Until now research lived in its own tab. Now it feeds the product. A Discover pass finds the accounts winning your niche, reads the hooks burned into their frames, reads the comments under them, and hands your app a voice pack: per hook family, the real winning hooks and what that kind of hook provokes (who asks for the app, who tags a friend, who pushes back), with real example comments; per niche, the phrases, questions and objections people repeat.
In every drop, research candidates compete against our bank of 500+ campaigns hook by hook, and a judge keeps the best. You set the share in Settings: 0% is bank only, 35% is the recommended default, 100% uses only your research. Each hook shows where it came from, and the comment generator writes in your audience's words instead of generic TikTok speak. Research is optional; a pass every 3-7 days keeps it on this week's winners, and the companion tells you which pass to run.
- Collections are now durable research: continue the same niche tomorrow, with a report that gets a new version every pass and shows what changed
- Comments read under the winners' posts (top-ranked, ~1,800 per pass), with seeded engagement flagged
- Research share slider per app (0-100%) and a 'research' badge on every hook that came from it
- New in the companion (0.6): research_menu, list_collections, open_collection, enrich_comments, audience_brief, voice_pack, versioned reports
- New guide inside the app: how to run research, with the exact prompt for Claude Code