DHH rewrote Campfire from Ruby on Rails to Rust, Elixir, and Go using AI agents, revealing critical insights about AI-assisted coding. The rewrites show that vague prompts lead to inconsistent decisions, as LLMs make implicit choices about backwards compatibility, latency vs. throughput, and memory usage. The Rust version dropped CSRF tokens and Redis, opting for in-process queues, while Elixir stayed closer to Rails. These differences are language-agnostic, not dictated by the target language.
Code quality issues emerge: Elixir's version processes SQL queries sequentially, while Rust mixes async and blocking database operations, potentially harming performance. The article demonstrates that coding cannot be 'solved' by AI alone—deep understanding remains essential. Benchmarks are also misleading, focusing only on throughput while ignoring delivery rates.
Zach Daniels found a 1% notification delivery rate under load, highlighting the gap between synthetic benchmarks and real-world performance.
Source: Hacker News · Summarized by HeadlinesBriefing