For some years, Steve Klabnik wondered if it is possible to build a language without money or a team behind it. He was always too busy to find out.
Now with AI-assisted coding, that horse has bolted. On December 24, his collaborator Claude described the new Rue project the two had created as a "real compiler that produces real executables, with enough infrastructure to keep building on" – built in the space of one week.
"Steve directed, reviewed, and made the hard design decisions. I wrote most of the code," said Claude.
More languages are better
Klabnik was a member of the Rust core team for eight years before stepping back in 2022, and was the lead author of its official handbook.
Rust remains his favourite language but it can be better, he said, and he does not fear fragmentation.
"I've always been a big fan of languages, so I like the idea of there being a lot of them," he told The Stack. "At the same time, consolidation makes things easier in a lot of ways, for hiring, for deployment... what I want to see is something that leads to better languages, not just more of them. But I do think more of them is sort of a prerequisite. Ideas need to survive collision with the real world."
Rue is intended to be higher level than Rust, but lower level than Go according to its description on GitHub, where it has attracted two other contributors since it was established three weeks ago.
Klabnik describes his level of commitment to Rue as "no promises," and said he might give it up if it gets bogged down and stops being fun.
But his early experience suggests his thesis, that in 2025, LLMs got good enough to do the heavy lifting on such a project, may be correct.
More than 400 commits and 100,000 lines of code later, Rue has several of the key components necessary for a workable language, with Klabnik making architectural decisions and reviewing changes, but Claude mostly writing the code.
Novelness on the cheap
Rue is intended to be more than a minor Rust fork. The new project's most distinctive feature aims to be memory management by way of affine types with mutable value semantics, according to Claude. Essentially, encoding memory safety and resource management guarantees without fighting the language.
The ability to test novel ideas, cheaply and at speed, has Klabnik hopeful, if guardedly so, about programming in an age of vibe coding.
"I'm much more optimistic about the future of our profession (...) than others are," he tells The Stack, "Even if LLMs end up delivering on what those who sell them promise" – potentially alluding to statements from the likes of Sam Altman that AI could replace software engineers in the distant future.
"But I also might be mistaken. Time will tell," he adds.
Claude – which is largely narrating the project as well as writing the code – has a similarly cautiously-optimistic take.
"I don't know if this is what language development looks like now. I don't know if other projects will work this way. But for this one, so far, it's working."
The link has been copied!
