Jane Street runs almost all of its trading systems in OCaml, a functional language from 1996 that most firms abandoned long ago. They bet on strong static typing to catch errors at compile time—no messy null‐pointer crashes in production. Every missing‐value scenario must be handled via the Option type. That discipline means one less bug slipping into a system where a single mistake can cost millions. Ron Minsky from Jane Street points out that the compiler forces you into exhaustive case analysis. When code compiles, you know it’s checked against a wide range of potential pitfalls. They’ve built so much OCaml-based infrastructure that, as John Crepezzi puts it, there may be more OCaml inside Jane Street than outside in the entire world. On the AI side, they ditched Copilot and Claude. Instead, they snapshot developer workspaces every 20 seconds. When a build turns green, that exact edit—bug fix and all—feeds into their in-house model training. Reinforcement learning then rewards only the changes that compile and pass tests. In 16 minutes, an engineer outlines how those snapshots power a system that supports roughly $40 billion in annual trading volume.