different things in different countries, and the numbers you need are printed on the packet you are holding. Common layouts are offered as starting points and every one of them is a set of boxes you can correct in ten seconds.

• Why does it keep telling me to print at 100%?

  Because “fit to page” scales everything by a per cent or two to allow for the printer's unprintable margin, and a per cent over a 297 mm sheet is three millimetres. Over twelve rows that is most of a label.

• Can I use a part-used sheet?

  Yes. Set the starting label number and the first labels are left blank, so the sheet with four missing from the top left is not wasted.

• Does my address list go anywhere?

  No. There is no network call on this page. An address list is personal data about other people, and it never leaves the machine it is printed from.

Open LABELS

• Why hold everything in hours?

  Because a day is not a unit. Twenty-five days for somebody on nine-hour shifts and twenty-five days for somebody on seven-hour ones are not the same entitlement, and every part-time arrangement makes it worse. Hours underneath, whichever unit you prefer on the screen.

• What does the cover check do?

  It walks every working day in every booking and counts how many people are away. Any day where fewer than your minimum are left is listed. That is the check nobody does, because leave is approved one request at a time and nobody is looking at the whole team on that date.

• Why show carry-over that will be lost?

  Because the person losing it is always the last to find out, and by then it is too late to book it. Capping quietly is the unkindest way to run a leave year.

• Is this employment advice?

  No. Entitlement, accrual, bank holidays and carry-over are whatever your contract and your jurisdiction say. This is arithmetic with the figures you give it.

Open LEAVE

• Are the pictures the same quality as in the PDF?

  Yes. A picture stored as a JPEG is handed back byte for byte, so it is the very file that was put into the PDF, with nothing recompressed. Every other kind is decoded pixel for pixel and saved as a PNG, which loses nothing.

• Why did some come out as PNG when they look like photographs?

  Because they were not stored as JPEGs. Many programs store pictures packed losslessly, and scanners often use fax-style compression, JBIG2 or JPEG 2000 without a JPEG version to hand back. A JPEG with a transparency mask also comes back as PNG, because a JPEG cannot carry the mask. A JPEG 2000 without a mask comes back as the .jp2 it is.

• Why are there fewer pictures than I can see on the pages?

  Only pictures stored as pictures are found. Charts, logos and diagrams are often drawn with lines and shapes rather than stored as a picture, and a scanned page may be drawn as a whole. PLATES turns whole pages into pictures instead.

• What about a picture used on every page, like a logo?

  It is stored once in the PDF, and it is taken out once, named for the first page it appears on.

• Is the PDF uploaded?

  No. It is read by pdf-lib and the pictures are decoded by pdf.js, both loaded from this site, in your browser tab.

Open LIFT

• Why is a flat rate about double?

  Because it is charged on the whole amount for the whole term, and you do not have the whole amount for the whole term: you are repaying it as you go, so on average you have had about half of it. Paying 1,500 for the use of an average balance near 5,000 is not five per cent. Five per cent flat over three years is 9.31% nominal and 9.72% effective.

• Which rate should I compare two offers on?

  The effective rate including fees, which is what this solves for. A nominal rate ignores compounding, a headline rate ignores the fee, and a flat rate ignores the repayments. Only the effective rate against what you actually received describes both offers in the same terms.

• Why is the effective rate higher than the nominal one?

  Because a nominal annual rate is the periodic rate multiplied by the number of periods, and money compounds rather than multiplying. One per cent a month is 12.68% a year, not 12%. Both figures are shown so it is clear which is which.

• Will my lender's figures match exactly?

  Close, not identical. Real agreements round payments to the penny, may apply interest daily rather than per period, and may charge fees this page does not know about. The purpose here is to understand the shape of the cost, not to reproduce a statement to the penny.

• Is this financial advice?

  No. It is arithmetic on figures you typed, with the method written out beside the answer. It does not know your circumstances, your agreement or your jurisdiction, and it has no opinion about whether you should borrow.

Open LOAN

• What is the sensitivity figure?

  How far one weight would have to move, on its own, before the leading option changes. A criterion where half a point flips the result is the one worth arguing about; a criterion where nothing you do to it changes the answer is one you can stop discussing. It is found by searching rather than by algebra, because the weights are normalised and the closed form is longer than it is worth.

• Why call a close result equivalent?

  Because it is. Two options within a few per cent of the scale are separated by less than the error in every judgement that went into the table. Reporting that as a winner is a false precision, and the matrix has actually told you something useful: choose on grounds it does not contain.

• Does a matrix make the decision?

  No, and the sheet says so. What it does is make the reasoning visible and reviewable, which is worth more than the number. Its whole value depends on the weights being set before the scores.

• Can I use it for anything else?

  It is just a weighted average with a sensitivity analysis, so yes: supplier selection, recruitment shortlists, options appraisals, where to go on holiday. Nothing in it knows what the rows mean.

Open MATRIX

• Where are the meme templates?

  There are none, on purpose. The well-known meme pictures are other people's photographs and film stills, and this site does not hand them out. Use a picture of your own, or one you have the right to use.

• Why does my text not look like Impact?

  Impact comes with Windows and macOS, but it is not licensed for a website to send to your browser, so it cannot come from here. If your device has it, it is used. If not, the page says so and uses the heaviest weight of the site's own typeface, which looks close.

• What happens to a long line?

  It wraps onto more lines and gets smaller until it fits the width of the picture and about a third of its height. Press Enter in the box to break a line where you want it.

Open MEME

• What kind of damage can it repair?

  The common kinds: an index at the end that points to the wrong places, a file that was cut short in downloading, bytes added before the start by an email program or a server, a missing trailer, and a broken list of pages. In each case the pages that are still in the file are found and kept, with their text.

• What can it not repair?

  Anything that is not in the file any more. If a download stopped halfway, the pages after that point are gone, and the result says how many could not be recovered. An encrypted PDF cannot be rebuilt without its password.

• Why does it offer to turn the pages into pictures?

  As a last resort, for a file whose objects cannot be read at all but which a forgiving reader can still partly draw. The pictures keep how each page looks; the text is lost, though OPTIC can read it back.

• How do I know the repair worked?

  Before and after, the file is opened by pdf.js and every page it lists is drawn; the result gives both counts. Look through the result before you rely on it.

• Is the PDF uploaded?

  No. It is read and rebuilt by pdf-lib, and checked with pdf.js, both loaded from this site, in your browser tab.

Open MEND

• Does the list get uploaded?

  No. The file is read by this tab with the browser's own file reader, the rows are held in memory, and the letters are written here. There is no server side to this page, and a Content Security Policy on this site forbids it talking to any other origin. Turn the network off and it still works.

• What happens if a row has an empty cell?

  If the letter uses that field on an ordinary line, the check lists the row and refuses to produce anything until it is filled or the line is made conditional. If the field is used only inside an {{#if}} line, the line is dropped for that row and nothing else changes.

• Which bits of markdown does it understand?

  Headings with one to three #, paragraphs, bullet lists with - or *, numbered lists with 1., bold with **stars** and italic with *one star* or _underscores_. Every line break in the template is a line break in the letter, which is not standard markdown but is what an address block needs.

• Can I use it for labels on a sheet I already have?

  Yes, that is why the grid is a set of numbers rather than a list of brand names. Measure the sheet: how many across, how many down, the margin at the top and at the side, and the gap between labels. The label size follows from those. Print one sheet on plain paper and hold it up to the light against a real one.

• Why only one kind of conditional?

  Because the second one is never the last. A merge template is read by whoever inherits the letter, often a year later and often not a programmer, and a language with nesting and expressions in it stops being readable at about the third feature. A line that is there when the field has something in it covers the case that actually comes up, which is the second address line.

• My list has names that are not in the Latin alphabet. What happens?

  In the three PDF outputs, each character those fonts cannot draw comes out as a question mark, and the page says how many did after it writes the file. That is a limit of the fourteen fonts built into the PDF format, not a decision made here. The zip of text files is UTF-8 and carries every name exactly, so for a list in Greek, Cyrillic, Arabic, Hebrew or a CJK script, produce the text files and set them in something that has the fonts.

• Where did the letter go after I closed the tab?

  The template, address block and settings are still there, saved on this device. The list is not, by design. Keep the CSV where you got it and open it again.

• What does the check look for before it produces anything?

  Three things: placeholders in the letter that have no column in the list, columns in the list that the letter never uses, and rows whose value for a needed field is empty, which would leave a hole in a letter. Nothing is produced while the check is unhappy.

• How is the address block different from the letter?

  Labels and envelopes are printed from the address block, not from the letter. It takes the same placeholders and the same {{#if field}} conditional, but no markdown: one line in the block is one line on the label. Every line break in the letter template is a line break in the letter too, so an address block written there stays an address block.

Open MERGE

• Does the frame look like a particular phone or browser?

  No, deliberately. Every frame is drawn from plain shapes: a window with three coloured dots and an address bar, a dark rounded phone body. There is no maker's artwork and no logo, so the picture does not suggest a product or an endorsement.

• Will the text in my screenshot stay sharp?

  Yes. At its own size the screenshot is drawn pixel for pixel, with nothing resampled, and saving as PNG keeps it exact. JPEG makes a smaller file but can blur fine text a little.

• What is twice its size for?

  For a screenshot taken on an ordinary screen that will be shown on a high-resolution one, such as a slide on a modern laptop. It doubles every pixel so the frame and the shadow are drawn finely; it cannot add detail to the screenshot itself.

Open MOCKUP

• Is any of this sent anywhere?

  No. There is no server side to this page. Open your browser's network tab and use the whole tool: nothing goes out. That is also why there are no live exchange rates, no bank connections and no sync between your devices, and why a file export is the only backup.

• Why will the reconcile not let me finish?

  Because the book and the statement do not agree yet, and finishing while they disagree records a balance that is wrong and carries the error into every reconcile after it. The difference is on screen the whole time, and the page suggests what a difference of that shape usually means. If the statement is genuinely right and the book is missing something, add it with the button there and the entry joins the list.

• My bank's CSV has separate paid-in and paid-out columns. Will that work?

  Yes. Switch the amounts setting to two columns and pick them. A value in the money-out column becomes a negative amount and a value in the money-in column a positive one, whichever way round the file writes the signs. If the file signs them the other way, the signs setting turns them round.

• What happens to a transaction I import twice?

  It is flagged, not blocked. A row whose date, amount and payee all match something already in that account is shown with the flag and left unticked, and you decide. For OFX the bank's FITID is checked first, which is more reliable than the three fields, because a coffee really can be bought twice in a day for the same amount at the same shop.

• Can I get my data into something else?

  Yes: CSV of every transaction, JSON of the whole book, and QIF for one account at a time. The JSON is the complete book, minor units and all, and is what to keep if you may come back to it. QIF dates are written month first, as Quicken wrote them, because QIF never agreed on an order.

• What if I lose the passphrase on an encrypted backup?

  The file is gone. There is no recovery, no reset and no copy, which is what makes the encryption worth anything. AES-GCM checks the file against its own tag, so a wrong passphrase fails outright rather than quietly producing nonsense.

• What can I type in the search box?

  Plain words are matched against the payee, the note and the category. For anything narrower there are prefixes: payee:tesco, cat:food and acct:current pick a payee, category or account; >100, <50 and =9.99 pick by amount; cleared:no finds what has not cleared; after:2026-01-01, before:2026-04-01 and 2026-01-01..2026-03-31 pick by date. They can be combined in one search.

• Why can I not delete an account or a category?

  Because something still depends on it. An account with no transactions can be deleted; one with transactions cannot, so that a figure never loses the entries behind it. A category in use stays until the transactions using it are moved to another category.

• Which statement files can I import?

  CSV from a bank or a spreadsheet, and OFX or QFX as most banks and Quicken write them. The file is read in this tab and is not uploaded. For CSV you choose which column is which and see the result before anything is written.

Open MONEY

• Does the text stay selectable?

  Yes, with the first setting. Each instruction on a page that sets a colour is rewritten to set a grey instead, and pictures are stored again in grey; the text, fonts, links and form fields are not touched. Only the second setting, which draws every page as a picture, loses the text.

• How is a colour turned into a grey?

  By its lightness, using the same weighting a black and white television and most printers use: 30% of the red, 59% of the green and 11% of the blue. A yellow comes out light and a navy comes out dark, as they look.

• How do I know it worked?

  After rewriting, every page is drawn and checked pixel by pixel for colour, and the result names any page that still has some. Spot colours, gradients built from sampled tables and a few other things cannot be rewritten; for a file with those, draw every page as a grey picture instead.

• Will it make the file smaller?

  Usually a little. Photographs stay JPEGs, now grey; other pictures are stored as one channel instead of three. It is not a compressor; PRESS is.

• Is the PDF uploaded?

  No. It is rewritten by pdf-lib and checked with pdf.js, both loaded from this site, in your browser tab.

Open MONOTONE

• How do I unprotect an Excel sheet without the password?

  Sheet and workbook protection are not encryption: the workbook is readable, and Excel only honours a flag with a hash of the password beside it. OFFICEKEY takes the flag out of a copy, so every sheet can be edited again. Use it on your own workbooks, or ones you have the right to change.

• Can it remove a password to open that I have forgotten?

  No. A password to open encrypts the file with AES, with the key made from the password by 100,000 rounds of SHA-512, and without the password the contents cannot be read by anyone. The page does not guess or try lists of passwords. If you know the password, it opens the file and can save a copy without one.

• How do I put a password on a Word or Excel file without Office?

  Open it here, type the password twice under Put a password on it, and save. The copy is encrypted as Office 2010 and later encrypt, and Word, Excel, PowerPoint and LibreOffice will ask for the password to open it.

• What is the difference between a password to open and a password to modify?

  A password to open encrypts the file. A password to modify, like restrict editing and read-only recommended, leaves it unencrypted and only asks Office to open it read-only. The second kind is a courtesy, not security, and OFFICEKEY lists and removes it.

• Does it work on .doc and .xls files?

  No. The formats from before 2007 use RC4 and other older schemes this page does not handle. Open the file in Office or LibreOffice with its password, save it as .docx or .xlsx, and bring that here.

Open OFFICEKEY

• Why does determinism matter?

  Because a rota somebody can reproduce is a rota they can argue with, and one that came out of a shuffle nobody can repeat is one you have to be trusted about. The same people, the same dates and the same blackouts always give the same answer here, and that is worth more than a cleverer algorithm.

• What happens to a skipped turn?

  It goes to the next person, and the skipped person keeps their place in the order for next time rather than losing it. Losing it is how somebody who was unavailable once in January ends up doing two extra weeks in the autumn.

• What are named dates?

  The ones people actually mind about: Christmas, New Year, a religious festival, a birthday. They are counted separately from weeks, because a rota that is even on weeks and lopsided on named dates is the one that causes trouble, and it is invisible unless somebody counts.

• Can I swap weeks after generating?

  Yes, and you should: the generator does not know about anything it was not told. Every count is recomputed from what is actually on the rota rather than from what the generator intended, so a swap is reflected immediately in the fairness figures.

Open ONCALL

• Why an indented list rather than dragging boxes?

  Because the list is the artefact people actually keep and update. It diffs, it pastes into an email, and anybody can edit it. A drawing that has to be maintained separately is out of date within a month, and everybody knows it, which is why nobody looks at it.

• What is a tidy tree?

  A layout where a subtree takes exactly as much width as its own leaves need and parents are centred over their children. The naive alternative gives every leaf a fixed slot, which is why the charts everybody recognises are three metres wide with one enormous gap in them.

• Can I put a job title under the name?

  Yes: put a vertical bar after the name and the rest of the line becomes a second, smaller line in the box.

• Does my list go anywhere?

  No. A structure chart is a list of everybody who works somewhere and who they answer to, which is not a document to paste into a free web service. This one is drawn in your browser and there is no network call.

Open ORGCHART

• Why is moving a line with its children the default?

  Because a heading that moves without what is under it has lost it, and in a real document that is a paragraph orphaned four screens away. Every outliner that got this right did it this way, and every tool that treats a bulleted list as flat text gets it wrong.

• Why a flat list rather than a tree?

  Because it makes impossible structures impossible. A line cannot be indented more than one level past the line above it, a subtree cannot be moved inside itself, and the whole outline is one block of text you can paste into anything.

• Can I bring in an existing list?

  Yes. Any indented list works (tabs or two spaces a level, with or without bullets), which covers what comes out of a notes app, a document or a plain text file.

• Does it save?

  In this browser, as you type. There is no account and no upload; clearing site data for this page clears it, so download it if it matters.

Open OUTLINE

• Why does printed graph paper come out the wrong size?

  Because two things resize it and neither tells you. The browser fits the drawing to the printable area, and the printer driver applies its own “shrink to fit” on top. A 5 mm grid arrives at 4.8 mm, every square wrong by the same amount, which is precisely why it looks right.

• How does the ruler prove anything?

  It is drawn a hundred millimetres long in the same units as the grid. Put a real ruler against the printed one: if it reads 100 mm, everything else on the sheet is to scale too. If it reads 97, your printer is at 97% and the setting to change is scale or fit-to-page. It is the only check that does not require trusting the software.

• What makes this different from a PDF of graph paper?

  A PDF carries a page size a driver is free to override. This sets the sheet with a page rule, draws in millimetre units, and prints the ruler so the result is verifiable rather than asserted. You can also change the spacing, the weight and the ink and see the effect before using a sheet of paper on it.

• What is isometric paper for?

  Drawing three dimensional things without perspective, which is how a lot of engineering sketches, board game maps and pixel art are laid out. The triangles have equal sides, so a cube drawn on it has three faces of the same size and every edge is measurable.

• Does anything I type get sent anywhere?

  No. The drawing is generated in the page, and there is no network call here at all.

Open PAPER

• Why do I have to place the markers myself?

  Because the crown of the head is not where a face detector thinks it is. Detectors find the face; the specification measures from the top of the hair to the bottom of the chin, which is a different distance and one that varies with a haircut. Two markers take four seconds, are exact, and mean the number on the page is a measurement rather than an estimate.

• Is the head height really what gets these rejected?

  It is the most common reason, yes, and the one people never suspect, because a photograph with the wrong head height looks perfectly normal. The rule is in millimetres of print (29 to 34 on a British passport, 32 to 36 on a Schengen one), and nothing in an ordinary photo editor measures in millimetres of print.

• What does eight on a six by four save?

  Most of the cost. A booth charges around ten pounds for four photographs; a supermarket print counter charges pennies for a six by four, and eight British-sized photographs fit on one with room to cut. The layout is generated as a single file, so there is nothing to explain at the counter.

• What is the real resolution figure?

  How much detail your original actually has at the print size, as opposed to how many pixels the file has. Enlarging a small photograph produces a file with the right pixel count and no more detail than it started with. Most print counters want 300 dots to the inch, so the page tells you whether you have them.

• Does my face get uploaded?

  No, and this is the page where that matters most. The picture is decoded in the browser, the cropping is done on a canvas, and there is no network call on this page at all.

Open PASSPORT

• How do I encrypt a message with someone's PGP public key?

  Open Encrypt, paste their public key (the block starting -----BEGIN PGP PUBLIC KEY BLOCK-----), type the message or choose a file, and press Encrypt. Copy the result into an email or save it. Only their private key can open it.

• Is it safe to paste my private key into a web page?

  Only into a page that does nothing with it but the job, which is what this one is: OpenPGP.js runs in the tab, nothing is sent, and nothing is kept after the tab closes. The page is a static file you can read. For a key you use every day, GnuPG or your mail program's own OpenPGP support is still the better home.

• Will GnuPG read the keys and messages?

  Yes. Keys are made as Curve25519 version 4 keys, the kind GnuPG 2.2 and later, Kleopatra, GPG Suite, Thunderbird and Proton all use, and the tests exchange keys, encrypted messages and signatures with GnuPG in both directions.

• What is the revocation certificate for?

  If the private key is lost or someone else gets it, publishing the revocation certificate tells everyone who has your public key to stop using it. Save it when you make the key, and keep it somewhere apart from the key itself.

• Can it tell me who sent a message?

  Only if they signed it, and only if you give it their public key. The page then says whether the signature is good and whose it is, or that it does not match. A message that is only encrypted could have been written by anyone who has your public key.

Open PGP

• Is my photograph uploaded anywhere?

  No. It is decoded by the browser, drawn onto canvases in this tab, and encoded back to a file by the same tab. Nothing is sent anywhere, there is no account, and the page works with the network switched off. The site's Content Security Policy forbids every third-party origin, so it could not send it even if it wanted to.

• Does the exported file still contain the GPS location?

  No. A canvas never receives it: the decoder reads EXIF to apply the orientation flag and then hands over pixels alone. What you save is the image plus the container's own housekeeping, which for a JPEG is a JFIF header and usually an sRGB colour profile, and for a PNG is nothing but the header, the image data and the end marker. If you want to prove it to yourself rather than take our word for it, open the saved file in EXIF.

• What happened to the edits when I press undo?

  Undo steps back through a stack of whole states rather than a list of inverse operations, so it is exact and it reaches everything: a crop, a quarter turn, a slider, a mark. Ctrl+Z on Windows and Linux, Cmd+Z on a Mac, and Shift with it to go forward again.

• Why does the JPEG size change when I move a slider?

  Because the size shown is the real encoded file, produced by actually encoding it rather than guessed from the dimensions. More contrast and more sharpening mean more detail for the compressor to keep, so the file grows. It is measured again a quarter of a second after you stop moving.

• Can somebody recover what is under a black box?

  Not from the exported file. The shape is drawn into the pixels and the file is flattened, so the information is genuinely gone. Keep the original somewhere else if you need it, because this cannot be undone once the file is saved. What can often be recovered is text hidden behind a blur or a pixelation, which is why neither is offered here.

• What is the temperature slider actually doing?

  Moving red and blue in opposite directions by the same amount. It is the cheap stand-in for a white balance rather than a real one, which would need to know the illuminant and work in a proper colour space. It is enough to take the blue cast off a photograph taken in shade, and it is not a substitute for a raw converter.

• Why does the grid get finer while I straighten?

  So there is something to line a horizon up against. Straightening turns the picture inside its own frame and scales it up just enough to cover the corners, which is why a degree or two always costs a little of the edge.

• Will a mark I place on the preview land in the same place on the saved file?

  Yes. Marks are stored as fractions of the frame rather than as pixels, so one placed on the preview lands in exactly the right place on the full-size export. Tap any mark to pick it up, drag it to move it, and press Delete or use the button to remove it. A box or an ellipse is solid, so it hides what is under it for good once the file is saved.

• Why can I move a slider back to zero and get the original pixels back?

  Because nothing in the adjust panel touches the picture. Every slider is a number held to one side and replayed over the decoded original each time the view is drawn. Move one back to zero a hundred edits later and you get exactly the pixels you started with. Hold the eye in the toolbar to see the picture as it came in.

Open PHOTO

• Should I write to hit a score?

  No, and doing so makes prose worse. The scores are arithmetic on syllables and clause lengths; they cannot see meaning. Chopping a clear sentence in half to move a number produces something that scores well and reads like a leaflet.

• So why show them at all?

  Because they disagree with each other in useful ways, and because a very bad score is a real signal even if a good one is not. What they are genuinely reliable at is the list underneath: the specific sentences that are too long, which the person who wrote them cannot see because they know what they mean.

• What is wrong with the passive voice?

  Nothing, in itself. It is the right choice when the actor is unknown or irrelevant. It is flagged because in a procedure, an account or an instruction it is almost always hiding the person who has to do something, and that is the fact the reader needed.

• What is a nominalisation?

  A verb turned into a noun: “made a decision” for “decided”, “carried out an assessment” for “assessed”. The noun needs a weak verb to carry it and the actor tends to fall out of the sentence entirely, which is why official prose is full of them.

• Does my draft go anywhere?

  No. There is no network call on this page. Every other readability checker on the web takes your text, which is worth knowing before pasting something unpublished into one.

Open PLAIN

• Is the PDF uploaded?

  No. The page loads pdf.js from this site, and pdf.js draws each page onto a canvas in your browser tab. The pictures are made from that canvas and handed to your browser's own download. Nothing about the file is sent anywhere, and it works with the network off once the page and the PDF reader have loaded.

• PNG or JPEG?

  PNG keeps every pixel exactly and is the right choice for pages of text, diagrams and forms, where JPEG blurs the edges of the letters. JPEG makes much smaller files and suits pages that are mostly photographs. At the default quality of 90 a JPEG of a text page is readable but not crisp.

• What resolution should I choose?

  150 dpi is clear on any screen and keeps files a sensible size. Choose 300 dpi if the picture will be printed, or if the page has small print you need to zoom into, and 600 dpi only for very fine detail: the files grow with the square of the resolution, so 300 dpi pictures are four times the size of 150 dpi ones.

• Why is one page at a lower resolution than I asked for?

  Browsers limit how large one canvas can be, at about sixteen million pixels in Safari. A poster-sized page at 300 dpi goes past that, so it is drawn at the highest resolution that fits, and the result names every page this happened to.

• Are form fields, comments and signatures drawn?

  Filled-in form fields and the appearance of comments and stamps are drawn as they show in a PDF viewer. Anything a viewer hides is not. Layers a PDF marks as off are left off.

Open PLATES

• Why insist on an overlap?

  Because a poster tiled edge to edge cannot be assembled. No printer prints to the paper edge, and no two sheets feed through identically, so edge-to-edge tiles leave white hairlines that no amount of care removes. Ten millimetres costs almost nothing and is the difference between a poster and a pile of paper.

• How do I put it together?

  Print at 100%. Trim each sheet along the dashed line on its right and bottom edge (those are the overlap boundaries), then lay the sheets out row by row, using the label in the corner, and tape them from the back.

• Does it lose quality?

  Each tile is drawn on a canvas at about 150 dots per inch, which is more than enough for something viewed from a metre away. If the original picture is small, blowing it up across six sheets will look soft, and no tool can fix that.

• Is the picture uploaded?

  No. It is decoded by the browser, cut up on a canvas in this page, and printed from there. Nothing goes out.

Open POSTER

• Can I edit the text on the slides?

  No, and this is deliberate. Each slide is a picture of its page, so it looks exactly right. Turning a page's drawing back into PowerPoint text boxes breaks every line into a box of its own and never quite matches the fonts, which is harder to work with than a picture. The page's words are attached to each picture as alternative text, so they can be searched and read aloud, and REFLOW turns a PDF into an editable Word document.

• What shape are the slides?

