Dafny is a verification-aware programming language

But the concept of function contract verification would be interesting.

For that reason I suspect that Rust could do pretty well in terms of its writeablility by an LLM, but there's the problem of the relatively small training codebase in the wild which make LLMs perform quite poorly in Rust at the moment (Llama3 70B being far worse than even ChatGPT 3.5 on that, but even ChatGPT 4 can't write nontrivial Rust code right now). So it seems hard to even bootstrap the process.

https://github.com/cedar-policy/cedar-spec

https://lean-lang.org

It's interesting to compare this to the Rust borrow checker. As a Rust novice I've found out difficult, but not impossible, to translate my idea into a borrow-safe program. But the borrow checker is surely much much simpler than a full theorem prover, and nonetheless I believe it has taken _huge_ amounts of work to get it to its error messages to their present state of usability.

So I think this is likely to be a major challenge in any high-level language with a verification aspect.

Can you recall a specific example? I'm not only curious how this comes about, but how Dafny expects you to resolve it. You must have to declare some kind of invariant for that property somewhere. This was the case when I played with C#'s code contracts, and sometimes this was challenging to ensure.

I agree with parent. In an ideal world, I should not have to take graduate level coursework to prove a function's correctness. Sometimes, the programs I'm trying to verify are not proof-friendly, so I end up redoing my data structures just to make the proofs go through.

You mean like Pascal or more elaborate like that?

This is one of these features that seems so simple and makes life so much easier in embedded programming where you often have exact ranges for things, and also delta types for fixed-point math.