Problem-intent board

This is an internal prospecting screen for a Supabase rep. The problem was described in plain words — teams stitch a database, auth, storage, realtime and vector search together from separate vendors and burn weeks on plumbing before shipping the feature a customer asked for — and the board shows the people who described that same problem themselves in the last 30 days, never the people complaining about a named competitor. The Board route ranks 34 authors on ICP fit out of 50 and pain clarity out of 50, and opens each one into a drawer with the quoted opening line and the posts behind it. The Post feed shows every post as Trayo returned it with the matched author beside it, over a posts-per-week chart. The Companies route is the same people rolled up by account.

Trayo supplied everything factual. Seven trayo_search_posts calls over a 2026-09-08 to 2026-10-08 window returned 175 posts; 37 survived a read for first-person pain, dropping vendors selling a competing backend, tutorials and coursework. trayo_add_people imported the 34 authors and created or matched 32 accounts, filling in titles, photos and employers from the LinkedIn profiles Trayo holds. trayo_enrich_emails was run on the 20 highest-scoring people and returned 13 verified work addresses; 7 lookups found nothing and are shown as such rather than guessed. Every name, title, company, logo, avatar, post text and email address on the screen came back from the API.

Three things are worth saying plainly. The two scores and the opening lines are mine, not Trayo’s — the API supplies no fit score, so the rubric is stated on each person’s drawer and should be argued with. Trayo attached the employer from LinkedIn for five people whose post carried a different company or none at all, and for one of them (Andrei Nita) the email lookup came back on a third domain again; those rows are flagged in the UI rather than quietly reconciled. And the post index runs about ten days behind publication, so the last two weekly buckets are genuinely under-collected — they are hatched in the chart and called out twice, because the drop from 13 posts to 5 to 0 is a collection artefact, not a quiet market. Two early post searches failed with post_search_failed on long query strings and succeeded once shortened. Salesforce, HubSpot, sequence and email-send buttons are previews that toast what would have been sent; nothing is integrated.