Overview

Reasonable AI builds AI systems for formal verification of mission-critical software. It combines language models with formal methods to generate, repair, and verify machine-checkable proofs alongside production code, using Verus. It aims to help engineers overcome the verification bottleneck by delivering proofs that scale with AI-assisted code, while analyzing failure modes to prevent cheating. Its approach emphasizes small, fast models that can approach larger systems' proof quality, enabling verifiable correctness for each line of code.

Launched Recently