Added

• A claim's details list every argument of yours that uses it. Below "Used in", a claim now lists your other arguments that use it — ones you own and ones you edit — with their version, whether each is a draft, and a note when one uses a different version of the claim. This shows on an argument's page as well as in your claims library.

• Your sources library includes arguments you edit. Sources cited in an argument you edit for someone else are now in your library, as your claims already were.

• You can block someone on the website, not only in the phone app. On another person's argument, Block author sits beside Report Argument in the argument's details. Each person listed under Participants has a Block button, which asks you to confirm first. Once you block someone, their arguments are hidden from you, and you can unblock them in Settings under Blocked users.

• You can report a claim. Select someone else's claim and its details end with Report and Block author, the same choices an argument offers. Before, only a whole argument could be reported on the website. Blocking a claim's author takes you out of the argument, as the phone app does, unless the argument is your own.

Changed

• Creating an argument looks and reads more clearly. Each step of the New argument window now shows how far along you are with a bar under its title, and Back and Next stay at the bottom of the window while the step scrolls above them. The two big choices, whose argument it is and how to build it, are cards you can click anywhere on. Only the step's main action is filled in; Back is quieter, and a Next that cannot be pressed yet stays visible with the reason beside it.

• Where a represented argument comes from is saved with Next. The citation step no longer has its own Save and Cancel: fill it in and press Next. If a box it needs is empty, it says which and stays on the step. Going Back and returning keeps everything you typed. When the argument is created, a tick marks it, and the two ways to build it sit side by side.

• Citation boxes and types are written in plain sentence case — "Page title", "Accessed on", "Book chapter" — like the rest of the site. The "Powered by Mozilla's readability" credit now sits under the link box it belongs to.

• The AI helper has one name, the Proposit Assistant. The panel, its manual and the menu item that opens it ("Open the Proposit Assistant") used to call it the Argument Builder.

• An argument's page header is calmer. "Add premise" and "Publish" are ordinary buttons in ordinary type rather than small capitals, and Publish is the one that stands out. An argument with no premises yet says so, with an Add premise button right there. The side panel's headings are in sentence case, and the note that a draft is not visible to others is plain rather than an orange warning.

• A button that cannot be pressed yet stays easy to see. It turns grey with a grey outline instead of fading almost out of sight.

• Clicking a button or choice no longer leaves an outline around it. The outline still appears when you move around with the keyboard.

• The review walkthrough keeps its buttons and progress in view. Back, Skip and Next stay at the bottom of the window, and the step and progress bar stay at the top, while each step scrolls between them.

• Nothing needs marking as unspoken any more. In an argument that represents a source text, a claim with no passage of that text linked to it now shows U for Unspoken, where it used to show R for Response, and a claim linked to a passage shows O for Original as before. So the "Mark as unspoken in the original" action, the separate Unspoken chip and the suggestions of what to mark are gone, and marks already set have been cleared. An argument with no source text shows no unspoken marker. An argument that responds to another still marks its own claims R.

• Only claims are linked to the source text now, not premises. A premise holds claims, and its title is only there to make it easier to read, so it has nothing of its own to quote. A premise's menu no longer offers "Link to the source text", and its header shows no Source chip. Links already set on premises have been removed; those on claims are unchanged.

• A claim taken from several places in the source text shows one Source chip, not one for each. The chip says how many, as "Source · 2". Opening it shows the first passage, with Previous and Next to move between them.

• Review appears only once an argument has something to review. An empty draft, such as an import that failed, no longer shows a Review panel with nothing in it.

• Accessed On is filled in for you on every kind of citation, and shown as a date to read rather than a box to type in. It used to be an empty box on social media posts, blogs, videos, podcasts and online documents.

• An imported argument keeps only the claims its premises use. A claim the import left out of every premise is now removed when the import is saved, together with any source or axiom that backed only that claim, so every claim you see traces back to the text.

• Proposit Scribe and the Proposit Assistant now use a newer OpenAI model, gpt-6-sol. Both used earlier GPT-5 models. You may notice different wording in the drafts Scribe makes from imported text, and in the Assistant's questions and the arguments it writes up.

• Adding a source to a claim backed by an axiom now asks first. The source replaces the axiom, which used to happen without a word. Choosing an axiom for a claim that has sources asks the same way, since the axiom replaces them. Cancel leaves the claim as it was.