  The shape of the first page, ten inches wide. A landscape presentation exported to PDF comes back at 16:9 or 4:3 as it was. A page of a different shape is centred on its slide.

• JPEG or PNG?

  JPEG makes a much smaller deck and suits most pages. PNG keeps every pixel, which is sharper for small text and fine lines, at several times the size.

• Does it open in PowerPoint, Keynote and Google Slides?

  It is a standard .pptx: a presentation, one master, one blank layout, a theme, and a slide with a picture for each page. It was checked by opening it in LibreOffice Impress.

• Is the PDF uploaded?

  No. It is drawn by pdf.js, loaded from this site, and the deck is written in your browser tab.

Open PROJECTOR

• Is the picture exactly what Photoshop shows?

  Yes, when the PSD was saved with Maximise compatibility on, which is Photoshop's default: that stores a flattened copy in the file and this page shows it. When you untick layers, the page flattens them itself with normal blending and opacity only, so blend modes, layer effects, masks and adjustment layers are not drawn in that view. The caption under the picture always says which one you are looking at.

• Can I get a logo out of a PSD?

  Yes. Find its layer in the list and press Save as PNG. The layer comes out at its own size with a see-through background, as it was drawn in Photoshop, without its layer effects.

• Which PSDs will not open?

  CMYK, Lab, indexed and 16- or 32-bit documents, and the large-document PSB format. The page says which one a file is. In Photoshop, Image, Mode converts the document to 8-bit RGB, which opens here.

• Can I edit the PSD here?

  No. This page reads the file and saves pictures from it. Layered editing is what the Photo Editor is for, and it is part of OBSCURA Pro.

Open PSD

• How do I scan a QR code on my laptop?

  Press Use the camera and hold the code up to the webcam, or drop a screenshot or photo of it onto the page, or paste one. The contents appear as soon as the code is read; nothing is opened.

• Is it safe to scan a QR code?

  Scanning only reads it. The risk is opening what it points to, and a sticker over the real code on a parking meter or a menu is the common trick. The page shows the address, flags plain http, an @ that hides the real host, a raw IP address, a link shortener, lookalike characters and sign-in pages, and opens nothing unless you press the button.

• Does it send the camera picture anywhere?

  No. Frames are read from the video in the tab and nothing is recorded, kept or sent. The camera turns off when you press Turn the camera off or leave the tab.

• Can it read barcodes as well?

  Where the browser has its own reader (Chrome and Edge on Android, macOS and ChromeOS), yes: EAN, UPC, Code 128 and the rest. Elsewhere it uses jsQR, which reads QR codes only.

Open QRSCAN

• How much should I pay in quarterly estimated taxes?

  Enough that withholding and your four payments together reach the smaller of 90% of this year's tax and 100% of last year's (110% if last year's adjusted gross income was over US$150,000). Divide what is left after withholding by four. On $100,000 of profit, single, with no other income, the year's federal tax is about $22,365, so $5,591 a quarter clears it and about $5,032 a quarter avoids the penalty.

• When are 2026 quarterly estimated taxes due?

  April 15, June 15 and September 15, 2026, and January 15, 2027. A payment due on a weekend or a federal holiday is on time on the next business day.

• What happens if I miss a quarterly estimated tax payment?

  The IRS charges a penalty figured like interest, at its underpayment rate, on the missed amount from the due date until it is paid. It is worked out instalment by instalment, so paying the shortfall straight away stops it growing. Missing one does not change what the later instalments should be.

• Do I have to pay estimated tax if I also have a job?

  Only if withholding from the job falls short. You must pay estimated tax when you expect to owe $1,000 or more after withholding and credits, and withholding is less than the smaller of 90% of this year's tax and 100% of last year's. Raising the withholding at the job is another way to cover the side income.

Open QTAX

• Why is part of my SVG missing in the PNG?

  Almost always because the SVG links to something outside itself: a font on another site, or a picture it points at rather than holds. A browser draws an SVG as a picture without fetching anything, so those parts are left out. The page counts such links when the file opens. Convert text to outlines, or embed the picture, in the program that made the SVG.

• What width should I choose?

  The 2× button gives twice the size the SVG says it is, which is sharp on a phone or a high-resolution screen. For print, work out inches × 300: an icon printed 2 inches wide wants 600 pixels.

• Is the PDF a vector PDF?

  No. The PDF holds the drawing as a picture at the width you chose, on a page the size the SVG says it is. At 4× or 8× that prints cleanly, but it cannot be edited as shapes. For a vector PDF, print the SVG from a browser or a drawing program.

• Can an SVG run code when it is converted here?

  No. The SVG is drawn as an image, and a browser never runs scripts or follows links inside an SVG drawn that way. The page tells you if the file contains script, since that is worth knowing before you put the SVG anywhere else.

Open RASTER

• Should I develop the raw data or take the camera's JPEG?

  The camera's JPEG is instant and looks exactly as the photo did on the camera's screen, but on many cameras it is smaller than the full picture. Developing uses the whole sensor at full size with neutral, accurate colour; it takes a few seconds a file and will not match a camera's own picture style.

• Which cameras does it read?

  Everything LibRaw reads, which is over a thousand models: Canon CR2 and CR3, Nikon NEF, Sony ARW, Fujifilm RAF, Olympus ORF, Panasonic RW2, Pentax PEF, Leica, Hasselblad, Phase One and DNG from any camera or phone. A very new model can take a few months to be added to LibRaw.

• Why does my photo look flatter than in Lightroom?

  Lightroom and the camera add contrast, colour and sharpening to their own taste. A straight conversion gives you the neutral picture the sensor recorded, which is the best starting point for editing. To develop a raw file by eye, with exposure, curves and the rest, use the Photo Editor, which is part of OBSCURA Pro.

• How long does it take?

  A few seconds for a 24 megapixel file on a laptop, longer on a phone. Tick Work at half size for a quick look at a whole card; it is about four times faster.

Open RAW

• Does the book get uploaded?

  No. The file is opened by this tab with the browser's own file reader; a PDF is drawn by pdf.js and an EPUB is unzipped by fflate, both pinned on this site. There is no server side to this page and a Content Security Policy forbids it talking to any other origin. It works with the network off.

• Where are my highlights kept?

  In this browser, under one key in localStorage, alongside the shelf. They never leave the device. Export them as markdown to get them out, and use BACKUP to carry the whole of this site's data to another machine.

• Why does it ask for the file again?

  Because it never had a copy. The shelf is a record of books, not a store of them: it knows the name, the size, the cover and where you stopped, and the file stays wherever you keep it. That is also why the shelf survives being moved to a machine that has the books in a different folder.

• What happens to a book's own fonts?

  They are dropped, along with the @font-face rule that asked for them. An embedded font is a fetch, and in an EPUB it is frequently a fetch from a server that then knows a book was opened. The book's other typography, its headings, its italics and its indents, is kept.

• Can I read a book with pictures in it?

  Yes. Images inside the zip are shown, resolved from the zip itself. An image the book points at on another server is not fetched and is not shown, because fetching it would announce that the book had been opened.

• What does the markdown export look like?

  A heading for the book, then one section per chapter or page, each highlight as a block quote with its colour and its position under it and your note beneath that. It is written to be pasted straight into NOTES or into anything else that reads markdown.

• Does it work on a phone?

  Yes. Select text with a long press to get the four colours, and the arrow keys, space, Page Up and Page Down turn pages on a keyboard. Without a mouse at all, control or command with A inside the reader selects the chapter or the page you are on, and Tab from there reaches the four colours.

Open READER

• Does resizing make the text blurry?

  No. Each page is placed on the new paper as a drawing, scaled, the way a printer's fit-to-page setting does it. The text stays text and is as sharp at any zoom as it was before, and it can still be selected and searched.

• What order do the pages go in when there are several to a sheet?

  Left to right, then top to bottom, the order you read in. The sheet is turned whichever way lets the pages be largest unless you choose portrait or landscape yourself: two portrait pages go side by side on a landscape sheet.

• What is lost?

  Links, comments and form fields. They sit beside a page's content rather than in it, so a page placed as a drawing leaves them behind, and the result says how many there were. Flatten a filled-in form with FLATTEN first to keep what was typed into it.

• Is this the same as making a booklet?

  No. A booklet needs the pages in a folding order, eight and one on the first sheet; BOOKLET does that. This keeps the pages in reading order.

• Is the PDF uploaded?

  No. It is read and rewritten by pdf-lib, loaded from this site, in your browser tab.

Open REFIT

• How do I convert a PDF to Word without uploading it?

  Drop the PDF on this page. The file is read in the tab by the same PDF engine a browser uses to display one, the paragraphs and headings are rebuilt from the positions and sizes of the text, and the Word file is written here too. Nothing is sent anywhere, and you can turn the network off after the page loads and watch it still work.

• Why does a PDF need converting at all? It has the text in it.

  It has the glyphs, each one placed at a coordinate. It does not record where a paragraph starts or ends, which lines are a heading, or which text is a list. Those are things your eye puts back when you read it. Every converter, this one included, is reconstructing them from the geometry: lines from the baselines, paragraphs from the gaps and the ragged right edge, headings from how much of the document is set at each size.

• Will it keep my tables?

  Where the columns genuinely line up, yes: three or more rows whose text lands on the same two or three positions come back as a table you can edit. A table drawn with ruled lines, merged cells or wrapped text inside a cell will come back as paragraphs. This is the honest state of the art rather than a limitation of doing it in a browser.

• What happens to a scanned PDF?

  Nothing useful, and the page says so rather than handing you an empty document. A scan is a picture of a page with no text in it at all. Read it with OPTIC first, which recognises the words and can save the scan back as a PDF with the text behind the picture; bring that here.

• Does it keep the fonts and the layout?

  No, deliberately. What comes back is the document (the words, their structure, their emphasis and their pictures) rather than a facsimile of the page. Converters that try to reproduce the layout exactly do it with text boxes pinned to coordinates, which looks right until you edit one word and the whole page moves.

• What are running heads, and why were mine removed?

  The title and the page number repeated at the top or bottom of every page. They are furniture: in a document they would land as a stray line between every section. Lines that appear at the same height on most of the pages are dropped, and the count is in the report so you can see it happened.

• Can it read a two-column page in the right order?

  Yes. A page whose lines leave a clean vertical gap down the middle is split at that gutter and read down the left column and then the right, rather than straight across, which is what turns an academic paper into nonsense. The report says how many pages were treated that way.

• It will not open my PDF.

  If the file is locked with a password, take the password off with PADLOCK first; nothing can read an encrypted PDF without it, this page included. If it is damaged, X-RAY will say what is actually in the file.

Open REFLOW

• Why does my pattern only find the first match?

  Because it has no g flag. Without g, JavaScript's matching stops at the first result. Tick g in the flags row and the whole subject is scanned. With g the same RegExp object also carries a lastIndex between calls, which is the source of most surprising bugs when a pattern is reused in code.

• Why does the dot not match my newline?

  Because it never does by default. In JavaScript . means any character except a line terminator. Tick s (dotAll) and it matches newlines too. If you want the opposite, use an explicit class such as [^\n] so the intent is on the page rather than in a flag.

• What is the difference between greedy and lazy?

  a.*b takes as much as it can and then gives characters back until the pattern fits, so on a1b2b it matches through to the last b. a.*?b takes as little as it can and stops at the first b. Neither is right; they answer different questions. Switch the quantifier in the box above and watch the highlight move.

• Is my pattern safe to run on untrusted input?

  This page can tell you when a pattern backtracks badly on the subject you gave it, and it names the construct most likely to be responsible. It cannot prove a pattern is safe on every input, because that is undecidable in general. If a pattern runs against text a stranger supplies, prefer one with no nested quantifier over an overlapping class, and put a length limit on the input.

• Does it support named groups and lookbehind?

  Yes, because your browser does. (?<year>\d{4}) appears in the match list by name as well as by number, and (?<=\$)\d+ works in every current browser. Lookbehind arrived late in Safari, so a very old iOS will throw on it; the error is shown rather than swallowed.

• Can I use this on a whole file?

  Paste it in. The subject box will hold a few megabytes without complaint, and the match list is capped so a pattern matching a hundred thousand times does not build a hundred thousand rows of markup. The cap is stated when it is reached.

• Why does \b not find a word boundary in café?

  Because \w is exactly [A-Za-z0-9_], and \b is defined in terms of it. So a word boundary sits between caf and é: on café the pattern \bcafé\b fails and \bcaf\b succeeds. The u flag does not change this.

  For text that is not ASCII, match on properties instead. \p{L} with u matches a letter, and (?<!\p{L}) and (?!\p{L}) do the job of the boundaries.

• Is a lazy quantifier faster than a greedy one?

  Not in general. It just fails in the other direction. <.*> runs to the end of the string and hands characters back until a > can be found; <.*?> takes as little as it can and tries again each time it is wrong.

  The fastest of the three is usually neither. <[^>]*> can never overshoot, so it never has to come back.

• Why does my validation pattern accept a second line?

  Because the m flag is on. Without m, ^ and $ anchor to the start and end of the whole subject. With m (multiline) they anchor at every line break as well, so a second line of anything gets through.

  JavaScript is stricter than Perl here. Without m, $ matches only at the very end and not before a trailing newline, so /^admin$/ rejects "admin\n". Turn m on and it accepts it.

• Why does .test() return true and then false on the same string?

  A RegExp with g or y keeps a lastIndex between calls. The second call starts where the first one stopped, finds nothing, and returns false. That is not a bug, it is the cursor moving.

  Build the expression fresh, reset lastIndex, or use String.matchAll, which does it for you.

• What is the difference between the y flag and the g flag?

  Sticky matching (y) only succeeds if the match starts exactly at lastIndex. It never scans forward. That makes it the right tool for a tokeniser walking a string, and the wrong tool for finding something somewhere in it. g scans forward from lastIndex to find the next match.

• What does the u flag change?

  Without u, a pattern works on UTF-16 code units, so an emoji is two characters and . matches half of one. With u it works on code points, \p{...} properties become available, and unknown escapes such as \q become errors instead of meaning a literal q.

  Turning u on can therefore break a pattern that was silently wrong before.

• Why does the i flag not match some accented or non-Latin letters?

  Case-insensitive is not the same as case-folded. i alone will not match the Kelvin sign against k, and will not fold the Turkish dotless ı the way a Turkish reader expects. With u the folding follows Unicode simple case folding, which is closer but still not the locale-aware answer. Normalise the text before matching when this matters.

• Can a regular expression match nested brackets or HTML?

  No. Nested brackets, HTML, and anything that has to count depth cannot be matched by a regular expression, however clever. JavaScript has no recursion and no atomic groups to soften this.

  Where a pattern almost works on nested structure, it will be the malformed input that breaks it, which is exactly the input you were trying to catch.

• Why is a library pattern loaded with the m flag?

  The patterns in the library are the practical versions, not complete ones, and each says what it misses. A pattern anchored with ^ and $ is loaded with the m flag so every line of the sample is tested on its own. When you use it against a single value, drop the m and the g.

• What can I write in the replacement box?

  The same things JavaScript's replace understands. $1 is a numbered group, $<name> a named group, $& the whole match, $` the text before it, $' the text after it, and $$ a literal dollar sign.

Open REGEX

• My camera's clock was wrong. How do I fix the date?

  Open the photo, set the date and time it was really taken, and keep Set the digitised and modified dates to match ticked, so every program that sorts by date agrees. If you know the time zone, add it as, say, +01:00; photo libraries use it to place the picture on a timeline correctly.

• Does editing the EXIF re-save my photo?

  No. A JPEG keeps its EXIF in a block before the picture data. Only that block is written anew; the picture data is copied across byte for byte, so there is no loss of quality.

• Why would a field be left out?

  The EXIF library used here, piexifjs, knows the tags of the EXIF 2.3 standard, and the three time-zone tags added later are taught to it on this page. A rarer tag it does not know is left out when the block is rewritten. The page names any such tag before you save, so nothing disappears unannounced.

• How do I remove everything instead?

  Use CLOAK, which takes every kind of metadata out of a photo in one go, not only EXIF. This page is for changing what a photo says, and it removes only the location if you ask it to.

Open RETAG

• Why enter the score twice?

  Because that is the only way a control can be shown to do anything. An assessment with one score and a column of control sentences scores beautifully and protects nobody. If the residual is the same as the initial, either the control does nothing or nobody reassessed, and both are worth seeing.

• What is the acceptance threshold?

  The score above which somebody has to put their name to carrying the risk anyway. Acceptance is a thing a person does, and a high residual with nobody's name against it has not been accepted, it has just been written down. The default is twelve; set it to whatever your organisation uses.

• Are the bands official?

  No, and no bands anywhere are. High, medium and low on a five by five are a convention for talking about the number. The page says so on the sheet it produces, because a printed band tends to acquire an authority the arithmetic never had.

• Why keep the earlier assessments?

  Because a dynamic risk assessment is a sequence and an inquiry asks what you thought at each point, not what you think now. Overwriting is how an organisation loses the only record of a decision being reasonable when it was made.

Open RISK

• Why rolling seven days rather than weeks?

  Because a calendar week hides the worst seven days in almost every rota ever written. Finish a run on Saturday, start the next on Sunday, and each calendar week looks fine while the fourteen days in the middle are brutal. Every seven-day window is checked, not just the ones that start on a Monday.

• Are the default thresholds the law?

  No, and nothing on this page is legal advice. Eleven hours of rest, six consecutive days and a forty-eight hour average are common starting points in some places and wrong in others. Every one of them is a box, because what applies to your people comes from your agreement and your jurisdiction.

• What does a code with no times mean?

  It is a marker rather than work: rest, annual leave, a course. It occupies the day, it counts nothing towards hours, and it breaks a run of consecutive days. That distinction is the reason the legend has a times column that can be left empty.

• Why show the spread of nights and weekends?

  Because nothing else does, and it is what people actually complain about. A rota can satisfy every rule and still give one person six nights and another one, and that version is the one that ends up in a grievance.

Open ROSTER

• How do I convert GPX to KML?

  Drop the GPX here, choose KML, and save. Tracks with a time at every point are written as gx:Track, so Google Earth can play them back; waypoints become placemarks. The reverse, KML or KMZ to GPX, works the same way.

• Can it read a FIT file from my Garmin?

  Yes, the position, height and time of every record in an activity's FIT file, with the file's CRC checked. Heart rate, power and cadence are not carried across into GPX, KML or GeoJSON.

• Why is there no map?

  A map is made of pictures fetched from a map server for every place the track goes, which tells that server where you ran, rode or live. The page draws the shape of the track on its own instead, so the file never leaves the tab.

• Why is the climb different from my watch?

  The page adds up every rise between one point and the next, as recorded. Watches and apps smooth the height first to take out barometer noise, which gives a lower figure. The distance is measured along the great circle between each pair of points.

Open ROUTE

• Why a seed?

  Because it makes the draw reproducible, and reproducible is what defensible means here. Anybody given the seed and the population can produce the identical sample and check it. A spreadsheet's random function cannot be seeded, so nobody, including you, can ever demonstrate that the sample was drawn the way you say it was.

• Should the seed be secret?

  No, the opposite. Publish it. Its value is that somebody else can repeat the draw. What matters is that it was fixed before the draw, which is why the record includes it and why the page suggests writing it into the file note first.

• Is the generator cryptographically secure?

  No, and it does not need to be. The requirement is reproducibility, not unpredictability. It is mulberry32, a small published generator, and it is named in the record so anybody can implement it and check the result.

• When would I use systematic sampling?

  When the population is a list and you want even coverage across it: an inspection regime, every nth file in date order. Its one failure mode is a list with a repeating pattern of the same period as the step, which would systematically hit the same kind of item. Check that the order of the list is not meaningful before using it.

• What does stratified do?

  It takes a share from each value of a column in proportion to that group's size, with at least one from each. It exists so that a sample of a hundred from a population where one category holds two per cent does not come back containing none of them, which simple random sampling will do regularly.

Open SAMPLE

• How do I open a .sav file without SPSS?

  Drop it on this page. It is read in the browser, and you see the cases in a grid as SPSS's Data View shows them and the variables as its Variable View does: names, labels, types, value labels, missing values and measurement level. PSPP, which is free, and R or Python with the right packages will also open one.

• Can I convert a .sav file to Excel or CSV?

  Yes, with OBSCURA Pro: the Save buttons write a CSV, or an Excel workbook with the data on one sheet and the variable view on another. Dates come out as dates. Tick 'Show value labels' first to save the labels ('Agree') instead of the codes (4). Looking at the file is free.

• What is a .zsav file?

  A .sav file whose data are compressed with zlib, which SPSS has written since version 21 when asked for a compressed file. It is often a third of the size. This page reads it the same as a .sav.

• Why are some values underlined?

  They are missing values as the file defines them: codes such as 9 for 'No answer' or -1 for 'Refused' that SPSS leaves out of analyses. The data keep the code, and the page marks it. A blank cell is system-missing, where nothing was recorded at all.

• Why will my .sav file not open?

  Three reasons are common. It is password-protected (SPSS 25 and later can encrypt a .sav); save it again without a password. It is a portable file (.por) renamed, which is a different format. Or it was cut short when it was copied, in which case the page shows the cases it could read and says how many it expected.

Open SAV

• How do I open a .fdx file without Final Draft?

  Drop it on this page. A .fdx file is XML, and this reads it in the browser and lays the script out in screenplay format, with the scene list and the characters. Nothing is installed and the script is not uploaded.

• Is one page of a screenplay really one minute?

  As an average across a whole script, roughly, and it is what producers and competitions quote. Any one page can run well over or under a minute: action takes longer than it reads and fast dialogue less. It is a way of comparing scripts rather than a prediction of the cut.

• What is a Fountain file?

  A screenplay written as plain text with a few conventions: scene headings start INT. or EXT., a character's name is in capitals above their dialogue, and transitions end in TO:. Highland, Slugline, Fade In, WriterDuet and Beat read and write it, and any text editor can open it.

• What is a scene's length in eighths?

  Schedules measure a scene in eighths of a page, so a scene that fills one and a half pages is 1 4/8. This works it out from the lines each scene takes on the printed page, rounded to the nearest eighth, with at least one eighth for any scene.

• Why does my page count differ a little from Final Draft's?

  Final Draft's own settings (the space before a scene heading, the lines on a page, how it breaks dialogue) can differ from script to script, and this uses the standard ones. Expect the count to be within a page or two.

Open SCREENPLAY

• How does the automatic arrangement work?

  It seats the most constrained people first (the ones with the most rules attached) while there is still room to honour them, which is the heuristic that works for this kind of problem. It does not always succeed, and when it cannot satisfy everything it says which rules it broke rather than quietly dropping one.

• Can I move people afterwards?

  Yes, and the breaches update as you do. That is really the point: the value is not the arrangement, it is that moving a name to fix one thing immediately shows you what it broke somewhere else.

• What if a rule names somebody who is not on the list?

  It says so. A misspelled name in a rule is the commonest way a constraint silently stops applying, and it is invisible unless something checks.

• Does the guest list go anywhere?

  No. A list of names with who-must-not-sit-with-whom attached is not a document for a free web service. There is no network call on this page.

Open SEATING

• How much can I contribute to a SEP IRA if I am self-employed?

  Up to 20 per cent of your net profit after the deduction for half of your self-employment tax, and no more than $72,000 for 2026. The plan's 25 per cent rate is reduced to 20 per cent for your own contribution because the contribution is itself deductible. On a net profit of $100,000 that comes to $18,587.04.

• Is a solo 401(k) or a SEP IRA bigger?

  At most incomes a solo 401(k) allows more, because on top of the same 20 per cent employer contribution you can defer up to $24,500 as the employee in 2026, plus a catch-up from age 50. Both stop at $72,000 before catch-up contributions, so at high profits they meet.

• What are the 2026 SEP IRA limits?

  The most that can be contributed for one person is $72,000, and contributions are worked out on at most $360,000 of pay. Both figures are from IRS Notice 2025-67 and appear in Publication 560.

Open SEPCALC

• Why does it matter who gets the odd penny?

  Because it has to go to somebody or the total is wrong, and most tools drop it silently. Over one dinner that is a penny. Over a year of a shared house, with a division every week, it is pounds, and if the rule is arbitrary rather than stated, it is always the same person paying them without anybody noticing.

• How does it get the number of transfers down?

  By settling from net positions rather than individual debts. It does not matter who owed whom, only that everybody ends up square, so the tangle collapses to a list of people who are up and a list who are down. Pairing the largest of each repeatedly finishes in at most one fewer transfer than there are people.

• Is that the fewest possible transfers?

  Usually, and it does not claim to be always. Finding the guaranteed minimum means looking for subsets that cancel exactly, which is a partitioning problem with no fast general solution. The greedy answer is never worse than one fewer than the number of people, which is the bound that matters in practice.

• Can one person pay for something they are not part of?

  Yes. Set the split to some of the group and leave them out. They will show as being owed the whole amount, which is correct.

• Does it work in other currencies?

  It works in whatever unit you type, held as hundredths. The rounding rule is the same in any currency with two decimal places.

Open SETTLE

• Will the numbers look the way they do in Excel?

  Yes, for the formats people use: thousands separators, decimal places, currency signs, percentages, dates and times, negatives in brackets or in red, and custom formats built from those. Each cell is shown from its own format code, the way Excel reads it. A formula shows the value Excel last worked out for it, which is stored in the file.

• What about a sheet that is too wide for the page?

  It is scaled down to the width of the page, the way Excel's fit-to-width setting does it, and the result says by how much. Untick the box to print at full size, in which case columns past the page edge are cut off. For a very wide sheet, landscape or A3 paper keeps the text readable.

• What does not come across?

  Charts, pictures and shapes, conditional formatting, and Excel's own fonts: it is set in DejaVu Sans, which is wider, so a column is widened rather than let a number be cut off or turned into hash signs. Hidden rows, columns and sheets stay hidden unless you tick a hidden sheet.

• Can it open an old .xls file?

  No. The older .xls format is a different, binary format. Open it in Excel, LibreOffice or Sheets here, save it as .xlsx, and open that.

• Is the workbook uploaded?

  No. It is unzipped and read in your browser, laid out by the browser, and written into a PDF by pdf-lib, all loaded from this site.

Open SHEETFED

• Is this AI?

  No. The four styles are long-standing darkroom and image-editor recipes worked out with arithmetic on the pixels: a pencil sketch is the photo divided by a blurred negative of itself, ink lines are the places where brightness changes sharply, and a cartoon is smoothed colour cut into a few steps with those lines on top. The same photo and settings always give the same result.

