Replying to
@failedLyndonLaRouchite@mas.to @ketan@climatejustice.social I don't think usefulness and spamming tons of subagents to iteratively verify and redo and verify and redo and verify and redo are conflicting in any way. The energy usage to get to the point where things are good enough could means massive consumption of energy/tokens for anything but very simple requests. It's like having a junior employee with an encyclopedic culture who never learns anything but to read more instructions every time before trying anything. Again and again and again.