Agents are live. Most fail quietly. How do we make them reliable, measurable, and efficient at scale?
That's Saturday, December 5, 11 PM – midnight Pacific · Sunday 2 – 3 AM Eastern · Sunday 7 – 8 AM London.
Errors compound across steps, cost grows per tool call, and offline benchmarks rarely predict production.
Not single outputs. Harnesses and simulators that reflect production; SFT and RL on traces; cost-aware routing.
When systems understand goals and take actions, recommendation shifts from predicting clicks to achieving objectives.
Will agents replace the retrieve → rank → recommend stack, or rebuild it?
Final 10 minutes: your questions, top-voted on Slido.
