Maintainable by whom - a human or an AI? If you are intending to maintain it yourself, it doesn’t matter how good the code is, you can’t maintain something that you don’t understand. So you need to read and understand the code at some point (with the agent’s help if necessary) and get at least a basic mental model in place.
One thing I’ve run into is that the agent’s choices or styles aren’t necessarily bad per se, but I would have made a different choice. One way to help with that is to start the project yourself and then tell it to follow the coding style.
Another thing you can do is write the tests yourself. You know better than anyone else what the intended functionality is. If an agent writes the code and the tests and you don’t look at them, you don’t know for sure that the product actually functions correctly.
Maintainable by whom - a human or an AI? If you are intending to maintain it yourself, it doesn’t matter how good the code is, you can’t maintain something that you don’t understand. So you need to read and understand the code at some point (with the agent’s help if necessary) and get at least a basic mental model in place.
One thing I’ve run into is that the agent’s choices or styles aren’t necessarily bad per se, but I would have made a different choice. One way to help with that is to start the project yourself and then tell it to follow the coding style.
Another thing you can do is write the tests yourself. You know better than anyone else what the intended functionality is. If an agent writes the code and the tests and you don’t look at them, you don’t know for sure that the product actually functions correctly.