• Cite a social media post by its username, its title or text, its website and the date you read it. A post by someone known only by a handle no longer needs an author, and a post's text can stand in for a title. The source is shown the way the IEEE citation style sets a post, for example J. K. Author [@user], "Title," Site, Aug. 12, 1995. Importing a post's link fills in its username, title or text and the day you read it, rather than putting the handle in the author's name.

• An author can be an organisation or a single name. Each author now has a Surname or Organization Name box and an optional Given names box. Leave the given names empty to cite "Reuters" or "Aristotle" exactly as written.

• Importing a Wikipedia article fills in its author as "Wikipedia contributors", the way IEEE credits Wikipedia, instead of leaving the author empty.

• An imported page title no longer repeats the site's name. A Wikipedia article's title now comes in as "Pascal's wager" rather than "Pascal's wager - Wikipedia", since the site already has its own Website Title box.

• Titles that a citation quotes now show their quotation marks, with the comma inside them, as in "Title," Journal.

• A source text is now chosen when an argument is created. Afterwards only its citation can be corrected, until the argument is published. A representative argument that was forked now tells readers that it kept the text and the claim to represent it from the argument it was forked from.

• Creating an argument is now a short series of questions. "New argument" first asks whether the argument is personal, proving your own point, or representative, setting out someone else's argument so you can respond to it. A representative argument asks where it comes from: paste a link to a post or article and its citation is filled in, and so is its text where it can be read. You then choose whether Proposit Scribe analyzes the text into a draft, or you skip that and give the argument a title. Once it is created you choose between building it with the Proposit Assistant and building it yourself.

• The analysis is now called Proposit Scribe, at a Quick or a Thorough depth. These replace the names "Scribe" and "Scholar". Importing text with Proposit Scribe is always offered.

• Arguments are no longer created from an X or Reddit post directly. To work from a post, create a representative argument and paste the post's link as its source.

• Paste a link to a post, a video or an article, and the source form picks the right kind of source for you. A link to an X or Reddit post fills in a social media source, a YouTube link fills in a video, and any other article fills in a website, with as many details as could be read. The link can be pasted before you choose a kind of source. If you had already chosen one, such as a newspaper article, an ordinary web page keeps your choice.

• The "Unclaimed Arguments" section is gone from your profile. Arguments can no longer be imported on someone else's behalf for them to claim later. An argument someone had already imported from your post that way now belongs to the person who imported it.

• Each claim says whether it is part of the original or a response. In an argument that sets out a source text faithfully, a small O marks each claim quoted from the source and an R marks the author's own reply. In a response to another argument, O marks what was carried over from that argument and R what the author added. Premises carry no mark; the claims in them do. Hover over or tab to the mark to read what it means.

• Say which passage a claim was quoted from while you write it. In an argument that represents its source text, the Add claim window has a Source Quotation section: choose Assign Source Quotation, select the passage, and it is saved with the claim. Editing a claim in Claim Details offers Reassign Source Quotation, which opens with the current passage already selected, and Remove Source Quotation. Nothing is saved until you press Save. If the claim saves but its quotation does not, you are told, and "Link to the source text" can add it.

• The Add claim window's Save button has moved to the bottom of the window, beside Cancel. The fields are labelled "Claim title" and "Claim body", and spaces around the body are trimmed, as they already were for the title.

• Choosing a source's type is quicker. The form now offers four common types first (Social Media, Website, Video and Newspaper Article); "Show all types" lists the rest, with a search box that also knows other names, so "YouTube" finds Video and "tweet" finds Social Media.

• The review walkthrough asks whether you agree with each claim. It used to ask "Do you believe this claim is true?" and offer True, False and Unknown, while answering the same claim on the argument page offered Agree, Disagree and Unsure. It now offers Agree, Disagree and Unsure, and the keyboard shortcuts follow the words: A, D and U in place of T, F and U.

• A review step's own heading no longer repeats the step number. The wizard's header says "Claim step 2 of 4"; a premise step now names only its role, "Conclusion" or "Supporting", and a claim step has no second heading.

• The review wizard says "Claim step 2 of 3" and "Premise step 1 of 2". It used to say "Step 2 of 3 — Claims Phase" and "Relationships Phase", while the phone said something else again. Both now use the same words, and screen readers hear the same label when a step arrives.

