Added

• Review a response, and then each argument it answers. Reviewing a response asks, for each of its points, whether you agree or disagree. Its results list your answers, what they fill in on the argument it answers, and anything they can't carry there, with the reason; "Continue to …" opens your review of that argument with those answers already filled in, each saying which point it came from, and asks only the rest. Your own answers always stay: if you have reviewed that argument before, you are told first, only what you haven't answered is filled in, and any answer of yours that differs is listed. If you change an answer on the response later, the older review offers to fill in again. If none of your answers can be carried, the results say why and Continue still opens that review, with nothing filled in. This works signed out too, in your browser, and moves to your account when you next open the argument signed in, once your account is set up.

• Cite a response as your reason. When you disagree with a claim or reject a step, you can point to a response that already makes your case. The choices appear under the reason, with "No cited response" first. Your results list the responses you cited and open them. A citation never changes your results, and changing your answer removes it. A citation made signed out moves to your account with your review, even if the response has been updated since; if it can no longer be cited, because it was hidden or removed or you have blocked its author, your review moves without it, your answer is kept, and you are told.

• "Cited by N reviewers" on each point of a published response that readers have cited in finished reviews.

• Reject the conclusion's own step. When the conclusion has a step of reasoning, such as "if A then C", you can now reject it. The argument is then shown as not establishing its conclusion, because you rejected the conclusion's own step; the conclusion's value is left as your claims make it.

• Bring your response up to date when the argument you answered changes. When an argument you responded to publishes a newer version, your response's header says so and shows how many of your links point at something that changed; the newest version shows the same offer beside your response. Review and update lists each changed link under its point, with what it said before and what it says now, and lets you keep it, point it at something else in the new version, or remove it. Removing a point's last reason removes the point, and the screen shows that before you confirm. The update changes your draft only: publish it when you're ready. Pressing Respond on a newer version than your response answers offers the same update first.

• Step between an argument and its responses with < and >. Above the premises of a response, or of an argument with responses, a strip names the argument you are reading and the one beside it. < takes you to the argument a response answers, at the version it answered. > takes you back to the response you came from, or, if you have not been further yet, to the argument's Responses section. Stepping back the way you came is instant, and the page keeps its place.

• "＋ Add the reason" under a counterargument that has no reason yet, as in the app. Pressing it lets you pick the claim that says why.

• Respond to any claim, or to a step of reasoning, without leaving the argument. Press the responses button on a claim card (with responses already there, Respond is at the top of the list it opens), choose Contradict or Affirm, and pick or write the reason that backs your answer. The same button on a premise's heading lets you Undercut or Reinforce the reasoning itself. Everything you answer goes into one response of yours to that argument, which "Your response" in the argument's actions opens. A claim you have answered is marked in your answer's colour on its responses button. You can also respond to a response. On a response, Respond on one of its points answers that point, with the form right inside its card. Before removing a point, the question names what the point answers and any other point that goes with it. A point's menu also deletes the reason shown under it.

• In your own draft, Respond writes a rebuttal straight into it. The new premise is titled "Contradiction of …" or "Affirmation of …", rests on the reason you picked, and opens in the argument's list of premises.

• Every argument now lists its responses, and a response shows what it answers, point by point. Under an argument, a Responses section lists the arguments that answer it, with how each one answers and which version. A claim or premise that responses answer shows how many, and pressing the count shows only those responses. Opening a response shows what it answers at the top, each of its points under its own heading, the words of the argument it answers, and whether each point holds up. A profile has a Responses tab beside Arguments. Copies of an argument made to rework it are now listed separately, under "Forks".

• Your account page shows how many responses you have created. A "Responses created" line sits under "Arguments created", with its own limit, which is the same size as your argument limit.

• Nine more Federalist Papers, and Alexander Hamilton. The curated arguments now include Federalist Papers #1, #14, #23, #39, #47, #51, #70, #78 and #84 beside #10, each with the full text of the paper, the passage each claim comes from, and a citation to where it was first printed. Hamilton has his own curated profile with his five papers.

Changed

• Clearer choices when you respond to a claim. Under Contradict and Affirm the form now says what each choice means for the claim, in its own words: "Q is false" or "Q is true". On a claim shown as false, Contradict reads "No: Q is true" and Affirm "Yes: Q is false". A counterpoint you add to your own draft is titled the same way, for example "Q is false".

