When AI Writes the Code, Who Proves It Works?

Formal verification is transitioning from a niche, labor-intensive practice to a scalable standard, driven by the emergence of auto-formalization. AI models now bridge the gap between natural language requirements and machine-checked proofs, allowing developers to verify software properties without manually coding every proof step. Lean, a programming language and proof assistant created by Leo de Moura, serves as the core infrastructure for this shift, enabling the creation of a "verified stack" for critical systems. While traditional testing identifies the presence of bugs, formal verification provides mathematical certainty of their absence. This evolution fundamentally alters software engineering and mathematics, shifting the focus toward precise requirements definition and the maintenance of formal models. As AI continues to automate proof generation, the value of human expertise shifts from manual coding to architectural design and the rigorous specification of system properties.