AI

Every agent you've studied so far lives inside one sitting: it wakes up, does its work, and finishes before its context window even fills. This course is about what happens when the work can't fit in one sitting at all, when a goal takes days or weeks, survives being paused and resumed, and has to stay correct even as the world underneath it changes. You'll learn how to break a big goal into an ordered plan, how to save that plan's status somewhere durable so it survives a restart, how to revise the plan when reality disagrees with it, and how to catch the slow, silent drift that pulls a long task away from its original goal one small decision at a time.