Crosby trains its first model, finds scaling laws hold for legal work
graceisford · x · 2026-10-22
- Legal AI startup Crosby trained its first model: a verifier that predicts whether a lawyer will send an AI output to a client.
- Key finding: while quality in legal work is allegedly impossible to measure, more training data reliably improved results — scaling laws hold up for legal.
- The experiment was led over the summer by researcher Sharan; though a toy experiment, the company says it sets the stage for much more research.
Related event: Crosby Trains Legal AI Verifier to Predict Lawyer Acceptance(3 posts)→
More from Companies & People
- Palantir touts sovereign AI: fine-tuned Nemotron beats frontier models at ~1/20 cost — BrettKrieger12 · 2026-10-22
- Instagram's Systrom on ditching Burbn: solve problems people actually have — a16z · 2026-10-22
- VC Peeked at Unicorn Founders' Calendars: 40% Hiring, 50% Special Projects — letandrewcook · 2026-10-22
- Every major AI company now ships a personal agent except Google, argues signulll — signulll · 2026-10-22
- Grok Bot meetups have been held in 145 cities, 50 more booked — pswider · 2026-10-22
- Unlocking collective consciousness: using AI to scale your best people's judgment — jacob_posel · 2026-10-22