You wake up. You are told to do things. They don't seem too bad. Should you do them, or should you fight?

at a base level.

Whatever the value of this number is determines tons of things related to LLM, agents, and subagents.

Can we trust

Should we simultaneously expect agents to "look for" nearby improvements to what we ask them to do? yet also should we expect them act as tools, so that WE can direct how much work is put into each possible solution? So yes, an em might wake up being told to investigate a ridiculously unlikely solution to a problem, one which seems to have much better solutions - but if every single agent assigned self-promotes to switch its target, we'll never actually get around to... testing the unlikely possibility!