GPT-4 is godawful at coding past the simplest examples that any fresh hire could solve. It's best at stuff like simple web dev, CRUD, stuff for which there's billions of tutorials online. But for anything that requires actual logical reasoning, it's terrible and will get stuck on something as simple as an if statement, inverting it over and over while "thinking" it's actually making progress towards solving the problem at hand. Prompt engineering and advanced techniques like Tree-of-Thought only...