Overview
&Truth is building a universal verification machine that reads code in any programming language and outputs mathematical proofs of safety, security, and correctness, with proofs that anyone can independently check. It uses the Lean proof system and open-source components like CSLib, combining AI with proof automation to analyze software, generate formal specifications, and produce machine-checked proofs through a single verification pipeline. The approach blends AI-assisted proof with established formal methods and emphasizes open-source foundations and a public-benefit orientation. Its goal is to enable provable software security and correctness at scale for real-world systems, helping improve cybersecurity and AI safety through independent, machine-checked proofs.
Work here?Claim your company
Simplify's take
What believers are saying
- November 12, 2026 &Atlas exposed numerous real-world defects, proving immediate security value.
- The company already verified 100,000 C lines, demonstrating execution beyond slideware.
- Anthropic's September and October 2026 Lean breakthroughs lowered verification costs, accelerating &Truth's roadmap.
What critics are saying
- &Truth still lacks a public product, paying customers, or disclosed revenue as of November 2026.
- Specification errors can create false confidence, and wrong requirements will ship wrong proofs.
- Open-source maintainer adoption can stall remediation; without enterprise wins, $70 million burns before traction.
What makes &Truth unique
- Lean-based proofs let third parties independently verify &Truth's results without trusting vendor claims.
- Clark Barrett, Ilya Sergey, and Chris Hsu combine theorem proving, verification, and commercialization expertise.
- &Atlas maps 100 million C lines across 3,000 codebases, turning verification into infrastructure.
Help us improve and share your feedback! Did you find this helpful?