Not a dig at the team, deepseek coder is one of the best coding models out there. What surprised me was the lack of training data for rust 0.08% compared to python which is around 15.12%

Is it just the fact that python and js like languages are more popular in OSS so they get a bigger share in in the datasets being generated for training these models? As a rust developer I would love to contribute towards creating more rust data out there (other than just writing rust code and open sourcing it)..

I think Rust is harder too.

On top of all the essential complexity of coding, Rust adds the accidental complexity of solving a logic puzzle to satisfy the borrow checker. If there is one thing LLMs can’t do it is solve logic puzzles, at least not if you want the right answer.

Thats true, maybe one of the reasons I like rust. But logical puzzle solving implies constraints and that's something LLMs can solve (in a constrained environment with the compiler telling you what's right and what's wrong..) I might be over ambitious about this but to me rust code would seem more solvable compared to say python or raw javascript (which have no such compiler based guidance)

Well that kind of constraint solving is NP-complete and we know a lot about algorithms that can do that and they don't look like an LLM in a loop with a small number of iterations.

I think people don't give garbage collection the credit it deserves for advances in software reuse. If you are creating libraries in, say, C, you face tough problems with memory allocation and deallocation. If the library wants to allocate memory do you just use malloc() or let the application pass a pointer to an allocator? When the library doesn't need a piece of storage can it assume the application doesn't need it anymore?

With garbage collection you can smoosh together N systems into 1 and always know memory management is sound because memory management is a self-contained system and isn't weaved intimately into libraries and applications.

I did a free trial of the Jetbrains assistant and I was blown away by its talent for science fiction chat (only knew one person who was better) but its programming ability was not so hot. It was remarkably good at "linear" tasks where you don't need to understand the relationships between things but it did not succeed when there was any depth in the problem.

For instance it could infer the SQL scheme for my database just by looking at the stubs that JooQ generated and it would make a good try at writing SQL queries. I really wanted help writing this kind of query

https://www.jooq.org/doc/latest/manual/sql-building/sql-stat...

and it could not get it right at all because of the reference circularity. Similarly, it did a good job of explaining code one line at a time but it could not say something like "this is a topological sort" or "this makes no sense at all".