Anonymous users get 3 live AI researches per rolling 24 hours. Cached topics — including trending chips — do not count.

That split matters: we want guests to feel speed on warm topics while capping OpenAI spend on cold ones.

What counts as “live”

Only a cache miss that triggers worker inference burns quota. Hit v2:research:{hash}? Instant, free, unlimited re-reads.

Identity

Next.js forwards an anonymous browser id (a UUID in localStorage) to the backend. Signed-in users get a higher tier, and that identity must come from a verified server-side session, never from a header the browser can set (ADR-005).

Storage

usage_limits.db on the Fly volume — separate from research.db so quota writes do not fight cache reads.

UX when blocked

429 → friendly copy on result page; nudge to try a trending topic or sign in later.

Same playbook as Invest AI’s free tier post — adapted for learning decisions instead of stock analysis.

Further reading: ADR-005 in xingai-research-ai/docs/adr/