Prediction: AI will make formal verification go mainstream — Martin Kleppmann’s blog (martin.kleppmann.com)

A recent prediction by Martin Kleppmann suggests that advances in AI will push formal verification—a process that mathematically proves the correctness of software—into the mainstream of software engineering. Traditionally viewed as a specialized and labor-intensive endeavor, formal verification relies on proof assistants and programming languages that can articulate and verify code specifications. Currently, these methods see limited use outside academia due to the expertise and substantial effort required, exemplified by the arduous verification of systems like the seL4 microkernel. However, with the rise of large language models (LLMs) capable of generating both code and proof scripts, the landscape is set to change dramatically. Kleppmann argues that as these AI tools improve, they will not only reduce the costs associated with formal verification but also create an increased need for verification due to the integration of AI in code generation. This could lead to a new era where developers can specify desired properties of software in a high-level language while the AI handles the implementation and proof processes. Ultimately, the move toward automated verification will likely challenge existing cultural perceptions of software development, making formal methods more accessible and practical.