

Yeah… this is the problem with generative AI. It’s trained to produce an end product. It doesn’t know or care about the process of getting to that end product, or what happens next. It knows how to make bits and pieces of code that look and function like the code it was trained on. If the problem you’re trying to solve has already been solved hundreds or thousands of times, you’ll probably be fine. But that isn’t why software engineers get paid good money. We get paid because we’ve spent years learning lessons about the advantages and disadvantages of different solutions and when to use them. AI doesn’t know or care about that. At best, it just knows “make thing work.” Trying to maintain or change AI-generated code is nearly impossible because there’s no reason or logic to how it made the decisions it made. Whenever it made a decision it just did whatever was the most statistically likely thing. If you want to make a change, the easiest thing to do is just modify your prompt and let the AI regenerate the entire codebase from scratch.



Why is Hazel going to the gym? Not getting enough play time at home?