I'm sorry, but you still have to think
Points and comments are a snapshot, not live.
AI agents cannot yet replace human judgment in code rewrites, as shown by a Rust port's flaws.
Piotr Sarnacki critiques DHH's AI-generated rewrites of Campfire Once from Rails into Rust, Elixir, and Go. The Rust version drops CSRF tokens and Redis for caching, uses non-async database operations and sync locks in async tasks, and has a 1% notification delivery rate under load due to a 256-capacity broadcast channel. Increasing capacity to 16384 raises delivery to ~90% but degrades latency to seconds. The Elixir version maintains 100% backwards compatibility but uses 1.8GB memory under load and times out on 23% of HTTP POST requests at 100 req/s. The author concludes that programming still requires understanding trade-offs and context, not just metrics.
What commenters are saying
Commenters largely agree that thinking remains essential, with some lamenting workplace pressure to use AI uncritically. A few defend AI's utility for porting tasks when paired with proper testing and expert oversight. The term 'clanker' is debated as either a useful descriptor or a signaling term that undermines serious discussion. One commenter notes that AI can identify and fix issues like broadcast channel capacity if told about them, but the author's point stands that trade-offs require domain expertise.