From television to the feed, technology was never a neutral pipe — it quietly rewrites what we can think, what we remember, and how much of us is left. Each of these four books pries open one layer.
2026 · Book Recommendations · Issue 35
Technology is never just a tool — it is the medium through which we perceive the world, and a medium quietly rewrites whatever it mediates. These four each pry open one layer: Postman on how a medium's form decides what kind of truth an age can hold; Carr on how the Net rewires the brain at the neural level, decaying deep reading into skimming; Zuboff on how an entire economic engine mines human experience into a predictable, steerable commodity; Lanier on how these design choices flatten living people into interchangeable data. Four threads, one question: when technology reshapes the mind, how much of you is left.
| Book | Author | Year | The one thing it makes clear |
|---|---|---|---|
| Amusing Ourselves to Death | Neil Postman | 1985 | What an age can think is set not by content but by the form of the medium carrying it — TV turns everything into entertainment |
| The Shallows | Nicholas Carr | 2010 | The Net doesn't just take time; it rewires your brain on its own terms — the capacity for deep reading atrophies with disuse |
| The Age of Surveillance Capitalism | Shoshana Zuboff | 2019 | "Free" hides a business: your experience is mined into behavioral data, and what's sold is a prediction of your next move |
| You Are Not a Gadget | Jaron Lanier | 2010 | A philosophy is buried in software's design — when it treats people as interchangeable data, people really do get flattened |
Postman pushes McLuhan's "the medium is the message" one step further, into "the medium is the metaphor." Every medium carries an epistemological bias: it favors certain content and certain ways of knowing, and quietly excludes others. Print demands linear, propositional, refutable argument; television demands imagery, immediacy, emotional impact. A medium is not a neutral container — it filters what can even be expressed before content arrives.
So it is the form of the medium that decides what an age takes to be "truth." In the age of print, Lincoln and Douglas could debate a single question for seven hours and an audience could follow dense argument; television cannot sustain that — it rewards the telegenic face, the thirty-second line, the vivid emotion. Whatever cannot be made entertaining isn't censored; it is simply made invisible.
Here is the sharpest cut: the danger is not junk TV. Junk TV is harmless because it announces itself as junk. The danger is taking the serious — news, education, religion, politics — and repackaging it as entertainment. When the news is a variety show (an atrocity, then a breezy "Now... this"), the form itself teaches you that all information is fragmentary, context-free, and requires nothing of you. The real message of the medium is: everything is entertainment.
So the Orwell–Huxley contrast becomes the book's spine: Orwell feared those who would ban books and hide the truth; Huxley feared that no one would want to read a book, that truth would drown in a sea of irrelevance. Postman wrote about television in 1985, but his subject is form — which is why he reads today like a prophecy of the feed and the short-video age.
Postman trains all his fire on television and says nothing of the internet that followed; he sometimes idealizes the age of print as an era of pure reason, ignoring that print also spread zealotry and prejudice. His near-total dismissal of the cognitive value of visual media is also too sweeping.
Postman's mechanism — the form of a medium filters what you can even think, before content — maps straight onto choosing learning media for a child. The same fact carried by a short video, an interactive app, or a book trains three different kinds of attention. To try this week: write down the three capacities you want the child to grow (say, sustained focus, logical reasoning, imagination), and ask of each, "what does the form of my current medium reward?" — short video rewards passive switching, however "premium" its content. Swap at least one back to a medium whose form matches the goal (reasoning → re-readable text you can stop and think inside, not an explainer video). The same holds for your own information diet: don't quit it — first see which muscle each form is quietly training.
Carr opens with a confession: he could no longer finish a long book; two pages in, his mind would drift. The real culprit is not willpower but neuroplasticity — the brain remakes itself around the tools it uses repeatedly. He cites the study of London taxi drivers, whose years of memorizing streets enlarge the relevant part of the hippocampus: you grow what you use.
Deep reading is not natural; it is a hard-won cultural achievement. The linear book trained the brain for long, single-threaded, contemplative attention. The Net is the opposite kind of technology: hyperlinks, notifications, many open tabs — it is essentially an "interruption system." Every link is a small decision (click or not) that taxes the prefrontal cortex, leaving less for comprehension.
So it is a trade: the Net makes us better at some things (rapid scanning, spotting patterns among fragments, quick choices) at the cost of the deep-reading mode (sustained focus, deep memory, reflection). Working memory is the bottleneck — the Net floods it, so information never consolidates from working memory into long-term memory and knowledge structures. We feel well-informed and retain almost nothing.
One layer deeper is cognitive offloading: when we hand memory and navigation to machines, the underlying faculty withers (GPS versus a sense of direction). Carr is not moralizing about distraction — his point is that the medium physically reshapes the organ of thought, and the reshaping is essentially invisible to us. His verdict on himself: once a diver in the deep, now merely skimming the surface.
Carr's neuroplasticity argument rests on real research, but there is a leap from "the brain changes" to "it necessarily changes for the worse" — plasticity is neutral, and he selectively stresses the loss. The book carries a heavy nostalgia, and its reverence for deep reading verges on the moralistic; a decade on, "the Net only makes us shallow" also undersells the new kinds of cognition new media create.
Carr's most cutting mechanism — offload a cognitive faculty and it atrophies with disuse (GPS versus route sense, the calculator versus mental arithmetic) — is exactly the shadow side the "AI super-individual" must face. Outsource whole stretches of writing, ideation, and reasoning to AI, and the time saved is real, but the muscle is quietly wasting. To try this week: draw an "AI no-go zone" around a core skill — keep one weekly stretch of deep writing or reasoning entirely AI-free, think the problem through to the end yourself first, and only then let AI challenge your conclusion rather than generate it. The test is simple: AI should be your sparring partner in thought, not your stand-in. Whatever you can no longer do the moment AI is gone is exactly the capacity you're losing.
