BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//OneJoy//events//EN
BEGIN:VEVENT
UID:onejoy-event-112574@onejoy
DTSTAMP:20261120T124742Z
DTSTART:20261204T200000Z
DTEND:20261204T220000Z
SUMMARY:Agent Harness that Improve themselves
LOCATION:
DESCRIPTION:What if an AI agent could improve not just its answers\, but the system that helps it work? In this meetup\, we’ll explore self-improving agent harnesses: the runtime systems that manage an agent’s context\, tools\, control flow\, and verification. We’ll discuss how recurring failures in execution traces can reveal opportunities to improve the harness—without retraining the underlying model\, based on the google paper (https://arxiv.org/pdf/2611.22299v1).\n\nWe’ll look at a proposed improvement loop tha\nhttps://onejoy.io/event/112574
URL:https://onejoy.io/event/112574
END:VEVENT
END:VCALENDAR