In a sense they do care. Anthropic / OpenAI care that your projects are successful because that means more revenue for themselves. Therefore, their models are designed to care that your products work.
It seems to me the incentive, if anything, is to make the codebase so complex and “write-only” that developers become entirely dependent on LLMs to make any change whatever. They care that you keep burning tokens, not that those tokens accomplish anything for you.