Claude generates 180 million lines of proof to check 123 code libraries

Developers of legacy software now have a faster way to confirm their code works as intended. Anthropic announced that its Claude model generated mathematical proofs to verify existing open-source libraries. The process found bugs that standard testing methods often miss.

Formal verification has historically required years of expert labor. It took specialists 20 person-years to verify seL4, a small operating system kernel. This cost kept the technique limited to high-stakes environments like Airbus avionics. Anthropic partnered with Theorem to automate this work using Claude.

In a matter of weeks, the AI model produced approximately 180 million lines of proof code. It analyzed crash tests from 123 open-source libraries within Google's OSS-Fuzz project. The model did not write new software. Instead, it converted crash tests into mathematical problems and analyzed compiled machine code to construct proofs. A separate program called Rocq automatically checks these proofs, requiring minimal human review.

The autonomous process uncovered hundreds of bugs. Researchers noted that many of these errors appear only under specific conditions, such as unusually large input files. While verifying a widely used PDF reader, the team found as many security bugs in one week as are typically reported for that software in an entire year. In another case, Claude identified a logical error that had gone unnoticed for more than a decade.

Anthropic emphasizes that formal verification does not guarantee software is error-free. The proofs hold only under defined conditions, including a specific processor architecture, single-threaded programs, and inputs smaller than 64 MB. Real-world deployments often use different hardware and configurations, which limits the scope of these guarantees.

Cost remains a factor in the process. Individual proofs consume hundreds of millions of tokens, and Anthropic has not disclosed total project costs. However, efficiency is improving. Opus 5.5 solves half the proofs using about 40 percent fewer tokens than its predecessor, Opus 5. The company argues that this efficiency will allow developers to trust AI-generated software more and spend less time on manual review.

Theorem's roadmap targets critical infrastructure libraries, including glibc, OpenSSL, and cURL. Anthropic expects formal verification with AI to become as practical as standard testing. The shift moves the focus from whether AI can program to whether it can prove its code is safe.