Building the right thing

the research assistant can cut and revise highlight reels in chat and search transcripts for quotes on any topic; study analysis reports are now available on every account. Improved: quote citations seek the recording to the cited moment; long searches no longer trigger a false stall notice. Fixed: partial responses kept out of analysis surfaces, skips recorded as real skips, required screen share restored mid-session, recordings no longer clipped at the start, multi-segment recordings joined in full, screen-share sessions play as video, failed reels marked failed, AI-moderated highlights add to reels, recording waits for consent, chat scroll steadied, and spoken 'continue' requests advance the session.

integrations can subscribe to a participation.booked webhook and read the moderator from the participant API; a combined recruitment and incentive spend report exports per-study fees, incentives spent, and unspent allocations in one CSV. Improvements: large participation CSV exports complete reliably; Google Meet and Microsoft Teams links are created before calendar invites go out; the AI moderated participant table shows live recording status. Fixes: bulk invite batches keep custom messages, duplicate participations from self-scheduling, recruitment credit refunds on study close, Edit booking on externally recruited sessions, transcript deep-links, public recording playback, livestream rescheduling, highlight reel processing, recording re-imports, manual interview locations, candidate ordering, and team assignment on new studies.

the study workspace adds dedicated Setup, Results, and Report tabs with a shareable report URL; report highlight reels can now stitch multiple clips into a single video; the researcher chat can read your study report. Improved: reports place charts and reels inline beside the findings they support, draw on your research plan, and are easier to scan with scaled headings, numbered citations, and an evidence carousel; AI-moderated sessions resume after a dropped connection; the sidebar remembers its collapsed state. Fixed: sub-steps can be removed from the workflow outline; moderation waits for prototypes to load; device-setup speech is kept out of transcripts; background moderation hiccups no longer mark sessions as no-shows; voice defaults off outside AI-moderated studies; report evidence slideouts focus the clicked theme; new chats show their title without a refresh; teammates' studies no longer replay the research goal as a first chat message.

GET /v1/studies/:id/participants now returns a moderator object (user_id, name, email) for each participant, so API and MCP consumers can see who is running a session without a second lookup. The field is nullable and returns null when no moderator is assigned. The user_id matches the id returned by GET /v1/interviews/:id/moderators, so you can join the two endpoints directly.