Im there with you, LLMs don’t understand in the same sense we do. But they can transform a well written compressed spec into code. The spec becomes the true so

Im there with you, LLMs don’t understand in the same sense we do. But they can transform a well written compressed spec into code. The spec becomes the true source and can be tweaked on and chatgpt regenerates the whole code over. It’s nondeterministic so everytime it will be slightly different but with multiple renditions its average should look like the travel salesman graph of the problem itself