• An argument is not created twice when the server's reply is garbled. If the argument was created but the reply could not be read, the New argument window now closes, says the argument was created and refreshes your list, instead of reporting a failure and inviting you to create it again.

• Your review says when it isn't saved to the server. Reviews save in the background. If the server keeps refusing, the review now says it isn't saved to the server yet and is kept on this device, and the message goes away once a later change saves.

• Adding a claim recovers from a garbled reply. If the server's reply goes missing part-way through adding a claim, the claim is now added anyway when it can be. When it cannot, the window says the claim was saved but not added, so you can find it rather than typing it twice.

• An import says it has started. If an import's reply could not be read, the window now says the import has started rather than that an argument was created, and your list shows the new argument as importing.

• A save is not repeated when its reply is garbled. Across the website, a change the server made but whose reply could not be read is now shown as made, by reloading or refreshing what is on screen, instead of being reported as a failure you might retry.

• Creating an argument while offline says so. The New argument window used to hang with no way out when the connection dropped; it now says it could not reach the server and can be closed.

• Adding a claim shows its length limits. The Add claim window counts the title and body against their limits and holds Save back when either is too long, and it stays open when you click beside it.

• Renaming shows the title's length limit. Renaming an argument or a premise counts the title against its 100-character limit and holds Save back while it is too long, instead of failing when you save.

• A window you are typing into stays open when you click beside it. Creating an argument, renaming an argument or a premise, adding a source, linking a passage of the source text and reporting content used to close on a click anywhere outside the window, taking what you had typed with them. They now stay open until you save, press Cancel or the close button, or press Escape. Windows that only show something still close on a click outside.

• The ⚙ menus, and everything about the source text, are drawn in the site's own icon set. The menu items carried typographic symbols — a pencil character, a negation sign, a wastebasket; each now shows the same kind of mark used everywhere else on the page, including a row-insert mark for adding a claim before or after one, and a struck-through magnifier for content the original left unsaid. The Source chip on a claim, the Source Text panel and the markers in the text all now show a magnifier in place of the quotation mark.

• Renaming an argument or a premise from its ⚙ menu opens a small window. "Edit title" used to turn the menu itself into a text box; it now opens a window with the title in it. Enter still saves, and Cancel or Escape leaves the title as it was. Screen readers also read each menu item by its name alone, without the symbol drawn beside it.

• The "this argument can be repaired" notice now looks like every other notice on the page. It used to draw its own outlined box; it now uses the same alert the rest of the site does, with the issue list inside it. The words are the same, the Repair button does the same thing, and Details still opens the list of what is wrong.

• Adding a source tells you which detail is missing before it tries to save. If a reference is short a title or a year, the form names that box and waits, instead of saving and coming back with "Couldn't save the source". A year or an identifier typed in the wrong shape is called out too, with the shape it expects.

• The reference type list works the same way on every device. It is the same control as the other pickers in Proposit now: the 33 types are still under the same nine headings, you can still type the first letters of one to jump to it, and the arrow keys and Escape still work. The one whose name is highlighted stays on screen as you move through the list.

• Each author now has a box for their given names and a box for their family name. One line of names had to be guessed at, and the guess was to break at the last space — so "Ludwig van Beethoven" was filed under "Beethoven" with "Ludwig van" as a first name, and "Ursula K. Le Guin" lost the "Le". Nothing said so, because a name split the wrong way still looks like a name. Use "Add author" for the next person. Pasting a link still fills the boxes in for you, and now you can fix them where it guessed wrong.

Fixed

• Imports count less of your monthly AI usage. Part of each import's usage was being counted twice.

• A Convert to Argument that failed is still there after you reload. It shows in the conversation where you asked for it, with the reason it failed. Reopening the page while a Convert is still running now shows the conversation it came from. Trying a Convert again is no longer offered when it cannot work, and when a retry is refused the Assistant says why. Convert to Argument now counts toward your monthly AI usage, as its control always said it would.

• A finished review stays finished when you look back through it. Going back to an earlier step of a review you finished no longer marks it unfinished, so it stays shareable and anyone reading it still sees it finished. Changing an answer still means you need to reach the results again.

• Back in a review always moves one step. Going back from the first relationship step used to need two presses.

• A claim you skip is answered Unknown when you finish. Skipping a claim doesn't count toward your review's progress. When you reach the results, each claim you skipped becomes Unknown, and shows as Unknown in the results and on the argument.

