As expected we’re getting closer and closer to the Trough of Disillusionment. That’s not a bad thing, because it leads to the Plateau of Productivity.[1]
Anecdotally for myself I’m finding that LLMs are great when I can give it a hyper specific target like a function to write. This isn’t because it can’t write an entire script. It can. It’s because the more I let it run wild, it feels like my understanding of the code gets exponentially worse.