Preregistrations now lock at first distribution
Surveyor's preregistration export has always been generated on demand — which meant a draft written after you'd already seen your results could look identical to one written before you collected a single response. That gap undercut the one feature that exists to stop p-hacking and HARKing.
Creating a project's first distribution now takes a SHA-256 fingerprint of the preregistration draft at that exact moment and locks it permanently — a second distribution can't re-lock it, and the draft's content can't be rewritten into the commitment after the fact. The preregistration export now always states where things stand: a locked draft shows the timestamp and a short proof prefix; an unlocked one says plainly that it isn't timestamped yet and that your first distribution will lock it.
Worth being precise about what this is: a commitment, not a notarization. It proves your study didn't change after Folio saw your first distribution go out — it doesn't prove anything to someone who doesn't trust Folio's clock, and it isn't a signature. Editing your project after the lock (the research question, the instrument) also isn't detected yet — that's next.
A real "Download proof" button on the certificate share card
Every anchored Integrity Certificate has had a complete, offline-verifiable proof file behind it since anchoring shipped — inclusion proof, timestamp root, the works, plus a raw .ots file for standard OpenTimestamps tooling. The only way to reach it was a single word, "Proof," buried mid-sentence in a paragraph on the /verify page. The share card — the place people actually go to do something with a certificate — didn't mention it at all.
The share card now has a Download proof button alongside Copy link, X, and LinkedIn, plus a quiet secondary link to the raw .ots file for anyone running their own verification tooling. Nothing changed about what's provable; it's just findable now.
Also new
- A published Reviewer Report now links straight to a journal match. If you've run the AI reviewer panel on a draft and published the report, the "report is live" confirmation now includes a link into journal matching for that same document — one less step between getting feedback and knowing where to send it.
- Tracking a topic now survives from a search result all the way into Radar. Read a full
/papers/topic page and sign up from there, and onboarding's "Track" option now pre-seeds your first Radar agent with that exact topic instead of dropping you into a blank one. The topic pages themselves also now offer to track new papers automatically, not just search once.
Fixed
- DOCX and PDF exports silently dropped part of a list item. A list item that held a nested sub-list (Tab to indent) or a second paragraph (Shift+Enter) exported those extra blocks in HTML and LaTeX, but not in DOCX or PDF — the two formats people actually hand in. Both now walk the full item, so nested lists and multi-paragraph items export completely, with no visible change to ordinary flat lists.
- DOCX and Print/Classroom exports ignored a figure's alignment and size. Right-aligning a figure and shrinking it to 25% in the editor already carried through to your PDF; the DOCX export and the Print/Classroom preview both ignored it and rendered every figure full-size and centered instead. All three now agree.
- A Surveyor instrument version bump could silently undercount your own results. Editing a question mid-collection creates a new version, but responses collected before the edit still reference the old question — and results/export was only reading the current version, so per-question stats quietly excluded valid pre-bump answers while the response count kept including them. Results and every export format now show a clear warning when this applies, so the gap is visible instead of hidden inside the numbers.
- The public transcript-verification page (
/transcript/[code]) wasn't localized. It's now fully translated and mirrors correctly for Arabic and Hebrew, joining the public Reviewer Report page on the same treatment. - Removing a student's degree track had no confirmation. One click on the track's remove button deleted it immediately, along with every waiver recorded against it — now a confirm step names the track before anything is removed. The remove, promote, and edit-ID buttons on that row also gained accessible labels for screen readers.
- Onboarding's option groups (role, academic level, citation style) had no accessible group name. A screen reader announced each choice on its own with nothing tying it to what you were choosing between; each group is now announced as a group.