The Missing Layer After Launch
summary
A production system runs agents that watch thousands of live conversations and close the loop: a log-monitoring agent scans hourly, deep‑dives traces and opens PRs that are often ready in half an hour; a review agent critiques and rejects edits and generates roughly ten times more PRs than the human team; a session analyzer scores every conversation, surfaces root causes and success rates, and powers a health dashboard; a computer-usage agent drives the browser to reproduce UI bugs — the harness works and catches silent failures like pricing and constraint regressions.
AI-written summary · about
chapters (10)
- 0:00Launch is just the beginning
- 1:23Monitoring agentic systems
- 3:32Hidden failures in agents
- 5:40Logs aren't enough for agents
- 7:44Automated PR agents for fixes
- 9:07Two loops for detection and insight
- 12:39Session analyzer for health
- 14:01User perspective agent
- 16:27Meta harness for visibility
- 18:56Closing the loop
related talks
21mRemote Agents, or: How I Fixed Our Page's Performance Without Really Doing Anything
26mAgents for the Inner Loop and Outer Loop
45mAI agents in practice - beyond the prototype
32mWhat Agents Do When You’re Not Looking
31mInternal Background Agent Systems: Architecture and Rollout
26mLet Them Eat Bugs: Practical Showcase of Agentic Issue Resolution
19mYour Biggest User Isn't Human: Designing DevTools for AI Agents
30mLive Panel with