A Couple Million Lines of Haskell: Production Engineering at Mercury

Mercury’s engineering team reveals they run a production bank with a couple million lines of Haskell, commanding 170 points and 66 comments — a demonstration of functional programming's viability in high-stakes finance. The codebase is 30% lighter than the runner-up consumer fintech stack in memory consumption, thanks to Haskell's strict type system. One concrete benchmark shows a 40% reduction in runtime errors after migration from a Scala-based predecessor, with compile-time checks catching 95% of null-pointer-like defects. The post sparks fierce debate: while pure functional purity reduces production bugs, its steep learning curve costs new hires an average of 3 weeks ramp-up, slower than the typical rival language like Rust, but the payoff is fewer outages over 18 months of operation.