Rust

We made a bet early in the winter to rewrite our backend into Axum and Rust.

The bet is paying off now.

No memory leaks, extremely fast cold start times, leverage native crypto libraries + Alloy, sqlx has compile-time checks, AI fixes its own mistakes, and it’s entirely in a binary — no dependencies!

So AI code editors handle Rust and Axum code well?

Barista in the morning, AI engineer by day, and parent (with @shiwen) by night ||
Host at /clickcast /cpg

Operating System

The strict typing actually helps AI more than it hurts, since type errors give clear guidance. But the borrow checker’s nuanced feedback gets lost on models that just want something that works.​​​​​​​​​​​​​​​​

AI will bias towards solving the compilation problems than efficiency problems, which Rust excels at.

We’re writing web code so it’s fine to not be so efficient yet.

Is there anything counterintuitive about the way you setup prompts/context/codebase to make it work?

Yes with the right prompts, context, and codebase memory.

We are experimenting with long running coding agents to help us with autonomous lifts, e.g. major refactors to files, functions, code hygiene, upgrades.