What is the YC Indexor app?

Describe a YC startup in any words and the matching logos float up out of a physics pile, each with how likely it is. All 6,241 companies are searchable by what they do, what their logo looks like, and what it says inside.

How does Jev factor in?

YC's own tagging is patchy, so Jev re-applies the 337 YC tags to every company, then scores up to 320 retrieval finalists in one parallel request and returns calibrated probabilities, which is why the on-screen percentages are real.

Why does it need an Apple Silicon Mac?

Every search embeds the query with MobileCLIP-S0, which is a Core ML model, and native/coreml_embed is a compiled arm64 binary; on Linux or Windows the first query throws. The README notes an ONNX export matches Core ML to a cosine of 0.999914 if you want to fix that.

Do I need an API key?

A TypeSafe key in .env.local as TYPE_SAFE_KEY is recommended. Without it the app starts and the pile still works, but Jev is never asked, ranking falls back to a local scorer, and results get noticeably worse with degraded: true and costUsd: null in the response.

How fast and how expensive is a search?

About one second per search and about a fifth of a cent, after a first start of roughly sixteen seconds while the sentence model loads. No hosted version is documented, so you run it locally with npm run dev on port 3000.

How do I add or rebuild company data?

Everything the app reads is committed, so you never need to. tools/yc-data holds the scripts that built data/companies.json and public/icons/ in the first place, with a README giving the order to run them in.

What about the optional motion feature?

npm run build:motion compiles a helper that reads the MacBook accelerometer and lid angle from HID devices, so tilting the laptop tilts gravity in the pile and shaking it throws the icons; without the helper the pile keeps ordinary gravity.

Jev directory

CommunityJev Demos & Experiments

YC Indexor

A local app that turns a worded query into a physics pile of floating YC logos: retrieval over 6,241 companies by meaning, logo image, OCR'd logo text, and Jev re-applied tags, with Jev scoring the finalists into calibrated probabilities.

See highlights Open github.com

Stars: 66
Language: HTML
License: MIT
Reading time: 3 min
Published by: Community
Author: aayan-dev
Last verified: Oct 2026

aayan-dev/aayans-yc-indexor

User repository on GitHub

View on GitHub

Describe a YC startup in any words and the matching logos float up out of a physics pile, each with how likely it is. 6,241 companies, searchable by what they do, what their logo looks like, and what it says inside.

Forks: 9
Last pushed: Updated Sep 2026

Repo stats from the GitHub API, cached Oct 2026.

Tagscommunity catalog-import madewithjev search logos apple-silicon

Works forJev for Ranking Jev for Classification

Used inJev for Software & DevTools

Languages: JavaScript
Platforms: macOS
Capabilities: Open sourceRuns locally

Highlights

Hard requirement: an Apple Silicon Mac, because every search embeds the query with MobileCLIP-S0 through a compiled arm64 Core ML binary.
Four retrieval signals: bge-small meaning vectors, MobileCLIP logo image vectors, Apple Vision OCR of logo text, and YC's 337 tags re-applied by Jev.
Retrieval narrows 6,241 companies to at most 320 finalists; Jev scores them all in one parallel request and returns calibrated probabilities.
First start takes about sixteen seconds while the sentence model loads; after that a search is roughly one second and costs about a fifth of a cent.
Without a TypeSafe key the app still runs, Jev is never asked, ranking falls back to a local scorer, and the response shows degraded: true.

Quickstart

bash

git clone https://github.com/Aayan-DEV/aayans-yc-indexor && cd aayans-yc-indexor
npm install
cp .env.example .env.local
npm run dev

Watch out

README reviewed 2026-10-15. The speed and cost figures and the ONNX/Core ML cosine match of 0.999914 are the author's own measurements in NOTES.md, and the repo notes that logos, YC data, and Apple models bundled with it are not its to license even though the code is MIT.

Questions about this entry

Short answers drawn from the linked project or page.

Reactions & coverage

Posts, threads, and videos about this entry from around the web.

X: LinkedIn bookmarks search

X: 1,066 reaction videos, searchable

X: Intent-based search in a Chrome extension

X: Margin, bookmark search

From the Jev guides

Original write-ups on what this entry is for.

Per-Tick Decisions in Games

Sep 28, 2026

A game tick is a state plus a legal action list. How Clash Royale, Minecraft, and the earlier Mario, Pokémon, StarCraft, and drone loops ask Jev one step at a time.

Jev Demos & Experiments#gaming#agents#control

Read article

From the Jev AI model community

Posts from builders shipping with Jev right now.

Follow @typesafeai