Reasonable AI

Reasonable is building a verification layer for mission-critical software development. Its work covers tooling, model training and agentic frameworks intended to make formal correctness guarantees accessible to software engineers. The team has fine-tuned NVIDIA Nemotron 3.5 Lightning to write machine-checkable proofs in Verus, as described in an August 2026 blog post.

The website describes the company as an applied AI research team and names no specific customers. It does not describe pricing or revenue model. The stated target users are software engineers building production systems.

The About page lists three founders: Ferenc Huszár (technical founder, Professor of Machine Learning at Cambridge), Craig Hughes (commercial and operations) and Anna Török-Böröcz (operations and people). Point Nine is listed as a backer. Open roles include Member of Technical Staff positions in San Francisco, London and Budapest, Forward Deployed Engineer roles in San Francisco and London, and an Executive Assistant role in Budapest. The founding year is not stated.

Product: Formal verification tooling, Verus proof-writing models and agentic frameworks for machine-checked software correctness.