Developer builds 3-tier risk system to progressively delegate decisions to AI agents
kevinkern · x · 2026-11-07
The author shares a hands-on workflow for managing multiple AI orchestrators: every few weeks, he has each orchestrator review his recent decisions and argue for a "promotion" — decisions he keeps approving become theirs to make autonomously next time.
A simple 3-level system based on risk:
- Decide on your own: low-risk work like merging critic-approved builds to main, cleanup, visual fixes
- Ask another model: medium-risk, e.g. gameplay rule changes or anything touching the online version identifier
- Ask me: high-risk decisions
An attached screenshot shows the review checklist he sends the agent (what it confirmed, where it overrode recommendations, its delegation proposal), making this a reproducible framework for progressively handing authority to agents.
More from coding & agent
- Solo founder claims $20K months running an entire team's work through Claude — PrajwalTomar_ · 2026-11-07
- Thoughtworks Technology Radar tabs mattpocock/skills as Trial for agentic coding — mattpocockuk · 2026-11-07
- OnCall-Memory: open-source incident response agent that learns from past production incidents — NewAlarm8213 · 2026-11-07
- Should MCP server operators trust agent identities issued by a third-party service? — zach-benson · 2026-11-07
- Build a 'strategist' agent: route every agent's accumulated context into key business decisions — EXM7777 · 2026-11-07
- Terrarium open-sources local VM sandbox giving AI agents explicit, auditable permissions — v_0009 · 2026-11-07