Claude Generates 180 Million Lines of Mathematical Proof to Verify Open-Source Code

Anthropic’s Claude model has successfully generated approximately 180 million lines of mathematical proof code to verify the correctness of 123 open-source libraries. This achievement, completed in a matter of weeks, marks a significant shift in how software reliability is established, moving beyond traditional testing methods to rigorous mathematical validation.

Formal verification, a process that mathematically proves software runs without errors under all circumstances, has historically been prohibitively expensive. For context, a specialized team required 20 person-years to verify the seL4 operating system kernel, limiting the practice to critical systems like avionics. By partnering with Theorem, Anthropic’s Frontier Red Team enabled Claude to automate this labor-intensive task. The model did not write new software; instead, it translated crash tests from Google’s OSS-Fuzz project into mathematical problems, analyzed the compiled machine code, and produced the extensive proofs. An independent tool named Rocq automatically checks these proofs, a process Anthropic describes as nearly free of human review.

The autonomous nature of the system allowed Claude to identify and fix hundreds of bugs that conventional testing had missed. Many of these defects only manifested under specific, rare conditions, such as unusually large input files or precise sequences of data access. In one notable instance, the team verified a widely used PDF reader and discovered as many security vulnerabilities in a single week as are typically reported for that software in an entire year. Additionally, Claude caught a logical error in another application that had remained undetected for more than a decade.

Despite these successes, the project has strict practical limitations. Anthropic clarifies that “formally verified” does not equate to completely error-free. The proofs are valid only under specific constraints, including a particular processor architecture, single-threaded execution, and input sizes under 64 MB. Since these libraries often run on diverse hardware and configurations, the verification does not cover every possible real-world scenario.

The computational cost remains a major hurdle, with individual proofs consuming hundreds of millions of tokens. However, efficiency is improving; newer models like Opus 5.5 solve half the proofs with roughly 40 percent fewer tokens than its predecessor. Anthropic suggests that if AI can verify its own code, developers can reduce the time spent on manual review, lowering costs across the entire development lifecycle.

The initiative suggests that software automation is evolving from simple code generation to guaranteed correctness. Theorem’s roadmap now targets critical infrastructure components such as glibc, OpenSSL, and cURL. Anthropic expects this approach to become as routine as standard testing, fundamentally changing how the industry trusts AI-generated software.