Agentic loops, the architecture that lets AI run continuously toward a measurable goal without human prompting at each step, are expanding beyond software engineering and into general knowledge work.
NLW and Nufar Gaspar break down the structural requirements that make this possible: loop design, graph engineering, and multi-agent coordination where agents check their own outputs and keep running until completion criteria are met. The session is worth reading for the loop design specifics alone, not just the high-level framing.
The practical question this raises is which knowledge work tasks can be made measurable enough to close the loop autonomously, and that is exactly what the full workshop session addresses.
[WATCH ON YOUTUBE →]