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

Software verification has historically been a bottleneck, reserved for high-stakes systems like aircraft avionics where a single error can be catastrophic. Today, that exclusive status is being challenged by artificial intelligence.

Anthropic announced that its Claude model, in collaboration with partner Theorem, has successfully generated formal proofs for a significant portion of the open-source codebase. Over the course of just a few weeks, the AI produced approximately 180 million lines of mathematical proof code. This effort covered 123 libraries from Google’s OSS-Fuzz security project, a massive undertaking that would have taken human specialists decades to complete.

The methodology shifts the focus from writing new code to validating existing software. Rather than creating new programs, Claude translated crash tests into mathematical problems and analyzed compiled machine code to produce extensive proofs. An independent verification tool called Rocq checks these proofs automatically. Anthropic describes the process as "fully autonomous," noting that it requires minimal human review, although the underlying formal foundations and tools were developed by human engineers at Theorem.

The immediate results have revealed vulnerabilities that conventional testing missed. During the verification process, Claude identified and fixed hundreds of bugs. Many of these issues only manifest under specific conditions, such as unusually large input files or particular access sequences, which often slip past standard testing suites. In one notable instance, while verifying a widely used PDF reader, the team discovered as many security bugs in a single week as are typically reported for that software in an entire year.

The AI’s capabilities extend beyond surface-level errors. In another application, Claude detected a logical error that had remained undetected for over a decade. Anthropic emphasizes that even when a complete proof fails, the attempt itself surfaces real bugs, making the process valuable regardless of the final outcome.

However, experts and the company note that "formally verified" does not equate to "error-free" in all real-world scenarios. The proofs are valid only under strict conditions, including a specific processor architecture, single-threaded programs, and inputs under 64 MB. Since these libraries often run on diverse hardware and configurations, the verification offers a high level of confidence but not a universal guarantee.

The computational cost remains a significant factor. Individual proofs can require hundreds of millions of tokens to generate. While Anthropic has not disclosed the total project costs, it suggests that if AI can verify its own code, developers can spend less time on manual review. This efficiency could reduce costs across the entire coding pipeline. Furthermore, the team noted that proofs are reusable and that newer models are becoming more efficient; Opus 5.5, for example, solves half the proofs with about 40 percent fewer tokens than its predecessor, Opus 5.

This development signals a potential leap in software automation. The industry is moving toward a future where AI not only writes code but also guarantees its correctness under defined conditions. Anthropic expects formal verification with AI to become as routine as standard testing in the near future. Theorem’s roadmap already targets critical infrastructure components such as glibc, OpenSSL, and cURL, suggesting that this technology is poised to move beyond experimental projects into foundational software systems.