Reco Rewrote JSONata in Go With AI in 7 Hours

Reco’s data engineering team used Cursor to port the JavaScript JSONata library to Go in seven hours, spending $400 in AI tokens and producing 13,000 lines of code with 1,778 passing tests. The rewrite eliminated expensive RPC calls between their Go policy engine and Node.js pods on Kubernetes, cutting $500K per year in compute – $300K from killing the RPC fleet alone. Simple field lookups got 1,000x faster, complex expressions 25-90x. They ran shadow mode for seven days with zero mismatches before promoting to production and even found bugs in the original reference implementation along the way. AI-assisted rewrites of well-tested libraries with comprehensive test suites are becoming a repeatable pattern, and the economics are getting hard for engineering managers to ignore.