Is it just got a big enough corpus of nation that it spits back to you impressive looking stuff? Or is it actually understanding the context in some test that's even matter? In some cases it does, it does understand the context in some cases. It just doesn't feel like it's just brute forcing its way into the problem. And so now it just feels like its doing something with understanding. Has it been trained on the entire corpus, on something like gitub or all open source repositories? Yes, so it has been trained on code, so it's able to write code. Buti still wod would say youre programming. Even if we get to the point where there