|
For example, if there's no consistent framework or architecture to the program, and if the AI ignores the plan and makes shortcuts and bodges all over the place, the program becomes unmaintainable because everything depends on everything and changing one thing breaks another in unpredictable ways. Sending the AI back on the job to do global changes to fix one part will introduce random errors in other parts because the AI does not understand the code either. There's no "inpainting" for code like that.
|