Gauntlet Loop Template: Builder, Harsh Critic, Real Bar

Goal: [YOUR GOAL - a site, an essay, a CLI tool, a research brief] Quality bar: [A NAMED, FETCHABLE reference - a specific live page, a specific published post, a specific repo. Not a category, not "award-winning". The critic must be able to open, read, run or screenshot it.] Instructions: 1. Split the goal into small, checkable pieces. Each piece must produce a real artifact, not a description of one. 2. For each piece, run a BUILDER agent that produces the artifact, and a SEPARATE CRITIC agent with fresh context that never sees how hard the builder tried. 3. The critic opens the actual output, puts it next to the quality bar with the labels stripped, and makes a binary pick: which one is better. No scores out of 10 - scores drift upward every round. 4. If the bar wins, the critic names the single biggest gap. The builder fixes that gap only, then the comparison runs again. 5. Keep looping until our work wins the blind comparison. Never exit after a fixed number of rounds - the exit is winning, or the human stopping the run.