Jane Street, a Wall Street trading firm, has shifted its stance on formal methods after 25 years of skepticism. The change follows growing evidence that agentic coding can lower the cost of rigorous verification. By integrating formal techniques into its OCaml‑based workflow, the firm aims to match the reliability gains that type systems already deliver for trading executions daily operations.

Previous experiments, like the formally verified seL4 microkernel, highlighted steep overheads: 25 person‑years to certify 8,700 lines of C, with each line demanding 23 proof lines and half a day. Jane Street argues that agentic models can reduce this burden by offering reusable proof skeletons and automated feedback, turning formal methods into a more scalable tool for critical codebases today.

With control over its language, Jane Street plans to embed modular specifications and type‑level constraints into OxCaml, leveraging tools like Lean and Dafny. The firm’s developer community, accustomed to rapid type‑system adoption, will test these innovations first. By marrying agentic coding with formal guarantees, the team aspires to deliver safer, faster software for high‑stakes trading environments and in real time.