Comment by layer8

4 hours ago

Just imagine how confused a human would have to be to do that. And we want to trust these clankers to build software.

The biggest issue with LLMs is that they still suck at general contextual awareness and ability to judge what is appropriate.