• Easier-to-read claims and premises. A claim's title and a premise's title are a little larger, a claim's description is the same colour as its title rather than greyed out, and the "is true" / "is false" band under a claim is a little larger.

• Argument lists show the date an argument was published. Each row now shows the day it was published, such as "Sep 5, 2026", and a draft that has never been published shows the day it was created, such as "Created Oct 8, 2026". A published argument with a draft open shows when the published version came out. The date no longer changes a moment after the page loads, and it reads the same as in the phone app.

• Every date about an argument shows the same day everywhere. The argument history, the forks list, the responses under an argument, the update screen for a response and your review's "Created … · Last updated …" line now all show the day, such as "Oct 4, 2026", the same day the phone app shows. The argument history and forks no longer show a time, and the review line no longer says "3 days ago".

• A representative argument is now your representation of someone else's argument plus your own response to it, published together. Responding on your representation's draft makes that draft what your response answers, and publishing either one publishes both, so your response always answers exactly the version readers see. Publish always tells you your response is published too. If either is refused, neither is published. When your edits to the representation changed, added or took away something your response answers, you see each of those points first, and keep or remove each one; a point whose last reason goes with a removal is listed with it. If the representation changes again after you chose, you are asked again rather than keeping a point on words you never saw. If you remove every point, your response isn't published: publish the representation on its own instead. People working on your representation can still publish it, and it is published on its own. If content screening refuses either one, you can appeal.

• Discarding your representation's draft looks after your response. If your response was answering that draft, the question before you discard now says which version your response goes back to and which of its points go with the draft, including when you first wrote your response on that draft. Discarding a representation never published says your response goes with it, and how many points it has.

• An argument's counterarguments sit in its one list of premises. The Proposition and Counterargument tabs are gone: each counterargument is listed and numbered with the other premises, the premise count includes them, and Jump moves you up or down the page to it and back.

• Going back to a draft you can edit always shows its latest version. Using Back or Forward to return to a draft you had edited could bring back the copy from your first visit, and your next edit then failed with "someone else edited this". The draft now loads again on the way back, and if it cannot, editing stays off until "Try again" succeeds.

• Respond no longer makes a copy of the argument. It used to fork the whole argument into a new draft and open a composer there; now your answer is written in one step and you stay where you were. Removing the only reason for something you answered asks first, and removes the answer with it.

• In an argument that represents its source text, every claim says whether the text states it. O (Original) when a passage of the source text is linked to the claim, E (Enthymeme) when none is. "Unspoken" is now called "Enthymeme", and the Response mark is removed. A claim keeps the same mark wherever it appears in the argument. A fork of an argument with no source text no longer marks its claims at all.

• A personal argument started from imported text no longer shows that text. Its marks, Source buttons, source-text section and reader, the link to the source text, the citation and the line saying where it was imported from are hidden, and copying it leaves the passages out. Nothing is deleted: the text is still kept with the argument.

• The argument details no longer say how an argument was made, unless it represents its source text. A hand-written argument's details used to read "Created manually" or "Built with the Proposit Assistant"; that line is gone, along with "Imported with Proposit Scribe" and the post links on a personal import.

• Failed requests say what went wrong. When something can't be saved or loaded, the message now says whether the server couldn't be reached ("Check your connection and try again") or ran into a problem ("Try again in a moment"), after saying what failed — for example "Couldn't save the source. The server ran into a problem. Try again in a moment." You no longer see "Internal Server Error", "Request rejected" or "Search rejected". A change made in a tab that had fallen behind now says the argument was changed somewhere else, instead of "Checksum mismatch", and a request made after your session ended says you're signed out.

• A vote or premise decision that fails to save says so. It used to quietly go back to what it was.

• Each claim says whether the argument holds it true or false. In the argument view a thin ribbon under every claim reads "is true" in blue or "is false" in red, in place of the NOT above a negated claim's title and its red stripe, so a premise reads top to bottom as a sentence. The connective between claims now reads "if" rather than "is true if". A screen reader says the same words straight after the claim's title.

• A claim reason has clearer words. "Proven false elsewhere" is now "Shown false elsewhere", described as "Credible prior work or analysis contradicts it." Answers you already gave with it keep it.