• Arguments you edit for someone else now say that only the owner can archive or delete them, instead of offering actions that could not work.

• Your monthly usage renews properly. After the renewal date, Settings shows the next one, and your first use of the month counts from zero rather than on top of last month's.

• More room for the argument on a phone. The site links now sit at the end of the argument instead of taking space at the bottom of the screen.

• Clearer dashboard and controls. Published arguments are marked in green and drafts in amber, as on the argument page; menu and graph buttons name the argument or premise they act on; and the search box says it searches your own content.

• Clearer names for how an argument was made. An argument's details now say "Imported with Proposit Scribe" or "Built with the Proposit Assistant", and the Assistant's help and warnings call it by its name and point to the Convert to Argument button.

• A claim backed by an axiom says so on its card, for example "Axiom: Stipulation", so you can see how it is supported without opening its details.

• Editing an argument is steadier. Double-clicking Add premise adds one premise. Respond opens one reply box, where you pressed it. Negating a claim that moves its premise to Counterargument, or back, takes you there and says why. The operator you pick on an empty premise stays picked after you fill its first claim. An untitled premise has one name everywhere, and the header's premise count matches the contents list.

• When the Assistant can't build your argument, it says why. Choosing Convert to Argument before you have described an argument now tells you so and what to do next, instead of offering a Retry that could never work. Other failures say plainly that the Assistant couldn't finish, rather than "An unknown error occurred."

• Your Tasks page is readable. Each task is named in words, shows the argument it was for and, if it failed, why. Its details page now shows the task too, instead of an empty page.

• You can discard a draft from the argument page. The gear menu on your draft now has Discard draft, which deletes only the draft and takes you back to the published version.

• Archive is offered only where it works. On an argument with no draft, Archive Permanently is shown as unavailable and says why, instead of failing.

• A finished review reads as finished everywhere. The Review panel, its mark and Share now agree: a review is finished once it reaches its results without a contradiction, before and after a reload. Going back from the results to look at a step shows the review as unfinished straight away, where before the panel still said completed until you reloaded and Share refused to make it public. A claim you skipped and then took "as unknown" now counts as answered, so the panel reads "5/5 claim steps".

• Each review step opens at its top, so the results open with the verdict in view rather than scrolled to the bottom.

• "You skipped 1 claim", no longer "1 claims".

• A shared review's text is written for whoever reads it. It no longer includes sentences addressed to you, such as "Your answers contributed to this.", run together with the conclusion; those stay on the review the link opens.

• A claim's reasons all fit on screen. Choosing why you agree, are unsure or disagree with a claim, or why you accept or reject a step, used to open a list that could run off the top of the window, so its first reasons could not be picked. The list now fits in the room there is and scrolls.

• Vote counts belong to the version you open. An argument in your list with a draft over its published version showed no votes, although the published version it opens had them. "Reactions received" on your profile now counts those reactions, and no longer counts the upvote publishing records for you.

• Your Sources and Your Claims list each thing once, with the right label. A source you added to a draft was listed twice and said it was used in two claims when it was used in one. Each version is now marked Published or Draft, as on the argument's page, where a draft used to show the published icon. Other people's copies of your arguments, and arguments hidden by a moderator, no longer fill Your Claims. When one claim is used in two of your arguments, its "Used in" names both.

• A citation's author boxes no longer land on top of each other on a narrow screen. Below a certain width the surname and given-names boxes, and other pairs of boxes such as Year and Edition, were drawn in the same place, so text typed into one went into the other. Each box now has its own line.

• Next in the New argument window stays where it is. Choosing an option used to remove the hint above Next and move it up; the hint's line now stays. When a citation is missing details, the list of every box still to fill in appears right above Next, instead of at the bottom of the form, out of view.

• A link that could not be imported is kept. If Quick import by URL cannot read a page, the link now appears in the URL box once you choose the type.

• Release notes read as ordinary paragraphs. Lines used to break part-way through a sentence, leaving short ragged lines down the right-hand side.

• The legal pages show when they last changed. The Privacy Policy now says it was last updated on September 9, 2026, and the Terms and Conditions and Community Guidelines on August 14, 2026, the dates their wording last changed. They all used to say July 13.

• Signing in says what you are agreeing to. Below the sign-in buttons, a line now says that continuing means agreeing to the Terms and Conditions, Privacy Policy and Community Guidelines, each a link, as the app already did.