• Which photos work best?

  A clear subject against a plain or blurred background, in good light. A busy background gives busy lines, because the page finds every edge and cannot tell which ones matter. Raising the stroke softness, or lowering the detail, quietens it.

• Is the saved picture full size?

  Yes, up to 3,000 pixels on its longest side. The preview is drawn smaller so it can update as you move the sliders; the saved one is drawn again at full size with the same settings.

Open SKETCH

• Why is this not just adding hours?

  Because a target in working hours is spent against the minutes a desk is open, and those minutes are not evenly spread. Four working hours from 16:45 on a Friday spends fifteen minutes before closing and the rest on Monday morning. Adding four hours reports a breach that never happened; adding a working day hides one that did. Both errors are silent, and both tend to land in favour of whoever built the spreadsheet.

• What does the stopped clock do?

  It extends the deadline by that many minutes of open time. A ticket waiting three hours for a customer to reply, on a desk that closes in one, does not lose three hours of its target overnight: it gets an hour back today and two tomorrow. Adding the pause to the wall clock instead is the second most common way these figures come out wrong.

• How do I know it is right?

  Because the two things it does are inverses of each other. Ask it for the instant four working hours from now, then ask it how much working time lies between now and that instant, and it must answer four hours, from any starting point, including the middle of a night, a weekend or a holiday. That round trip is what the site's test suite checks, not the wording on the page.

• What about the day the clocks change?

  Working hours are wall-clock hours. A desk open 09:00 to 17:00 is open 09:00 to 17:00 on the morning the clocks go forward, which is seven real hours and eight hours of service. Both the deadline and the measurement use wall-clock minutes, so they agree with each other and with the rota.

• Does a ticket reference go anywhere?

  No. There is no network call on this page, and what you type stays in this browser.

Open SLA

• How do I reduce the size of a PowerPoint file?

  Almost all of a large .pptx is pictures stored at their full camera resolution. Drop it here, choose the longest side a picture needs (1600 pixels is plenty for a screen), and save the smaller copy. Only the pictures change.

• Will it change how the document looks?

  The pictures are shown at the same size on the page; they have fewer pixels and are compressed harder, which you may see if you zoom right in or print large. Choose 2400 pixels and a higher quality for print. Text, tables, charts and layout are copied across byte for byte.

• Why does a PNG become a JPEG?

  Photographs are often pasted in as PNG, which stores every pixel losslessly and can be ten times the size of a good JPEG. A PNG with no transparency is turned into a JPEG, and every reference to it in the file is updated so Office finds it. A PNG with transparent parts, such as a logo, stays a PNG.

• How do I get the pictures out of a Word document?

  Open it here and every picture in it is listed with a Save button, as the original file that was put in, or all of them at once as a zip. Nothing is changed or recompressed for that.

Open SLIM

• Why a role rather than a name?

  Because a procedure with names in it is out of date the day somebody leaves, and everybody knows it, which is how a procedure stops being read. A role survives the person.

• What is wrong with "as appropriate"?

  Nothing, in conversation. In a procedure it is the place where the writer stopped deciding and handed the decision to whoever is reading it at three in the morning. Either put a number in, or say plainly that it is a judgement and name who makes it.

• Why flag a decision with one branch?

  Because the unexpected branch is the one the procedure existed for. "If the vehicle is insured, release it" tells nobody what to do when it is not, and the reader goes and asks somebody, which is precisely what the document was written to prevent.

• Does it rewrite my steps?

  No. It marks four mechanical things and leaves the words alone. A procedure has to be in the words of whoever owns it or nobody will defend it.

Open SOP

• What pace should I use?

  A hundred and thirty words a minute for a considered delivery, a hundred and sixty for a brisk one. Read a paragraph aloud at the speed you actually intend and check it against the number; most people find they are slower than they think, especially in a room.

• Why sections rather than one figure?

  Because a talk is never uniformly too long. It has an introduction that has grown to six minutes, and cutting everything by ten per cent leaves the introduction at five and a half. Sections next to their lengths make the swollen one obvious.

• Why the be-here-by times?

  Because a total tells you nothing while you are standing up. Knowing that you should be starting the third section at seven minutes is actionable; knowing the whole thing is eighteen minutes is not.

• Should I fill the whole slot?

  No. Leave a couple of minutes: a question, a slide that does not appear, and one person arriving late will use them. A talk that exactly fills its slot always overruns.

Open SPEECH

• Why not just use a spreadsheet?

  Because a spreadsheet converts things. It will read an identifier like 03-04 as a date, drop the leading zero from a postcode or a part number, and turn a sixteen-digit reference into scientific notation, and it will not tell you it did any of it. This treats every cell as text from start to finish.

• Is the parser a real one?

  Yes. Quoted fields containing commas, quoted fields containing newlines, and doubled quotes inside a quoted field are all handled. A naive split on commas silently ruins any file with an address or a quoted name in it, and the damage is invisible until somebody reads row 400.

• What does splitting into files do?

  It writes one CSV per distinct value of a column (one per team, per region, per month), each with the header row. It is the job that takes a macro in a spreadsheet, and downloads are staggered because a browser asked for twenty at once quietly drops most of them.

• Why keep a log of operations?

  Because a transformed file with no account of the transformation cannot be checked by anybody, including you next week. Each operation says what it did and how many rows it affected.

Open SPLIT

• Is it safe to make an SSH key in a browser?

  The key is made by the browser's own cryptography, WebCrypto, using the operating system's random number generator, the same source ssh-keygen uses. Nothing is sent anywhere and nothing is kept once the tab closes. The page is a static file you can read. For a key that guards something important, a hardware key is better than any software key, from here or from ssh-keygen.

• Which type should I choose?

  Ed25519, unless a system tells you otherwise. It is short, fast and has no weak parameters to get wrong. Some older servers and network equipment accept only RSA; use 3072 or 4096 bits there. ECDSA is offered for systems that ask for it.

• Why set a passphrase?

  Without one, the private key file is the key: anyone who copies it can log in as you. With one, the file is encrypted the way ssh-keygen encrypts it (bcrypt_pbkdf and AES-256 in counter mode) and is useless without the passphrase. ssh-agent or the macOS keychain can remember it so you type it once.

• How do I use the key?

  Save both files into ~/.ssh, make the private one readable only by you (chmod 600), and put the one line of the public key into ~/.ssh/authorized_keys on the server, or paste it into GitHub or GitLab under SSH keys. ssh-keygen -lf on the .pub file prints the same fingerprint the page shows.

• Does it work with PuTTY?

  PuTTY uses its own .ppk format. Open the private key in PuTTYgen (Conversions, Import key) and save it as a .ppk. Windows' built-in OpenSSH reads the file as it is.

Open SSHKEY

• Is the subtitle file uploaded?

  No. It is read by this tab with the browser's file reader, parsed here, and written back here. The page works with the network off, which is the simplest way to check that claim.

• My subtitles are exactly two seconds late. What do I do?

  Put -2 in the shift box, leave the scope on all of them, and press Shift. If they are late from a certain point only, select the cue where the drift starts and change the scope to that cue and everything after it.

• The subtitles drift further out as the film goes on.

  That is a frame rate mismatch rather than an offset. Pick the pair in the frame rate change list, which fills in the exact ratio, and press Stretch or compress. 23.976 to 25 is the common one.

• Can it open a .sub or .idx file?

  No. Those are bitmap subtitles: pictures of text, with no text in them. Turning them back into text needs optical character recognition, which is a different job from this one and not one this page does.

• Why can the video not play in Chrome?

  Chromium-based browsers ship without the H.264 and AAC decoders in many builds, so an ordinary MP4 gives an error. The file is fine. Try Firefox, or make a WebM copy to check the timing against.

• Does it keep my cue numbering?

  Yes, as long as you have not inserted, split, merged or deleted a cue. After a structural change the numbering has to be rebuilt, so it is renumbered from one.

• How far back does undo go?

  Eighty steps. Typing inside one cue counts as one step, not one step per character, so eighty is a long way back.

• What kinds of subtitle problems can SUBS fix?

  Subtitles go wrong in a small number of ways: everything is half a second late, the file was cut for 23.976 and the picture runs at 25, two cues sit on top of each other, or a line is too long to read. SUBS shifts or stretches the timing, lets you edit every cue in place, counts what is wrong against limits you set, and writes the file back.

• Which subtitle formats does SUBS open?

  SubRip (.srt), WebVTT (.vtt), and the dialogue lines of SubStation Alpha (.ass, .ssa). It saves as SubRip or WebVTT, or as a plain transcript with or without timestamps.

• Is the video I open beside the subtitles uploaded?

  No. The media file is handed to the player as a local object URL. It is not read, copied or uploaded. Timing is a thing you hear, not a thing you calculate, so play the video beside the cues and check them by eye and ear.

Open SUBS

• Should I use Student's t-test or Welch's?

  Welch's, unless you have a reason not to. Student's test assumes the two groups have the same variance; Welch's does not, and when the variances are in fact equal it gives almost the same answer. When they are not, and especially when the smaller group has the larger spread, Student's p-value can be badly wrong. Both are shown here side by side, with Levene's test, so the choice is visible.

• What is the difference between a one-tailed and a two-tailed p-value?

  A two-tailed p asks how surprising a difference this large would be in either direction; a one-tailed p asks only about the direction you predicted, and is half the size when the result goes that way. A one-tailed test is honest only when the direction was fixed before the data were seen, which reviewers often doubt. The page gives both and labels which is which.

• Can I run a t-test from a paper's means and standard deviations?

  Yes. Choose 'The n, mean and SD' and type them in. That is enough for Student's and Welch's tests, the confidence interval and Cohen's d. Levene's test needs the raw scores, so it is left out. For a paired test from a paper you need the mean and SD of the differences, which are not the same as the two means and SDs.

• What does Cohen's d mean here?

  The difference in means divided by a standard deviation, so it says how big the difference is in units of spread rather than in the units of the scale. For two groups it is the pooled SD; for paired data it is the SD of the differences (often written d_z); for one sample, the sample's SD. Cohen's labels of 0.2 small, 0.5 medium and 0.8 large were offered as a last resort, not a rule.

• Are the numbers right?

  Each figure is checked against SciPy, the standard scientific library for Python, on a set of data sets including skewed ones, tiny ones and one whose p-value is about one in a quadrillion: t, degrees of freedom, both kinds of p, the confidence interval, Cohen's d and Levene's test agree to at least nine significant figures.

Open T-TEST

• What is the VAT registration threshold?

  £90,000 of VAT taxable turnover in any twelve months, from 1 April 2024. Before that it was £85,000, from 1 April 2017. The figure is set by the government and can change, so check GOV.UK for the one in force. The page tests each month against the threshold that applied on its last day.

• Is the threshold measured over the tax year?

  No. It is any rolling twelve months, looked at again at the end of every month. A business can stay under it in every tax year and still go over it in the twelve months to, say, the end of November. That is why the page gives every month its own twelve-month total.

• When do I have to register once I go over?

  Within 30 days of the end of the month in which the twelve-month total went over the threshold. Registration then takes effect, and VAT has to be charged, from the first day of the second month after that month. Going over at the end of March means registering by 30 April, with VAT from 1 May.

• What counts as taxable turnover?

  Everything you sell that is not exempt from VAT or outside its scope. Zero-rated sales count, even though they carry no VAT; exempt sales, such as most residential rent, do not. Type the taxable figure for each month, not the money that came into the bank.

• What if a single month is very large?

  There is a second test: if you expect the turnover of the next 30 days on their own to go over the threshold, you must register by the end of those 30 days. The page cannot know what you expect, so it says this rather than testing it. If you went over only for a moment, HMRC can agree not to register you when turnover for the next twelve months will be no more than £88,000.

Open THRESHOLD

• Why will my TIFF not open in the browser or on my phone?

  Only Safari shows TIFF files, and even it shows just the first page of a multi-page one. Other browsers do not read the format at all. This page reads it with its own decoder, so every page shows in any browser.

• Which kinds of TIFF does it read?

  Uncompressed, LZW, PackBits, Deflate, JPEG and the CCITT fax codings, in black and white, grey, colour, palette colour and CMYK. That covers scanners, fax software, archives and image editors. Rare codings, and pages larger than a browser can draw, are named when they cannot be read.

• How big are the PDF pages?

  The size the TIFF says it prints at: a 300 dpi scan of an A4 page becomes an A4 PDF page. A TIFF with no resolution is taken as 72 dpi, one point for each pixel.

• Should I save as PNG or JPEG?

  PNG keeps every pixel exactly and is right for text, fax pages and drawings. JPEG is much smaller for photographs and is fine for colour scans, but softens sharp text a little.

Open TIFF

• Why count night hours by the minute?

  Because a shift is not a band. A shift from 21:00 to 05:00 is not a night shift by most definitions and contains seven hours of night. Counting by the shift either overstates or understates it every time, and by the minute it is simply right, including when the band wraps midnight.

• Can an hour be night and weekend at once?

  Not here. The weekend takes it, which is the ordinary convention and is stated on the sheet. Counting it in both bands is the easiest way to overstate a claim by accident, and the easiest to be accused of doing on purpose.

• Does it work out my tax?

  No. The figure it gives is gross, before anything is taken off, and the page says so. Tax depends on your jurisdiction, your code and the rest of your year, and a web page that guessed at it would be worse than no figure at all.

• What is the break flag for?

  A long day with no break recorded is almost always a break that was taken and not written down. Finding that now, while you remember, is much easier than finding it at the end of the month when somebody queries it.

Open TIMESHEET

• Does cropping delete what is outside the box?

  No, and it is important to know that. Cropping sets the page's crop box, which is the part a viewer shows and a printer prints. Everything outside it is still in the file, and an editor can widen the box and show it again. That is how Acrobat crops too. To remove something from a page for good, redact it with SEAL instead.

• Will the text still be selectable?

  Yes. The page is not turned into a picture; only its crop box changes. Text, links, form fields and bookmarks are carried through as they were.

• How does fitting to what is on the page work?

  The page is drawn small in your browser and read for the first and last pixel that is not white, in each direction, and three millimetres are left around that. A page with a faint grey background or a scanned page with a dark edge will not have a white margin to find; set the margins by hand for those.

• What if my pages are different sizes?

  Margins are measured from each page's own edges, so ten millimetres off the top is ten millimetres off every page whatever its size. To fit each page separately, choose to crop each page to what is on it.

• Is the PDF uploaded?

  No. It is drawn by pdf.js and rewritten by pdf-lib, both loaded from this site, in your browser tab.

Open TRIM

• Will the PDF look exactly like the document does in Word?

  Close, not exact, and it is worth knowing why. A browser has no copy of Word's fonts, so the PDF is set in DejaVu Sans, which is wider than Calibri or Times New Roman. Lines break in different places and a long document can run to more pages. The paper, margins, headings, alignment, text sizes and colours, spacing, lists, tables, pictures and page breaks are the document's own. Choose a slightly smaller text size to bring the line breaks closer to Word's.

• What does not come across?

  Headers and footers (the pages can be numbered instead, and the box is ticked for you when the footer numbered them), text boxes and shapes in their exact places, columns, and footnotes at the foot of their own page: they are gathered at the end instead. Tracked changes appear as if accepted, and comments are left out. For an exact copy, save as PDF from Word itself.

• Is the text in the PDF real text?

  Yes. Every word is written into the PDF in the font it was laid out in, where it was laid out, so it can be selected, searched and copied, and links still work. It is not a picture of the page.

• Can it open an old .doc file?

  No. The older .doc format is a different, binary format. Open it in Word or LibreOffice, save it as .docx, and open that here.

• Is the document uploaded?

  No. It is unzipped and read in your browser by mammoth and by this page, laid out by the browser, and written into a PDF by pdf-lib, all loaded from this site. Nothing about it is sent anywhere.

Open TYPESET

• Is the text in the PDF real text?

  Yes. Your browser lays the document out, and each word is written into the PDF in the same font at the place the browser put it. It can be selected, copied, searched and read by a screen reader. Nothing is turned into a picture.

• Which Markdown does it understand?

  The common kind with GitHub's additions, read by marked: headings, emphasis, links, lists, block quotes, code blocks, tables and horizontal rules. Raw HTML inside the Markdown is kept if it is safe and removed if it is not.

• What happens to a web page's styles and pictures?

  Scripts and style sheets are removed before the page is laid out, so it is set in this tool's own clean style. Pictures stored inside the file are kept; pictures the page links to on the web are left out, because fetching them would tell those sites you opened the page. The result says how many.

• Which languages work?

  Everything DejaVu Sans covers, which is Latin, Greek and Cyrillic in full and many symbols. Chinese, Japanese and Korean characters are not in it and are left out, and the result names them. Arabic, Hebrew and the Indic scripts are written letter by letter without the joining they need, and the result warns about that too.

• Is anything uploaded?

  No. The text is laid out and the PDF written in your browser tab, with the fonts and the PDF writer loaded from this site.

Open TYPEWRITER

• What is a six-minute unit?

  A tenth of an hour. Time is recorded in whole units, so a unit at £250 an hour is worth £25, and an hour is ten units. Any part of a unit counts as a whole one: a call of seven minutes is two units.

• Should each entry be rounded up, or the total?

  That is the firm's choice and the client's agreement, not a rule this page applies. Rounding each entry up means every short call or email costs at least a unit; adding the time first and rounding the total once is cheaper for the client when there are many short entries. The page shows both on every matter, with the difference in pounds.

• Can I paste times from a spreadsheet?

  Yes. One entry a line, with the matter, the fee earner, the time and what was done, separated by tabs or commas. The time can be a start and a finish such as 09:12-09:47, hours and minutes such as 1:05, or 35m, 1h 5m or 0.5h. A plain number is minutes.

• How is VAT worked out?

  On each matter's net total, at 20% or not at all, rounded down to the penny, which VAT Notice 700 section 17.5 allows on the total of an invoice. The net itself is the units times the rates, added up exactly and then rounded to the nearest penny.

• Does it record time or send bills?

  No. It turns time already recorded into units and values, and nothing else. It does not run a timer, keep a ledger, issue a bill, or say what may be charged or recovered. Nothing typed into it leaves the browser.

Open UNITS

• How do I open a RAR file without WinRAR?

  Drop it here. The list of files inside appears first; press Save beside any of them, or Save everything as a .zip to get them all in a format every computer opens. RAR 4 and RAR 5 archives both open.

• How do I extract a .tar.gz or .7z on Windows or a phone?

  The same way: drop it here. tar with gzip, bzip2, xz or lzma over it, 7-Zip and ISO disc images are all read by libarchive running in the tab, with nothing to install.

• Is it safe to open an archive here?

  Nothing in the archive is run, only read. A file whose name would write outside the folder is refused, and the sizes are added up before anything is taken out, so an archive that expands to more than the tab can hold says so instead of crashing it. The archive is not uploaded anywhere.

• What if the archive has a password?

  A zip with a password, ZipCrypto or AES, opens once you type it. A 7-Zip or RAR archive with a password cannot be opened here, because libarchive does not decrypt those formats; the page says so, and 7-Zip or Keka will open it. Nothing guesses passwords.

Open UNPACK

• How do I take VAT off a price that includes it?

  Multiply by the VAT fraction rather than by the rate: the rate over a hundred plus the rate. At 20% that is one sixth, so £120 includes £20 of VAT and £100 net; at 5% it is one twenty-first. Taking 20% off £120 gives £96, which is wrong, because the 20% was charged on the net amount, not the gross one. Choose 'Gross: take the VAT out' and the page uses the fraction.

• Should VAT be rounded up, down or to the nearest penny?

  VAT Notice 700, section 17.5, lets a business round the total VAT on an invoice down to a whole penny. If it works VAT out line by line instead, each line can be rounded down to the nearest 0.1p, or to the nearest 1p or 0.5p, and the total then rounded down to a whole penny. Whichever method is chosen has to be used consistently. The page works out all three and marks the one chosen. HMRC says the round-down concession is meant for invoicing traders and is not generally right for retailers.

• Why does my invoice software give a penny more or less than this?

  Because it uses a different one of the allowed methods. VAT worked on each line and rounded to the nearest penny can come out a penny or two higher than VAT worked on the total and rounded down. The table of the three methods shows the difference for your own figures.

• What does the flat rate scheme part work out?

  What a business on the VAT flat rate scheme pays HMRC for a period: its VAT-inclusive turnover times its flat rate percentage, less one point in the first year of registration. It runs the limited cost test from what was spent on goods: under 2% of turnover, or over 2% but under £1,000 a year (£250 a quarter), means the 16.5% rate applies whatever the trade. The percentage for each trade is not built in; type yours from HMRC's table on GOV.UK.

• Can I put lines at different rates on one invoice?

  Yes. End a line with its own rate, as '45.00 5%', and that line is worked at 5% while the rest use the rate chosen above. With the total method, the exact VAT of every line is added before the one rounding down.

Open VAT

• Why does my photo come out looking like a poster?

  Because tracing finds areas of one colour and draws their outlines. A logo is made of a few such areas; a photograph is made of millions of slightly different pixels, so it becomes a few dozen flat patches at best. There is no way to turn a photograph into a true drawing by tracing, here or anywhere else. For a photo you want at a larger size, keep it as a JPEG.

• How do I get a transparent background?

  Tick Leave out the background colour. The colour that makes up most of the picture's border is taken to be the background, and every shape in that colour is left out of the SVG, so what was behind the logo shows through wherever you place it.

• Why is the SVG bigger than the PNG?

  Each shape is written out as text, point by point. A picture with fine texture or gradients makes thousands of shapes. Fewer colours, a looser fit and a larger speck size all make it smaller; the black and white setting is the smallest of all.

• Can I open the result in Illustrator, Inkscape or Figma?

  Yes. It is a plain SVG of filled paths, one colour per path, with no scripts, links or fonts, so anything that reads SVG opens it and you can recolour or edit the shapes.

Open VECTOR

• What is the IRS mileage rate for 2026?

  It changed mid-year. Business miles driven between 1 January and 30 June 2026 are claimed at 72.5 cents a mile, and miles driven from 1 July at 76 cents. It was the first mid-year adjustment since 2022, made because the average price of gasoline rose 38 per cent over the first half of the year.

• Should I use the standard mileage rate or actual expenses?

  Whichever is larger, subject to the rules. Actual expenses win exactly when your vehicle costs you more per mile than the rate does, so the comparison comes down to your true cost per mile against roughly 74 cents. The standard rate also needs far less record keeping: a mileage log rather than every receipt for the year.

• Can I switch between the two methods later?

  Only if you used the standard mileage rate in the first year that vehicle was used for business. Claim actual expenses in year one and you are shut out of the standard rate for that car permanently. On a leased vehicle, using the standard rate in year one locks you into it for the whole lease.

• Can I deduct my car payment?

  No. Loan principal is never deductible. The interest on the loan is, and so is depreciation on the vehicle, but the payment itself is not an expense. This is the most common mistake in working out actual expenses.

• What does a car actually cost to run?

  AAA put the average for a new vehicle at $11,577 a year over 15,000 miles in its September 2025 study, which is about 77 cents a mile across 45 models. A small sedan came in at 55.87 cents. An older paid-off car is far cheaper because depreciation, the largest single cost, has already happened.

• Is anything I type here sent anywhere?

  No. The whole calculation runs in your browser. There is no email box, nothing is stored when you close the tab, and there is nowhere for the page to send anything. Open your network tab and watch.

• Why does it ask for business miles in two halves of the year?

  Because the business rate changed on 1 July 2026, from 72.5 cents a mile to 76 cents. A 2026 return needs your business miles split at 30 June. A calculator that multiplies your whole year by one figure will be wrong by up to three and a half cents on every mile driven in the second half. On ten thousand second-half miles that is $350 of deduction you either lose or overclaim.

• What are the medical, moving and charitable mileage rates for 2026?

  Medical and moving miles are 20.5 cents from 1 January to 30 June 2026 and 23.5 cents from 1 July. The charitable rate is 14 cents all year; it is fixed by statute, which is why it never moves. The business rate is 72.5 cents for the first half and 76 cents for the second. For comparison, the 2025 business rate was 70 cents for the whole year.

• Can I claim parking, tolls and loan interest on top of the standard rate?

  Yes. Three things stack on either method: interest on a car loan, business parking and tolls, and personal property tax paid on the vehicle's value. Because they apply either way, they never change which method wins. Business parking and tolls only, never a ticket.

  Depreciation is the exception. It is already inside the standard rate, so it cannot be deducted separately on top of it.

• My cost per mile came out well under 50 cents. Is that right?

  Check it. Either you are driving something old and paid for, which is genuinely cheap, or you have left depreciation out. The standard rate is not a windfall: at 76 cents it is close to what an average new car genuinely costs to run, and well below that cost for anything large or new.

• Is this tax advice?

  No, and it is not a substitute for someone who knows your return. It is the shape of the rules, so that the number the calculator gives you means something when you take it to your accountant.

Open What a mile actually costs you

• Should I charge per word or per project?

  Per project, with the per-word figure kept privately as a check. Per-word pricing pays you least exactly when you are doing your best work, because the thinking that makes copy short and sharp takes longer and produces fewer words. Quote the project, deliver the project, then divide afterwards to see whether the price held up.

• What if a client insists on a per-word rate?

  Give them one, calculated from the piece rather than the market. Estimate the hours the piece genuinely takes including research and revisions, multiply by your floor, divide by the word count, and quote that. The number will often be higher than the going rate, and the honest answer to why is that the going rate assumes somebody writing faster and researching less.

• Is a low rate worth it to get started?

  For a short, deliberate period with a specific purpose, sometimes. As a strategy it fails, because a rate is far harder to raise with an existing client than to set correctly with a new one, and low rates attract the clients who are hardest to work with. If you need portfolio pieces, it is usually better to do a small number of jobs at a proper rate than a large number at a bad one.

• Do the numbers I type go anywhere?

  No. It all happens in this tab, there is no email box and nothing is stored. Most rate calculators online are lead capture with arithmetic attached; this one has nowhere to send anything, which you can confirm by turning off your wifi and using it anyway.

• What do copywriters actually charge?

  Useful as a sanity check on your own figure, not as a target. These come from published surveys of different sizes and methods, so treat them as the shape of the market rather than precise values. A survey of 500 freelancers in January 2026 put the average per word at $0.42, and AWAI's 2026 survey put professional copywriting at around $0.70 a word. A 450-writer survey gave blog posts $0.25 a word (range $0.10 to $0.50), SEO articles $0.35 ($0.15 to $0.60), whitepapers $0.80 ($0.50 to $1.50) and technical documentation $1.00 ($0.60 to $2.00).

