AI Was Supposed to Free My Time. It Consumed It. — Katie Parrott

Parrott documents her experience with her OpenClaw agent Margot: what started as a few-hour Friday afternoon configuration session turned into a 12-hour marathon of writing essays, rebuilding her website, and adding app features until 1 a.m. New research shows AI does not reduce work — it makes people want to do more of it.

The mechanism is threefold: (1) task expansion — AI makes adjacent tasks feel achievable, so completed work generates more work; (2) blurred boundaries — ambient AI erases the separation between work time and personal time; (3) dopamine loops — “one more prompt” operates like a slot machine, with variable reward schedules that make it psychologically difficult to stop.

Parrott argues this is not just a personal discipline problem but a structural feature of how AI tools interact with human psychology. The piece cites research and includes perspectives from others experiencing the same pattern.

RDCO Mapping

Important counterweight to pure productivity narratives. Our own agent operations run 24/7 by design, but this article highlights the risk for the founder side — the temptation to stay in the loop when the agent can handle things autonomously. Reinforces why our no-babysitting principle exists: the agent should work independently, not create an engagement trap.