signalGitHub Trending2026-09-21
cactus-compute/needle
Needle is a 2-bit automation foundation model compressed into an 8-29 MB binary, built on a Laddered Simple Attention Network (Monarch Hadamard MLP, GQA with causal conv, engram n-gram memory) that beats models 10x its size on mobile tool calls and matches 2-3x larger models on structured extraction, while the 121M model does the arithmetic of a 50M one. It handles tool calling with exact-match accuracy, typed JSON extraction via grammar-constrained decoding, and text embeddings, with calibrated confidence scores per response - deployable via `pip install cactus-needle` on phones, wearables, robots, cars, and microcontrollers.
- why now
- New on-device AI model released - grab it for edge apps.
- topic
- AI Tech & New Models
- source
- GitHub Trending
score
score 8 out of 10. 0-10: how dense the facts are, multiplied by how much you can do with them after reading. 8+ means the topic's evidence bar is met: benchmarks and availability for a new model, amount and investors for a funding round, revenue figures for a solo-money story. Below 5 an item does not enter the digest. A press release scores 3 or less, a reprint loses 2, anything older than 14 days loses 1, a headline that misleads loses 3.
read the source