Developers tend to overestimate correctness of AI code, particularly when code structure appears familiar and complete.
How do you trust an LLM to write code when you don't know what it does? Everyone is doing it. I use 3rd party libraries. Yet I can't trust LLM generated code.
