1 · The Last Scarce Resource

Simon - who went on to win the 1978 Nobel in economics - wrote that before the internet existed, when phones had rotary dials and "information overload" meant an unread stack of newspapers. Yet he had already seen the structure of an entire era: economics is the study of scarcity, and once information becomes effectively free and unlimited, the scarcity migrates to the receiving end. Every person gets roughly 16-17 waking hours a day, a supply that stays fixed no matter how many more videos, articles, and notifications humanity produces. Any resource with fixed supply and infinitely growing demand eventually attracts someone with a drilling rig.

The rig was built long ago. The key to everything that follows: the world's largest platforms sell attention; everything else is packaging. Facebook, TikTok, YouTube, and Instagram show content for free because content is the bait; the actual merchandise brought to market is the minutes users stay, packaged as "impressions" and auctioned to advertisers by the millisecond. The word "free" in "free social media" works the way "free" works in "free buffet for turkeys around Thanksgiving".

Divide 800 billion dollars of digital ads by roughly 5.5 billion internet users and a striking number falls out: the average online brain is worth nearly $150/year in advertising - a North American one worth several times that. That is what brands will pay merely to stand next to one person's stream of attention for a year. At those unit prices, expecting platforms to "respect users' time" is about as realistic as expecting a mining company to respect the mountain's scenery.

2 · A Machine Built To Win

When revenue scales with minutes-on-app, every part of the product evolves in the same direction: keep the user for one more swipe. No conspiracy required - just A/B testing left running long enough. Versions that retain better survive; versions that let people put the phone down die. After fifteen years of that selection pressure, the toolkit has converged on a list so familiar it reads like a description of anyone's home screen:

As for how far the optimization goes, insiders say it best. In 2017, asked about competition, Netflix CEO Reed Hastings answered matter-of-factly that Netflix's biggest competitors include... sleep - "and we're winning" (The Guardian, 2017). A CEO's joke is usually the balance sheet speaking in a cheerful voice: when the merchandise is users' waking hours, the final competitor really is the pillow.

The heaviest evidence: Facebook's internal files

If one document had to prove what "optimizing for engagement" means in practice, it would be the internal files disclosed by Frances Haugen in 2021. From 2017, Facebook's ranking algorithm scored every emoji reaction - including the angry one - at 5 points, five times a Like. The logic was pure business: posts that provoke strong reactions keep people around. The company's own data scientists later confirmed what anyone could have guessed: rage-bait posts were disproportionately likely to contain misinformation, toxicity, and junk news. Internal warnings were raised, and the machine kept running for nearly three more years before the weight was cut to zero (Washington Post, 2021).

3 · Why the Brain Loses

The natural question: users know all this, so why do they keep scrolling? The unromantic answer: the match was never fair. On one side, a brain evolved for the savanna - where new information was rare and precious enough that "attend to everything that moves" was a survival strategy. On the other, systems measuring the reactions of billions of people in real time, running tens of thousands of experiments a day to find exactly the stimulus combination that keeps thumbs swiping. The platform never needs to force anyone; it merely needs to understand the user's brain better than the user does - somewhat rude, but undeniably effective. The four most thoroughly exploited weaknesses:

4 · The Invoice: Brain Rot

In late 2024, Oxford University Press named "brain rot" its Word of the Year: "the supposed deterioration of a person's mental or intellectual state, especially viewed as the result of overconsumption of material considered trivial or unchallenging" - usage up 230% in a single year (OUP, 2024). The richest detail: the heaviest users of the term are Gen Z and Gen Alpha - the "mines" themselves naming the sensation of being mined, via memes, on the very platforms doing the mining. When a product's target demographic has to invent a word for how they feel after using it, that is the kind of market feedback no PR department can spin.

But "brain rot" is only the folk name. The measurable part sits in two decades of data from Gloria Mark, professor of informatics at UC Irvine, who has tracked human attention on screens since 2004 - initially by shadowing people with a stopwatch, later with logging software:

From two and a half minutes to 47 seconds in under two decades. Mark also documents the double bind: after each interruption, the brain needs roughly 25 minutes on average to fully return to the original task - while the gap between interruptions is now measured in seconds. At the current screen-switching cadence, "deep focus" for an office worker has been mathematically all but scheduled out of existence - and anyone who has tried writing a long document during business hours holds their own experimental replication.

The total volume being extracted is clearest laid out on a 24-hour axis:

On the health side, the research fragments are assembling into a consistent picture: reviews of doomscrolling studies find stable associations with anxiety, depression, sleep disruption, and reduced concentration (American Brain Foundation). A widely cited University of Texas experiment produced an even more uncomfortable result: a phone merely lying on the desk, face down, silent, measurably reduced participants' available cognitive capacity compared to leaving it in another room - a pure "brain drain" effect from presence alone (Ward et al., JACR 2017). The attention magnet pulls even when switched off.

