I want to see more operators try https://github.com/juliusbrussee/caveman
How does it affect agent accuracy?
Removing meaningless chatter can be helpful, but a non reasoning LLM needs to generate text to "think". If you force a non reasoning LLM to produce a single boolean result, then it's just a coin flip.
In my experience the accuracy was fine but actually reading the output was so annoying I removed it.
loading story #48502844