AI agents shouldn't run in single player mode.
By Allie K. Miller · June 25, 2026 · Curated by George's Blog
AI agents shouldn't run in single player mode.
So we changed it.
Most people work in teams, but the more individualistic AI usage gets, the more gatekept its power becomes. So we built agent workforce loops — now my teammates can prompt each other's AI agents (image below).
How it works (we’re still iterating!):
- Each core teammate gets a "loop" Slack channel (ex: loop-allie)
- Teammates + Claude Code/Codex are in it
- Anyone can drop a task in the chat
- Low-risk: my AI Chief of Staff routes it, completes it, verifies, reacts ✅
- Higher-risk (sending emails, pulling files): it preps but stops, drafts only
- I approve by reacting ✅
- The whole workforce runs on a 15-min loop
Essentially: teammates can prompt another teammate's AI agent workforce.