Fixed
- Two-column export only worked in DOCX. The editor's two-column layout toggle rendered correctly in DOCX but was silently dropped in PDF and never even reached
.texexport — same document, same setting, three different results, with nothing telling you..texnow genuinely renders two columns. PDF still can't (no multi-column text flow in the renderer we use), so instead of pretending, it now shows a toast explaining that and confirms DOCX and.texboth kept the layout. - The AI-assist disclosure toggle wasn't enforced on the public certificate page. The "AI-assist detail" toggle is your promise that the per-feature AI-assist breakdown stays private from anyone holding the certificate link. It was correctly checked in the professor-facing batch checker, but
/verify/[code]showed the full breakdown regardless of the toggle. Both surfaces now agree. - Shared
/verify/[code]links previewed as a generic "Folio Integrity Certificate" card, even for a fully published Folio Page. Link previews now show the document's real title and author, matching how other share surfaces already look. - A discovery agent briefing that failed to send could be lost. If the email failed to go out (a transient provider hiccup), the agent was still marked as having received its briefing, so it silently skipped to the next cycle. A failed send is now retried instead.
- Figure Library could crowd out the chart builder on narrow screens. On phone-width layouts, opening the Figure Library in Studio squeezed the chart builder into a sliver. It now opens as a bottom sheet on small screens instead of sharing space with the builder.
- Hardened the public directory's pagination so it keeps listing every opted-in profile as Folio grows past 1,000, instead of quietly capping at the first page.
Also new
- Instructors get a read-only integrity signal in the Classroom grader. The same source-integrity check a writer sees on their own manuscript — retracted or orphaned citations — now shows up (read-only) when grading a submission. It's deliberately not a score: the checks that need a live editor or the writer's own library can't run here, and instead of guessing or reading as "clean," those sections say plainly that they weren't checked.