Fast and Hard Code

In this blog post, Armin Ronacher reflects on the meme that 'programming is solved' and argues that LLMs have reduced the friction of learning new languages, making language choice less important. This has led to a trend where developers, even those unfamiliar with systems programming, are choosing 'hard languages' like Rust and Zig for performance-critical projects. He cites examples like Cloudflare's Artifacts service using Zig and Vercel's fx agent. Ronacher also notes that LLMs enable developers to work with previously gatekept technologies (e.g., DWARF, eBPF, custom crypto), potentially increasing the number of developers focused on fast and small software.