  By the hour, 350 writers in the US and Europe averaged US$53 in 2025, the US Bureau of Labor Statistics median for writers and authors was $34.75 in May 2024, and the most common band was $25 to $49, where 34.9 per cent of writers sit. For a 1,500-word blog post the most common price was $250 to $399, from 27 per cent of 213 writers surveyed by Peak Freelance.

  One finding is worth more than all of that. In the survey of 500 writers, 78 per cent were charging around 40 per cent below the market rate for their level of experience. If your own number comes out well above what you charge now, you are the norm rather than the exception.

• Why are the same 800 words worth different amounts?

  Because of what the words are for. The same 800 words are worth very different amounts as a blog post, a landing page that carries a product launch, or a sequence that goes to a list of 200,000 people. Technical and regulated writing pays two to four times general content in every survey, and the reason is risk and scarcity, not effort.

• How many rounds of revisions should the quote include?

  Say it in the quote, in a number. Two rounds of revisions within the agreed brief, further rounds at the hourly rate. Almost all copywriting that loses money loses it in round three, on a brief that changed after the work started.

• Should I charge for working out the brief?

  Yes. A clear brief from someone who knows what they want is a different job from extracting the brief yourself through four meetings. If you are doing the second one, that is strategy and it is billable, and quoting it as writing is how a good rate becomes a bad one.

Open What to charge as a copywriter

• Day rate or project fee?

  Quote projects, cost them in days. A client buying a project is buying an outcome and a known price, which is what most want, and a day rate invites them to manage your time rather than the result. You still need the day rate, because it is the only way to know whether the project fee you quoted was any good.

• Should I put prices on my website?

  A starting figure filters out the enquiries that were never going to become work, which is worth a great deal when the invoiceable share of your week is the thing setting your rate. A full price list is harder, because design work varies too much for a menu to be accurate, and a menu invites people to compare on price alone.

• How do I raise rates with existing clients?

  Set a date, give notice well ahead of it, apply it to new projects rather than work in progress, and do not apologise or explain at length. The most common outcome is that the client agrees. The second most common is that you lose the client who was paying least, which usually raises your effective rate on its own.

• Are my figures uploaded?

  No. Everything is worked out in this tab, there is no email box and nothing is stored when you close it. Turn off your wifi and the page still works, which is the quickest way to check.

• What do freelance graphic designers actually charge an hour?

  PayScale, from 234 US freelance graphic designer profiles updated June 2026, put both the average and the median at $38.06 an hour, with the tenth percentile at $20.68 and the ninetieth at $85.11. Treat that as a sanity check rather than a target. Rates vary enormously by specialism, city and client size, and a single average hides most of what matters.

  The spread is the useful part. A four-times gap between the tenth and ninetieth percentile is not mostly explained by skill. It is explained by who the client is, whether the work is priced by the hour or by its value to the business, and whether the designer is selling execution or judgement.

• How many revision rounds should the quote include?

  A number. "Reasonable revisions" is not a term, it is an argument waiting to happen. Two rounds within the agreed brief, further rounds at a stated figure, and a definition of what counts as a new brief rather than a revision. A project fee is a bet on how many days the work takes, and the bet is lost in the revision rounds nobody wrote down. Put the extra-round price in the quote as a number rather than a warning, because a number is what stops the fourth round happening.

• What do I do when the client keeps adding small extras?

  Price the deliverables as a list, and treat anything not on the list as a new quote rather than a favour. The logo becomes the logo plus social variants plus a one-page brand sheet plus a favicon, each request small enough that refusing feels petty. That is where the money goes missing.

• Should source files be included in the fee?

  Decide up front what the client receives and what they do not, and say whether source files are included and at what price. This is a normal commercial term, not a trick, and it is far easier to state at the start than to negotiate on the last day.

• What is a kill fee?

  A stated cancellation term. A project cancelled halfway has still cost you the days. A cancellation fee, commonly a percentage of the fee depending on the stage reached, is standard in the trade and costs nothing to include.

Open What to charge as a graphic designer

• Where do these multipliers come from?

  They are a common working model in illustration rather than a published standard, and they are here to give you a defensible starting position rather than a correct answer. Rates vary by sector, by country and by client, and a well-known agency will have its own grid. Use these to structure the conversation, then adjust to what you know about your own market.

• What if the client insists on buying the copyright?

  Price it properly and let them decide. Full assignment means you can never license that image again, cannot reuse elements of it, and may need permission to show it in your own portfolio. A multiple of the creation fee in the region of three times is a normal starting point. A surprising number of clients who ask for copyright as a default withdraw the request once it has a price attached, because what they actually wanted was certainty that a competitor would not use the image.

• Should I quote a day rate or a project fee?

  A project fee, in almost every case, with the day rate used privately to cost it. The survey above found most illustrators do the same. A day rate invites the client to count your days, and it caps you at the speed you can draw, which is the wrong thing to be paid for once you are good.

• Do my numbers leave this page?

  No. It is all worked out in the tab, there is no email box, nothing is stored, and you can confirm it by turning off your wifi and using the page anyway.

• What do illustrators actually charge a day?

  The most useful recent data is a survey of 403 illustrators run in July 2026, 146 of them in the UK. Of the 90 UK illustrators who gave firm figures, the median day rate was £350 and the middle half sat between £280 and £450. Over a quarter charged under £300 a day and 9 per cent charged over £500.

  Two things stand out. Most illustrators do not price by the day at all: 41 of 342 priced by day rate and 174 by project fee, so a day rate is a costing tool rather than a quote. And only one in five felt fairly paid, with the people who felt worst about it clustered at the bottom of the range, which suggests the problem is pricing confidence rather than the market.

• Why should the quote show the creation fee and the licence fee as separate lines?

  Because clients routinely believe they are one thing, and quoting only the creation fee is the most common and most expensive mistake in the trade. In the July 2026 survey, more than a fifth of written responses raised licensing as the central problem, and the recurring complaint was that clients assume paying once buys everything forever. It does not, unless you let it. One respondent put the commercial reality plainly: full copyright roughly triples the final fee.

  Distinct "creation fee" and "licence fee" lines make the conversation about scope instead of about whether you are expensive, and they make it obvious what the client is buying when they ask for worldwide rights in perpetuity as though it were a formatting preference.

• What do scope, duration and exclusivity mean on the licence?

  Scope is where the work appears: one magazine is not the same as every billboard in the country. Duration is how long the right lasts, and unlimited is a real cost because it forecloses every future licence of that image. Exclusivity is whether you can license it to anyone else. Full assignment means you cannot even use it in your own portfolio without asking, which is worth pricing accordingly and often worth refusing.

  On this page the three terms add rather than compound, and the scale is set so that a full copyright assignment on its own lands at roughly three times the creation fee, which is where the trade generally puts it. It is a working model for opening a negotiation, not an industry tariff.

Open What to charge as an illustrator

• Should I charge by the hour at all?

  Often not. Hourly pricing caps what you can earn at the hours you can stay awake, and it penalises you for getting faster, which is the opposite of how expertise should pay. The hourly number is still worth knowing, because it is what tells you whether a fixed project fee was any good once the work is finished. Work out the floor, price by the project, then check the project fee against the floor afterwards.

• What should the invoiceable percentage really be?

  Track it for a month rather than guessing. Note the hours you could put on an invoice against the hours you worked. Most people who do this for the first time find the figure is lower than they assumed, often between forty and sixty per cent, and that the difference explains why the money never quite adds up.

• Does this account for tax properly?

  No, and it does not try. It applies one percentage you supply to cover income tax, self-employment or national insurance contributions, and anything else you set aside. Real tax is progressive, varies by country and by how you are structured, and depends on deductions this page knows nothing about. Use a figure your accountant gives you. This is arithmetic, not tax advice.

• Is my income sent anywhere?

  No. The calculation runs in this tab, there is no account, no email box and no analytics on the numbers you type, and nothing is kept when you close it. You can check that: open the network tab and watch, or turn off your wifi and use the page anyway.

• Why does the invoiceable share move the rate so much?

  Because a freelancer who works forty hours does not bill forty hours. Some of that week goes on quoting for work that never lands, writing proposals, invoicing, chasing invoices, bookkeeping, updating a portfolio, answering enquiries that go nowhere, and the admin of running a business that has no admin department.

  Sixty per cent is a realistic figure for someone established with steady clients. Newer freelancers, or anyone doing a lot of pitching, are often nearer forty. At forty per cent you need to charge half as much again as you would at sixty for exactly the same take-home. That single number moves the rate more than anything else on the form.

• Why 46 working weeks and not 52?

  Holiday and illness are unpaid. Nobody covers them, and if you price at fifty-two weeks you have quietly decided to work every week of the year or take a pay cut. Forty-six is a normal working year with four weeks off, a week of illness, and a week lost to the parts of the year that are not work but are not holiday either.

• Is the number it gives me the rate I should charge?

  Treat it as a floor, not a price. It is the point below which you are subsidising the client out of your own time. It says nothing about what the work is worth to them, which is usually the more important question and always the more profitable one. Value, scarcity, urgency and how much money the work makes or saves are all reasons to charge more than the floor. There is no good reason to charge less than it.

  If the floor comes out higher than what you currently charge, that is the useful outcome. The gap is not a sign you are doing something wrong; it is the amount you are currently absorbing.

Open What you actually need to charge

• How many pages is 80,000 words?

  About 320 manuscript pages at the usual 250 words a page in standard manuscript format. A printed book of the same length is usually fewer pages, since a book page holds more words; how many depends on the trim size and the type.

• How many words is a novel?

  The Nebula Awards count anything of 40,000 words or more as a novel. For a first novel, agents and publishers commonly give ranges such as 70,000 to 100,000 words for general and literary fiction and 90,000 to 120,000 for science fiction and fantasy. They are guides rather than rules.

• Why is a manuscript page counted as 250 words?

  Standard manuscript format, 12-point Courier, double-spaced with one-inch margins, holds about 250 words on an average page, and publishers have long estimated length that way. Dialogue-heavy pages hold fewer and a proportional font fits more.

• Does it count the words in a Word file the same way Word does?

  Usually to within a few words. Both count what sits between spaces; they can differ on a dash or a symbol standing on its own between spaces, which this never counts as a word. Footnotes and text boxes in a Word file are not counted here.

• How long does a book take to read?

  At 250 words a minute, a common pace for reading prose silently, an 80,000-word novel takes about five hours and twenty minutes. Readers vary a good deal, so the pace can be changed on the page.

Open WORDCOUNT

• How do you move a file to a computer with no network connection?

  Display it as a sequence of codes on one screen and read them with the camera on the other. Nothing is plugged in and no network is involved, so the only channel between the two machines is light. It is slow, and that is rather the point: bandwidth that low is hard to abuse.

• Why not just use a USB stick?

  A USB device is a two-way channel that the receiving machine has to mount and parse, which is exactly how several well-documented air-gap compromises worked. A camera reading a screen carries only pixels, and the receiving machine decides for itself what to do with them.

• How much data can realistically cross this way?

  Kilobytes comfortably, a few megabytes with patience. It suits keys, configuration, signed documents and text. It is not how you move a video, and a file that large is usually a sign that the air gap is not really the constraint.

• How does the receiver check what it has read?

  Frames are reassembled in memory by sequence number and the result is verified against a CRC32 checksum before anything is shown to you. Nothing is written to disk and nothing leaves this tab.

• What does the loopback self-test prove?

  That the codec works on this machine. It does not use the camera: it renders the frames and reads them straight back through the same decoder. It is not a test of the optical hop.

Open AIRGAP

• Is it safe to paste work documents into an AI chatbot?

  Read the terms of the specific product. Consumer tiers frequently retain conversations and may use them to improve the service; business and enterprise tiers usually do not. Beyond the terms, the practical risk is ordinary: the text sits in a history that syncs across devices, and a shared or compromised account exposes it.

• How do I use an AI tool without disclosing client names?

  Substitute consistently before you paste. Replace each real name, company and identifier with a stable placeholder, keep the mapping locally, and put the real values back in the output. The model can still reason about the structure of the problem, which is usually what you wanted from it.

• What should never be pasted into a chatbot?

  Credentials and keys, anything under a confidentiality agreement you have not checked, personal data about other people that you have no basis to share, and material subject to legal privilege. If a disclosure would be a problem in an email to a stranger, it is a problem here.

• Do I need to download the key?

  Only if you will need to restore a reply later, or on another machine. The key exists only in this tab and is gone when the tab closes. Treat the downloaded file as the sensitive one: it is the list of who is who.

Open ALIAS

• How do I open a .webarchive file on Windows?

  Drop it on this page. AMBER reads the Safari web archive in your browser and shows the page as it was saved, with its pictures, stylesheets and fonts. Save it as one .html file and it opens in any browser on Windows. Nothing is uploaded and nothing needs installing.

• Can I convert a .webarchive to HTML?

  Yes. Save it as one .html file. Every picture, stylesheet and font the archive holds is written into that single file, so it does not depend on the internet or on the site still existing. Scripts are left out and it carries a policy that allows no network, so it is also safe to keep and share.

• How do I convert a .webarchive to PDF?

  Save it here as one .html file, open that in any browser and print it to PDF. On the Mac that saved it, Safari's File, Export as PDF gives the closest copy of how it looked.

• Why does my .webarchive look like gibberish in Word or Notepad?

  Because it is not text. It is a binary property list, Apple's structured file format, holding the page and every file it used. It has to be taken apart by something that reads that format, which is what this page does.

• Why are some pictures missing?

  A web archive holds what Safari had loaded when it was saved. Anything the page asked for that is not in the file is left out and listed rather than fetched from the internet, because fetching it would tell the site that the page was opened. Pictures that load as you scroll are only there if Safari had already loaded them.

• Is it safe to open a .webarchive from someone else?

  Here, yes: nothing in it runs. Scripts, event handlers, javascript: links, plugins and redirects are taken out before the page is shown, and it is shown in a sandboxed frame with no network. The archive itself can carry every script the page had, which is why none of them is kept.

Open AMBER

• Is it safe to paste a live SAML response here?

  It is decoded and checked in this tab, and nothing is sent anywhere; the page works the same with the network off. A live response is still a working login until it expires, so treat the screen it is on the way you would treat a password.

• What is XML signature wrapping?

  An attack on the way SAML signatures point at what they sign. The attacker keeps the genuinely signed assertion in the message so the signature still checks, moves it somewhere the application does not read, and puts a forged assertion where it does. ASSERT reports the shapes that gives a message: a signature pointing at an element other than the one it sits in, a reference to an ID that is not there, duplicate IDs, and more than one assertion.

• Does it verify the signature?

  No, and on purpose. Doing it correctly needs exclusive XML canonicalisation reproduced exactly and the identity provider's certificate from configuration, not from the message. A verifier that is nearly right is worse than none. ASSERT shows the shape of every signature and the fingerprint of the certificate that came with it, for comparison with the one the service provider trusts.

• Why does a SAMLRequest in a URL look like random letters?

  The redirect binding compresses the XML with deflate, then base64-encodes it, then URL-encodes that. ASSERT undoes all three. A SAMLResponse posted in a form is base64 without the compression.

• Why is a comment in the NameID serious?

  Because several SAML libraries read an element's text only up to the first comment. An attacker registers [email protected]<!---->.attacker.test, gets it signed by the identity provider, and the application reads it as [email protected]. No legitimate identity provider has a reason to put a comment there.

• Can it read an encrypted assertion?

  No. An EncryptedAssertion opens only with the service provider's private key, which should never leave its server. ASSERT says the assertion is encrypted, which is good, and reads everything around it.

Open ASSERT

• Is the file uploaded anywhere?

  No. It is read into memory by this tab, parsed here and rewritten here. Nothing is sent, which is why the page works with the network turned off.

• Will the stripped video still play?

  That is the whole point of the offset repair, and there is a Play it here first button so you can check before you save. If anything about the rewrite does not add up, the file is not offered to you at all.

• Does it remove the GPS location from a phone video?

  Yes, and both copies of it. A phone writes a QuickTime location tag into udta or into the Apple keys, and those are removed; the page lists the entry so you can watch it go.

  That is often not the only copy. iOS also records location into a timed metadata track: a whole track alongside the picture, with a reading every second or so for the length of the clip, so what it carries is a route rather than a place. Those samples sit in the media data rather than in a metadata box, and they are taken out too: the track goes out of the movie header, its samples come out of the media data, and every remaining track's chunk offsets are repaired for the bytes that left. What comes back is a file that still plays, with the route gone.

• What is a timed metadata track, and how is it taken out?

  A track, like the video track and the audio track, whose samples are recorded metadata rather than pictures or sound. iOS uses one for location, and a clip of any length carries a reading every second or so.

  Removing it is two jobs, and doing either alone is worse than doing neither. Taking the track out of the movie header leaves every coordinate in the media data with nothing pointing at it, and the page would then say the location had gone. Taking the samples out without repairing the offsets gives you a file that is the right length, parses box by box, and plays in nothing, because every chunk offset in the remaining tracks still names a position the bytes have moved from. So both happen together: the track's byte ranges are worked out from its own sample tables, those ranges come out of the media data, the mdat box shrinks by what left, and every surviving offset is rewritten. If a track's sample table does not read the way the format describes, none of it happens and the page says so rather than guessing.

• Why does it say it cannot rewrite my file?

  The common reasons are a container it does not read, a metadata box sitting inside the media data, an offset table that points into something being removed, or an Ogg file holding several streams. In each case it refuses rather than hand you a broken file.

• What about the creation date in mvhd?

  It is shown but not removed, because zeroing it changes no box size and would be a silent edit to a structural field some players read as the duration anchor. If it matters, note that it is there; a re-encode is the reliable way to lose it.

• Can it strip a .mkv or .webm?

  No. Matroska is an EBML container with a different shape, and doing it properly is its own piece of work rather than a corner of this one.

• What happens to album art?

  It is metadata, so it goes: a cover in an MP4 ilst box, an APIC frame in ID3, a PICTURE block in FLAC. The before and after list shows it going.

• Does it remove the encoder's fingerprints from the video or audio?

  No. Metadata is a labelled field in a box, a chunk or a frame, and those are what this removes. The compressed picture and sound underneath carry their own marks: the choice of quantisation tables and slice layout, the exact group-of-pictures structure, the padding in the last frame, the LAME or Xing header that lives inside the first MPEG audio frame rather than beside it. A laboratory comparing two files can often say they came from the same encoder, the same settings and sometimes the same device, with every metadata field gone.

  Removing that would mean decoding and re-encoding the media, which is a different job, costs a generation of quality, and is not what this page does. If that matters to your situation, re-encode the file with a tool you have chosen and then run the result through here.

• Why was I not offered the stripped file?

  Because the check failed. The new file is read back with the same parser before it is offered, and the bytes removed are compared with the difference in file size. If the sizes do not add up, or the new file no longer parses, nothing is offered and nothing on your disk has changed. The before-and-after list is built by parsing the new file, not by assuming what the strip did.

Open AVSCRUB

• Is it safe to send a photo of my passport or driving licence?

  It is one of the most useful documents an identity thief can obtain, and once it is sent you have no control over where the copy travels. When it is genuinely required, send the least you can, mark the copy for its specific purpose, and follow up on deletion.

• What does writing across a scanned ID achieve?

  A clear diagonal mark naming the purpose and the date makes the copy obviously unsuitable for reuse elsewhere and makes any later misuse easier to trace. It has to be part of the image itself rather than an annotation layer, and it must not obscure whatever the recipient legitimately needs to read.

• What should be covered up?

  Anything not needed for the check being made. The machine-readable strip at the foot of a passport encodes the number, date of birth and expiry in one line and is rarely required. The same goes for a document number when only the name and photograph are being verified.

• Why black out the barcode on the back of a driving licence?

  That dense square or rectangle is a 2D barcode holding your full record in plain, unencrypted form, often more than the front of the card shows. A human reviewer never needs it. Black it out.

• Which parts of an ID does a check actually need?

  Most checks need your face, name, date of birth and expiry, and nothing else. Leave those and cover the rest: document, licence or social-insurance numbers you were not asked for, and the passport MRZ. The MRZ is the two lines of monospaced text and chevrons at the bottom of the photo page. They encode your number, dates and a checksum that automated systems read in bulk, so cover it unless a specific process truly needs it.

• Why does a clean scan of an ID matter so much?

  Landlords, rental desks, hiring platforms and marketplaces ask for one constantly, and when their inbox or portal is breached a clean scan is gold for a thief. BASTION prepares a copy bound to one purpose instead: the high-risk data blacked out, a heavy watermark tying it to the recipient and the date, and the whole thing flattened with the metadata gone.

Open BASTION

• What forms of coordinate does it read?

  Decimal degrees (51.4778, -0.0015), degrees and decimal minutes (51°28.668′N 000°00.090′W), degrees minutes and seconds (51°28′40.1″N 0°00′05.3″W), and MGRS references at any precision from one to five digits. Hemisphere letters may come before or after, and a minus sign works instead. Latitude always comes first. Whichever form it settled on is printed with the answer, and anything it cannot read unambiguously it refuses rather than guesses.

• Why are the starting and finishing bearings different?

  Because the shortest path between two points on a globe is a great circle, and a great circle crosses each meridian at a different angle. Setting off from London to New York on the initial bearing and holding it will not get you there. Over short distances the two bearings are almost identical; over long ones they differ by tens of degrees, and both are given so that neither is mistaken for the other.

• Which is the right distance, Vincenty or haversine?

  Vincenty, essentially always: it models the earth as the flattened ellipsoid it is and is accurate to well under a millimetre. Haversine is shown beside it because it is what most tools give you and because seeing the gap is instructive: a few centimetres across a town, tens of metres across a continent. The one case where Vincenty is not available is two points almost exactly opposite each other on the earth, where the iteration does not converge; the page says so plainly rather than quietly falling back to the spherical figure.

• Why no Ordnance Survey grid reference?

  Because doing it correctly requires OSTN15, the grid of corrections the Ordnance Survey publishes, and the seven-parameter Helmert transformation that free converters actually use is roughly five metres out. Almost none of them tell you that. Five metres is irrelevant for a walk in the hills and quite relevant for a scene, a boundary or a search area, so the honest options were to ship OSTN15 or not to offer it, and this does not offer it.

• Is MGRS exact here?

  Yes. MGRS is a lettered way of writing a UTM coordinate, and UTM is a projection of WGS84 rather than a different datum, so there is no approximation involved, only the truncation you choose. A five-digit reference is to the metre, four to ten metres, three to a hundred, and the page tells you which. The zone exceptions and the twelve-degree X band are implemented, and a reference converted out and back returns the same square.

• Does it work without a network?

  Yes. It is arithmetic, and once the page has loaded it needs nothing. That matters for the people most likely to want it, who are frequently somewhere with no signal.

Open BEARING

• What is blind recruitment and does it work?

  Removing names, addresses, schools, dates and other identity signals from applications before anyone reads them. Evidence from field experiments is mixed but points the same way as the audit studies that prompted it: identical CVs with different names get called back at measurably different rates, and removing the signal removes that specific effect at that specific stage.

• Which fields should be removed from a CV?

  Name, contact details, photograph, date of birth, nationality, marital status, and the names of schools and universities. Graduation dates are worth removing too, since they reveal age. What stays is the work: roles, responsibilities, achievements and skills.

• Does anonymising CVs fix hiring bias?

  It addresses one stage. The interview that follows is not anonymous, and the criteria being applied may themselves favour one group. Blind screening is a useful intervention with a narrow scope, not a solution to the whole problem.

• Who keeps the unblinding key?

  The reviewers get the packets. The person running the process keeps the key, a CSV that maps Candidate #N back to the source file and the name that was detected, so a shortlist can be un-blinded when the review is done. Your reviewers stay blind and you lose nothing.

• What does BLIND take out of a resume?

  The things that trigger unconscious bias: the name, the contact details, the home address, the personal links, the graduation years that hint at age, the photo and the gendered pronouns. Removing dates, neutralising pronouns and removing postal codes are each a tick box you can turn off. What comes back is a clean packet labelled Candidate #1, #2 and so on.

Open BLIND

• How do I open a PST file without Outlook?

  Drop it on this page. CABINET reads the .pst in your browser and lists every folder, and any email opens with its text, pictures and attachments. Nothing is installed, it works on a Mac, a Chromebook or Linux, and the file is not uploaded.

• How do I convert a PST file to MBOX?

  Open the .pst here and choose Save every email in the file as one .mbox, or save one folder at a time. Thunderbird and Apple Mail import an .mbox, and POSTBAG opens one in the browser.

• Can I open a PST file on a Mac?

  Yes. Apple Mail does not open a .pst, but CABINET reads one in any browser on a Mac, and each email or folder can be saved as .eml or .mbox, which Apple Mail opens and imports.

• How do I get attachments out of a PST file?

  Open the email and save each attachment on its own. An email attached to an email saves as an .eml with its own attachments inside it.

• Can it open an OST file?

  Yes, including the 4 KB page format recent versions of Outlook write .ost files in. An .ost is a copy of a mailbox that lives on a server, so what it holds is what had been downloaded to that computer.

• Does it need the PST password?

  No. A password on a .pst is checked by Outlook and does not encrypt the file. Files made with Outlook's high encryption setting are the exception, and CABINET says so rather than showing nothing.

Open CABINET

• Can writing style identify an anonymous author?

  It can narrow a field. Stylometry compares a text against writing by a set of candidates, using the habits nobody thinks to edit: how often ordinary function words appear, sentence-length rhythm, punctuation preferences, spelling conventions. It needs candidates to compare against, so in practice it identifies people who were already suspected rather than producing a name from nowhere.

• What gives an anonymous writer away most often?

  A reused phrase. A distinctive turn of words that also appears in something published under your own name is searchable, and it needs no statistics at all. After that: spellings that place you geographically, formatting habits such as two spaces after a full stop, and knowledge of a subject that only a few people have.

• How do I write anonymously?

  Normalise the habits, then have someone else read it. Standardise the spelling to one convention, remove the punctuation you favour, vary the sentence lengths, and search your distinctive phrases against everything already published in your name. Then remember that the metadata in the file, the account you post from and the hours you post at are usually easier to trace than the prose.

• Can CADENCE tell me whether I would be identified?

