The Attention Ledger

The Exhausted Hour

You have more free time than your grandparents did. Fewer working hours, more machines doing the chores, more entertainment available at zero marginal cost than any human could consume in a thousand lifetimes. And you are more tired than any of them ever were at the end of an easier day.

That's not a character flaw. It's an accounting error, and not yours.

The technologies that gave us hours back demand attention to use. The dishwasher freed twenty minutes; the phone that came free with it takes donations around the clock. So leisure quietly became the most attention-intensive part of the day. Nobody's tired from lack of hours anymore. We have free time without free attention.

The Proxy

Let me be precise about the claim, because this genre loves a theory of everything and I'm not selling one. Attention doesn't explain the economy any more than oxygen explains the conversation. The claim is narrower and harder to escape: attention is an integral input to everything we do as humans. Every email triaged, every decision weighed, every line written, every conversation actually listened to: whatever else those acts run on, they are all partly paid for in the same currency. Herbert Simon said it in 1971: a wealth of information creates a poverty of attention. Fifty years later, economists have formalized what everyone feels. Loewenstein and Wojtowicz's "The Economics of Attention" (Journal of Economic Literature, 2025) treats attention as a productive resource, scarce, depletable, and now the binding constraint on both labor and leisure, and names the flood of attentional externalities from digital technology as a frontier problem.

And attention has a property almost nothing else we run on has. Nearly every other resource is fed to us from outside: calories, capital, information, tools, even our hours are issued by the calendar whether we like it or not. Attention is one of the only resources a human generates internally. It cannot be imported. You can hire someone else's time; you cannot hire someone else's focus and have it count as yours. There is no external supply line, which is exactly why the extraction industry has to operate at the source. There's nowhere else to mine.

It also behaves like nothing on the time side of the ledger. It varies in quality hour to hour. It depletes with use. And it does not renew within the day: you can be handed a free evening and arrive at it broke.

Time is the unit on the invoice. Attention is what actually gets spent.

The Arbitrage

Now look at what the economy actually prices. Wages are per hour. Subscriptions are per month. Video is measured in watch-hours, podcasts in listen-time, engagement in minutes per day. Everything is denominated in time, while the thing being consumed is attention.

The gap between those two denominations is not a rounding error. It is the business model. The platform economy buys human focus at time-prices and sells it at attention-prices, and the spread is one of the great arbitrages in commercial history. Tim Wu chronicled the extraction industry's rise in The Attention Merchants; James Williams called distraction a threat to the will itself. What I want to add is the mechanism: the entire trade only works because the resource being strip-mined never appears on anyone's books. They found the one resource the accountants forgot to put on the ledger, and they strip-mined it.

And because attention is what your future self was going to think with, the extraction has a precise name. Attention extraction is wage theft from your future self.

The proof that this is concrete economics and not a metaphor shows up somewhere unglamorous: your conversion rate. Attention expenditure is compelled at work and sovereign in commerce: an employer can require cognitive spend; a seller cannot. A prospect arrives at your landing page already overdrawn, spending a guarded, discretionary reserve. Every ounce of cognitive work the page demands (every "figure out what we do," every "watch our story first") is a price increase that never appears on the invoice. High-traffic pages with nil conversion usually don't have a traffic problem or a persuasion problem. They have a pricing problem in the attention currency. The old direct-response people understood this before anyone had the vocabulary: Eugene Schwartz mapped awareness stages in 1966, and the usability tradition from Krug onward treats every unnecessary thought as friction. The rule that falls out of the ledger is simple: every stage of awareness gets its own shortest path, and the sin is charging cognitive tolls that don't match what the visitor already knows. Taxing someone who arrived holding money is theft twice. Educating someone who arrived knowing nothing isn't a detour. For them, education is the shortest path, measured from a different starting point.

Residue

Fragmented attention isn't less attention. It's a different substance.

Sophie Leroy's research, the work Cal Newport built "attention residue" into common vocabulary from, shows that each task switch films the next task with a residue of the last one. You don't get eight hours of attention in an interrupted day; you get two clean ones and six coated ones. Then we wonder why the tank is empty on a light day.

The Second Shift of Leisure

Which brings us back to the exhausted hour. Recovery time is now contested territory, optimized by the best-funded engineers alive against the one thing rest actually requires: disengagement. The feed is designed to be spent on, not restored by. So people spend their free time and arrive at Monday having never closed the register.

There's a name for the pattern. When a resource gets cheaper to use, consumption of it explodes: a Jevons paradox, and applying it to attention is by now a shared observation, not a discovery (Newport ran a version of the argument for knowledge work just this year). Every efficiency that made attention cheaper to claim multiplied the claims. The hours were liberated and the focus was colonized, and the second transaction ate the first.

The Flood Asymmetry

Then 2023 happened, and the arithmetic broke completely.

Generative AI collapsed the cost of producing a claim on your attention to approximately zero. The cost of evaluating each claim remains fully human: seconds of focus, spent one claim at a time, by a brain that still runs at biological speed. Machine-priced supply, human-priced filtering. That equation does not balance, and no amount of personal discipline makes it balance.