• One address for reaching us. The Report page now gives support@proposit.app, the same address as everywhere else on the site, instead of a second one.

• The terms are called the Terms and Conditions everywhere, including the sign-up checkboxes and the Child Safety Standards, which used a different name for the same page.

• The premise graph now says NOT on a negated claim. A claim a premise denies used to appear in the graph under its own words, marked only by a red border, so the graph read as the opposite of the premise. It now carries the same NOT label the argument's text shows, and a negated group of claims is labelled the same way.

• The review's Back, Skip and Next fit on one row on a phone. On a narrow screen the buttons drop their keyboard hints, which were what pushed Next onto a second row. An unavailable Next is now grey rather than faded, so it can still be read, and the buttons match the new-argument wizard's.

• Clicking a control no longer leaves a focus ring on it. The ring still appears when you move through the page with the keyboard.

• Archiving an argument now takes it off the Arguments count at the top of your profile as well as off Published, so Published and Drafts always add up to Arguments.

• Add claim finds claims from your other arguments. Searching in the Add claim dialog now looks through the published claims of your other arguments as well as this one's, marks each with "From another argument" and the argument's title, and places the one you choose. It used to search only the argument you were editing.

• A published claim can be found after its argument is edited again. Claim search used to look only at each argument's newest version, so a claim published in an argument that now has a draft on top was missed.

• Moving through a Source chip's passage with the keyboard now carries on where the chip is. Tab past the last button closes the passage and moves to what follows the chip, and Shift+Tab from the first button closes it and returns to the chip. Tab used to jump to the top of the page.

• Acting on a premise or claim someone already removed says so. If a premise or claim was deleted in another tab or on another device, making it the conclusion, renaming it, negating it, removing it, adding a claim to it or marking it unspoken now says "That premise has already been removed." or "That claim has already been removed." and the argument refreshes. It used to say "Please try again" and keep showing what was gone.

• Working in an argument that was deleted elsewhere says so. If you delete an argument in another tab or on another device while it is still open, the next change you make there says "This argument has been deleted." and takes you to your profile, instead of keeping a change that could not be saved on screen. Changing an operator in a premise or claim someone already removed now says so too, where it used to say "Please try again".

• Adding a claim to a premise someone already removed says so and closes. "Add a claim here" on a premise deleted in another tab used to say "Couldn't add the claim. Please try again." and keep the dialog open. It now says "That premise has already been removed." and closes the dialog, as adding or wrapping a claim elsewhere now does too.

• The counts at the top of your profile update as soon as you delete or archive an argument from the list. They used to stay as they were until the page was reloaded.

• A refused save in the New argument window no longer outlives what it was about. After a refusal such as "Source text is too long", changing the citation or the text now clears the message, instead of leaving it on every step while describing text that is no longer there. Trying again shows any new refusal.

• Resetting an argument now says it is done: "The argument was reset. Its unpublished changes were discarded."

• Selecting a claim in your draft with the keyboard keeps your place. Pressing Enter on a claim used to move the keyboard focus back to the top of the page.

• A long source address now wraps onto the next line in Citation Details, your Sources and the source text, instead of running off the edge of the panel.

• What you add to an argument stays there. A premise or claim you added could vanish when the page was reloaded, and the argument went on showing how it was before, even in a new tab.

• You can add a claim from another of your arguments. Picking one used to fail.

• The chips under a claim line up. The True/False/Unknown value, Source, the three reaction buttons and Respond are now one height and shape on one line. The value used to sit higher and smaller than the rest, and the reaction buttons were taller and square.

• A social media post cited without a title shows the start of its text in quotation marks, as a title would be: [@user], "Whole tweet text," X. A long post, such as a Reddit comment, is cut to its first twenty words with "…", where it used to print every paragraph in full. The post's text itself is kept whole.

• Closing the New argument window partway through now asks first. If you have entered a citation or any text, closing it asks whether to discard the argument. Keep editing takes you back to where you were with everything still there; Discard closes it, and it starts empty next time. Once the argument has been created, closing never asks.

• Choosing a premise from the Contents now shows its whole title. It used to land partly hidden under the Contents bar at the top of the page.

• Quick Import says what went wrong when a link can't be read. A link that isn't an article, such as a PDF or a site's home page, now says so and asks you to fill in the details yourself, instead of "Failed to parse webpage"; a problem reaching Proposit gets its own message. A site's home page or a Wikipedia disambiguation page is no longer filled in as if it were an article.

