With a Harness
LLM Aggressiveness
=
LLM Reasoning
+
LLM Subagent
+
Subagent Response
+
LLM Reasoning
+
LLM Output
Human Input
=
5
With a harness, LLM Aggressiveness increases because you've got more happening for every human input. Five LLM steps for one prompt — 5 / 1 = 5.