Harvey's 'Moneyball' playbook: world-class AI research without a frontier-lab budget
FinanceYF5 · x · 2026-08-22
Legal AI firm Harvey shared its "Moneyball" approach to building world-class AI without a frontier-lab budget: have domain experts guide synthetic data generation and build legal eval sets; partner with multiple emerging labs for internal post-training; and use model routing, automatic fallback, and SLAs to serve 60 countries reliably. The core idea isn't more money — it's spending a limited budget on the few things that matter most.
More from Companies & People
- Databricks switches to GLM, achieving 92% pass rate at 1/4th cost — philfung · 2026-08-22
- 50k Users, Tiny Revenue: Founder Reveals Harsh Conversion Reality — emmanuelomn · 2026-08-22
- DHH Launches $8M Omacom Foundation Backed by Lütke, Collison, Dell for Linux Desktop — JeremyCMorgan · 2026-08-22
- Former Meta AI Engineer Harper Carroll Launches Free 'AI Basics' Series — HarperSCarroll · 2026-08-22
- Breadcrumbs suggest OpenAI is building the entire layer around its models — imjustnewatai · 2026-08-22
- Discussion on academic evaluation reform citing Goodhart's Law — MaxUnfried · 2026-08-22