The LegalQuants x Replit Hackathon (LQ002) results are in
Attest was a co-winner!
Forty in-house counsels built working software in a hackathon. Not prototypes that demo well and die — apps with real backends, real UIs, real problem-solving behind them. And the judging criteria weren't "prettiest demo." They were: how did you think through the problem? Did your prompt history show genuine architectural reasoning?
I tied for first with Attest — a governance protocol that seals agentic AI decisions into hash-chained certificates of action. The premise: if your agent just made a consequential call, you need more than logs. You need a tamper-evident record that proves what policy governed the decision, whether the agent exercised diligence, and whether a qualified human was in the loop.
The other two winners built equally sharp tools. Sergey Kinchin's ClauseBench gives you market comparators and negotiation-ready rewrites from a single pasted clause. Adam Masser's Contract Cortex layers AI classification and semantic search on top of Dropbox without moving files. Both solve real workflow problems that in-house teams hit daily.
What stood out about LegalQuants' approach: they scored thought process, not polish. The judges — Anna Guo and Michael T. Brown (who won the Global Claude Code Hackathon) — read our Replit prompt histories to evaluate how we reasoned through build decisions. That's a better signal of legal-technical thinking than any finished product alone.
The gap between first and runner-up was half a point. Five more teams were within 1.5 points of that. This cohort was competitive in a way that says something about where in-house counsel is heading — building, not just buying.