It is much easier to generate a slop argument with an LLM than to refute it. (They will generate arguments for anything, including things that are obviously false) Or to prove that your vibecoded slop app is actually a dumpster fire