• A claim's responses, reactions and citations are icons with counts. Each claim shows its title with its menu, then its text, then one line of small buttons. On the left: reactions, responses and citations, each an icon with how many there are. On the right: the source passage and, on a claim that has been rebutted, the crossed swords that take you to the rebuttal and back; then your value for the claim as its first letter ("T", "F" or "U", with a small arrow for its menu), and the claim's origin. Hover over or focus any of them for its name; the reaction and response buttons say just "React" and "Respond" (or the count), since the card already shows which claim they belong to. The citations count shows even at 0, and a claim that still needs support shows its citations icon in the warning colour with a small warning mark, with the reason in its tip.

• Pressing a claim's responses count opens its responses, with Respond at the top. Respond and the count of responses are now one button, on claim cards, on a premise's heading and on a response's points. With responses, it shows just those responses under the argument, with Respond above them, and your answer is written right there. With none yet, it opens the form to write one.

• React to a claim from one button. A claim's reactions start as one thumbs-up button. Press it to show Agree, Unsure and Disagree, and press one to react; each still asks for your reason. Your reaction stays visible on the folded button, and your reason is shown under the claim. Press your reaction again to take it back.

• See how others reacted once you have reacted yourself. After you react to a claim, or decide a premise while reading, you see how many readers chose each answer. Before that you see only which way readers lean and whether they are divided, as before, so your own answer comes first. The argument's author always sees the numbers.

• Arguments that describe violence can now be published. Publishing used to be refused automatically when an argument described violence vividly, even when the description was the point, such as how animals are slaughtered. Those arguments now publish. Content can still be reported for a person to review.

• Every Federalist paper is titled by its number. Curated Federalist papers are now titled "Federalist Paper #N: …", so #10 is "Federalist Paper #10: The Union outcontrols state factions". Its text is now the Project Gutenberg edition, and its citation names the newspaper that first printed it, the Daily Advertiser. Each paper's "Full text" link goes to that paper in the Project Gutenberg edition.

Fixed

• Accept and Reject match the buttons beside them. On a selected premise, the Accept and Reject buttons were taller and square-cornered next to the responses button. They are now the same small round buttons.

• A premise's name gets room on a narrow screen. Where a premise's buttons left its name no room, such as the conclusion on a phone-width window, the buttons now move to a line of their own just below the name, which gives the name the whole width of its line.

• A response quotes a "not" over a group. Answering part of an argument such as "not (A and B)" quoted it as "A and B", which says the opposite. A quote now keeps every "not" and every group where the argument has them, and reads a step as "if A then B".

• A conflict in a review shows what you answered. A claim you set False that a step you accepted then settled both ways was listed only as "Contested"; it now reads "False → Contested", as the argument's own view does.

• Copying an argument keeps a "not" over a group. The copied text of "not (A and B)" read the same as "A and B"; the "and" line now says the group around it is negated.

• An empty draft offers the Proposit Assistant. Its message suggested building the argument with the Assistant, but the only way there was the settings menu. "Build with the Proposit Assistant" now sits beside Add premise, marked as a metered feature, and opens the Assistant on the draft.

• A conflict in a review names every part of the step. Where a part was itself "not" something, or a group, the message read "the not below it" instead of saying what it was; it now says, for example, "not Taxes should rise" or "(A and B)".

• A cited claim you marked false no longer shows as contested. After a review, the argument could show such a claim as "False → Contested" while the review's results found nothing to settle. It now shows the value you gave it, as the results do.

• A button label that wraps onto two lines stays centred.

• Converting a conversation in two places at once converts it once. If you pressed Convert to Argument with the same draft open in two tabs, or on the website and the app, each press started its own conversion, and the one that finished last replaced what the other had written. Now the second press shows the conversion already under way, and you get one result in both places. While a conversion of a different conversation, or someone else's, is running on the draft, Convert says "This draft is already being converted. Try again once that has finished." Pressing Retry on a failed reply or conversion in two places, or again later, also runs it only once.

• Buttons at the end of a list row no longer crowd the edge. In the participants list and the other lists with a button or menu at the end of a row, that control sat about half as far from the right edge as the row's text sits from the left. Both sides now have the same space.

• A contradiction's explanation reads naturally when a premise has no title. It said "a premise you accepted, which you accepted" or put the words "a premise you accepted" in quotation marks as if they were a title. It now says "a premise you accepted settles it the other", "both values come from premises you accepted" or "when you accepted one of the premises".

