Learnings from 100K lines of Rust with AI (2025)

A developer used AI coding agents to build a modern, production-grade Rust implementation of Azure's multi-Paxos consensus engine in approximately three months, generating over 100K lines of code while achieving a 13x performance improvement (23K to 300K operations per second). The project demonstrates significant productivity gains through AI-assisted development, with key techniques including extensive testing (1,300+ tests), AI-driven code contracts for correctness verification, and spec-driven development practices. The modernized system addresses limitations in the decade-old original RSL by adding support for pipelining, non-volatile memory, and RDMA hardware optimization.