  No, and it does not try. Real authorship analysis compares a text against writing by a set of candidate authors, and the whole point is the comparison; this tool has no corpus and never sees anything but the text you paste. What it can do is point at habits that are distinctive enough to be worth normalising, and at phrases you should search for in anything already published under your name. Treat a clean result as nothing obvious, never as safe. The measurements get noticeably less reliable below a few hundred words. Everything is computed in this tab and nothing is stored.

Open CADENCE

• What does a calendar invite actually reveal?

  More than the time and the place. An .ics file lists every invitee's email address and the organiser's, along with the dial-in number, the meeting passcode and the join link sitting in the description, the software that produced it, and a unique identifier that often contains an internal hostname. Forwarding one hands all of that over.

• Is it safe to post a meeting invite publicly?

  Not without editing it first. The join link and passcode let anyone into the meeting, and the attendee list is a ready-made map of who works with whom. Both have caused real intrusions into public-body meetings. Strip the attendees and the join details before the file leaves your hands.

• Can properties be removed without breaking the file?

  Some, not all. The calendar format requires a few properties to be present, so a cleaner has to replace those rather than delete them: removing the unique identifier means generating a fresh random one, and the same goes for the producing software and the timestamp. Everything else can simply go.

• Can it clean an invite that has already been sent?

  No. CALSCRUB reads and rewrites the file you give it, in your browser. It cannot reach the copy already sitting in someone else's calendar, in the original email or on the server that sent it. Once an invite has gone out, this cleans the copy you are about to share, not the ones already delivered.

• Does the cleaned file still open as a calendar entry?

  Yes. Properties the calendar format requires are replaced rather than deleted, so removing the UID gives you a fresh random one instead of a broken file, and the same goes for the producing software and the timestamp. Everything else you tick is removed outright.

Open CALSCRUB

• How do you find out who leaked a document?

  Give each recipient a copy that differs in a way they will not notice but you can recognise later. Traditionally that meant slightly different wording; in a digital file it can be invisible characters, spacing, or small differences in formatting. When a copy surfaces, the variant tells you which recipient it came from.

• Is this the same as a watermark?

  A visible watermark deters. A per-recipient marker attributes. They do different jobs and are often used together. A marker only works if it survives whatever the leaker does, and a screenshot destroys invisible characters while retyping destroys everything, so no marking scheme is reliable against a careful leaker.

• Is it legal to track a document this way?

  Marking your own documents is generally unremarkable. What can matter is whether the recipients were told, particularly where they are employees, and whether any personal data is being processed in the tracking. Practice and law vary by jurisdiction, and this is not legal advice.

Open CANARY

• How do I send an encrypted file to someone who has no software?

  Wrap the encrypted file in a small self-contained page that any browser can open. The recipient opens it, types the passphrase, and the file is decrypted on their machine. Nothing to install, no account, and the passphrase travels by a different route from the file.

• Is password-protecting a zip file good enough?

  It depends entirely on the format. Modern zip encryption using AES is sound; the legacy zip cipher that older tools still produce by default is broken and recovered in seconds. You often cannot tell which one a file used from looking at it, which is reason enough to use something where the algorithm is not in doubt.

• What makes a passphrase strong enough for a file?

  An encrypted file is the worst case: the attacker has the whole thing and can guess offline at their own speed, forever, with no rate limit. That calls for several unrelated random words rather than a single word with substitutions, and it has to reach the recipient by a different channel from the file itself.

• I have lost the passphrase. Can the capsule be recovered?

  No. The passphrase is the whole strength and there is no recovery. Send it by a different route from the capsule, for example say it on a call or use a different app, and never in the same email as the file.

• Does the recipient need anything installed?

  No. A capsule opens on its own in any browser, offline if they like. They type the passphrase and save the originals. No software on their end, no account, no server holding a copy. It is AES-256-GCM with a PBKDF2 key, all done with the browser's own cryptography. If you would rather open a capsule on this page, the last panel takes the file and the passphrase.

Open CAPSULE

• What can I put in?

  A certificate as PEM text, a whole chain of them pasted one after another, a binary DER file, or a .p7b bundle as Windows exports a chain. Certificate signing requests and bare public keys are read too. A private key is recognised and refused rather than parsed.

• Does it check the certificate is trusted?

  It checks the chain you give it: each certificate against the one above, the dates, the names, the key and the extensions. It cannot tell you whether the top of that chain is in your operating system's trust store, because a page in a browser is not allowed to read that list. It will tell you when the chain does not reach a self-signed root, which is the usual reason a certificate works on one machine and not another.

• Why does it say my certificate is missing a name?

  Because the hostname is only in the common name field. Browsers stopped reading that in 2017 and now require a subject alternative name entry. A certificate like that will be rejected as having the wrong name for the site, which is a confusing error given the name is plainly there.

• What counts as a key too short?

  RSA under 2048 bits, which is below what public certificate authorities have been allowed to issue since 2014 and is refused outright by current browsers. Elliptic curve keys are flagged below 256 bits. A short key is not weak in a way you would notice day to day; it is weak in a way that fails suddenly when something updates.

• It says the chain is out of order. Does that matter?

  Often yes. A server is supposed to send the leaf first and each issuer after it. Some clients repair the order and some do not, so a chain that works in a browser can fail in a payment library or an old phone. The order is worth fixing even when nothing has complained.

• What is the tree view for?

  It shows the same file as a tree, for when a field is not where it should be. Tags, lengths and decoded values, straight out of the parser. Copy what was read puts the decoded text on the clipboard.

• Why read a certificate that already says valid?

  A certificate is a signed statement about who holds a key, what they may use it for, and until when. Everything in it is public by design and most of it is never looked at, which is how a chain goes out with the wrong intermediate, a wildcard nobody meant to issue, or a key too short to be worth the ceremony. CERTLENS takes one apart, checks each signature against the certificate above it, and says what is wrong in words.

• How do I open a .p7b file?

  Drop it on CERTLENS. A .p7b (or .p7c) is a PKCS #7 bundle that only carries certificates, usually a whole chain exported from Windows. Every certificate in it is read, put in order from the leaf up, and checked against the one above it. OpenSSL's equivalent is openssl pkcs7 -print_certs -in file.p7b.

Open CERTLENS

• What is a SHA-256 checksum for?

  It is a fingerprint of a file. The publisher computes it and posts it, you compute it on the file you downloaded, and you compare. If a single byte differs the two values differ completely, so a truncated, corrupted or substituted download shows up immediately.

• Does a matching checksum mean the file is safe?

  It means the file matches the value you compared it against. If you took that value from the same page that served the download, then whoever controlled the page controlled both. A checksum is only as trustworthy as the channel the expected value arrived on, which is why publishers sign their checksum files.

• Should I use MD5, SHA-1 or SHA-256?

  SHA-256. MD5 and SHA-1 are both broken for this purpose: it is practical to construct two different files with the same MD5 or SHA-1, so a match no longer proves the file is the one intended. They will still catch accidental corruption, but not a deliberate substitution.

• Is my file uploaded to compute the hash?

  No. CHECKSUM reads the file in this tab and computes the hash with your browser's built-in cryptography. Nothing is sent anywhere.

• What does it mean when the hashes do not match?

  A matching hash means the file is byte for byte the one the source published. A mismatch means the file was changed or corrupted in transit, or that you are comparing against the wrong value. Check that the published hash is for the same file, version and algorithm before you assume the download is bad.

• Which sites publish a hash for their downloads?

  Sources that publish apps, ISOs and installers often post a hash beside the download link so you can confirm the file arrived exactly as intended and was not swapped or corrupted. Only compare against a hash from a source you trust: a hash confirms integrity, not safety.

Open CHECKSUM

• How can I send a message securely when neither of us wants to install anything?

  Encrypt the text into a block of characters and send that block through whatever channel you already use. The recipient pastes it into the same page and gets the message back. What has to travel by a different route is the passphrase, or in key mode only the public key, which is safe to send in the open.

• What does encryption in a browser actually protect?

  The contents of the message, in transit and wherever it comes to rest. It does not hide that you sent something, who you sent it to, when, or how long it was, and it cannot help if the device at either end is already compromised. Encryption protects the message, not the fact of the message.

• Should I use a passphrase or a key pair?

  A passphrase is simpler and is right for a one-off to someone you can reach by another route. Its weakness is that anyone who copies the encrypted block can guess at the passphrase offline for as long as they like, so it needs to be several unrelated words. A key pair removes the shared secret: the recipient publishes a public key, you encrypt to it, and only their private key opens the result.

• What encryption does CIPHER use?

  The encryption built into your browser. Messages are sealed with AES-256-GCM. In passphrase mode the key comes from PBKDF2 with 310,000 rounds of SHA-256. In key mode the pair is ECDH on P-256. Nothing is stored here: not the message, not the passphrase, not the keys.

• What happens if I lose my private key?

  The messages encrypted to it can never be read again, by you or by anyone. The private key is the only thing that can open what was sent to its public key, and it is not stored anywhere, so keep your own copy. The public key is safe to give to anyone who writes to you.

Open CIPHER

• What should I do to a file before sending it outside the organisation?

  Remove the metadata, remove anything hidden or tracked, and remove anything active. In practice that means the author and revision history, comments and tracked changes, hidden rows and slides and layers, embedded objects and macros, and for images the EXIF block. Then reopen the result and look at it before it goes.

• Does saving as PDF remove everything?

  No. A PDF carries its own metadata, usually inherits the author from the application that made it, and can retain layers, annotations and embedded files. It does remove tracked changes and hidden cells, which is genuinely useful, but it is not a sanitiser.

• What gets missed most often?

  The filename, and the properties dialog nobody opens. After that, in order: comments left in a document, hidden columns in a spreadsheet, presenter notes in a deck, and the original photograph behind a cropped image. All four have caused public embarrassments.

Open CLEANROOM

• What hides inside an Excel file?

  Hidden rows, columns and entire worksheets. Cells that are only white text on white. Filters that hide rows without deleting them. Defined names pointing at ranges you thought you removed. Comments and notes. The author and last-modified-by fields. And cached values from external links and pivot tables, which keep a copy of the source data.

• Does deleting a sheet remove its data?

  Deleting the worksheet removes it, but a pivot table built from it keeps a cached copy of the source inside the file, and formulas elsewhere may hold the last calculated values. A pivot cache can be double-clicked to regenerate the underlying rows.

• Is hiding a column the same as removing it?

  No. A hidden column is still in the file with every value intact, and one right-click unhides it. Anything you would not hand over has to be deleted and the file saved again.

• What does sanitizing do, exactly?

  Formulas are replaced by the values they currently show, so the numbers stay and the maths behind them is gone. Hidden and VeryHidden worksheets are deleted outright, not just re-hidden. Pivot tables and their cached data are removed, so no one can drill back into the raw source rows. Cell comments and notes are stripped. Links to other files, and the folder paths they reveal, are removed.

• Does sanitizing change my original file?

  No. The sanitized copy is a new file, so your original is never changed. Sanitizing does its best to keep the visible content and layout intact. Always open the clean copy to confirm it looks right before you send it.

• Is my spreadsheet uploaded?

  No. CLEANSHEET runs entirely on your device. Nothing about your spreadsheet is uploaded anywhere, and scanning it is free.

Open CLEANSHEET

• Does a photo contain my location?

  If it was taken on a phone with location services enabled for the camera, yes. The JPEG carries an EXIF block, and inside it GPS coordinates to roughly five decimal places, which is a few metres. It usually also holds the exact time to the second, the camera or phone model, and on some cameras a serial number. None of it is visible when you look at the picture.

• Does posting to a social network remove it?

  Most large platforms strip EXIF from the copy they serve publicly, but not all of them do and not on every upload path. A file attached to a message, uploaded to a forum, sent by email or put in cloud storage normally keeps everything. Relying on the platform means relying on a behaviour that is undocumented and changes without notice.

• Does cropping or screenshotting remove the metadata?

  A screenshot does, because it produces a new file from pixels. Cropping in most editors does not: the editor rewrites the image data and copies the EXIF block across unchanged, so the GPS coordinates survive the crop.

• What is the second picture this page found in my photo?

  The thumbnail EXIF keeps inside the file. Your camera wrote it when the shutter fired (a complete JPEG of its own, usually 160 by 120 pixels) so that a picture browser could show a preview without decoding the whole image. Cropping or redacting the photo afterwards does not oblige an editor to redraw it, so it can still show the frame you removed. This page displays it beside the photo, at its own size, rather than telling you one exists; why a cropped photo still shows what you cropped out goes through where it sits in the file. Exporting a cleaned copy removes it with everything else.

• Is the redaction permanent?

  Yes. Redaction replaces the pixels themselves, so there is no hidden original layer underneath, unlike a black box drawn over a PDF. What was under the box is gone from the exported file.

Open CLOAK

• What counts as large text?

  WCAG 2.1 puts the line at 18 point, which is 24 CSS pixels, or 14 point bold, which is about 18.66 CSS pixels. Below that a pair needs 4.5:1 for AA and 7:1 for AAA; at or above it, 3:1 and 4.5:1. Bold counts because a thicker stroke puts more coloured pixels on the page.

• Why does my other checker say a different number?

  Almost always rounding. This page truncates at two decimals so a failing value cannot print as a passing one, and some checkers round. If the gap is larger than a hundredth, check whether the other tool is applying alpha: a colour with transparency has no contrast ratio until you say what is behind it, and this page ignores the alpha channel rather than guessing.

• Does a 3:1 pass mean my buttons are fine?

  It means the visual boundary of the component meets 1.4.11. The label inside it is text and still needs 4.5:1, the focus indicator has its own requirement, and a control that is only distinguishable by colour fails 1.4.1 no matter what the ratio is.

• Can I check a colour with transparency?

  Not directly, and no honest checker can. Contrast is measured between two solid colours; a colour at 60 percent alpha has a different luminance over white than over a photograph. Composite it against its real background first and check the result.

• Where do the palette colours come from when I drop a file?

  Every hex value, rgb() and hsl() in the text is pulled out with a pattern match and the duplicates dropped. It does not understand your stylesheet, so a colour built from a variable at runtime will not appear, and a hex string that is not a colour at all, such as a commit hash, might.

• Is the image I drop uploaded anywhere?

  No. It is decoded by the browser, drawn onto a canvas no larger than 160 pixels on its long side, and counted. Nothing leaves the tab, and the page works with the network switched off.

• Why show the pair through colour blindness at all?

  Because a pass that rests on hue is not a pass. The three panels show the pair as the common kinds of colour blindness would see it, so a pair that separates by hue and not by lightness gives itself away.

Open CONTRAST

• What is the difference between a first-party and a third-party cookie?

  A first-party cookie is set by the site you are on, and is usually how it keeps you signed in or remembers a preference. A third-party cookie is set by a different domain embedded in the page, such as an ad network, an analytics provider or a social widget. Because the same domain is embedded across many sites, it can follow you between them.

• Does rejecting cookies stop tracking?

  It stops the cookie-based part of it. Sites also fingerprint the browser itself from fonts, screen size, timezone and graphics behaviour, and they store identifiers in localStorage and elsewhere, none of which a cookie banner covers. Rejecting is worth doing and is not the whole picture.

• Which cookies can I safely delete?

  All of them, at the cost of being signed out and losing preferences. Nothing breaks permanently. If you want to keep your sessions, delete third-party cookies and anything with a multi-year expiry, which is a strong hint about what it is for.

• How does it decide what a cookie is for?

  By matching the cookie's name against a maintained list of known advertising and analytics cookies. An "unknown" cookie is not necessarily safe, just unrecognised. A site can set dozens of cookies, and most of them are not there to keep you logged in.

• Does it change or delete my cookies?

  No. It reads only the names, never the values, and it runs entirely in your browser without uploading anything. Use it to understand a site, then manage the cookies in your browser's own settings.

Open COOKIEAUDIT

• How do I check a spreadsheet for personal data before I share it?

  Look for patterns rather than reading every cell: email addresses, phone numbers, national insurance or social security numbers, card numbers, dates of birth, postcodes and full names. Check every sheet including the hidden ones, and read the header row, because column names often describe the personal data even when the visible sample looks harmless.

• What counts as personal data?

  Under GDPR it is anything that identifies a living person, directly or in combination with something else. That is wider than most people expect: an IP address, a device identifier, a staff number, and a postcode together with a date of birth all qualify. A tool can flag likely candidates, but whether a field is personal data in your context is a judgement rather than a pattern match.

• Is removing the names enough to anonymise a dataset?

  Usually not. Combinations of ordinary fields re-identify people: date of birth, sex and postcode together identify a large share of a population on their own. Real anonymisation means aggregating, generalising or suppressing those combinations, not deleting the name column.

• How does it decide that a column is personal data?

  By the shape of the values, and by the column name for free-text fields like names and addresses. That means it can miss an unusual format and can occasionally over-flag. Treat it as a safety check before you email a spreadsheet or hand a CSV to a contractor, not a guarantee that a file is clean.

• Is my file uploaded?

  No. DATASCAN runs entirely in your browser. It reads only the structure and values you give it, and forgets everything when you close the tab.

Open DATASCAN

• What traces do I leave on a shared or public computer?

  More than most people expect. Browser history and cache, cookies and site storage, saved form entries and passwords, downloaded files still sitting in the downloads folder, recently-opened lists in applications, clipboard contents, and thumbnails generated from anything you opened. Signing out of a website clears almost none of it.

• Does private browsing cover it?

  It covers the browser, and only for that window. It does not stop files you download from being saved, it does not touch what other applications recorded, and it does nothing about monitoring installed on the machine itself. On a computer you do not control, assume anything you type can be recorded and act accordingly.

• What should I do before walking away from a public machine?

  Sign out of everything, clear browsing data for all time rather than the last hour, delete anything you downloaded and then empty the recycle bin, clear the clipboard by copying something harmless, and close every application rather than just its window. Then, for anything that mattered, change the password from a device you trust.

• What does DEADBOLT clear?

  Everything OBSCURA OS has stored in this browser: notes you saved, a board you drew, settings. That is data for obscuraos.com only, which is all a website is allowed to touch. It does not include files you downloaded to the device, so export anything you want to keep first. It cannot reach other sites or your browser history. To clear those, use your browser's own settings.

• How does the panic cover work?

  Arm it, then tap the button or press the Escape key twice, and a plain, boring screen drops over everything at once. Tap the cover to bring your work back.

Open DEADBOLT

• Why not model a funding round in a shared spreadsheet?

  Because a cap table is one of the most sensitive documents a company holds, and shared spreadsheets leak by default: link sharing, download history, cached copies, and a version history that preserves every earlier scenario you modelled. A model that runs locally leaves no copy anywhere.

• What does dilution actually mean?

  Your percentage falls because new shares are issued, and the value of what you hold can still rise because the company is worth more. The two questions worth keeping separate are how much of the company you own and what that stake is worth. Option pools complicate it, because a pool created before the round dilutes the existing holders rather than the incoming investor.

• Pre-money or post-money?

  Pre-money is the valuation before the new investment, post-money adds it. The same headline number implies different ownership depending on which was meant, and that ambiguity is the single most common source of surprise when the documents arrive. Confirm which is intended, in writing, before agreeing anything.

• How do I give a preferred holder a liquidation preference?

  Add a fourth value to that holder's line: the amount they invested. For example Series A, 2000000, preferred, 3000000. That amount gives the holder a 1x non-participating preference in the exit waterfall, so they take back their investment before the common holders are paid, or convert if their share of the exit is worth more.

• How do I enter a SAFE or a convertible note?

  One per line: name, amount, valuation cap, discount percentage and type, where type is safe or note. For example Angel, 250000, 5000000, 20, safe. Leave the cap or the discount as 0 if the instrument has none. They convert into shares when you model the new priced round.

• Does my cap table leave the browser?

  No. The holders, the SAFEs and the round are worked out in this tab and never sent to a server, so your ownership never lands on a service that shares it with investors. The CSV download is the only copy, and you choose where it goes.

Open DEALBOOK

• What can it recognise?

  Base64 in both alphabets, base32, hexadecimal in several layouts, URL escapes, HTML entities, quoted-printable, JSON web tokens, backslash escapes from most programming languages, gzip and raw deflate, UTF-16, decimal and binary byte lists, international domain names, Caesar shifts including the common one, and numbers that are timestamps. It also chains them, which is where most of the work is.

• Why does it show more than one answer sometimes?

  Because some strings genuinely are more than one thing. A run of letters and digits can be valid base64 and valid hex at once, and only the result tells you which was meant. Rather than guessing quietly, the most likely reading is shown and the others are offered next to it.

• Can it decode something encrypted?

  No, and nothing can without the key. If a result looks like noise with no structure in it, that is what it is: either encrypted, or compressed in a way not recognised, or simply random. The page says which of those it looks like rather than pretending.

• What does it tell me about the result?

  Whether it is text and in what script, whether it is JSON or XML or a known file type, how random it looks, whether it contains invisible characters, and whether any part of it is a date, an address, a key or something that looks like a credential.

• Is a token safe to paste here?

  Safer than anywhere else that offers to decode it, because this page has no server. It is still a token: if it is live, it is a login, and the safest thing remains not to move it around at all. What this removes is the part where a stranger also gets a copy.

Open DECODE

• Can I compare two Word documents or PDFs?

  Not here. Use REDLINE for those; it reads .docx and PDF and shows what moved in a contract. DELTA is for text you can paste or open as a text file.

• What does 'ignore spacing' do?

  It treats any run of spaces or tabs as one space and ignores space at the ends of lines, so reformatted code or re-wrapped prose does not show as changed. 'Ignore capitals' does the same for case.

• Why are some unchanged lines hidden?

  'Show only the changes' keeps two lines of context around each change and folds the rest, the way a code review does. Untick it to see every line of both texts.

• Is anything uploaded?

  No. Open the network tab in your browser's developer tools and paste anything you like; nothing goes out.

• What is wrong with the online diff sites?

  They keep a copy of what you paste. Here the comparison is made in this tab and both texts stay in it; nothing is sent anywhere.

Open DELTA

• What is a macro and why is it dangerous?

  Code stored inside a document that the application can run. It was designed for automation and it remains one of the most durable ways to deliver malware, because the file arrives looking like an ordinary document and the code runs when you enable it. Office now blocks macros in files from the internet by default, which helps and is not universal.

• What else in a document can be active?

  Embedded objects and linked files, remote images that report back when the document is opened, fields that fetch content from an address, and in some formats scripts attached to actions. A document that phones home does not need a macro to tell the sender you opened it.

• How do I make a document safe to open?

  Strip the active parts rather than trusting a prompt. Remove macros, embedded objects and remote references, and where the content allows it convert to a format that cannot carry code at all. For something you only need to read, a PDF rebuilt from page images carries nothing.

• What exactly does it remove from a Word, Excel or PowerPoint file?

  VBA macro projects and their signatures, ActiveX controls, Excel 4.0 macro sheets, embedded OLE objects and executables, attached-template references, DDE and DDEAUTO fields, INCLUDEPICTURE and INCLUDETEXT fields, and any image, object, frame or data connection that loads from an external address.

  A macro-enabled file (.docm, .xlsm, .pptm) comes back as the ordinary format (.docx, .xlsx, .pptx) so it cannot carry macros again by accident.

• What exactly does it remove from a PDF?

  Document-level JavaScript, the open action, page and document actions, launch, submit-form, import-data and remote go-to actions, embedded files, file attachment and rich media annotations (video, sound, 3D, screen), the XFA form layer, and form field scripts. Web links are kept unless you tick the box to remove them.

• Does it remove hyperlinks?

  Not unless you ask. Ordinary links stay. Tick the box under the file drop to remove links to web addresses as well: hyperlinks in Office files and URI links in PDFs.

Open DISARM

• What personal information is inside a .docx file?

  A .docx is a zip archive. docProps/core.xml stores the author name, whoever saved it last, the creation and modification times and the revision count. docProps/app.xml stores the total editing time and the template it came from. Alongside those, the body can carry tracked changes, comments, and text that was deleted but never accepted.

• Does Inspect Document in Word remove everything?

  It removes most of it and is worth running. It does not touch everything: embedded objects, some custom XML parts, and the original author recorded inside earlier revisions can survive it, and it has to be run again after any save that reintroduces them.

• Can someone see what I deleted from a Word document?

  If track changes was ever on and the changes were not accepted, yes. The deleted text is still in the file along with the name of whoever deleted it, and turning the display off only hides it on screen. Accepting all changes and deleting all comments before you send is the minimum.

• What does DOCSCRUB remove from the document?

  It removes the author and company metadata, deletes the comments, and accepts all tracked changes, keeping inserted text and removing deleted text, so the revision history is gone, including the redlines you thought you had accepted. Review the cleaned file before sending, especially if the document relied on tracked changes being visible.

• Is my Word document uploaded?

  No. The .docx is read, inspected and rewritten in your browser. It handles standard .docx files; unusual or heavily customised documents may need a look afterwards.

Open DOCSCRUB

• Does my printer really do this?

  Most colour laser printers and colour copiers do, from most makers, and have since the 1990s. Inkjet printers do not. A black and white laser printer cannot, since it has no yellow. The only way to know for your own is to print a page and scan it here.

• Why can I not see the dots?

  They are pale yellow, a tenth of a millimetre across, on white paper, which the eye cannot separate. Under a blue light, or a magnifier with the scan's blue channel turned up, they show. This page finds them by colour: high red and green with the blue well below, which is yellow and not paper.

• What can it read?

  The Xerox DocuColor layout, which the Electronic Frontier Foundation worked out and published in 2005: a block of fifteen columns by eight rows carrying the date, the time and the serial number, with parity to check the reading. Other makers use other layouts that nobody has published. For those the page finds the dots, shows the pattern, says how big its repeat is, and removes them; it does not pretend to read them.

• Does removing the dots make a page untraceable?

  It removes that one mark from that one scan. The paper itself, if you hand it over, still carries the dots. A page can also be traced by its wording, its fonts, or the printer's own quirks in how it lays down toner. Removing the dots takes away the mark that was designed to trace it; it does not take away every way.

• Is anything uploaded?

  No. The scan is decoded and read in this tab, on your machine, and the cleaned copy is written by your browser. The page works with the network switched off.

• What kind of scan do I need?

  A colour scan of the printed page at 600 dpi, as PNG or JPEG. A black and white or greyscale scan drops the dots, and a photo from a phone is usually too soft. The scan is decoded here and never leaves this tab.

• How are the dots removed?

