Loop Engineering Meme Revives Interest in Second-Order Cybernetics

Loop Engineering shifts AI work from manual turn-by-turn prompting to designing autonomous systems that discover tasks, assign work, verify outputs, persist state, and iterate until goals are met.

This is cybernetics revisited. First-order cybernetics (Wiener) gave external feedback for control and stability in observed systems. Second-order (von Foerster, Maturana/Varela) includes the observer: self-referential loops where systems observe and regulate themselves, enabling autopoiesis, reflexivity, and co-evolution of boundaries and models.

Modern agent loops embed meta-observation and adaptation—the designer sets purpose while the system runs internal feedback. Taking second-order cybernetics seriously means building agents that construct and maintain their own coherence, not just execute scripts. A natural evolution for robust agentic AI.