Agentic design workflows have broken something specific: the feedback loop between hand, tool, and thought. A designer and developer since 2008, the author identifies two distinct cognitive states in design work. Making-feel is the generative state where spatial manipulation in canvas tools like Figma produces thinking as a byproduct of gesture. Result-feel is the evaluative state produced by prompting an agent, where you describe an outcome and a system resolves the specifics. The shift from one to the other is not a productivity question. It is a neuroscience question.
The argument is grounded in real research. A 2025 Life journal study confirms handwriting recruits broader motor, sensory, and cognitive networks than typing. The author extends this logic one step further: prompting is further removed than typing, closer to dictating to someone who paraphrases you. Polanyi's observation that 'we know more than we can tell' is the structural problem. Tacit spatial knowledge, the kind that lives in gesture, cannot be fully translated into language without loss. That translation cost is where the depletion comes from. Previous tool shifts, drawing board to Photoshop, Sketch to Figma, preserved gestural continuity. The agent is the first shift that breaks it entirely.
The author lands on a distinction worth sitting with: some of the fatigue is a mismatch problem, reaching for the agent when the problem needed the canvas, and that part should improve with experience. But some of it is a permanent structural tax on any workflow that requires converting tacit knowledge into explicit instruction. That second cost does not get cheaper with practice. Read the full piece for the extended mind thesis connection via Clark and Chalmers, and for the author's precise breakdown of why the friction in direct manipulation was never waste. It was the thinking itself.
[READ ORIGINAL →]