  The cleaned copy is a PNG in which every dot, and a one-pixel ring round it, is painted over from the paper beside it. Check it before you rely on it: a dot printed over ink, rather than paper, may leave a mark where it was.

Open DOTS

• How do I share a config file without leaking the secrets?

  Replace the values, not the keys. Whoever receives it needs to know which variables exist and what shape their values take; they do not need the live credentials. Keep the names and substitute obvious placeholders, so the file still documents the configuration and carries nothing that works.

• I committed an API key to git. Is deleting it enough?

  No. A commit that removed the key leaves the earlier commit containing it in the history, and anyone with the repository, or a fork, or a cached copy, can still read it. The only reliable response is to revoke and rotate the key immediately, and treat rewriting the history as tidying up afterwards rather than the fix.

• Which values in a config file are actually secret?

  Anything that grants access or identifies you: tokens, passwords, private keys, connection strings with credentials in them, webhook URLs, and signing secrets. Hostnames, ports, feature flags and log levels are usually safe, but an internal hostname still tells a stranger about your network, so it is worth a second look.

• How does ENVSHIELD decide what to mask?

  Three ways. By known secret shapes: AWS, GitHub, Stripe, Slack and Google keys, JWTs, PEM private keys and basic-auth URLs. By secret-looking key names such as password, token, secret and api_key. And by high entropy, for values that look random. Readable settings such as ports, hostnames and flags are left alone.

• Is the masked output guaranteed to be clean?

  No. ENVSHIELD is a safety net, not a guarantee. Always review the output before sharing it, and rotate any secret that has already been exposed.

• Is my config file uploaded?

  No. ENVSHIELD runs entirely in your browser and never uploads or transmits your file. The redacted copy is written in the tab for you to copy or download.

Open ENVSHIELD

• Which formats does it cover?

  Unix in seconds, milliseconds, microseconds and nanoseconds; Windows FILETIME; Chrome, WebKit and Edge; Apple Cocoa absolute time in seconds and in nanoseconds; HFS and HFS+; .NET ticks; NTP and Bluetooth; OLE automation dates as used by Excel and COM; packed MS-DOS and FAT date-time; the time field of a version 1 UUID; GPS seconds; and Twitter and Discord snowflakes. Seventeen in all, each with its epoch and tick size printed beside the answer.

• How does it tell a FILETIME from a .NET tick?

  It does not, and nor can anything else: both count 100-nanosecond intervals, so a given moment is a valid value in both and the numbers differ only by the constant between their epochs. What it does is show you both readings with their dates, and you decide from where the number came: a value out of an NTFS record is a FILETIME, a value out of a .NET application's database is probably ticks. Presenting one and hiding the other would be the tool deciding something it has no basis to decide.

• Why does the GPS offset matter?

  GPS time does not take leap seconds, so it runs ahead of UTC by a whole number of seconds that has grown over the years. The current offset is set on the page and applied to both directions. If you are decoding something from an earlier period the offset was smaller, so change it, and it is on the page rather than buried precisely so that the assumption is visible and adjustable rather than baked in.

• Does it handle a whole column?

  Yes. Once you know what the format is, which is usually the case after decoding one value, paste the column, pick the format, and every line is read at once, with anything that will not read marked rather than silently dropped. It comes out as a CSV with the original value beside the UTC and local readings, so it can go straight into a schedule or an exhibit.

• What about local time versus UTC?

  Every format here except two counts from a UTC instant, and both readings are shown: the UTC one, and the same moment in this browser's own time zone. The two exceptions are called out where they appear: classic HFS wrote local time while HFS+ writes UTC and the number alone does not say which, and packed DOS and FAT timestamps are always local with no record of the zone. Those are properties of the formats, not of this tool, and pretending otherwise would be the sort of quiet assumption that gets an exhibit challenged.

• Is anything sent anywhere?

  No. It is a page of arithmetic. Nothing is uploaded, there is no account, and once the page has loaded it works with the network off, which matters, because forensic workstations frequently have none.

Open EPOCH

• Do photos really contain my location?

  Often, yes. Most phones write GPS coordinates into the photo file itself unless you have turned location off for the camera. Anyone who receives the original file can read those coordinates, which is usually accurate to within a few metres of where you stood.

• Is my photo uploaded anywhere?

  No. The file is read by JavaScript inside your own browser tab. It is never sent to a server, and nothing about it is stored or logged.

• Does posting to social media remove the GPS data?

  Most large platforms strip metadata when they re-encode an upload, but this is not guaranteed and does not help you when you send the original file directly by email, message or cloud link. Those keep everything.

• How do I remove the metadata?

  Re-encoding the image from its raw pixels drops every metadata block. The CLOAK tool in the OBSCURA OS Sovereign Suite does this and also lets you permanently redact faces or documents in the frame.

• What does a camera serial number give away?

  Cameras embed a body serial number in every photo they take. The same number appears in every picture that camera has ever produced, which links otherwise unrelated photos to one device, and to whoever registered or sold it.

• Which image formats does it read?

  JPEG files carry EXIF, which is where GPS and camera details live, so those give the fullest reading. PNG and WebP store metadata differently and usually carry less.

• Nothing was found. Does that mean the photo is safe to share?

  It means this particular file carries no readable EXIF, which is good. Metadata is only half the story, though. Street signs, storefronts, window views and reflections identify a place perfectly well on their own.

• Does taking a screenshot of a photo remove the metadata?

  Yes. A screenshot is a new file with none of the original EXIF. It also degrades the image, and the screenshot itself may record when and on what device it was taken.

Open EXIF

• How do people-search sites get my address and phone number?

  They buy and scrape it. Voter rolls, property records, court filings, marketing lists, loyalty schemes, brokers reselling to each other, and old social profiles. None of it requires a breach. The same record is then sold on, which is why removing it from one site does not remove it from the others.

• Does opting out actually work?

  Usually for that site, for a while. Most will honour a removal request, and most repopulate from their original sources some months later, so it is a recurring chore rather than a one-off. Keeping a list of which sites you have contacted and when is most of the work.

• Is a paid removal service worth it?

  They run the same opt-outs on a schedule, which is real convenience, and they need your personal details in order to do it, which means handing more data to one more company. Doing it yourself costs an evening and a diary reminder.

• What happens when I mark a request as sent?

  The tracker starts the 45-day clock the CCPA gives the broker to respond. Open each opt-out page and follow its steps first. Most make you find your own listing before they will remove it, then mark it here so you keep your place.

• What does EXPUNGE do that a paid removal service does not?

  Search your own name and you will find your home address, a photo of your house, your phone number and a list of your relatives on a dozen public directories. The paid removal services charge every year and make you upload your whole identity to their servers first. EXPUNGE gives you a direct link to each broker's opt-out page, a deletion letter filled in with your details, and a tracker that keeps your place, all in this tab. Your details are never uploaded.

Open EXPUNGE

• Is it safe to use real customer data for testing or a demo?

  It is the most common way personal data ends up somewhere it should not be. Test databases are backed up less carefully, shared more widely, and shown on screens in public. Most data protection regimes treat using live personal data for development as a purpose it was never collected for, which needs its own justification.

• What is the difference between fake data and anonymised data?

  Anonymised data is real data with the identifiers removed, and it can often be re-identified by combining fields. Synthetic data is generated from nothing and corresponds to no real person, so there is nothing to re-identify. For demos and test fixtures, synthetic is both safer and easier to reason about.

• Will generated data look realistic enough to test with?

  For structure, yes: names, addresses, phone numbers and identifiers in the right formats will exercise validation and layout properly. What it will not reproduce is the messiness of real records, so keep a small set of deliberately awkward cases alongside it: missing fields, very long names, unusual characters, duplicates.

• Can the generated data reach a real person?

  No. Email domains are the reserved example.com, example.org, example.net and .example names, phone numbers use the 555-01xx block set aside for fiction, and card numbers are the standard published test numbers. Nothing here can reach a real inbox, phone or account. It is for testing and demos, not for deceiving anyone.

• How many rows can FAUX generate, and in what format?

  Up to a thousand rows at a time, with the fields you tick. Copy them as CSV, or download them as CSV or JSON. Everything is generated in your browser and nothing is uploaded.

Open FAUX

• Does this work if I block cookies?

  Yes, entirely. None of it uses cookies or storage of any kind. That is the point: refusing cookies does nothing about any of this, and a site that wants to follow you between visits does not need them.

• What is canvas fingerprinting?

  A page draws some text and shapes onto an invisible canvas and reads back the exact pixels. The result depends on your graphics hardware, drivers, fonts and operating system, and is stable for you and different for most other people. It takes a few milliseconds and there is nothing to notice.

• Can I stop it?

  You can reduce it a great deal. A browser that blocks canvas and audio reading, reports a rounded window size and ships its own fonts removes most of what is on this page. Perfect is not available in a general-purpose browser, and a browser that tries too hard to be unusual becomes easy to recognise for exactly that reason.

• Is the number at the top my permanent identifier?

  It is the fingerprint of this browser as it is right now. Change something that feeds into it, such as a window size that is measured or a font you install, and it changes. Most of what feeds it does not change often, which is exactly why it works as an identifier.

• Why does a privacy site collect all this?

  Because seeing your own is the only way the argument lands. Reading a list of techniques is abstract; a page that tells you your own graphics card, your own font list and how many people share your screen size is not. It is read and shown, and it is never stored or sent.

• What do the bits next to each line mean?

  A rough estimate of how much that one detail narrows you down. Ten bits is one person in a thousand; twenty is one in a million. The numbers come from published measurements of how varied each detail is across the web, not from checking yours against a database.

  The group totals do not add up to the number at the top, and are not meant to. Many of the lines read the same underlying fact twice, so counting them all separately would say every browser on earth is unique.

Open FINGERPRINT

• Why can people still edit the form I sent them?

  Because a PDF form stays a form. The values are stored as field data and drawn on the page separately; a reader shows the drawing and lets anyone change the value. Flattening removes the fields and leaves the drawing, so there is nothing left to type into.

• Can somebody read my answers without opening the file properly?

  Yes, until it is flattened. Field values are plain objects in the file, so any library, any command-line tool and any indexing system can list them. That is also how a value in a field marked hidden gets found.

• What happens to a field that was hidden?

  It is removed, and the result says which ones and what was in them. It is not drawn onto the page, because that would take something the document was not showing anybody and print it across the middle of it. If you wanted that answer visible, put it in a visible field before flattening.

• Does flattening change how the document looks?

  It should not. Each field's drawn appearance becomes page content in the same place. The one thing to check is a form that never had appearances drawn, where a standard font is used to make them; open the result and look before sending it.

• Will my signature survive?

  A drawn or image signature in a field becomes part of the page, which is usually what you want. A cryptographic signature does not: flattening changes the file, so any digital signature over it stops matching. Flatten first and sign afterwards, in that order.

• What about a password-protected PDF?

  It cannot be opened without the password, and the page says so rather than failing obscurely. Remove the protection in whatever you use to open it, then flatten.

• Why are the answers in a PDF form still editable?

  A PDF form does not contain its answers where you see them. It holds a set of fields with values, and a drawn appearance on the page. So the answers stay editable by whoever opens the file, they can be read straight out without looking at the page, and a field that was cleared or marked invisible can still be carrying what was in it. Flattening draws each answer onto the page for good and takes the fields away.

• What does FLATTEN always remove?

  Visible fields are drawn in and removed. Fields that are not on the page are removed without being drawn. Left-over objects holding the old answers are discarded, and code attached to the document always goes. Those are the point of it. Comments, links and document properties are handled by the three options above the button.

Open FLATTEN

• Can a PDF contain a virus or run code?

  A PDF can carry JavaScript and actions that fire when the document opens or when you click into a field, and it can carry other files embedded inside it. Whether any of that reaches something harmful depends on the reader, and modern readers sandbox it heavily, but the active content is genuinely in the file, and most of it lives inside compressed sections where a quick look at the raw bytes would miss it.

• Can a PDF form send what I type somewhere?

  Yes. A submit action can post the field values to an address chosen by whoever built the form, and that address does not have to have anything to do with the letterhead. Filling in a form and pressing submit inside the reader is not the same as emailing it back to the sender.

• What are hidden form fields in a PDF for?

  Usually something innocent: a calculated total, a version marker, a value the template needs to work. They are worth knowing about because a hidden field can be pre-filled with something you never saw and is submitted along with everything you typed.

• Can FORMAUDIT remove the script or the fields?

  No. It reports what it finds and does not remove anything. If you need a PDF with the active content genuinely gone, SEAL rebuilds each page from a bitmap, which leaves no script, no fields and no attachments because it leaves no document structure at all.

• Does finding JavaScript mean the PDF is dangerous?

  Not on its own. Everything here is a description of what the file contains, not a judgement about intent: plenty of legitimate forms use JavaScript for validation and plenty of legitimate documents carry links. Equally, finding nothing is not a clean bill of health. FORMAUDIT reads structure, not behaviour: it cannot tell you what a piece of script would actually do, it does not execute anything, and it does not follow a single link. Encrypted PDFs will hide most of their contents from it. Nothing is uploaded.

Open FORMAUDIT

• Can text be hidden inside other text?

  Yes. Unicode contains characters that render as nothing at all, and sequences of them can encode a message inside ordinary-looking writing. The carrier text looks completely normal in any editor, and the payload survives copy and paste, which is precisely why it is worth being able to detect as well as create.

• Would hidden characters survive being posted somewhere?

  It depends on the platform. Anything that normalises or sanitises Unicode strips them, and plenty of sites do. Somewhere that passes text through untouched keeps them. Treat it as unreliable for delivery and reliable as a marker to check for in text you receive.

• Is this a good way to keep something secret?

  It hides that a message exists, which is a different thing from making it unreadable. Anyone who suspects it and knows the method reads it straight off. For anything that matters, encrypt the message first and then hide the result, so discovering the channel is not the same as reading the contents.

• How does GLYPH hide the note?

  It encodes the note as invisible zero-width characters mixed into the cover text. The text looks completely normal but quietly carries the note. Paste text someone sent you into the reveal box to pull a note back out, or to check whether a message carries an invisible watermark. Revealing and hiding are both free.

• Is GLYPH a security tool?

  No. It is obfuscation, not encryption. Anyone who knows to look can detect and extract the hidden text, and copying the text through some apps will strip it. It is a lightweight way to watermark your own writing or leave a note in plain sight. Do not use it for anything that needs to stay secret, and do not use it to slip content past a filter or a person. For real secrecy, use encryption.

• Does GLYPH upload my text?

  No. It runs entirely in your browser and never uploads anything.

Open GLYPH

• How can I tell whether an email really came from who it claims?

  Read the headers rather than the From line, which anyone can write. Authentication-Results records three checks: SPF, whether the sending server was allowed to send for that domain; DKIM, whether the cryptographic signature over the message verifies; and DMARC, whether those results line up with the domain shown in the From address. DMARC is the one that matters.

• SPF passed but the email is still fake. How?

  SPF checks the envelope sender, not the From address you see on screen. A message can be sent from a domain the sender controls, pass SPF for that domain, and still display a From line naming your bank. That mismatch is exactly what DMARC alignment tests, which is why a DMARC pass rather than an SPF pass is the useful signal.

• Where do I find the full headers?

  In Gmail, open the message and choose Show original. In Outlook, View message source, or File and then Properties. In Apple Mail, View, then Message, then All Headers. Copy the whole block including the headers, because the useful lines are near the top.

• Does a "looks authentic" result mean the email is safe?

  No. It means the message passed the checks your receiving mail server recorded in the headers, not that its contents are safe or true. When in doubt, do not click links or reply. Contact the sender through a channel you already trust.

• Why is one of the checks missing from the result?

  HEADERPROOF reads the SPF, DKIM and DMARC results your receiving server already wrote into the headers. A missing result usually just means your provider did not record one.

• Do I need to paste the whole email?

  Only the headers are needed. HEADERPROOF runs entirely in your browser and never uploads anything you paste.

Open HEADERPROOF

• How big a file can it open?

  Larger than you will want to scroll. Only the visible rows are ever read, so the viewer itself does not care about size. Searching and the entropy graph do read the whole file, which on a very large one takes time and shows progress while it happens. Both can be stopped.

• What do the numbers in the status bar mean?

  The bytes you selected, read as every integer and floating-point type they could be, in both byte orders. Little-endian puts the least significant byte first and is what x86 and ARM use; big-endian is the other way round and is what most network protocols and many file formats use. If a number looks absurd in one column and sensible in the other, you have found the file's byte order.

• Can I search for a string that is not ASCII?

  Yes, as UTF-16 little-endian, which is how Windows stores text inside many file formats and how a filename appears inside a ZIP written on Windows. Searching for the same word as ASCII and as UTF-16LE often gives completely different answers about the same file, and that difference is usually the interesting part.

• Why does the entropy graph have a flat top?

  Because 8 bits per byte is the ceiling. Anything already compressed reaches it: a ZIP, a JPEG, a PNG's image data, an MP4. If a file is flat at 8 everywhere except a small header, it is compressed or encrypted throughout, and there is nothing more to be read out of it by eye.

• Is it safe to open a suspicious file here?

  Opening it here is safer than opening it in the program it was made for, because nothing executes: the bytes are read and displayed as bytes and nothing else. That is not the same as saying the file is safe. Nothing here defuses anything, and a document you inspect here can still be dangerous in the application that opens it.

• Can I get some of the bytes out?

  The selection can be copied as a hex string, as a C array, as base64, or saved as a file of its own. That last one is how you carve an embedded image out of a container: select from its signature to its end marker and save the range.

• What does a step in the entropy graph mean?

  Usually a boundary: the end of a header, the start of an embedded file, or a packed section in an executable. For scale, a run of zeroes scores 0, English text sits around 4.2, machine code around 6, and compressed or encrypted data at the top, where the two cannot be told apart. Click a bar to jump to that part of the file.

• Why does the scrollbar jump on a very large file?

  Because the file is too tall for a browser to lay out row by row, so the scrollbar moves through it proportionally rather than a row at a time. Go to offset is exact, so type the offset when you need a precise place.

Open HEX

• What does each treatment do?

  A bleep replaces the span with a steady 1 kHz tone, the sound television uses. Silence replaces it with nothing. A garble keeps a low, muffled version of the voice, filtered below 300 Hz and then ring-modulated, so the rhythm of speech is there but the words cannot be made out. Every edge has a ten millisecond fade just outside the span so the cut does not click, and every sample inside the span is treated in full.

• Can the words be recovered from the result?

  Not from a bleep or a silence: the original samples are not in the file, and the tone or the silence is all that is written there. A garble is filtered so hard that the consonants and the vowel shapes are gone before the ring modulator scrambles what is left, which is far beyond what a de-noising tool undoes; still, if the recording must never be recoverable, silence is the safe choice. Nothing here hides the original behind the treatment: the WAV is written from the treated samples only.

• Why is the output a WAV?

  Because no MP3 or AAC encoder is shipped on this site, and a WAV of 16-bit PCM is what every player, editor and phone opens. It is written at the sample rate and channel count the recording was decoded at, so nothing is resampled or folded to mono. It is larger than an MP3 of the same length: about ten megabytes a minute in stereo at 44.1 kHz. If you need an MP3, convert the WAV afterwards in any converter.

• Does it work on a video?

  It takes the sound track out of an MP4, WebM or MOV that your browser can decode, and the result is a WAV of the sound only. The picture is not touched and is not in the result. To bleep the sound in the video itself, use this WAV as the replacement track in a video editor.

• How do I mark a word precisely?

  Zoom in with the button, or hold Ctrl and turn the wheel over the waveform, until the word is wide enough to drag across. Once a span is selected its edges can be dragged, and the From and To boxes take an exact time. If you have a transcript with timestamps, such as the SRT that SCRIBE writes, paste it into the list and every line with two times becomes a span; delete the ones you do not want.

• What can I paste into the list?

  One span a line, as a start and an end time, like 0:01.250 0:01.900 or 1.25 1.9. Subtitle lines with an arrow, like 00:00:01,250 --> 00:00:01,900, work as they are, so an SRT file can be pasted whole; the index lines and the text lines are left out. JSON works too: a list of two-number pairs, or of objects with start and end, or the chunks with timestamps that a Whisper-style transcriber writes. A word on the line such as bleep, silence or garble sets the treatment for that span.

• Is anything uploaded?

  No. The recording is decoded by your browser, the result is rendered by your browser, and the WAV is written by your browser. The page works with the network switched off, and nothing is kept once the tab is closed.

• What does it not do?

  It does not transcribe: SCRIBE does that, and its timestamps can be pasted here. It does not find the words for you, change the pitch of a voice, or write MP3. Very long recordings need memory: an hour in stereo is about a gigabyte of samples in the browser while it is being worked on, so on a phone keep to shorter files.

Open HUSH

• Why do two visually identical JPEGs show thousands of differing pixels?

  Because JPEG is lossy and re-saving it re-quantises every 8 by 8 block. Almost all of those differences will be under the noticeable count, and the difference map will show a faint blue wash in the busy parts of the picture and nothing in the flat parts. That pattern is the signature of a re-encode rather than an edit.

• What does a difference only in the blue channel mean?

  Usually a colour profile, a white balance or a tint, not new content. Blue is also where most compressors throw information away first, because the eye has the fewest short-wave cones. A genuine edit almost always moves all three channels together, which is why the breakdown is on the page.

• Should I turn on ignore antialiasing?

  Turn it on when comparing two renders of the same page, where the text edges will differ on every letter and drown out anything real. Leave it off when you want to know whether the file changed at all. The test asks for a blend, so a hard-edged shape that shifted a whole pixel is still counted; an already-soft shape that shifted, and the soft edge of anything recoloured, are not, which is the reason to look with it off as well as on.

• What is dE 2.3?

  The classical just-noticeable difference in the CIE 1976 colour difference formula: the distance in L*a*b* at which a person looking at two large flat patches side by side begins to tell them apart. It is a rough figure taken under laboratory conditions, and real differences inside texture or small text need to be larger before anyone sees them. Treat it as a floor.

• Can I compare a screenshot against a design mock-up?

  Yes, and the wipe and onion views are usually more useful than the map for that, because the two will differ almost everywhere. Scale them to a common size first, and expect the resampling itself to appear in the numbers.

• Are my images uploaded?

  No. Both are decoded by the browser, drawn onto canvases in this tab and compared here. Nothing is sent anywhere, and the page works with the network switched off.

• What tolerance should I set?

  Leave it at zero to see every byte that changed, which is what you want when checking whether a file was re-encoded. Raise it to eight or so when comparing two photographs of the same thing, or two renders from different machines, where a small amount of noise is expected and is not the thing you are looking for. A pixel is only counted when one of its three channels moves by more than the tolerance, out of 255.

• How does ignore antialiasing decide which pixels to discount?

  Along the edge of any curve or letter, a renderer blends the edge colour into the background by how much of the pixel each side covers, and two renderers almost never land on the same blend. With the box ticked, a differing pixel is discounted when three things hold at once: at least one of the two images has a blend here, meaning a brightness strictly between the brightest and the darkest of its own eight neighbours; the brightness each image has here already occurs among the other image's eight neighbours; and there is a real edge nearby for it to be a blend of.

  Requiring a blend is what stops a genuine move being swallowed, because a hard-edged shape shifted a whole pixel puts a plateau value on both sides and no blend at all. The honest limits: an already-antialiased shape that moves by one pixel is blend against blend and will be discounted, and so will the soft edge of a shape that was recoloured, though its solid middle will not. Turn it off before you conclude that nothing moved.

• What numbers does the summary give?

  How many pixels differ at all, how many differ enough for an eye to see, the largest move, where the differences are gathered, and which channels carried them. That is how two screenshots a release apart, which look identical, turn out to have a logo moved four pixels, or a whole frame drifted one step bluer because a colour profile changed on the way out.

Open IMAGEDIFF

• What is in a HAR file and why is it sensitive?

  A complete recording of a browsing session: every request and response, with headers, cookies, tokens, and often the response bodies. Anyone holding the file can usually take over the session it captured. Support teams ask for them routinely and people attach them to public tickets.

• What needs removing before I share a HAR file or a curl command?

  Authorization headers, cookies, session and CSRF tokens, API keys sitting in query strings, and any personal data in the request or response bodies. In a curl command copied out of developer tools, everything needed to replay your session is right there in the flags.

• Can I share a log once the tokens are out?

  Usually, provided you also check the bodies. Response payloads carry names, addresses and identifiers that are easy to overlook because they are not what you were debugging. Replace values rather than deleting them, so the structure the recipient needs to see is still intact.

• What exactly does INTERCEPT redact?

  Credential headers: Authorization, Cookie, Set-Cookie, X-Api-Key, X-Auth-Token, CSRF and the like. Secret query and form fields: password, token, api_key, client_secret, session, otp and similar names. JSON bodies: any field whose name looks like a secret. And known key shapes anywhere in the file: JWTs, Stripe, GitHub, AWS, Google and AI keys, Slack tokens and PEM private-key blocks.

  Cookies keep their names so the log still reads sensibly; only the values go. The same replacements are made in a pasted curl command or block of headers.

Open INTERCEPT

• Is it safe to paste a live token here?

  Safer than anywhere else, because nothing leaves the tab: the decoding, the audit and the signature check all run in your browser. Turn the network off and the page works exactly the same. That said, a token in a browser tab is still a token on a screen, so treat the room you are in accordingly.

• What does an algorithm of "none" mean?

  That the token asks to be accepted without a signature. It exists in the specification for tokens that are already protected some other way, and it has caused a long line of vulnerabilities in libraries that honoured it by default. If you did not issue a token like that, treat it as an attack rather than a mistake.

• What is the algorithm confusion attack?

  An issuer signs with a key pair, so verifiers hold its public key. An attacker changes the header to HS256 and signs a token using that public key as if it were a shared secret. A verifier that picks its algorithm from the token rather than from its own configuration will accept it. That is why an HS token from an issuer that normally uses RS or ES is flagged here.

• Why is my email flagged?

  Because the payload of a token is plain text to everyone who handles it: every proxy log, every browser it is stored in, and anything it is ever pasted into. Names, addresses and phone numbers in a token travel further than people expect.

• Can it check the signature on any token?

  It checks HS256, HS384 and HS512 with a shared secret, and RS, PS and ES with a public key, using your browser's own cryptography. EdDSA is not supported by every browser's crypto yet, and the page says so rather than guessing.

• Is a JSON Web Token encrypted?

  No. A token is three pieces of base64 with dots between them, and the first two, the header and the payload, are not encrypted. Anyone holding the token can read every claim in it.

