Polished speed is not what we should ask of AI — or ourselves
The real costs of fast, polished, thought-free output
There’s a passage from Mike Barlow‘s new book, The Thinking Mirror, that’s been sticking with me:
AI is exceptionally good at smoothing things out—filling gaps, offering plausible continuations before you’ve fully realized what you want to say. That smoothness can feel like help. But it also removes the small points of resistance where reflection normally happens. When friction disappears, decisions still get made. They’re just made faster, and with less awareness.
That idea brought me back to something I’ve long played with: strategic inefficiency. Not the unintentional kind—the convoluted process, outdated equipment, or poorly trained staff. And not the purposeful bureaucratic stalling that slows people down to wear them out.
I mean strategic inefficiency as a deliberate choice: the clinician who takes time to talk with patients, the flood plain left undeveloped to absorb a storm surge, the local producer who mitigates systemic risk. It’s intentional elasticity—deciding in advance where you want the system to flex when you need it to. Financial reserves. Ample staffing. Redundant systems. “Extra” resources sitting mostly idle—and seemingly surplus, until the one day they’re not. It’s not as strange as it seems. Your fire department runs on this logic.
One argument for strategic inefficiency is the cost of cutting it close. Generative AI’s early promise has largely been about automating friction away entirely: lay off the copywriter, skip the meeting, let the chatbot churn out a coherent proposal before you’ve finished typing the meeting invite to discuss it. But those hyper-lean, just-in-time supply chains looked brilliant until COVID shut down global trade and there was no fat left to absorb the shock. Bananas are under threat for similar reasons. In 2008, the economy collapsed from ever faster and ever more opaque transactions that obscured underlying risk.
What raw efficiency really costs us, though, is more personal than any of that. Great art and breakthrough ideas rarely come from efficiency—they come from mental meandering, from the “default mode network” that only lights up when your mind is free to wander: on a run, in the shower, or sitting in the woods. It’s true of good copy, innovative strategies, and providing thoughtful feedback. That quality time is under threat in a world built for doing rather than being. And when a task gets automated, the time it frees up rarely turns into a walk in the park. It just turns into more tasks. More pings.
We are all Lucy and Ethel in the chocolate factory, scrambling faster and faster as the conveyor belt speeds up—while the attention economy stands by, ready to fill any remaining silence with “snackable” content. Potato chips for the brain.
AI isn’t the villain in this piece—I used Claude to help research it. The question isn’t whether we use these tools, but how we choose to be with them: whether we build in the friction, the slack, the strategic inefficiency that lets us stay human while we do.
Read the full piece on Substack for more examples, and the full case for daring to be inefficient—strategically, of course.
Want the rest of the story? I’m now writing longer pieces on Substack. It’s all free, so check it out here.