• Importing a second link no longer mixes two sources. Details the first import filled in are cleared, and those you typed yourself are kept unless the new link fills them. A post's text in the source text follows the new link too, and is left alone when the link can't be read. Importing a link while changing a saved citation replaces the saved details the same way.

• Forking an argument now shows that it is working, and pressing Fork again in the meantime no longer makes a second copy.

• A fork of a published argument now carries only that version's sources. Sources the author had added or removed in an unpublished draft were copied into the fork's data too, though they were never shown.

• An import that fails now says so. The argument's page shows "Import failed" with the reason and a Retry button, rather than an empty argument and a row of ticks, and your arguments list reads "Import failed" instead of "Draft". A cancelled import says "Import cancelled", with the same Retry.

• An import that could not work out how the claims connect now says that, instead of an unexplained "stage failed" message. When this happens, the analysis is now also tried again before it gives up.

• A premise with a long title from an import can be saved without shortening it. Saving it from the phone was refused, even when the title was left as it was.

• An import that fails after being tried again now says so straight away. Until now it could keep showing as in progress for several minutes before reporting the failure.

• An argument you started importing on your phone now shows "Importing…" on your web profile. Returning to your profile could list it as an untitled draft while the import was still running.

• An argument you have open while it is being imported now shows its title when the import finishes. The premises and claims appeared, but the heading still read "(No title)", the browser tab still read "Untitled argument", and the import progress with "Cancel import" stayed on screen until you reloaded the page. If the import fails or is cancelled, the page now shows that instead of staying in progress.

• Removing a premise's only claim now says the premise goes too. The confirm adds "This is the premise's only claim, so the premise will be removed too." Removing the last claim of an argument's only premise keeps that premise, empty, since an argument needs at least one premise.

• Choosing an axiom for a claim that has sources now replaces them. The sources used to stay listed on the claim, and adding a source afterwards failed.

• Removing a claim from a premise now asks "Remove this claim from this premise?" rather than naming an internal term.

• Clicks just after an argument opens now count. For a moment after an argument loaded, the page quietly redrew itself: a list of citations you had just opened closed again, and a click on a claim, citation or source could do nothing until you clicked a second time. The page now stays as it first appears.

• Only one "Contents" bar shows while you scroll back to the top of an argument.

• Adding a claim that is already in the argument shows its usual letter, rather than a symbol such as "P4", and Claim Details lists each premise that uses it only once.

• Sources added to a draft stay in the draft. Adding or removing a source on a claim in a draft also changed the published version, and resetting the draft did not undo it. The published version now keeps exactly the sources it was published with.

• Busy moments no longer fill a source in with blanks. If too many sources were read at once, the form said it had succeeded and left every box empty. It now says the page could not be read, so you can try again or type the details in.

• Resetting a draft no longer removes sources from earlier versions. Resetting a new draft of a published argument used to remove the sources cited on its claims from the published version too. Now a reset removes only the sources added in the draft.

• Resetting a response's first draft keeps it a response. Comparing a response with the argument it answers used to show every part as removed and added again after its first draft was reset, even with nothing changed. The reset also brought back the original's title instead of the "Re:" title, and added the original's author to the response. Now a reset leaves the response exactly as it was when first made.

• A keyboard shortcut pressed right after moving on can no longer change the answer you just left. In the review walkthrough, pressing a letter the moment after Enter could reach the previous step instead of the new one — so the answer you had just given could be quietly replaced, and the step in front of you left unanswered. Every shortcut now acts on the step on screen.

• The review walkthrough fits a phone's screen. On a narrow screen the button to finish a step ran off the edge of the window. The buttons along the bottom now move onto a second row when they do not fit, and an answer too long for its space wraps onto a second line rather than being cut short.

• The review results no longer overlap. The panel reporting the conclusion and the argument drew its findings over the explanation beneath it; each finding now sits inside the panel.

• A profile page for a username that does not exist no longer claims to be that person's page. The page itself said "User not found" and always had, but the browser tab above it read "someone's arguments" — naming an account nobody holds. That wrong name was what got saved if you bookmarked the page, and what showed in your history afterwards. The tab now says the same thing the page does. Typing the address in was already correct; following a link to it was not.

• Report Argument no longer appears on your own argument. There was nothing to report, and your own claims already left it out.