Release notes
Auto-extracted sources are clearly marked. When the analyzer pulls a source out of your text but hasn't yet turned it into a full reference, the source now shows the extracted text plus a clear "needs…
June 23, 2026
Reply with a counterargument in one click. When you're reading a published argument, a new "Create Counterargument" button sits at the top of the argument — click it to spin up your own editable copy…
June 22, 2026
Building a premise by hand is now direct: instead of choosing "Add first claim" or "Add inference" up front, an empty premise shows fill-in "skeleton" nodes you complete by tapping a "+". Selecting a…
June 21, 2026
Moderators and admins can delete published arguments. A user holding the argument:delete system role (or admin:full-access) can now hard-delete a published argument — previously no one could, and a pe…
June 21, 2026
Edit an argument's description. The argument actions menu (the ⚙ button) now has an "Edit description" option alongside "Edit title". Your edited description is saved and sticks — previously this coul…
June 21, 2026
Faster failure diagnosis for staff. When an import fails, staff with debug mode on now see a single consolidated diagnostics panel on the task page that summarizes everything that went wrong — the ove…
June 21, 2026
Import failures now explain themselves while you watch. If an import you're watching fails, the page now shows the actual reason ("Import failed: …") the moment it happens — you no longer have to refr…
June 19, 2026
No user-facing changes. A maintenance release: faster local semantic search (the embedding cache now keeps its vector indexes model-agnostic, so they stay effective as the local model setup changes),…
June 18, 2026
No user-facing changes. This release swaps the local development LLM backend from Ollama to a local llama.cpp server. Production argument import and semantic search are unchanged — they continue to ru…
June 17, 2026
When you import an argument from raw text, you can now pick how deeply it gets analyzed:
June 17, 2026
Fixed another case of the "Checksum mismatch" error when editing a premise twice in a row. After the v0.22.1 fix, some older arguments could still hit a spurious "checksum mismatch" (409) on the secon…
June 17, 2026
Fixed a spurious "Checksum mismatch" error when editing the same premise's title a second time. Editing a populated premise's title twice in a row could fail the second edit with a "checksum mismatch"…
June 17, 2026
Citations pulled from the text of an argument now display reliably wherever they appear, including references that don't have a web link (books, court cases, footnotes).
June 16, 2026
Copy an argument to your clipboard. Open the argument menu (the gear button at the top of any argument) and choose Copy to clipboard to grab the whole argument as clean, readable text — its title and…
June 15, 2026
New "Your Tasks" page. Your profile now lists every import and argument-building job you've started, newest first — with quick actions to cancel a running job or restart one that was cancelled or fail…
June 15, 2026
Your Claims page loads much faster. It used to load the full logic details for every one of your claims up front, which made the page slow when you had a lot of them. Now the list appears right away a…
June 15, 2026
Sign in with X works again. Logging in with X (Twitter) was getting stuck on the X login screen — you'd sign in, but never get brought back to Proposit. X sign-in now completes and logs you in.
June 14, 2026
Publishing is now one click. The Publish button sits right next to "Add Premise" in the argument header instead of being tucked away in the gear menu, so sharing a finished argument is front and cente…
June 14, 2026
Start with a blank argument or the AI builder — your choice. When you create an argument, choosing "Assisted" opens the Proposit Argument Builder right away, while choosing "Empty" gives you a clean s…
June 13, 2026
Signing in with X (Twitter) works again. A recent change to support Sign in with Apple unintentionally broke the X sign-in flow on iPhone and other Safari-based browsers — after logging in with X you…
June 12, 2026
Importing an argument from text now runs entirely on our servers, so it finishes on its own even if you close the tab, switch apps, or lose your connection. Re-open the import any time and you will se…
June 3, 2026
Imports now show a full stage-by-stage progress indicator from the moment the import starts, with each stage listed as pending before its LLM call begins. You no longer see a blank list that fills in…
June 3, 2026
Imports settle on a conclusion more reliably. When an imported argument had several claims that could each be the main point, the builder used to sometimes give up and produce nothing. It now picks th…
June 3, 2026
Refreshed Profile & Explore. Your profile and the Explore page have a cleaner, more modern look: arguments now appear as easy-to-scan cards, the pages work better on phones, and starting a new argumen…
June 2, 2026
Stop an import you no longer want. While an argument is being built from your text, the import-progress page now has a Stop button. Confirm, and the import winds down cleanly at the next step instead…
June 1, 2026
_This release contains behind-the-scenes testing and stability improvements. No user-visible changes._
May 31, 2026
Cleaner claim wording in imported arguments. When you build an argument from text, each claim now reads as a plain statement — for example, "Rain wets the ground" — instead of being wrapped in "The au…
May 29, 2026
Fixed: arguments with web-link citations now open reliably. Some arguments that cited a source by its web link could fail to load. Those arguments now open as expected, and newly added link citations…
May 28, 2026
Pipeline status reliability fixes — Fixed two bugs in the pipeline status view: clicking a stage that hadn't started yet no longer throws a 500 error, and completed stages no longer hang in a "Loading…
May 28, 2026
Improved progress UI for AI-assisted imports
May 27, 2026
Argument imports now allow more headroom for long-text segmentation based on your account tier (Tier 2 gets a higher cap than Tier 1). Long-text imports that previously hit an internal cap mid-generat…
May 27, 2026
See a detailed per-stage view of any argument import at /task/<id> from the task progress link. Click any stage for outcome, timing, and per-stage diagnostics — including any warnings or errors surfac…
May 26, 2026
Argument imports now use our improved multi-stage AI pipeline by default for better extraction quality on long or structured inputs. Imports also enforce your monthly token budget throughout the run —…
May 25, 2026
Argument imports now record per-stage progress in the background so we can show you a detailed view of how each import was processed (coming soon). No change to import behavior; no user action needed.
May 24, 2026
Selecting a claim that appears in more than one place no longer jumps your view to a different part of the argument. The selected claim closest to where you're already looking stays put.
May 22, 2026
Selecting a claim or premise in an argument view is now instant. Tapping or clicking a claim card or a premise title used to take a noticeable beat in production — the page was running a server round-…
May 21, 2026
Claim Details edit controls are now hidden when you can't edit the argument. The Add Source button, the axiom assignment picker, the per-citation remove icon, and the axiom-card remove icon in Claim D…
May 20, 2026
Search keeps working when our AI provider is down. When the semantic search service is unavailable, search now falls back to keyword matching across your arguments, claims, premises, and sources inste…
May 16, 2026
Fixed argument view text overflowing the right edge on narrow viewports / mobile. Claim cards and the conclusion-premise header now stay within the visible screen on phone-sized displays.
May 12, 2026
Citations now save reliably across all claim types. Inserting a new claim immediately after or before another claim now lands in the correct position. A handful of polish issues in the argument view a…
May 9, 2026
Reading an argument now starts in text view. The text outline of an argument is the default when you open one — easier to skim, easier to share. Citations show up right next to the claims they support…