YC Indexor
Search 6,241 Y Combinator companies by meaning, logo, visible text or tags, then have Jev score the shortlist.
Added to Jevfast
How it uses Jev
The app sends up to 320 company finalists to Jev after local retrieval combines text, logo-image, OCR, and tag signals. Jev judges each candidate against the query; candidate questions are chunked at 120 per call and the chunks run in parallel. A separate data-preparation script reapplies YC tags to company records.
What Jev decides
Illustrative Jev requestExample answers · not a recorded Jev response · Source ↗
Question 1 · c0
YOUR APP
INSTRUCTION
Does this candidate fit what looking_for describes?
STATE
A person’s query, one candidate’s written company information and measured logo colors; optionally today and newest-first YC batches.
JEV · NOUL
YesNo
App workflow
- Embed the search
Encode the query against company descriptions and logos; use OCR text and tags as additional retrieval signals.
- Narrow the candidate set
Combine the signals to select at most 320 of the 6,241 company records.
- Score finalists
Split shortlisted companies into batches of up to 120 candidates and call Jev for each batch in parallel.
- Show ranked matches
Render companies and their probabilities in the interactive pile; without a key, use the local fallback scorer.
Why it is interesting
Multiple local retrieval channels narrow a large catalog before Jev evaluates a bounded shortlist. The result pile exposes probabilities, and the code retains a degraded local ranking when a required Jev call does not answer.