Designing agents that don't loop forever — companion videos

AI Engineer

Barry Zhang on three rules — don't build an agent when a workflow would do, keep the loop as simple as possible, and "think like your agent" (sit in its context window and notice that it is making decisions in the dark between screenshots). The simplicity argument and the "is this task even worth an agent" checklist are exactly the discipline the article asks for.

What you should get from this: Design simpler agent loops with clear stopping rules, task boundaries and human control points.

Watch or know first: Have built or at least specced an agent; the talk assumes you know why loops run away.

Anthropic

Three Anthropic engineers walking through the most common pitfalls they see — agents that don't know when to stop, over-prompting in the system prompt instead of fixing the environment, the cost of multi-agent designs nobody actually needed. Useful right after Barry's talk; you will recognise the same patterns from a different angle.

What you should get from this: Recognize common agent-building pitfalls before adding multiple agents, complex prompts or hidden state.

Watch or know first: Watch the primary pick first — this assumes the same baseline and adds field pitfalls.