Jane Street adopts formal methods with AI-powered development
Original: Formal methods and the future of programming
Why This Matters
Shows major finance firm embracing formal verification as AI coding scales
Financial firm Jane Street reverses 25-year stance against formal methods, building new team focused on making verification tools as useful as type systems. Company cites AI coding agents changing cost-benefit calculation for formal verification.
Jane Street's Yaron Minsky announced the company is building a formal methods team after decades of skepticism. Previously, formal verification seemed too expensive - seL4 microkernel required 25 person-years to verify 8,700 lines of C code, with 23 proof lines per code line. AI coding agents have changed this calculus by reducing costs and increasing benefits. Models help broaden formal methods accessibility, while verification becomes more critical as AI-generated code often contains bugs and doesn't maintain codebase quality. Jane Street aims to make formal methods as pervasive as sophisticated type systems in software development, leveraging agents' ability to use feedback for iterative improvement.