• A constraint is called a constraint everywhere. A constraint is a premise that states something must hold rather than drawing a step such as "if A then C". Your review's results already listed these as constraints, but the premise's own review step, its details, its heading when it has no title, and the list of formulae called it "Supporting". They now all say "Constraint", and an untitled one is named "Constraint premise".

• A dropped connection no longer loses your message to the builder. If the connection dropped just after you pressed Send, your message used to vanish even though the Assistant had started replying. Now it stays and the reply appears under it. If the app can't tell whether it was sent, it says "Not sent — the connection dropped." with a Retry that won't send it twice.

• You can reach a button you can't use yet with the keyboard, and see why. Before you finish registering, the vote, report and block buttons on an argument, and the vote buttons on its claims and premises, are off. Tab used to skip them, so the note saying why only showed to a mouse. Now Tab stops on each and shows the note; pressing it still does nothing.

• Screen readers announce the reason list and the source passage by name. Opening the list of reasons after a vote or a decision, or a claim's source passage, used to announce an unnamed dialog as well. Now the reason list is read as one menu named by its question, and the passage as one dialog.

• You're told when your browser won't let Proposit copy a link. Copying an argument's sharing link, a section's link or an invitation link used to fail silently if the browser refused. It now says the link wasn't copied, and the sharing link stays selected so you can copy it yourself.

• The home page's opening sentence no longer runs words together on a phone. On a narrow screen it read "argumentwith" and "system —and".

• A folded button acts on your second press, however quickly it follows the first. Where actions show as icons, the first press shows the button's words and the second used to be ignored while the words were still appearing, so Publish could take a third press.

• The Proposit Assistant picks up where it was after a dropped connection. If the connection dropped while the Assistant was answering or converting a conversation into an argument, the chat kept showing it working until you reloaded the page. It now reconnects on its own and shows the answer when it arrives. If the answer can no longer arrive, the message says the Assistant couldn't finish and offers Retry, on the website and in the app, instead of waiting for good.

• Your message stays in the chat when the Assistant's reply fails. It used to disappear, leaving only the failure and its Retry. The message box also empties once the message is sent, so pressing Send again can't ask the same question a second time; use Retry to ask again.

• When the Assistant can't take your message, the chat says why. The message used to disappear from the chat with no explanation; it is still given back in the message box, and now the reason is shown above it.

• Opening the same conversation twice no longer asks the Assistant the same question twice. Reloading the page, or opening the conversation somewhere else, while the Assistant was answering could start the same request again, using up more of your monthly token limit.

• Going Back from an argument to a list shows the list as it is now. After you blocked the argument's author, or the argument was deleted or taken down by a moderator, Back used to show the list as you left it, with that argument still in it. The list is now loaded again.

• A claim that couldn't be added no longer leaves a hidden piece behind. When adding a claim failed partway, for example because the premise had been removed in another tab, part of it could stay in the argument unseen.

• Arguments fit a narrow phone screen. On a screen 320 points wide, the sidebar's buttons, each premise's graph button and menu, and every claim's buttons and menu now stay on screen, and a claim's menu no longer covers its title. Choosing a citation opens its details at any screen width.

• The page no longer shifts sideways when a dialog opens. On a computer whose scroll bars stay visible all the time, opening a dialog such as Add claim, Edit title or New argument used to make the page behind it slightly narrower, so it jumped left, then jumped back when the dialog closed. The page now stays exactly where it was.

• An argument a moderator has hidden can no longer be published or responded to. Publishing it shows a message saying the argument is hidden, with a link to appeal; the draft is left as it was, and you can keep editing it. Forking or responding to it shows a message saying it is hidden.

• Publishing now says why it was refused. When an argument cannot be published, for example because its premises cannot all be true at once, a message now says so. Before, pressing Publish could appear to do nothing.

• A citation can be dated to its year, its month or its day. When you cite a source, its date is now entered as a year, then a month and a day if you know them, and the form shows how it will be cited: "1787", "Nov. 1787" or "Nov. 22, 1787". Before, the date had to be a whole day picked from a calendar. A book, a journal article and the other sources dated by year can be left undated, and are then cited "(n.d.)".

• A citation's date shows the day it names, wherever you are. A citation's date is printed as the same day for every reader, and shows as that day when the citation is opened again. Before, a reader in a different time zone from the author could see it printed a day early or a day late. Citations already saved keep the day they were given.