Comment by fastball

Yes, that is in fact how models get better at coding.

Such a ridiculous stance: "I want LLMs to code for me, but I want them to be trained on other people's code, not mine, duh".