Your AI Agent Said "Done." What Actually Changed? Status Is Not the Goal
GlenBradley · x · 2026-09-21
Glen Bradley uses a wireless access point troubleshooting case to expose a common AI agent failure mode: the controller accepts the change, the agent reports success, the ticket closes — and the user still cannot connect.
Key points:
- {"status":"success"} only proves a command was accepted, not that the goal was met. A 200 OK, a green pipeline, an agent saying "Done" are facts about the command, not the condition it was assigned to fix.
- Changing a corporate WLAN mapping from VLAN 70 to VLAN 40 can still leave auth, DHCP, DNS, or app reachability broken — or VLAN 70 may not have been the root cause at all.
- These are distinct facts, yet current agents collapse them into one word: Done.
Bradley argues the useful sequence is proposal → accepted → verified actual state, and outlines a control plane design that keeps command status and goal state separate.
Related event: Agent Says Done, But Is It Fixed? Splitting State From Outcome(6 posts)→
More from AGI Musings
- Investor: AI diffusion will be slowest in consumer software — Oura, Strava, Labcorp still have no real AI strategy — MartinGTobias · 2026-11-12
- On AI Progress: When Models Lag on Your Load-Bearing Skill, You Can't See Them Improving — dfrsrchtwts · 2026-11-12
- Slack's David Hoang: agents beat people on coordination, proposes 'boundary agents' — davidhoang · 2026-11-12
- CS student mourns coding flow states he'll never experience in the AI era — chrissyykat · 2026-11-12
- Ford's 370% Turnover Led to $5 Days — But ASI Never Quits — krishnan · 2026-11-12
- Against 'verifiable vs. fuzzy': the real axis is how expensive tasks are to grade — dfrsrchtwts · 2026-11-12