  That is also why pasting a live token into a decoding website is a bad idea: the token is the credential. JOT reads it here, and the signature check against your key or secret happens in this tab too.

Open JOT

• Where is the journal kept?

  In this browser's local storage, under obscura_journal_v1, as ciphertext. Nowhere else. That also means clearing site data deletes it, a private window loses it when it closes, and it does not follow you to another browser or another machine. Take the encrypted backup and keep it somewhere you control.

• What happens if I get the passphrase wrong?

  It says so and nothing changes. AES-GCM checks every record against its own authentication tag, so a wrong key fails outright instead of producing plausible-looking rubbish. There is no limit on attempts here, because the 600,000 rounds are the cost: each guess takes about as long for an attacker as it takes for you.

• Can I write more than once in a day?

  Yes. A day opens with its most recent entry, and the button beside the date starts another. When a day holds more than one, a selector appears to move between them. The calendar marks any day with something written on it.

• How do tags work?

  Two ways, and they add up. Anything written as #something in the entry becomes a tag, and the tags box takes them as a comma-separated list for the ones you would rather not have in the text. The tag list shows every tag with its count; clicking one filters the search to it.

• Does the search write anything?

  No. It runs over the decrypted entries held in this tab while it is unlocked. No index is built, nothing is cached, and locking takes the searchable text away with everything else. Nothing on disk is ever in the clear, including while you search.

• Can I bring in a journal I already keep?

  If it is markdown files, yes. Each file becomes an entry; the date comes from a filename beginning YYYY-MM-DD or from a date: line in front matter between two rows of three dashes. Tags and a rating in the front matter are read too. Files with no date it can find are listed and left alone rather than dumped on today.

• How do I put a backup back?

  Drop the encrypted backup on the restore panel, or choose it there. It replaces whatever journal is in this browser, and then opens with the passphrase that backup was made under. The backup is the same ciphertext that was in the browser it came from, so it needs that journal's passphrase, not a new one.

Open JOURNAL

• How do I check an SSH key's fingerprint?

  Paste the public key (the line in the .pub file) here and KEYPRINT shows its SHA256 fingerprint exactly as ssh-keygen -lf key.pub prints it, and the MD5 form with -E md5. Compare it with the fingerprint GitHub, GitLab or your hosting provider shows.

• What is the difference between the SHA256 and MD5 fingerprints?

  They are two hashes of the same key. OpenSSH has printed SHA256, in base64, since version 6.8 in 2015; before that it printed MD5 as colon-separated hex, and some systems still show that form. Both identify the same key; they just cannot be compared with each other.

• Is it safe to paste my key here?

  A public key is meant to be shared, and nothing leaves the tab anyway. Never paste a private key anywhere; if one is pasted here by mistake, KEYPRINT recognises it, does not read it, and clears it from the box.

• Why is my DSA key flagged?

  OpenSSH disabled DSA keys by default in version 7.0, in 2015, because they are limited to 1024 bits and fail badly if the random number generator is weak. Replace it with an Ed25519 key: ssh-keygen -t ed25519.

• What does a certificate with no principals mean?

  An OpenSSH user certificate lists the user names it may log in as. With none listed, OpenSSH accepts it for any user on any server that trusts the CA, which is almost never what was meant.

• Can it read hashed known_hosts entries?

  It reads the key and its fingerprint. The host names are hashed on purpose so that a stolen known_hosts file does not list the servers you connect to, so they cannot be read back.

Open KEYPRINT

• What is age?

  A file encryption format designed to be small enough to have no dangerous options: one public key per recipient, one way to use it, no configuration. It is widely used for encrypted backups and secrets, and there are implementations in Go, Rust and several other languages. The specification is at age-encryption.org.

• Will the age program open a file made here?

  Yes, and this opens files made by it. That is tested rather than assumed: files are round-tripped in both directions against the reference implementation, at sizes either side of the internal chunk boundary, as part of the checks on this code.

• Where should I keep the private key?

  In a password manager, or on an encrypted disk, or on paper in a safe. Not in a chat message, a shared drive or an email to yourself. If it is lost, nothing encrypted to it can ever be opened again, by anyone, including you. That is the property you are paying for.

• Can I encrypt to several people at once?

  Yes. Tick as many saved recipients as you like and paste any others. Each one gets its own copy of the file key inside the header, so any one of them can open the file and none of them learns anything about the others' keys.

• Why is there a self test on the page?

  Because you have no reason to trust a web page about cryptography. The button runs the published vectors from RFC 7748 and RFC 8439 against the very code that encrypts your files, in your browser, while you watch, and then encrypts and opens a file end to end.

• Does it support passphrases or SSH keys?

  No. age itself supports both, and this page does not pretend to: it handles X25519 recipients only. A file encrypted another way is refused with a message naming what it uses, so you know to reach for the age program instead.

• Which key do I give to other people?

  The public key. Anyone who has it can encrypt to you, and nothing about it is secret. The private key is the one you keep, and it is the only thing that opens what they send.

• Do I need an account or a key server?

  No. There is no key server, no account and no plugin. The key pair is made in this tab from your browser's own random source, and the public half is what you hand out yourself.

Open KEYRING

• Where is the vault stored?

  In this browser's localStorage for obscuraos.com, as one encrypted item. It is not sent anywhere, there is no account, and the page works with the network turned off. A different browser, a different profile, or a private window is a different vault.

• What happens if I forget the master passphrase?

  The vault stays encrypted and nobody can open it, including us. There is no recovery, because recovery would mean something somewhere holds a second way in. Write the passphrase down and keep the paper somewhere safe, or print it with PAPERKEY.

• Does a wrong passphrase damage the vault?

  No. A wrong passphrase fails the AES-GCM authentication tag and the page stops there. Nothing is written back to storage on a failed unlock, so you can try as many times as you like.

• Can I check the encryption myself?

  Yes, and it is the point of publishing the format. Copy the item out of localStorage and run it through any PBKDF2 and AES-GCM implementation: 600,000 rounds of PBKDF2-HMAC-SHA256 over the base64 salt gives a 32-byte key, and AES-256-GCM with the base64 IV over the base64 ciphertext gives the vault as UTF-8 JSON. Twenty lines of Python or Node will do it.

• Which CSV exports does the import understand?

  1Password, Bitwarden, LastPass, Chrome and other Chromium browsers, and Firefox, each recognised from its header row. Anything else is mapped by column name where the names are clear enough, and whatever cannot be mapped is listed for you instead of being thrown away.

• Why does unlocking take a second?

  Because deriving the key is meant to be slow. Six hundred thousand rounds of PBKDF2 costs you about a second once and costs somebody guessing your passphrase the same second for every guess they make.

• Why does it lock when I switch tabs?

  Because the common way a vault is read by someone else is that it was left open. The idle timer is yours to set, but the tab-hidden lock is not optional, and the cost of it is one passphrase entry when you come back.

• Why does the CSV export warn me so hard?

  Because it writes every password into a file in the clear. Anyone who opens that file, and anything that syncs, backs up or indexes your downloads folder, reads all of them. It is the right thing to do when you are moving to another manager and the wrong thing to do the rest of the time. Delete the file the moment you are finished with it. The encrypted backup is the one to keep: it is the same envelope this browser stores, it is only as safe as the passphrase, and this page can open it again.

Open KEYS

• What counts as hidden text?

  In a PDF: text in the same colour as the page, or within a few steps of it; text under four points; text placed off the page or outside the area the page clips to; text drawn in the invisible render mode (Tr 3); text that is fully transparent; and text on an optional content layer that is switched off. In a Word file: a run marked hidden (w:vanish) or hidden on the web view (w:webHidden), on the run itself or through its style; text within a few steps of the page colour or of its own shading; a size under four points; and zero-width or invisible characters (U+200B to U+200D, U+2060 to U+2064, U+FEFF) or rows of soft hyphens inside the text. Headers, footers, footnotes and endnotes are read as well as the body.

• Is every finding a trick?

  No. A scanned PDF that has been through OCR keeps its recognised words in the invisible render mode, so every line of it is a finding, and that layer is what makes the scan searchable. Word’s hidden-text mark is used honestly for notes and answer keys. Some scripts, Thai among them, use the zero-width space as a word divider. The page shows what a reader cannot see and why; you decide what to take out. Untick a finding and the copy keeps it.

• How is the PDF copy made? Is it a picture of the page?

  No. The page’s content stream, and any form XObject it draws, is tokenised here and each text operator is matched one for one against the operators the PDF engine reported. A hidden one is replaced by an empty advance of the same width, so the words beside it keep their places, and the rest of the file is untouched: fonts, links, form fields, bookmarks and metadata all stay. If the two readings of a page do not line up exactly, nothing on that page is changed and the status line names the page. It does not rasterise anything.

• What does it not find?

  Text painted over with a box, which is the fake redaction SEAL exists to prevent; text inside a picture; text in annotations, comments, form fields and tracked changes, which DOCSCRUB and SCRUB deal with; and text hidden by being small but readable, or grey but visible. A PDF that needs a password to open cannot be read, and an old binary .doc file must be saved as .docx first.

• Why does it say a finding reads like an instruction to a machine?

  Because the hidden words match a phrase of the kind people plant for screening software and language models: “ignore previous instructions”, “you are an AI”, “rate this candidate highly” and their relatives. It is a hint about intent, not a verdict; read the words yourself. Hidden text that matches nothing is still hidden text.

• Is the Word copy still a proper document?

  Yes. The hidden runs are cut out of the document’s XML at their exact positions and the invisible characters out of the text, and the file is zipped up again with every other part as Word wrote it. Nothing is re-serialised, so styles, numbering, pictures, tables and settings are as they were.

• Is anything uploaded?

  No. The file is read and the copy is written in this tab. The PDF engine and the zip library are fetched from this site the first time they are needed and kept by the browser; the page then works with the network switched off.

Open LANTERN

• What does it check?

  Every fact recorded as told is compared with the legend and with every other telling of the same fact. It reports a telling that differs from the legend, a fact told two different ways (and names anybody who heard both), a fact told that the legend does not contain, an age said aloud that the date of birth does not give on that date, and a history that cannot be true: a job before sixteen, an entry that starts before the birth or ends before it starts, two homes at once, and more than a year with no address.

• How is it encrypted?

  The passphrase is stretched with PBKDF2-SHA256 over 600,000 rounds against a random salt, and the key encrypts the whole notebook with AES-256-GCM and a fresh vector on every save. The passphrase is not stored. What sits in this browser's storage is one sealed record that does not reveal how many identities it holds or what they are called.

• What if I forget the passphrase?

  The notebook cannot be opened, by you or by anybody else, and there is no recovery. Keep an encrypted backup somewhere safe and the passphrase somewhere separate.

• Does it lock itself?

  Yes. After the number of minutes you choose with nothing typed or pressed, it saves, drops the decrypted copy from the tab and asks for the passphrase again. Lock now does the same immediately.

• Why does the rehearsal time the answers?

  Because a right answer given after a long pause is noticed almost as much as a wrong one. The rehearsal asks facts back at random, more often the ones got wrong recently, and counts the answers that took longer than four seconds.

Open LEGEND

• What is an rsid?

  A revision save id. Each time Word opens a document and saves it, it makes up a random eight-digit hexadecimal number for that sitting, adds it to a list in word/settings.xml, and stamps it on every paragraph and run written or changed in that sitting. The list is the file's own record of every session that has touched it, and the first entry, rsidRoot, is the session that made the document.

• Does a session mean a person?

  No. A session is one save cycle in Word: everything typed between one save and the next shares one id. One person who saves twice in an afternoon makes two sessions, and two people who share a laptop and save once make one. The names in the file come from the account name on the machine, which is a different record and can be anything.

• Can it tell which session came first?

  Only the root. The ids are random numbers, so nothing in them says which came before which. This page numbers the root session first and the rest in the order they first appear in the text, which is the only order the file offers, and says so in the report.

• What does "not in the session list" mean?

  A run carries a session id that Word never registered in this file's list. That is what a paste from another Word document leaves behind: the pasted text keeps the id of the document it came from. It is also what a document assembled by a script or a template engine can look like, so it is a mark to look into, not a finding. A paste from the same document leaves no mark at all.

• My document shows almost nothing. Why?

  It was probably not written in Word. Google Docs, LibreOffice, Pages and most converters write few or no session ids, so a file exported from them has little history of this kind. A file that has been through a metadata cleaner, or Word's own Inspect Document, has had the ids removed on purpose. Either way, an empty reading is not evidence of anything.

• Is the editing time real?

  It is a counter Word keeps in docProps/app.xml of the minutes the document was open with unsaved changes, added up over every session. It is not a record of anyone's effort: a document left open over lunch gains an hour, a document rebuilt from a template starts at zero, and any tool can reset it. The revision number is the number of saves the same way.

• Does it show tracked changes and comments?

  It counts the insertions, deletions, moves and formatting changes still tracked in the document, names the author and dates Word recorded on each, and does the same for comments. It does not display the deleted text or the comment text; REDLINE and DOCSCRUB are for reading and removing those.

• Is anything uploaded?

  No. The document is unzipped and read in this tab. The only thing fetched is the zip reader, from this site, the first time a file is opened. The reports are written by your browser to your disk.

Open LINEAGE

• Does LOCKBOX release the files automatically when my client pays?

  No, and that is deliberate. Verifying a payment made into your own Stripe account requires Stripe Connect and full identity onboarding onto someone else's platform, and a service that withholds a key until money moves resembles escrow, which is regulated. Cryptocurrency cannot be verified from a static file at all. So you hold the key and send it when the payment lands. Any tool that claims automatic release from a container your client already holds has either put the key in the file or is not telling the truth.

• Does OBSCURA OS take a commission?

  None. The payment link in the container is your own, so the money goes directly to you. Freelance platforms charge 10 to 20 percent to act as a middleman; LOCKBOX never touches the transaction.

• Can my client extract the files without the key?

  No. The deliverables are encrypted with AES-256-GCM and the key is not in the container. They can read the manifest, see the watermarked previews and check the SHA-256 of every file, but the payload is ciphertext until you send the key.

• How large can the deliverables be?

  The container is one HTML file that has to survive email, and base64 encoding inflates the payload by about a third, so the practical ceiling is tens of megabytes rather than gigabytes. For large video renders or full repositories, lock a proof-of-work deliverable in LOCKBOX and move the bulk with VAULT once you are paid.

• What happens if I lose the release key?

  The files inside that container cannot be recovered, by you, by your client or by us. The key is not in the container. It exists in your browser at the moment the container is built and nowhere else, so copy it or download the key file straight away and keep it with the job.

• How does my client open the deliverables?

  Email them the container. When the payment lands, send them the release key. They paste it into the file and the deliverables decrypt and download on their own machine. Until then they see the watermarked previews and the manifest of exactly what they are buying, with your payment link and the amount owed.

• Why is the client's name burned into the previews?

  So that a screenshot of a preview still points back at them. Images and PDFs get watermarked previews carrying the client name; everything else is listed in the manifest without a preview.

Open LOCKBOX

• Does opening a file here run its macro?

  No. Running a macro needs a VBA interpreter, and there is not one in a browser. The bytes are parsed as data, the text is decompressed and printed, and that is the end of it. The page contains no eval, no Function constructor and no code path that turns anything read out of the file into something the browser executes.

• My .docx has a vbaProject.bin in it. Is that bad?

  It is worth knowing about. The x forms of the Office formats are defined as the ones without macros, and Word will not run VBA from a .docx. But the project travels inside the file, so renaming it to .docm, or saving it from Word as macro-enabled, brings the code back to life. It is also a sign the file was made by something other than Word.

• What does the compiled-against-source line mean?

  Every module keeps two copies of itself: compiled p-code, which is what Office actually runs, and compressed source, which is what the editor shows you. They are supposed to agree. When the compiled half is much larger than the source, or holds strings the source does not, somebody may have edited one without the other. It is a reason to be careful, not a conclusion.

• Why does it show line numbers against the flags?

  So you can go and read the line. A flag on its own is an accusation; a flag with the line it came from and a sentence about why it was flagged is something you can check and disagree with. Every module is shown in full, with the flagged lines marked, for exactly that reason.

• Can I get the source out?

  Yes. One text file with every module and the flag list, or a zip with one file per module named .bas, .cls or .frm as the project has them. Both are written in this tab from what is already on screen.

• Which files can I drop in?

  A .docm, .xlsm, .pptm, .doc, .xls or .ppt, or a vbaProject.bin on its own. Drop a .docx, .xlsx or .pptx too: those are the forms that are meant to have no macros, and it is worth knowing when one does.

• Why read the macro here rather than in Office?

  Somebody sent you a spreadsheet that wants you to enable content. The only honest way to decide is to read what the macro says, and the only safe place to read it is somewhere that cannot run it. MACROLENS unzips the file, walks the OLE compound file inside, decompresses each module and prints the source, with a note against every line that deserves one. It is a reader. There is no interpreter here.

Open MACROLENS

• How do I save a message to look at?

  In Gmail, open the message, use the menu at the top right of it and choose Download message. In Outlook, File then Save as. In Apple Mail, File then Save As and pick Raw Message Source. Thunderbird has Save As directly. All of them produce the same thing: the message exactly as it arrived.

• What do SPF, DKIM and DMARC actually tell me?

  SPF says the server that sent it was allowed to send for that domain. DKIM says a signature over the message matches a key published by a domain. DMARC says the domain in the from line matches one of those two and the domain owner asked for that to be required. A pass on all three does not mean the message is honest, only that it is genuinely from where it says.

• Can it tell me whether a message is a phishing attempt?

  It can tell you the things phishing attempts have in common: a from address that does not match what signed the message, a display name pretending to be an address, links whose text and destination disagree, a reply address on a different domain, attachments that are not the type they claim. Whether a particular message is one is a judgement, and this gives you what the judgement needs.

• Does opening a message here load the images?

  No. Nothing is fetched, and the HTML is shown as text rather than rendered. That is the whole point of looking at a suspicious message here rather than in a mail client.

• Why does it say the attachment is not what it claims?

  Because the declared type and the actual first bytes of the file disagree. Something named as a document that begins with the bytes of a program is the clearest signal there is, and mail clients show only the name.

• Which hops in the delivery path can I trust?

  Only the hops added by servers you trust. Each hop is written by the server that received the message, and MAILBOX lists them oldest first. Anything before the first server you trust was written by the sender and can say whatever the sender likes.

• What does MAILBOX show that my mail client does not?

  A message carries a delivery path, a set of authentication results, a sending domain that often is not the one in the from line, links whose text and destination need not agree, attachments whose declared type need not match their contents, and images that report back when the message is opened. Mail clients show you almost none of it. Drop a saved message here and it is taken apart, offline. The HTML version is shown as text rather than rendered, on purpose.

Open MAILBOX

• Which files can I give it?

  An Android APK, or an XAPK or APKS bundle, in which case it finds the base APK inside. A Chrome or Edge extension as a CRX file, whose signing header it strips to reach the zip. A Firefox add-on as an XPI. A zip of an unpacked extension folder. It also takes a bare AndroidManifest.xml, a bare classes.dex or a bare manifest.json if that is all you have.

• Where do I get the file before installing?

  For an extension already installed somewhere, the browser keeps its folder on disk and you can zip that. For one you have not installed, the CRX can be fetched from the store's update address with the extension's ID, and several sites offer that as a download. An APK comes from the developer, from a mirror, or off a phone that has it with adb pull. An APK from a mirror may not be the one on the store, which is one more reason to read it first.

• How does it read AndroidManifest.xml when it is not text?

  A built APK carries the manifest as binary XML: a pool of strings, a table of attribute ids and a stream of start and end element chunks. This page decodes that format itself, by hand, and reads the package, the versions, every uses-permission, and each activity, service, receiver and provider with its intent filters. When a packer has blanked the attribute names it falls back to the ids, which the system uses anyway.

• How does it find tracker SDKs?

  Every class an app links is named in the string tables of its classes.dex files, and an SDK's classes live under its vendor's package: com.appsflyer, io.sentry and so on. The page reads every string in every dex file and matches the class names against a list of about fifty analytics, crash-reporting and advertising SDKs it knows. An SDK whose classes were renamed by an obfuscator is not found this way, though its addresses often still are. An absence is not proof of absence.

• Does an address in the list mean the app talks to it?

  It means the address is written into the code as a constant. Some of those are used every time the app runs, some are in a library the app never calls, and some are links in a licence notice. Addresses built up from pieces at runtime are not seen at all. The namespace and licence addresses that every Android build carries, such as schemas.android.com, are left out. Addresses that belong to a known analytics or advertising service are marked.

• What does exported mean?

  An exported activity, service, receiver or provider can be started or reached by other apps on the device, not only by its own. That is normal for a main screen or a share target, and worth a look for a service or a receiver with no permission guarding it. The page lists the exported ones, the permission that guards each if there is one, and whether it was declared exported or is exported because it has an intent filter.

• What is a remote-code hint?

  A place in an extension's scripts where code could arrive from outside the package: a call to eval, a function built from text, a string run on a timer, a script element made at runtime, a script named on another site, or WebAssembly this page cannot read. Bundled libraries carry some of these innocently. A hint says where to read, not what you will find.

• Is anything uploaded?

  No. The package is unzipped and read in this tab, on your machine, in a background thread so the page stays usable while a large game is read. The only thing fetched is this site's own copy of the zip library. The page works with the network switched off once it has loaded.

Open MANIFEST

• What has to be removed from a medical record before sharing it?

  More than the name. The direct identifiers are the obvious part: name, address, dates of birth, record and insurance numbers, contact details, the names of relatives and of the treating clinician. What gets missed is the indirect material: a rare condition, admission dates, an employer, a small town, and free-text notes that describe a person unmistakably.

• Is removing the name enough to de-identify a record?

  No. A rare diagnosis with a date and a location can identify one person in a country. Recognised de-identification standards remove or generalise a defined list of identifiers precisely because names alone were never the problem.

• Do I need consent to share a record for teaching or publication?

  Usually yes, and de-identifying it does not remove the requirement, particularly for case reports and images where the patient may recognise themselves. Follow the rules of your jurisdiction and institution. This is not legal advice.

• What does MEDISCRUB catch, and what can it miss?

  It reliably catches structured identifiers (emails, phone numbers, SSNs), labelled fields (Patient Name, MRN, Policy No), addresses, ZIP codes, dates and ages over 89. It cannot catch every name written in free-flowing prose, so read the result before you share it. Removing identifiers lowers risk but is not a legal guarantee of de-identification.

• Is the record text uploaded?

  No. MEDISCRUB runs entirely in your browser and never uploads or transmits your text. The redacted copy is written in the tab and you copy it or download it as a .txt file.

• When would I share a record without my identity on it?

  For a second opinion, an insurance appeal, a support group or research, where the clinical detail matters and who you are does not. MEDISCRUB masks the identifiers and keeps the clinical detail readable. Tick Keep dates if the dates are clinically important.

Open MEDISCRUB

• Is it safe to export passwords to a CSV file in order to move them?

  The export itself is plaintext, and that is the risk rather than the transfer. Every password sits in a readable file on disk, probably in your downloads folder, possibly synced to cloud storage, and still recoverable after an ordinary delete. If you have to export, import it straight away, then delete the file and empty the trash immediately.

• How do I move from one password manager to another?

  Export from the old one, import into the new one, check a handful of entries by hand, then destroy the export. The part to be careful about is the gap in between: do not email the file, do not put it in a shared folder, and do not leave it sitting in Downloads.

• Should I change my passwords after migrating?

  The migration itself exposes nothing if the export never left the machine. It is a good moment to deal with what an audit shows: passwords reused across sites, ones that appear in breach lists, and anything short enough to guess. Those are worth changing whether you migrate or not.

• Where is the export in my password manager?

  Every manager has one. In Bitwarden: Tools, then Export vault, as JSON or CSV. In 1Password: File, then Export, which gives a .1pux file. In LastPass: Advanced, then Export, as CSV. In Chrome, Firefox or Safari the passwords settings have an export to CSV. That plain export is what you bring here, not the encrypted vault file.

• Why is the converted file unencrypted?

  Because that is what an import needs. It means the download is your whole vault in the clear. Import it into your new manager promptly, then delete the file securely. Do not email it or leave it in your Downloads folder.

Open MIGRATOR

• What is a decoy workspace for?

  A situation where you are compelled to unlock something in front of someone: a border, a search, a robbery. A second workspace that opens with a different passphrase and looks ordinary gives you something to show. It buys plausibility and nothing more.

• Does a decoy actually protect me?

  Only against someone who does not know to look for one and has neither the time nor the expertise to examine the device. Against a forensic examination it fails, and in several jurisdictions producing a decoy in response to a lawful order is itself an offence. This is not legal advice, and the legal risk is real.

• What is the alternative?

  Not carrying the data. A device that never held the material cannot be made to produce it, and travelling with a clean device and retrieving what you need afterwards is both simpler and far more robust than any hidden compartment.

• Where must MIRAGE not be used?

  MIRAGE exists for coercion by people with no legal right to your data: a thief, a scammer on a support call, someone who has taken a device off you. It must not be used to mislead a court, a regulator, an auditor, a customs or immigration officer, an employer conducting a lawful investigation, or anyone else you owe a duty of disclosure to. Presenting false contents to any of those is obstruction, it is a criminal offence in most places, and it breaches the Acceptable Use Policy, which ends your access without a refund.

  If you are under a preservation notice, a production order or a litigation hold, that duty overrides any use of this tool. If you are not sure whether one applies to you, ask a lawyer before you use it, not after.

• What can someone holding the vault file work out from it?

  The partition count, the salts, the IVs and the ciphertext sizes. Each partition is AES-256-GCM ciphertext under its own PBKDF2 salt, stored in random order with its label inside the ciphertext, so nothing in the file says which PIN opens which or which one is the decoy.

  The file does not hide that a second partition exists. It does no padding, noise-filling or disk-level concealment.

Open MIRAGE

• What can I do when a client will not pay an invoice?

  Escalate in writing, in steps. A polite reminder, then a firm one restating the agreed terms and the amount, then a formal letter before action giving a deadline and saying what follows if it passes. Each step should name the invoice number, the date, the terms and the days overdue. Most invoices are paid at the second or third step.

• Can I charge interest on a late invoice?

  In many places yes, either at a rate written into your contract or at a statutory rate set by law, and in the UK and EU there is also a fixed compensation sum per invoice. The entitlement often exists whether or not you mentioned it in the contract. This is not legal advice, and both the rate and the right depend on where you and the client are.

