GitHub · by Aayan

YC Indexor

Search 6,241 YC companies in about a second, by meaning, look or logo text.

Aayan

@aaayandev

Jev (@typesafeai) is so insane & cheap for search!! > 6000+ @ycombinator Startups indexed. > Sub 1 second search results. > 90M tokens & $2.7 in total testing costs. Search any startup in a second, in any way! - Color - Niche - Your Competitor - Age - Image - etc... > watch the entire video, it's so freaking cool omg! > this is the coolest thing i have ever built for fun! (worked on it for 2 days straight!)

Sep 21, 2026 · 1.4K likesOpen on X

Aayan indexed 6,241 Y Combinator companies so you can search them in any words: by colour, niche, competitor, age or image. Jev ranks the shortlist; bge-small handles meaning, MobileCLIP-S0 and Apple Vision OCR handle what logos look like and say, and Jev re-applied YC’s 337 tags to every company. He put the testing at 90 million tokens and $2.70, with searches landing in about a second.

Retrieval is four signals at once over the whole index. Meaning comes from bge-small sentence embeddings of each company’s write-up, looks from MobileCLIP-S0 image vectors of the logo, letters from what Apple’s Vision OCR read inside it, and tags from YC’s own 337 labels, re-applied by Jev because YC’s tagging is patchy. That narrows 6,241 down to at most 320 finalists. Jev then scores every finalist in one parallel request and returns a calibrated probability, which is why the percentages on screen are real.

A search after the models are warm is roughly one second and about a fifth of a cent. Apple Silicon is a hard requirement: every query embeds with MobileCLIP-S0 through Core ML. Without a TypeSafe key the pile still works, but ranking falls back to a local scorer and the results get worse.