That's great for now. It's not clear to me, though, where LLM's will get their training data from here forward without ingesting lots of LLM generated code and answers and eating it's own tail.
Didn’t you get the memo? LLM’s either already are capable of or just a step away from being able to reason so no need for human generated training data in the future.
Or at least that’s what 3/4 of HN commentators believe and all AI CEOs want you to believe.
That's only now and in the near term future. If AI is actually successful, every year the amount of human written code will decrease. That's the whole point of this.
