LLMs are quite capable coders. Now, people say they are limited by programming languages designed for humans.

I think it is the opposite: they are enabled by programming languages designed for humans.

You cannot just design a new programming language for LLMs because where would the training data and project examples come from?

It’s also interesting how, for that reason, Python acts as a reasoning trace for LLMs. I.e., ask to code up something in Python, then port it to CUDA or Rust (versus coding it in CUDA or Rust) directly.