I Run a Fleet of AI Agents Across Three Machines. Here's What Broke.
summary
Runs a fleet of AI agents across three machines and reports concrete failures and fixes: a hierarchical CEO VP manager worker design that reduces human attention to one context, state moved to workspaces on disk so resets and crashes do not lose progress, a review gateway that blocks plans until approval and a single Discord bot router that collapses multimachine inboxes. Numbers and failures matter: five break modes, three machines, one control plane, and a move toward Kubernetes for scheduling, secrets, and tools.
AI-written summary · about
related talks
32mWhat Agents Do When You’re Not Looking
22mThe Agent Control Plane: Managing Thousands of AI Agents
24mAI Agent Swarms Are Amazing
18mCollaborative AI Engineering: One Dev, Two Dozen Agents, Zero Alignment
17mFighting AI with AI
20mThe Three Ways AI Agents Fail. Here's How to Fix Yours.
15mThe Responsible Path to Hyperproductivity with AI Agents
34mThe Hard Truth About AI Agents in Production