Legora CTO on building legal agents you can't see: evals, 100 legal engineers, drafting-as-search
Latent Space · rss · 2026-11-03
Latent Space interviews Legora CTO Jacob Lauritzen on building legal agents when client data is invisible. Key points:
- Legora lets lawyers draft Word docs, review 10,000 contracts in one table (Tabular Review, copied industry-wide), and run due diligence; it used its own product for its acquisition diligence.
- He estimates models already do 60% of legal work well; the rest needs smarter models.
- 3 engineers + 3 lawyers build RL-style test 'worlds' modeled on customer work; one release took citation accuracy from 70% to 99% without Legora changing anything.
- Why they resisted a model picker for years, why Opus/Fable feel too chatty for drafting, why everything runs on Azure, and why they skip ElevenLabs.
- The '300 Spartans' slide argued 20 engineers would suffice; they're now 100 and still shorthanded.
- Drafting long contracts is a search problem; advice against default embeddings; MapReduce vs one big context window.
- Will model labs cut off legal tools like they cut off coding tools? Plus the future of billable hours and junior lawyer jobs.
Legora is hiring in New York, Stockholm, and London.
More from coding & agent
- 10,000 AI agents burned his API credits; 500 agents in 50 waves worked — Beneficial-Cow-7408 · 2026-11-03
- Walrus MCP Server connects AI assistants to decentralized storage — modelcontextprotocol · 2026-11-03
- News Volume Trends MCP queries news volume growth for any topic — modelcontextprotocol · 2026-11-03
- Pen ships a canvas plugin for Codex, adding design powers to ChatGPT — paw_lean · 2026-11-03
- Elicit engineers all use Ramp's Inspect — a credit card company builds a top coding agent harness — jungofthewon · 2026-11-03
- VibeArmory launches macOS private beta: monitoring and firewalls for AI coding agents — agf2007 · 2026-11-03