KIT lets AI coding agents formally verify code as they generate it
LingmingZhang · x · 2026-10-22
Grigore Rosu's research lab released KIT, a toolkit that lets AI coding agents run formal verification automatically as they generate code, with every verdict backed by the mathematical ground truth of the programming language. It works with Python and Solidity/EVM today, with Rust, C and JavaScript on the way. Install is two commands as a Claude Code/Codex plugin; KIT extracts intent into a formal spec (K framework), proves it with a Prover, audits the proof, and ranks verified concerns P0–P3. The team plans to run it across open-source systems to find real bugs and is soliciting target libraries.
More from coding & agent
- Grok Bot Lands in X DMs: An Autonomous Assistant That Connects to Your Apps — mark_k · 2026-10-22
- 5 ways AI makes engineering leaders more effective — beyond writing code — rseroter · 2026-10-22
- ToM2Vec: continuous theory-of-mind vectors help LLM agents coordinate with unseen partners — EliasEskin · 2026-10-22
- LiDAR scan + LLM workflow produces dimensionally accurate 3D room models — OpenAIDevs · 2026-10-22
- Working with agents: give them principles, not rules — _AustinCalvert_ · 2026-10-22
- Quant-Arena: 9 LLM agents trade real A-shares, only 4 profit and none beat equal-weight — mhdfaran · 2026-10-22