Worse, the filtering burden was quietly privatized. The editors, schedulers, and curators who once absorbed it professionally have dissolved into every individual's evening. You are now your own acquisitions department, unpaid, after hours.

One consequence, though, runs the other way: when the flood is infinite, the thing that can't be flooded (genuinely earned attention, willingly given) appreciates. I made that argument from the maker's side in my last essay. This is the economics underneath it.

The Canary

I've been living in your future for forty years. Let me tell you what works there.

I have ADHD, which means I ran involuntary fieldwork on this economy decades before it went general-admission. Fragmentation as the default state. Residue on everything. Exhaustion that arrives without any hours being taken. The world didn't get kinder to ADHD. It universalized ADHD's operating conditions, and now everyone's asking questions people like me have been forced to answer since childhood.

The answers were never willpower. They were structural: externalize the memory so the head doesn't have to hold it; fence the time so the decision is made once instead of continuously. My calendar is not a schedule, it's a prosthetic. My notes app is not a productivity habit, it's where my working memory lives, because the onboard version was never going to hold the queue. A lifetime of building compensatory systems turns out to be the survival kit for conditions everyone now shares.

The Branch-Capture Hypothesis

Here is the part I can't cite my way to, so I'll label it exactly what it is.

Since large language models became part of my daily work, the change I notice is not more attention or less attention. It's less tension. And I think I know the mechanism. An ADHD mind generates associative branches faster than working memory can hold them: every thought spawns three forks, and the forks decay while you're servicing the first one. That mismatch is experienced as tension: the racing, the fear of losing the good branch, the hum of unexplored forks dying. What an LLM does, in my experience, is branch capture: every fork can be chased to resolution in seconds instead of dying unexamined. The loops close. Less open-loop inventory, less hum.

The distinction that matters: this is not offloading the thinking. It's offloading the queue. The cognitive-debt worry, the one the general-population studies keep circling, is about AI replacing cognition. Branch capture is AI absorbing the working-memory overflow of an associative mind so the cognition can actually happen.

What does the literature say? Nothing directly, and the adjacent evidence points the right way. Purpose-built ADHD chatbots have shown symptom improvement: Jang and colleagues' four-week trial of the Todaki chatbot reduced attention-deficit symptoms, with improvement tracking usage; Shou's meta-analysis of six randomized trials found online interventions improved attention deficit with a solid effect size; a 2025 systematic review by Pergantis and colleagues reports chatbots improving concentration, emotional self-regulation, and memory capacity, and a 2026 review describes AI as a "dynamic cognitive and emotional scaffold."

And the honest counterweight, because an argument without one is content: Selaskowski's group in Bonn ran a randomized trial of chatbot psychoeducation for ADHD adults and found no advantage over a conventional app, with the authors raising the possibility that constant interaction itself taxed learning. So the interaction can be attentional load. The open question, which I'd rather pose than pretend to resolve: does the tool absorb your branches, or spawn new ones? Both happen. Design and usage pattern decide which.

No one has studied everyday, general-purpose LLM use by ADHD adults doing real work, and no one has measured the tension dimension at all. So: the mechanism predicts it, the adjacent trials are consistent with it, and forty years of fieldwork keeps confirming it. I'll say it as a hypothesis and mean it. AI, used as branch capture, reduces the background tension of an associative mind. That's the essay's offering. I've designed the experiment to test it on myself, and I'd love for a real lab to steal the idea.

The Regressive Tax

Notice, meanwhile, who gets to opt out. The wealthy buy their attention back: ad-free tiers, assistants, human filters, silence as an amenity. "Free" services are paid for in focus, which makes attention extraction a regressive tax, heaviest on the people with the least ability to buy relief. A decade ago privacy went from a nerd concern to a consumer right. Attention sovereignty is next, and it will follow the same path: first ridiculed as precious, then priced, then regulated.

The Guardian Question

The uncomfortable symmetry is that the technology flooding the commons is the only technology that can defend it. Something has to filter, triage, and absorb cognitive load at machine speed, because the claims now arrive at machine speed. The design ethic that separates a guardian from another extractor fits in one line: the tool does the labor; the human keeps the judgment. And any guardian that itself demands attention has failed its only job.

I won't pretend the tension away. Delegation can restore attention or atrophy it, and the same tool does both depending on how it's built and how it's used. That's a design responsibility, not a settled fact. But "the flood is also the levee" is the only structural answer on the table. Everything else is telling individuals to bail faster.

The Ledger Balances

Almost every book in this genre ends the same way: meditate, minimize, log off. Individual discipline against an industrialized extraction economy: bringing a habit tracker to an arbitrage fight. The way out is structural. Price attention honestly. Build technology that guards it instead of mining it. And if you make things for a living, make the third choice, the one fully in your control: if attention can no longer be efficiently stolen, the alternative isn't shouting louder. It's making something worth the spend. The campfire, one level up. The commercial version of the same ethic: price attention honestly and the customer can finally afford you.

I made a short that literally tells the audience to pay attention. This essay is about what that request costs, and why the only decent way to ask is to make the paying worth it.