Grace Hopper's Revenge
Summary
Explores how language design and verification influence AI-assisted software development, arguing that English-like specifications and functional programming improve auditability over traditional OO approaches. The piece reviews benchmarks (SWEBench, TerminalBench, AutoCodeBench) and finds Elixir, Kotlin, and Racket among top languages, asserting that locality, immutability, and explicit semantics aid verification. It predicts a future where LLMs write code while humans specify, audit, and verify, emphasizing clear contracts and verifiable logic.