Attention as a craft, not a resource

There's a quiet shift happening in how we describe our days. We say we 'spent' the morning on a task, that we 'invested' an hour in a meeting, that we 'saved' time by skipping lunch. The vocabulary of attention has collapsed into the vocabulary of money.

I've started to think this metaphor is doing real damage. Money, when you spend it, is gone. Attention, in the right shape, is the opposite — it returns to you stronger. The hour I gave to writing this essay didn't drain a fixed reserve. It deepened the riverbed.

The accountant's view

The accountant in your head wants every minute logged. Productivity culture is built on this — track it, optimize it, defragment it. But the accountant has never written anything good. Good work happens in long, undisturbed troughs of time, the kind that look wasteful on a calendar.

The hour I gave to writing this didn't drain a fixed reserve. It deepened the riverbed.

What practice looks like

If attention is a craft, then it has a practice. The practice, for me, has three parts: long sittings with one thing, deliberate boredom, and the courage to reread what bored me yesterday.

Long sittings are the obvious one and the hardest to defend. The first hour with a hard problem is mostly throat-clearing — the mind keeps trying to leave. The second hour is where it gets good, and you only ever reach the second hour by refusing to spend the first one somewhere else. Every interruption resets the clock. A craft measured in deep, uninterrupted blocks is the one thing an open-plan calendar is designed to prevent.

Deliberate boredom is stranger, because it feels like doing nothing. But the gap between two thoughts is where the next thought comes from, and the reflex to fill that gap — reach for the phone, open a tab — is the exact muscle the craft is trying to train. Sitting in the gap on purpose is the rep.

And rereading what bored me yesterday: the page that put me to sleep is often the page I wasn't ready to read. Attention is partly the willingness to return to something that gave me nothing the first time, on the bet that the failure was mine and not the book's.

None of these are radical. None of them require an app. The hardest part is believing that the slow thing is the fast thing, and that the river will find its bed if you let it.

The craft you can lose

Here's where the craft framing earns its keep, and where it starts to worry me. A resource, once spent, is simply gone — but it doesn't decay from neglect. A craft does. The muscle you stop using doesn't wait at its peak for your return; it softens. If attention is a skill, then not practicing it isn't neutral. It's a slow un-learning.

Which makes this a strange moment to be writing about attention at all, because we have just built the most efficient machine yet for not practicing it.

When I draft something with a model open in the corner of the screen, the work gets done and the riverbed doesn't deepen. A team at MIT's Media Lab put instruments on that feeling. In their 2025 study, Your Brain on ChatGPT, participants who wrote essays with an LLM showed the weakest brain connectivity of any group on EEG, while those who wrote with nothing but their own heads showed the strongest. They called what built up "cognitive debt" — the bill that comes due later for the thinking you skip now. A separate survey of 666 people, published the same year, found heavier AI users scored lower on critical thinking, and that the effect ran through exactly the mechanism you'd expect: cognitive offloading. The young, who offload most, scored lowest.

Both findings are early and easy to over-read — one is a small preprint, the other a survey that can't prove which way the arrow points. They rhyme with something I already believed before I read them, which is the part I have to be honest about. But they describe a riverbed silting up. Stop sending water down the channel and the channel fills in. The water still gets where it's going; it just takes the shallow route.

What I'm not sure about

Here's the objection I can't fully answer. Every tool that ever freed up attention was also accused of destroying it. Socrates thought writing would ruin memory. He was right — we don't memorize epics anymore — and it also didn't matter, because writing let us hold thoughts no memory could. Offloading isn't always atrophy. Sometimes it's what clears the desk so a harder kind of attention can begin.

So maybe the craft isn't attention in general. Maybe it's the narrower, more stubborn skill of knowing which things to keep for yourself — which sittings not to delegate, which boredom not to fill, which difficult page to reread instead of summarize. The accountant in your head will always vote to offload, because offloading shows up as time saved on the calendar. The craftsman has to override that vote by hand, every time, with no proof it was worth it.

I don't have the rule for which is which. I have a suspicion that the question matters more now than it did two years ago, and that the people who never bother to ask it will look up one day to find the river has quietly moved on without them.

Sources

• Your Brain on ChatGPT: Accumulation of Cognitive Debt when Using an AI Assistant for Essay Writing Task — Nataliya Kosmyna et al., MIT Media Lab, June 2025. media.mit.edu
• AI Tools in Society: Impacts on Cognitive Offloading and the Future of Critical Thinking — Michael Gerlich, Societies 15(1), January 2025. mdpi.com