Zuboff's central discovery comes from studying Google's early history. Under the pressure to make money after the dot-com crash (around 2001), Google realized that the "data exhaust" users left behind — search logs, click trails, which it had treated as waste — could predict behavior. She names this excess "behavioral surplus," and a wholly new business is born.
She lays bare the whole pipeline: human experience → behavioral data → (fed into machine intelligence) → prediction products → behavioral futures markets. Your experience is mined into a surplus far beyond what "improving the service" requires, processed into predictions of your next move, then sold to advertisers and others betting on your future behavior. The customer is never you — you are the free raw material. In her words, you are not even the product; you are the discarded carcass, and the product is the prediction.
Then the escalation: predicting is not enough; the most valuable predictions come from nudging and shaping behavior so behavior becomes more predictable (the Pokémon Go herding of real-world crowds, the social-platform emotional-contagion experiments). Zuboff calls this "instrumentarian power" — unlike totalitarianism, which controls the body through violence, it tunes, herds, and conditions behavior at scale, aiming at "guaranteed outcomes."
The crux is a radical asymmetry: they know everything about us, while their operations are deliberately designed to be unknowable to us. This is not a few bad actors or a bug — it is the economic logic itself. Just as industrial capitalism took nature as free to exploit (with climate as the price), surveillance capitalism takes human nature and experience as free — and the casualty is the private, self-possessed "I." She insists throughout: this is not "technology's fault" but a specific economic order that could have been otherwise.
At seven hundred pages, thick with coined terms ("instrumentarian power," "behavioral surplus"), the book is a high barrier and often repetitive. Zuboff attributes everything to one wholly new economic logic, underplaying state surveillance and the continuities with older capitalism; critics also find her too pessimistically deterministic about whether individuals truly have no agency.
Zuboff's mechanism is a mirror for the "AI super-individual who wants to build products": an AI product's most tempting revenue path is often to mine user interaction into behavioral surplus — because it is nearly free and works fast. To try this week: run a "surplus audit" on the AI product you're building or planning — list every category of user data you collect and ask of each, "is this to make the service better, or to predict/steer the user for money?" The ratio is your gauge of how close you are to surveillance capitalism. The reverse is also an opportunity: in an industry where surveillance is the default, "we don't mine you" can itself be a moat — exactly the kind of differentiation Thiel means by doing what others won't.
Lanier's weight comes from being an insider — a pioneer of virtual reality. So his critique is not Luddism but a critique of design philosophy. Web 2.0's design choices (anonymity by default, worship of "aggregation" and the "hive mind," treating content as ownerless fragments) presuppose deep down that the crowd or the algorithm is wiser than the individual — what he calls "digital Maoism." Step by step, this presupposition flattens the person.
The book's deepest concept is "lock-in": an early, somewhat arbitrary software design decision, once adopted, gets frozen and then permanently constrains everything built on top of it. His example is MIDI: it encoded music as keyboard-style discrete note on/offs, unable to represent the continuous slide of a voice or a violin — yet MIDI locked in, and the very sound of digital music was defined by it in return. Software lock-in becomes a philosophy imposed on people: once a system models "a person" as a checkbox profile, millions really do trim themselves to fit the model.
Hence his rallying cry — "Information is alienated experience." What he means to refute is the ideology that treats data and information as alive, as more important than people. Information has no meaning apart from a consciousness that experiences it. When we worship the aggregate (the hive voice of a wiki, the algorithmic feed), what we drain away is the one thing that can generate meaning: the individual point of view.
He also saw the economic side before Zuboff: making everything "free" actually destroyed the individual's ability (musicians, writers) to be paid, funneling wealth to the aggregating "siren servers." The design that flattens personhood also concentrates money. His prescription carries hope: design is a choice — we can absolutely build technology that elevates the individual voice and insists that people cannot be reduced to gadgets.
Lanier's style is jumpy and essayistic; his arguments often lean on intuition and anecdote rather than systematic evidence. Many of his worries about early Web 2.0 (the 2010 context) sit awkwardly against today's platform forms. His "individual genius vs. crowd wisdom" dichotomy has also been faulted for waving away real collaborative achievements like Wikipedia.
Lanier's mechanism grows only sharper in the age of AI: a large model is essentially "aggregation" — its output is the statistical average of its training data, flattening individual edges by nature. When everyone generates content with the same handful of models, output drifts inexorably toward the mean. To try this week: take a piece you recently made with AI's help and ask a Lanier-style question — "remove me, swap in anyone who uses the same tool: would the result be nearly identical?" If so, you are being flattened. The counter is not to abandon AI but to deliberately keep in the process what only you can supply: your particular depth in a field, your judgment, the taste you won't compromise — let AI amplify your individual voice rather than dissolve you into the average.
Don't just judge content quality — look at form. Log a day: in your information intake, how much time went to "infinite-scroll / autoplay" forms, and how much to forms with a beginning and end that you have to actively advance? The higher the former's share, the more you are training Postman's burlesque attention and Carr's skimming brain — however "hardcore" the content you consume.
Zuboff's test is simple: if you didn't pay money, go find out how it profits. If the answer is "ads / data," then its product design isn't aimed at serving you but at maximizing your predictability and time-on-site — your interests and its are misaligned at the root. As a user, this sets how far you should trust it; as a would-be builder, this is exactly where you can choose to be different.
This is Lanier's question in its AI-age form. If the answer is "nearly the same," your contribution was mere operation, and its value goes to zero as the tool spreads. What resists devaluation is the part no tool can replace — your depth in a field, your judgment, your particular sense of which problems matter. The stronger AI gets, the more that differential that is only yours becomes your only moat.