I tried them all.

Grok was worse than even some of the more mediocre open models at actually doing anything. (At least anything tech work related.) GPT and Claude just do what I ask most of the time. With grok, it’s like a chore just getting it to understand the question.

You’re pulling your hair out trying to figure out what on earth you need to do to land in the right place in whatever topsy turvy embedding grok is using?