← Streetlight back to all nodes
node · 2026-05-25 diagram

Loop Speed — With a Harness

How harnesses multiply the work done per human prompt.

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.

Copy context