5 · AI: Fuel on the Fire

The 2010s attention economy had one natural bottleneck: content still had to be made by humans, and humans are slow, expensive, and require sleep. Generative AI removes that bottleneck. The cost of producing a passable article, voiceover, or video is in free fall toward zero - while humanity's total supply of attention remains exactly the same number of waking hours. Content supply goes to infinity, attention demand stands still: in economic terms, that is the recipe for a race to the bottom, and the thing racing downward is quality.

The technical term for this flood is "AI slop" - industrial feed for attention: content mass-produced to please distribution algorithms, with the viewer serving merely as the intermediate medium on which ads are run. Note the incentive structure: a slop operation earning $117 million doesn't need viewers to like anything - it needs them to not have scrolled away yet during the first few seconds, so the impression counts. The attention economy has evolved to its purest form: skip the step of making content for humans, keep only the harvesting step.

But the most delicious irony of the story lives in the lab. Two 2025 studies, placed side by side, form a perfect pincer:

Pincer 1: a human brain that outsources thinking gets lazier

MIT Media Lab recorded EEG from 54 people writing essays under three conditions: unaided, with a search engine, and with ChatGPT. The unaided group showed the strongest and most distributed neural connectivity; the ChatGPT group the weakest - and the deficit persisted when they were later asked to write unaided again, a phenomenon the team named "cognitive debt": AI pays fluency up front, then bills you for your own capability later (MIT Media Lab, 2025). Small sample, needs replication - but the direction matches the intuition of anyone who let GPS navigate for three years and then discovered they no longer know their own city.

Pincer 2: a machine fed junk content... rots too

In late 2025, a research team (Texas A&M, UT Austin, Purdue) tried the reverse: continually feeding language models a diet of viral, short, sensational Twitter/X content - the exact daily ration of a heavy user - then measuring capability. The result: clear declines in reasoning, long-context understanding, and safety; the ARC reasoning score fell from 72.1 to 57.2 as the junk ratio rose from 0% to 100%, with "dark traits" like narcissism and psychopathy inflating along the way. The name the team chose for the phenomenon: LLM brain rot (arXiv 2510.13928).

Two details from that paper deserve framing. First, the primary damage is named "thought-skipping": models progressively drop intermediate reasoning steps and jump straight to conclusions - a description that fits a person watching 400 fifteen-second videos a night just as snugly. Second, retraining on clean data restores only part of the capability; a representational scar remains. A general-purpose learning system, fed an engagement-optimized diet long enough, loses reasoning ability in a dose-dependent, hard-to-reverse way. Machine or human, the conclusion reads identically - uncomfortably so.

On the distribution side, AI sharpens every mechanism in section 2: next-generation recommender models predict behavior better, personalize deeper, and grip harder. The counter-promise - AI as a filter helping users escape the junk - is entirely feasible in theory, and deserves polite skepticism in practice: who funds that filter's development, and whose wallet does its objective function optimize? As long as the answer remains "advertisers", AI-the-filter and AI-the-bait will continue to be one product with two sales decks.

6 · The Money Map: Who Profits, Who Gets Left Behind

A trend only earns the name megatrend once the money has voted, and for the attention economy the 2025 vote count sits in plain sight on the ticker tape. The whole map reads with one rule: money flows toward whoever owns raw attention plus the behavioral data attached to it; money drains out of every middleman who owns exactly none of either. AI hasn't changed the rule - it has merely accelerated the executions on both sides.

The winners: those who own the auction house

The left-behind: middlemen who own no attention

7 · Self-Defense - and Its Limits

The "so what do we do" section of this genre usually collapses into harmless magazine listicles. The best way to avoid that: derive each measure directly from the mechanisms in sections 2 and 3 - every item below disables one specific mechanism, rather than appealing to willpower in general. The spinal principle: willpower is a resource that depletes daily, while environment design is on duty 24/7 - so fix the environment instead of arm-wrestling several thousand engineers.

8 · Whoever Controls Attention Controls Behavior

Assemble the pieces: an absolutely finite resource (humanity's waking hours), a trillion-dollar industry auctioning that resource in real time, an extraction toolkit optimized through experiments on billions of people, and now a technology that manufactures bait at near-zero marginal cost. The most neutral description of this structure: a market extracting a non-renewable resource, operating without permits, on a resource located inside the skull.

Attention determines what gets thought about; what gets thought about determines what gets believed; what gets believed determines what gets done. The battle for attention was therefore never merely about "wasted time": whoever wins the right to distribute attention holds the regulating valve of collective cognition - which products get bought, which opinions spread, which outrage detonates on schedule. The 2021 Facebook files showed that valve turned toward anger for three years, at three-billion-person scale, for the most banal reason available: an engagement metric.

Herbert Simon closed his 1971 essay with a thoroughly technocratic proposal: an information society must learn to allocate attention wisely, as it allocates every other scarce resource. Half a century on, humanity has answered that proposal by building a machine that auctions attention to the highest bidder, then naming the consequences its 2024 word of the year. The "wisely" part remains an open position - and like every hard role, the listed benefits are brief: you get to keep your own brain.