Zef Hemel<p>I'm going to make make Rust my first AI-only language. I learned some <a href="https://hachyderm.io/tags/rust" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>rust</span></a> but have never built anything significant with it BY HAND. I intend to keep it that way, and apply CHOP (CHat Oriented Programming as coined by Steve Yegge) only with anything I produce in Rust.</p><p>We had a discussion today about the future of programming languages with LLMs becoming good.</p><p>Do we need "AI Native" languages and what would they look like? My intuition would be that Rust is actually is a good fit, for a few reasons:</p><p>1. Safety<br>2. Performance<br>3. Easy to read, but hard to write. And if the LLM does the writing, that's actually not a problem.<br>4. Very strongly typed, which I think benefits LLMs a lot because it gives itself a very quick feedback loop — if it compiles, it's probably correct.</p><p>I've spent the better part of the day with Claud Code writing a program that otherwise I would probably have hacked in TypeScript or Python. Instead, I had Claude Code generate Rust code and it went really well. The result feels very robust, and it is fast.</p><p>Although I don't know Rust intimately, I do assume general programming principles apply like in any other language. I can read the code the LLM writes and to challenge it with general engineering practices. On many occasions have I asked it to write tests, refactor the code, take a completely different implementation approach. I may not know the subtleties of Rust as a language, but I'm not sure that it matters.</p><p>And picking a language I can't easily write myself has proven to be a very effective strategy to fight the urge to step in and do it myself.</p>