Bend 2 and the AI-written compiler controversy
Bend 2.0 launched as a language designed to prevent AI coding mistakes through formal proof, which is a compelling premise. The problem: the README openly states the compiler is 99% AI-written and has not been fully audited. A companion thread called 'Bend 2 and the Vibe-Coding Trap' went further, pointing out that the repo had its history squashed to a single commit for the v2.0.4 release, which makes code review and trust verification essentially impossible.
The community reaction split cleanly. Some people are genuinely excited about the underlying theory, citing Victor Taelin's interaction combinator work and noting active university research building on HVM. Others pointed out the obvious irony: a tool meant to catch AI mistakes, built with unaudited AI code, is a punchline about the current state of the field.
The pattern here is a real one. Vibe-coded tools are shipping as serious infrastructure. The compiler is the trust root of the entire toolchain. Shipping an unaudited AI-generated compiler for a language designed around correctness guarantees is exactly the kind of second-order risk that gets missed when the demo looks good.
So what?
If you are evaluating any new programming language or compiler tool, check whether the implementation itself was AI-generated and whether there is a meaningful audit trail. An AI-written compiler with squashed git history is not a foundation to build on. The underlying research may be interesting, but shipping timing matters less than shipping something you can actually trust.