The Pragmatic Engineer

Sean Allen on programming languages, testing and software correctness

Key points

Key takeaways from The Pragmatic Engineer podcast with Sean Allen, Distinguished Engineer at Antithesis — a conversation that never says “AI” even though it is about the hardest parts of developing with AI (November 2026):

  • Language roots.: Allen traces today’s reliability culture through the ML family: OCaml, the functional language popular at Jane Street largely without its object features; Ada, the defence-driven language pitched as “you like Rust, how would you like some more guarantees?”; F#, Don Syme’s OCaml for.NET, born partly from the Java-Sun lawsuit that pushed Microsoft onto its own runtime; and Pony, the open-source language Allen works on, built for fast, safe concurrency.
  • The compiler as test suite.: Sum types, pattern matching and exhaustive matching push validation into the compiler: change one variant of a type and the compiler finds every spot that must handle it. Allen, who spent years developing C discipline to avoid crashes, says he would rather have a compiler help him than rely on human rigour.
  • Tests are a means to an end.: His essay argues tests are “a means to an end; the means are unimportant”, that no tool can guarantee correctness and that teams inevitably find themselves testing in production; technical depth only hurts when it has to be paid back, and debt taken before product-market fit can still be valid.
  • Red-team perspective.: On Antithesis’s Red Team, Allen deliberately uses the product only as customers would — with no internal shortcuts — and works on agent skills, teaching and the evidence that lets teams trust AI-written code, as quality shifts from a gate to continuous confidence.

More top stories