Niko Matsakis on why Rust fits agent coding, and what to change

Rust language designer Niko Matsakis argues Rust is a strong fit for agentic development because it is fast by default, exposes performance knobs, and its type system gives agents stronger guardrails than tests alone. He wants Rust to double down on "if it compiles, it works," make proof tools like Verus easier to integrate, and aim for a north star where unsafe code can be proven free of undefined behavior. He also says design and specification remain the open problem, with markdown-driven agent workflows still weak. On Hacker News (70 points, 58 comments), there is no consensus. The biggest camp says compile times hurt agent loops that rebuild often; others say that is fine if you tune the harness so agents do not rebuild constantly, or you put a machine-wide cargo lock across parallel agents. Some prefer Go for faster iteration. Another camp says type-heavy languages help agents, including Haskell anecdotes. Skeptics say any language can be described in a prompt, or that generating source code is only a temporary step. A smaller group rejects agentic programming as "slop." Sorted take: if your team already ships Rust, invest in agent harness rules around builds and proofs rather than switching languages on vibes.