Researchers used the Claude AI model to produce 180 million lines of formal mathematical proofs across 123 open-source software libraries. This breakthrough automates software verification at the machine code level, eliminating vulnerabilities and finding bugs that traditional testing missed.