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

Hacker News users discussed an AI-assisted Rust workflow where one model writes a spec, another critiques it, and both may help implement and review the code.

The thread debated whether starting a fresh model context improves reviews, or whether it just repeats the same blind spots.

Commenters questioned how well LLMs can handle Rust lifetimes, cloning, and idiomatic architecture in large codebases.

The discussion highlighted both the promise of multi-pass AI coding and the current limits of using LLMs for maintainable Rust at scale.