I know excatly how bad the output they give is, because I ask for output that I can understand, debug and improve. People misusing tools don't make tools usele

I know excatly how bad the output they give is, because I ask for output that I can understand, debug and improve. People misusing tools don't make tools useless or bad. Especially since LLMs designers never claimed the compressed information inside models is spotless or 100% accurate, or based on logical reasoning. Any serious engineer with a modicum of knowledge about neural networks knows what can or can't be done with the output.