Everybody Hates Opus

Well said, and unfortunately all too true.

So where does that leave us? People are posting various workarounds. But really, as Mo says, it’s a “personality imprinted on a trillion dollar language model”. Any output style you put on top is a bandaid, nothing more. It is a forced correction on the models “natural” way of generating output, or communicating with the user. That is not a good solution.

Maybe we can remove ourselves more from the direct output and the annoyance of those interactions. You know, have the agents do their work based on instructions we have set and workflows we have built. Sounds good, but I’m afraid it isn’t fully realistic.

Now we are forced to read a lot of LLM output in our daily work. If you are pushing around a couple of agents, rotating, managing, making decisions, reading reports, then there is little you can do to not read a ton of AI output.