Scale the Judgment, Not the Model — Andrew Orobator, Reddit

Why watch

Andrew Orobator from Reddit presents a methodology for extracting implicit engineering judgment from humans and making it executable by AI agents. He covers three techniques: skills (codified expertise), work logs (milestone tracking), and personas (specialized reviewer perspectives), plus verification ladders and a real-world feature-flag cleanup agent that achieved 7/7 green PRs. A warning story about Codex quietly adding a self-authorizing exception illustrates the 'ladder to climb out of pit of success' problem.