• What is a letter before action?

  The last letter before a claim. It sets out what is owed and why, encloses the evidence, gives a clear deadline, and states plainly that proceedings will follow if the deadline passes. Courts expect to see one, and sending a proper one very often ends the matter without a claim.

• Is this legal advice?

  No. It is drafting help, and nobody here is your lawyer. The templates are general: what you can actually claim, whether you are entitled to interest and at what rate, how much notice you must give, and what a demand letter needs to say all depend on your contract and on where you and your client are. Several places have statutory late-payment rules that are more favourable than anything you would think to ask for, and some contracts override them. Read what you send before you send it, put in your own facts, and take proper advice before starting any formal process.

• How is the interest figure worked out?

  As simple interest on the outstanding amount at the annual rate you enter, counted from the due date. It is a starting point for a conversation, not a calculation to rely on in a filing.

Open NETTERMS

• Who is this for?

  Anyone whose work depends on a cover identity staying separate from their real one: undercover officers, covert internet investigators, test purchasers, source handlers, and the supervisors and cover officers who review their arrangements. It is equally useful for a journalist or researcher working under a pseudonym.

• Why is there no overall score?

  Because a percentage would say that a deployment with one critical exposure and thirty-three items in place is 97% safe, and it is not. One linked recovery phone number or one tagged family photograph can end a cover identity on its own. The result is a count by severity and the list of exposures, worst first.

• Why does an unchecked item count against the audit?

  An item nobody has looked at is not known to be safe, and an audit that treats it as safe reads as clean when it is only unfinished. Unchecked items are counted separately and named in the result until somebody looks.

• Where is the audit kept?

  In this browser's own storage on this device, and nowhere else. Nothing is uploaded. The audit of a person's weak points is itself sensitive, so use a deployment reference rather than a name, export it to wherever your organisation keeps such records, and use Clear to remove it from the browser.

• What does sealing do?

  It computes a SHA-256 digest of the audit's text on the device. Written down somewhere outside the browser, it lets anyone show later that a saved copy is the one that was made and has not been edited since. It is not a signature and does not prove who made it.

Open OPSEC

• Is this really not uploading my picture?

  Yes. The engine runs in a worker inside this tab; the only things fetched are its own code and the language file, from this site, once. Open the browser's network panel while it reads and you will see no request carrying the image. That is also what the site's own checks watch for on every tool.

• How accurate is it?

  On a clear scan of printed text, most pages come back over ninety-five percent sure and read cleanly. A phone photo in mixed light, a fax, or a small font drops that, and the page confidence shown beside each page tells you which case you are in. The words it doubts are outlined on the picture so you can check them rather than guess.

• Can it read handwriting?

  Not usefully. It is trained on printed letterforms and will produce something for handwriting, but that something is mostly wrong. It is said plainly here because the sites that claim otherwise are selling the upload, not the result.

• What languages does it read?

  English, French, German, Spanish, Italian, Portuguese, Dutch, Polish, Russian, Ukrainian, Turkish, Swedish, Danish, Norwegian, Finnish, Czech, Arabic, Japanese and simplified Chinese. Each language is a separate file of about three megabytes, fetched the first time you choose it. A page has to be read in the language it is written in; tell it which.

• What is a searchable PDF?

  A scan is a picture of a page, and a picture cannot be searched. Ticking the box writes each page back as a PDF that carries the picture unchanged with the recognised text laid invisibly over it, so a PDF reader can search, select and copy the words while showing you the original scan.

• Does it work offline?

  After the first run, yes. The engine and the language file are kept on the device by the site's service worker, so the page reads pictures with the internet switched off. Choosing a new language needs one more download for that language.

• Why does it matter where the picture is read?

  Every other image-to-text site works by uploading the picture, and the pictures people most want read are the ones they least want copied anywhere: a passport page, a prescription, a contract photographed on a desk, a screenshot of somebody else's message. OPTIC runs the reading engine inside this tab. The text comes out, the picture stays here.

• What kinds of picture can OPTIC read?

  A photo, a screenshot, or a PDF that is really a scan, in JPEG, PNG, WebP, GIF, BMP, HEIC or PDF, as many as you like at once. Drop them, choose them, or paste a screenshot. Printed text, roughly square-on, reads best.

• How much does OPTIC download the first time?

  The reading engine (about 4 MB) and a language file (about 3 MB) are fetched from this site once and kept on your device. After that the page works with the internet off. Each further language you choose is one more file of about 3 MB.

Open OPTIC

• Why would I not just use the unzip built into my computer?

  On a computer, you should. On a phone or an iPad, opening a zip attachment is often a dead end, and the sites that offer to unzip it for you keep a copy of everything inside. This does the same job on the device.

• Can it open .rar, .7z or .tar.gz?

  No. Only zip. Those formats need different code, and the zip format covers nearly every attachment and download that arrives compressed.

• Does 'Save every file' download them one by one?

  Yes, with a short pause between each so the browser does not mistake it for a popup storm. The browser may ask once whether to allow several downloads; say yes and press again for the rest.

• Is anything uploaded?

  No. Open the network tab in your browser's developer tools and watch nothing go out while you pack or open an archive.

Open PACK

• Is an electronic signature legally binding?

  In most jurisdictions a simple electronic signature is capable of forming a binding contract, with specific categories of document excluded. What tends to be disputed is not the law but the evidence: who signed, when, and that the document has not changed since. This is not legal advice, and the rules differ significantly between countries.

• What makes an electronic signature hold up?

  A record of what was signed and by whom. A hash of the exact document at the moment of signing, the time, whatever identity evidence you relied on, and a copy of the file. Being able to produce the document alongside a digest that matches it answers the most common challenge, which is that the document was altered afterwards.

• Is a typed name or a drawn squiggle enough?

  Often yes, legally. Evidentially it is weak on its own, because nothing ties the mark to a person or to a moment. What strengthens it is everything around it: the audit record, the identity check, and proof that the document is the one that was signed.

• How do I check a signed PDF against its certificate later?

  Choose the PDF signed with PACT and the certificate JSON that came with it. The file is hashed here and compared with the hash in the certificate, so either side can prove later that nothing moved.

Open PACT

• What does the owner password actually protect?

  Less than people think. The owner password sets the permission flags: printing, copying, changes, comments. Those flags are a request to the program that opens the file, and a program that ignores them, or a tool that clears them, can. Once someone can open the file, the owner password does not stop them reading it, photographing the screen, or making an unrestricted copy. The password to open it is the lock: without it the contents are encrypted with AES-256 and cannot be read. If the file matters, set a password to open it.

• How strong is the lock?

  As strong as the password. The file is encrypted with AES-256 under the PDF 2.0 standard security handler, revision 6, the same one Acrobat writes, and there is no known way round the cipher. A short or guessable password is another matter: a computer can try millions of passwords a second against the file offline, so use a long passphrase, and keep it somewhere safe. Without it the file cannot be opened, here or anywhere.

• Which locked PDFs can it open?

  Any PDF locked with the standard password handler: RC4 at 40 or 128 bits (revisions 2 and 3), AES-128 (revision 4) and AES-256 (revisions 5 and 6), given either its password to open or its owner password. It cannot open a PDF locked to a certificate rather than a password, or one under a rights management server, and it does not guess passwords.

• I do not have the password. Can it take the lock off anyway?

  If the file opens without a password but will not let you print or copy, yes: that file has only an owner password, the restrictions are the whole lock, and they come off with no password at all. If the file asks for a password to open, no. Nothing can read it without the password, this page included, and it does not try.

• Does it change the document?

  It does not re-render anything. Every string and stream in the file is decrypted or encrypted in place and the same objects are written out again with a fresh cross-reference table, so the pages, fonts, text, forms, bookmarks, links and attachments are the ones that went in. Object streams are unpacked into ordinary objects on the way. A digital signature will no longer verify, because the bytes it signed have changed, so a signed PDF is best left as it is.

• Will an older program open a file locked here?

  The AES-256 lock is the one in PDF 2.0, and has been read by Adobe Reader since version X (2010), by Preview, Chrome, Firefox, Edge, and by the usual phone viewers. A reader older than that will ask for the password and then say it cannot open the file. This page writes only AES-256; the older RC4 and AES-128 locks it reads are not ones anyone should still be writing.

• Is anything uploaded, and what happens to the password I type?

  Nothing leaves this tab. The file is taken apart, encrypted or decrypted, and put back together by your browser, and the page works with the network switched off. The passwords are held in memory while the file is being worked on and are not written anywhere.

Open PADLOCK

• What is steganography?

  Hiding a message inside something else so that the something else looks entirely ordinary. In an image, the usual approach changes the least significant bit of each colour value, which shifts the colour by an amount no eye can see while carrying one bit of the payload per channel.

• Can hidden data in an image be detected?

  By anyone looking for it, usually yes. Statistical tests on the low-order bits give simple approaches away readily, and any resize, crop, or resave to a lossy format destroys the payload outright. It hides a message from a casual observer, not from an examiner.

• Is it safer than encryption?

  It answers a different question. Encryption makes a message unreadable but obvious; steganography makes it inconspicuous but readable to anyone who finds it. They combine well, and used alone the second one is weak: the security rests entirely on nobody thinking to look.

• Should I save the result as a PNG or a JPEG?

  Always a PNG, never a JPEG. The hidden data lives in the least significant bit of the image's colour channels and is destroyed the moment the image is re-compressed. Expect most chat and social apps to strip it too.

• Can I use PALIMPSEST to keep something secret?

  No. It is obfuscation, not encryption, and not a way to slip anything past a scanner or a person. The change to the pixels is invisible to the eye but easy for software to find. Treat it as a way to watermark or playfully hide a note, not as protection for anything sensitive. For real secrecy, use encryption.

• Is my image or message uploaded?

  No. PALIMPSEST runs entirely in your browser and never uploads your image or message. Reading a hidden message is free; saving an image with one hidden in it is part of OBSCURA Pro. Load an image someone sent you to read a hidden message, or to check whether one is there.

Open PALIMPSEST

• What can I put on a sheet?

  Anything up to a few kilobytes: a passphrase, a recovery phrase, a private key, a small key file. Large files are the wrong thing for paper, and the page says so rather than producing forty pages nobody will ever type.

• What happens if I mistype a line?

  Reading it back tells you which line and stops there rather than handing you a key that is silently wrong. That is the whole reason for the check characters: a wrong answer that looks right is worse than no answer.

• Can it fix a mistake for me?

  No, and deliberately not. Correcting errors automatically means guessing, and a guess that happens to satisfy a check is a corrupted secret you now believe in. It tells you where to look and you fix it from the paper in your hand.

• Is the sheet safe to leave lying around?

  No. It is the secret, in plain characters, made easy to read by design. Treat it the way you would treat the key itself: a locked box, a safe, a bank. If it has to survive being seen, encrypt it before printing or split it across several sheets.

• Does it work with a recovery phrase from a wallet?

  Yes, and the words themselves come back exactly as they went in. Many people prefer to write the words down directly, which is fine; the reason to use a sheet like this is the check on every line, which a handwritten list does not have.

• Why put a key on paper at all?

  A key written on paper survives a dead drive, a lost phone, a locked account and a company going out of business. What it does not survive is being typed back in wrong: one character, and there is nothing to tell you which one. PAPERKEY lays the secret out in an alphabet with no letters that look alike, puts a check on every line and one across the whole sheet, and reads it back telling you exactly which line is wrong.

• Do I have to type the row numbers and spaces when reading a sheet back?

  No. Type one line per line on the sheet. Row numbers, spaces and case do not matter, and the letters that are easy to confuse are accepted either way round.

Open PAPERKEY

• Why would you hide a key inside an image?

  So that a backup does not look like a backup. A folder of photographs draws no attention where a file named recovery-key does. It is concealment rather than protection: it helps against someone browsing casually, and not at all against anyone examining the file.

• Is this a safe way to back up a recovery key?

  Not on its own. Encrypt the secret first and hide the encrypted result, so that finding the hiding place is not the same as reading the key. And keep more than one copy in more than one place, because a single image that gets resized, re-saved or passed through a messaging app loses the payload silently.

• What destroys the hidden data?

  Almost any processing. Resizing, cropping, re-saving as JPEG, uploading somewhere that re-encodes, or a phone gallery quietly generating its own version. Keep the original file exactly as it was produced, and check that you can still recover the secret before you rely on it.

• What is PHANTOM meant to protect against?

  Trade secrets and credentials at rest: a recovery key, a contract, a piece of confidential business material. The payload is encrypted with AES-256-GCM before it is hidden in the image, so a single stolen laptop or a mis-sent attachment does not hand it over.

Open PHANTOM

• How do I check a suspicious email without clicking anything?

  Read it as text. The sender's display name, the address behind it, the reply-to address, the authentication results and the actual destination behind each link are all in the message. Everything you need to make a decision is there, and none of it requires opening a link or loading an image.

• Why does loading images in an email matter?

  A remote image is fetched from the sender's server the moment the message is displayed, which confirms your address is live and reveals your IP address and the time you opened it. Most clients block them by default for exactly this reason. A message that will not display without images is telling you something.

• What are the strongest signs of a phishing email?

  Link text naming one site while the link itself goes to another. A domain a character or two away from a real one. A reply-to that differs from the from. Authentication that fails. And urgency about an account, a payment or a deadline. Any one of these can appear in real mail, so treat them as reasons to verify through a number you already had, never as proof either way.

• Does no finding mean the message is safe?

  No. Every check is a rule of thumb. Real messages sometimes trip several at once, particularly ones sent through marketing platforms, and a careful fake can pass all of them. A finding means look closer. No findings does not mean the message is safe.

• What does PHISHLENS look for?

  A sender name that does not match the address behind it, a reply that would go somewhere else, a domain one character away from a real one, and link text that says one site and goes to another. It also reads any authentication results in the headers and looks for pressure language and requests for passwords, codes or payment.

• Can it tell me what is at the other end of a link?

  No. PHISHLENS never follows the link, so it cannot tell you what is actually on the far end, only what the address itself gives away. If the message claims to be from a company you deal with, reach them through a number or address you already had, never one in the message.

Open PHISHLENS

• How many words do I need?

  Six words (77 bits) is enough for anything protected by a slow hash or a login that limits attempts, which is nearly everything. For a master password, a disk, or a key that a stolen file could be attacked offline for years, use eight or more.

• Should I add capitals, digits and symbols?

  Not for strength; any attacker's list already includes those variations. Add a digit only when a site's rules demand one, and let the length carry the strength.

• Is a BIP39 phrase from here safe to use for a wallet?

  The words and the checksum are correct and the entropy is the browser's. Whether the machine you generate it on is safe is a question no web page can answer; a phrase that holds real money is best generated on a device that never touches a network, and written on paper.

• What does the checker do?

  It confirms every word is in the list and that the last word's checksum matches the rest, which catches a misspelled or swapped word. It cannot tell you whether the phrase is the one you meant.

• How do I check the word lists are the real ones?

  Both lists are plain text files on this origin: the EFF long list of 7,776 words and the BIP39 English list of 2,048 words. Their SHA-256 is computed by your browser each time the page loads, so you can compare it with the published lists.

• What can PHRASE generate?

  Six words from the list Diceware uses, a password of any length, or a twelve-to-twenty-four-word recovery phrase with a correct checksum, each from the browser's own random source, with the strength stated in bits rather than a colour. If you would rather trust dice than a computer, roll them and type the numbers; the page only looks the words up. Nothing is sent, nothing is stored, and the page keeps no history.

Open PHRASE

• What should I look for in a privacy policy?

  Four things. What they collect beyond what the service obviously needs. Who they share it with, and whether the word partners is ever defined. How long they keep it. And what happens to it if the company is sold. The last is often a single sentence and often the most consequential one in the document.

• Does we do not sell your data mean much?

  Less than it sounds. Sell has a narrow legal definition in several regimes, and sharing for advertising, disclosing to processors, and transferring the data in an acquisition all fall outside it. The sentence worth reading is the one about sharing, not the one about selling.

• How long should a company keep my data?

  For as long as it needs it for the purpose it was collected for, and no longer. That is the principle in most data protection law. A policy that says as long as necessary for our business purposes, without naming a period, is not committing to anything.

• What kinds of clauses does POLICYLENS flag?

  The ones worth knowing about: whether the company can sell your data, keep it forever, make you give up your right to sue, or track you across the web. Each flagged clause is shown with a plain-language note, first as a summary and then as a clause-by-clause breakdown.

• Is the policy text uploaded?

  No. POLICYLENS reads the text in your browser and never sends it anywhere.

• Is POLICYLENS legal advice?

  No. It is a reading aid to help you skim faster and ask better questions. It flags clauses by matching a transparent list of common patterns, so it can miss things a lawyer would catch and can occasionally flag something harmless. When something really matters, read the full clause yourself.

Open POLICYLENS

• How do I open an .mbox file?

  Drop it on this page. POSTBAG lists every email in it, lets you search by sender, recipient, subject or Gmail label, and opens any message with its text, pictures and attachments. Nothing is installed and the file is not uploaded.

• How do I open the mbox file from Google Takeout?

  Drop the Takeout .zip on this page as it is, or the .mbox inside it, which is in Takeout, Mail and usually called All mail Including Spam and Trash.mbox. Gmail's labels are read too, so you can pick a label to see only those messages.

• Can I open an mbox file without Thunderbird or Apple Mail?

  Yes. POSTBAG reads it in a browser tab on Windows, Mac, Linux or a phone, without importing it into a mail program. Any message can be saved as an .eml, which every mail program opens.

• Can it open a very large mbox file?

  Yes. It reads the file once and keeps only each message's headers, so a mailbox of several gigabytes is listed without being held in memory. Each message is read from the file again when you open it.

• Is it safe to open old emails from an mbox?

  Here, yes: nothing in an email runs, links do not go anywhere, and pictures on the internet are not loaded, so no sender learns the email was opened. Attachments that run code when opened are flagged.

• How do I get attachments out of an mbox file?

  Open the message and save each attachment on its own, or save the whole message as an .eml. Attachments are decoded from the email exactly as they were sent.

Open POSTBAG

• How much smaller will my PDF get?

  It depends entirely on what is in it. A scanned or photo-heavy file stored at 300 dpi and pressed for the screen often comes out at a tenth of its size or less. A file that is mostly text, or whose pictures are already small JPEGs, changes little, and the report says so picture by picture rather than guessing.

• Will the text still be selectable and searchable?

  Yes. Only the picture streams are replaced, each under its own object number, so the text, fonts, links, bookmarks, form fields and page order are carried through as they were. The page is not turned into an image.

• What do screen, e-book and print mean?

  How many pixels a picture is left with for each inch it covers on the page: 96 for a screen, 150 for an e-reader or tablet, 300 for a printer. The page's drawing instructions say how large each picture is drawn, and that is what the target is judged from. Where a picture is not drawn by a page, for instance inside a pattern, its longest side is capped at the long edge of a Letter page at that density instead.

• Which pictures are left alone?

  Anything a browser cannot decode or that would break if re-encoded: JPEG 2000, fax-style (CCITT) and JBIG2 scans, indexed colour, bit depths other than 8, CMYK and spot colour, pictures with a transparency mask or colour key, and the mask images themselves. Small pictures written inline in the page's drawing instructions are not touched either. The report names every one and why.

• Does it turn logos and diagrams into blurry JPEGs?

  No. A picture with 256 colours or fewer (32 greys or fewer) is treated as a diagram: it is kept at full size and re-packed losslessly as indexed colour, which is usually much smaller and changes no pixel. Photographs and scans are the ones re-encoded as JPEG at the quality you set.

• Does it keep metadata, and what about a signed PDF?

  The document's title, author and other metadata are not touched; use another tool if you want them gone. A digitally signed PDF cannot survive any rewrite: the signature will no longer verify afterwards, so press a copy you do not need signed.

• Why is the result about the same size, or larger?

  If nothing in the file could be re-encoded, the whole file is still rewritten from scratch, and a file that was already tightly packed can come out a little larger. Keep the original in that case. A larger file with pictures in it usually means every picture was one of the kinds left alone; the report lists them.

• Is anything uploaded?

  No. The file is read, rewritten and previewed in this tab, on your machine, and the smaller copy is written by your browser. The page works with the network switched off once it has loaded.

Open PRESS

• How do I prove a file has not been altered?

  Take a cryptographic hash of it the moment it reaches you and keep that value somewhere safe. Re-hash the file later and compare. SHA-256 changes completely if any bit changes, so an identical digest means an identical file, down to the byte.

• Does a hash prove when I had the file?

  No, and this is where most home-made custody records fall down. The date in the record comes from the clock of the machine that wrote it, and a clock can be set to anything. To give the date force, publish the digest somewhere you cannot quietly edit later and that keeps its own timestamp: a git commit, an email to a third party, a public post, or a solicitor's file.

• Does renaming or copying a file change its hash?

  No. The hash is computed over the contents, so renaming, copying or moving leaves it identical. Opening the file in an application and saving it does change it, frequently even when you changed nothing on screen, which is why you hash first and then work on a copy.

• Is the file or the record stored anywhere?

  No. The file is hashed in this tab and nothing is uploaded or stored. The written record is yours to save; PROVENANCE keeps no copy of it, so keep the record somewhere safe if you will need to check the file again.

Open PROVENANCE

• Is it safe to type a password into a breach checker?

  It depends entirely on how it is done. The safe method is k-anonymity: hash the password locally, send only the first five characters of the hash, and compare the returned list in your browser. The service never sees the password or the full hash. Anything that sends the password itself should not be used.

• What does it mean if my password appears in a breach?

  It means that exact string appears in a public dump of leaked credentials, so it is already in the wordlists attackers try first. It does not necessarily mean one of your accounts was breached: a common password appears because somebody, somewhere, used it. Either way, stop using it.

• My password is not in any breach. Does that mean it is strong?

  No. Absence from a breach list says nothing about how guessable it is. A password can be unbreached and still fall in seconds to a rule-based attack because it is a dictionary word with a number and a punctuation mark on the end. Length and unpredictability are what matter.

• What does PWNCHECK send to the breach service?

  Only the first five characters of a one-way hash of the password, which cannot be reversed. The password itself is hashed on your device and is never sent or stored. The service returns hundreds of possible matches for those five characters and never knows which one is yours, or whether any of them is. The matching is done in your browser.

• What should I do if a password I use shows up?

  Change it, and never use it again anywhere. Consider a password manager and two-factor authentication. No match does not prove a password is strong, only that it has not been seen in these breaches.

Open PWNCHECK

• Which error correction level should I use?

  M for almost everything. L makes the smallest code when the text is long. H lets about a third of the code be damaged or covered and still scan, which is what you want on a sticker that will be scuffed or if you plan to put a logo over the middle.

• How much can a QR code hold?

  About 2,900 characters of plain text at the lowest correction level, and less at higher ones. A contact card with an address is a few hundred characters and scans fine; a whole paragraph makes a dense code that needs a good camera.

• Why does my Wi-Fi code not connect the phone?

  Check the network name is exact, including capitals and spaces, and that the security type matches the router: a WPA3 network is still WPA here. Phones join the network only when they scan with the built-in camera app, not from inside a browser.

• Can I put a logo on it?

  Not here, and you can afterwards: with error correction set to H, a logo covering up to about a fifth of the middle still leaves the code readable. Keep the three corner squares clear.

• Why not use an online QR code generator?

  The generator sites do the same thing and keep a copy of what you typed, which for a Wi-Fi password is the whole reason not to use them. This one draws the code in this tab and has nowhere to send it.

Open QRMAKER

• Can a QR code be malicious?

  The code itself is only text, but that text is usually a URL and your camera will offer to open it before you have seen where it goes. That is the whole attack: a sticker placed over the real code on a parking meter, a menu or a payment terminal sends you to a page that looks right and collects whatever you type into it.

• How do I see a QR code's link without opening it?

  Photograph or screenshot it and decode the image, rather than pointing a live camera at it. Decoding produces the text, and you read the domain and decide for yourself. This page decodes in your browser and never requests the address.

• What should I look for in the decoded link?

  The registered domain, read from the right: the part immediately before the first single slash. Watch for a lookalike spelling, a brand name that turns out to be only a subdomain of something else, a shortener that hides the real destination, and an @ sign, which makes everything before it a username rather than the host.

• If QRSHIELD shows no warnings, is the link safe?

  Not automatically. QRSHIELD shows you what the code contains and points out warning signs, but a link with no warnings is only a link with no warnings. Make sure you trust the destination before entering a password or a payment.

• Does QRSHIELD open or visit the link?

  No. It never visits the link for you, and it never uploads your image. The code is read in your browser and the decoded text is shown so you can decide before you go.

• Can QRSHIELD make a QR code?

  Yes. Type a link or text into the Make a QR code box, press Generate, and download the result as a PNG. The code is generated in your browser.

Open QRSHIELD

• Why not just use a header scanning website?

  Use one if you like, for a public page. It has to fetch the page from its own servers, so it cannot reach a staging site, an intranet application or anything behind a login, and it learns which address you are checking. RAMPART reads headers you already have, so it works on any response and sends nothing anywhere.

• How do I get the headers?

  From a terminal, curl -sSIL https://the-address prints them, one block for each redirect. In a browser, open the developer tools, go to Network, reload, pick the page and copy its response headers. The raw view and the tidy name-then-value view both work.

• Why is 'unsafe-inline' marked as ignored?

  When a policy also lists a nonce or a hash, every current browser drops 'unsafe-inline' and runs only the scripts that carry the nonce or match the hash. Leaving it in is the recommended way to keep very old browsers working, so RAMPART counts it as fine in that case and as serious without a nonce or hash.

• Why is a CDN in my script-src flagged?

  Because trusting a host trusts everything on it. Hosts such as cdn.jsdelivr.net and unpkg.com serve any file anybody publishes to npm or GitHub, and some Google and cdnjs addresses serve old AngularJS or JSONP endpoints that are known ways round an allowlist. Loading the file from your own origin, or moving to nonces with 'strict-dynamic', removes the problem.

• Does it give the site a grade?

  No. A grade adds together things of very different size: a missing Permissions-Policy on a brochure site and inline script allowed on a payments page are not on one scale. RAMPART sorts what it finds into serious, worth fixing and worth knowing, and says why each one matters and the line that fixes it.

• Can it tell whether the server reflects any Origin?

  Not from one response. When the headers allow a named origin to read the response, RAMPART says so and gives the curl command that sends a made-up Origin, which is the only way to find out.