Dan Wang
Technology isn't written on the blueprint — it lives in people's hands, and a country can outsource those hands and then forget what it once knew how to make

1. One strange thing

In 2009, Mother Jones reported that the United States government had forgotten how to build a hydrogen bomb.

More precisely, how to build one component of one. That component used a material codenamed Fogbank, discontinued in the 1980s. Twenty-odd years later it was needed again, and the process was gone — not because files had been destroyed, but because how to do it had never been written down in full. It existed in a specific set of hands, and those hands retired. The government spent millions relearning the craft.

Dan Wang picked this old story up while studying semiconductors. He picked it up because it is a counterexample, and a sufficiently absurd one: a country capable of building hydrogen bombs can forget how to build hydrogen bombs.

What he asks from there sounds almost like a tautology: when we say a country "has" a technology, what is it that it has? Blueprints? Factories? Or something else?

His most extreme statement of it is a thought experiment that runs two thousand years in each direction:

We would be capable of making few meaningful advancements if a civilization from 2,000 years in the future were able to dump blueprints on us, just as the Pharaohs and Caesars from 2,000 years in the past would have been able to do nothing with the blueprints of today.2019 letter, 2020-01-01

If that sentence is right, then the entire public conversation about technological competition — patent counts, paper counts, R&D spending, export-control lists — is measuring the thing that doesn't matter.

2. Who he is

Dan Wang's parents both have deep roots in Yunnan; the family emigrated to Canada when he was seven, and he grew up in Ottawa and Toronto, where as a teenager he was in the Royal Canadian Army Cadets. He graduated from the University of Rochester in 2014; his About page gives his subject as philosophy, and he also writes of having sat four years of econ classes. He has written about the questions he was "too dumbfounded to pose in four years of econ classes" — the second proposition in section 4 of this page grew directly out of those unasked questions.

There is one item in his CV that does not look like a technology analyst's: the Philadelphia Museum of Art. Then Shopify, Flexport, and the Beijing-based research firm Gavekal Dragonomics. In 2017 he moved to China and stayed six years — Hong Kong first, then Beijing and Shanghai, then a flight to Yunnan two days before Shanghai locked down in 2022, and several months in Dali. In 2023 he returned to the US and finished his first book, Breakneck, at Yale Law School's Paul Tsai China Center; in 2025 the historian Stephen Kotkin recruited him to the Hoover History Lab at Stanford, for a reason he gives himself: to be closer to AI. The book was published in August 2025 by W. W. Norton in the US and Penguin in the UK, and was a New York Times #3 bestseller.

His site's subtitle is the secure transport of light. On his About page he explains why it is one of his favourite phrases: it refers both to optic cables, which make modern communications possible, and to semiconductors, which make modern electronics possible — and we can thank Alexander Graham Bell for allowing us to speak from one side of the Atlantic ocean to the other, "through coils of sunbeams under the seas." Then he adds: "Isn't that a wonderful image?" A man who names his blog that has announced his subject twelve years in advance: the objects that make the world run and that nobody looks at.

He has not stopped writing, but he stopped once. In July 2023 he wrote on the blog: "My 2022 letter will probably be my last. Now I have to devise another use for this site." Then he went off to write the book, and published no letter at all in 2024. On 1 January 2026 the 2025 letter appeared, with an explanation at the end: the Bay Area is so stimulating that he decided to restart.

3. What he has been wrestling with

The question he has not let go of in twelve years can be asked in one sentence: why has the digital world sprinted while everything else stopped — and whose fault is that?

How he came to it is documented, because his blog was doing this from its first posts.

In September 2014, one of the site's earliest pieces is him doing evidence-collation for Peter Thiel. The title is "Why is Peter Thiel pessimistic about technological innovation?" and the body is Thiel's own evidence, gathered from his writing and talks and sorted by sector: energy, transport, medicine, space. Wang barely speaks; he works as a research assistant. One of the passages he copied down reads (these are Thiel's words, not his):

It's being used to throw angry birds at pigs; it's being used to send pictures of your cat to people halfway around the world; it's being used to check in as the virtual mayor of a virtual nowhere while you're riding a subway from the nineteenth century.Why is Peter Thiel pessimistic about technological innovation? (quoting Thiel), 2014-09-10

That is someone a few months out of college organising another man's argument (he graduated in 2014; this went up that September). What is interesting is what he did next. He did not defend Thiel, and he did not rebut him. He went looking for a data source that could test the claim independently.

In September 2015 he read an odd little book: The Wonderful Future that Never Was, the science fiction author Greg Benford's collection of technological predictions from Popular Mechanics between 1903 and 1970, laid out with hand-drawn futurist art. His favourite is a 1951 personal helicopter: "This simple, practical, foolproof personal helicopter coupe is big enough to carry two people and small enough to land on your lawn. it has no carburetor to ice up, no ignition system to fall apart or misfire…" What he loves, he says, is not the helicopter but the earnestness with which ease is promised.

Then he counted. The conclusion fits in a line: most predictions made before 1950 were realised, most made afterwards were not, and the exception is anything to do with communications. Then comes the question that framed the following decade of his work:

"High-tech" now refers almost exclusively to communications technologies, and the "tech industry" has become shorthand for software businesses centered around Silicon Valley. Isn't that linguistic evolution a concession that the other fields have stagnated?Technological predictions of the twentieth century, 2015-09-08

This is the first time he has evidence of his own. In the same piece he raises a follow-up he returns to for years: what would the world look like if just one more field had succeeded the way communications did? His candidate answer is energy — many of the failed mid-century predictions would work out if energy really were too cheap to meter.

At this point the problem is clear but he is still missing one thing: a mechanism explaining why it stopped. He found the mechanism in the semiconductor industry, which is the first proposition below.

One clarification is essential to reading this man correctly: he is not writing American declinism. He says explicitly that he doesn't think manufacturing has special moral worth, and acknowledges that most manufacturing work is unpleasant and even hazardous. What he wants is something else:

I'm interested in industry because I see the maintenance of an industrial base as a precondition to building the science fiction technologies of the future.How technology grows, 2018-07-24

He lists what he wants without embarrassment: energy too cheap to meter; colonies on Mars and beyond; re-forestization of our deserts; nanotechnology that lets us print basic materials. His score for where humanity currently stands: less than 3 percent of where it should ultimately be.

4. His ideas, told as how he found them

Proposition one: technology lives in hands, which is why it can be forgotten entirely

Around 2018 he was studying semiconductors. Semiconductors are a convenient specimen because in that industry all three forms of technology are visible to the naked eye. TSMC, Intel and Samsung each push more than ten billion dollars a year at the frontier, and what they are buying is three things: tools, instructions, and process knowledge.

Tools and IP, he says, are the easy ones to observe. What is process knowledge concretely, in a fab? His examples are not mystical at all: how to store wafers, how to enter a clean room, how much electric current should be used at different stages, and countless other things.

This kind of knowledge is won by experience. Anyone with detailed instructions but no experience actually fabricating chips is likely to make a mess.How technology grows, 2018-07-24

Then comes the kitchen analogy he knew was good and that has been quoted everywhere since. It works because it explains in one breath why technology transfer is so hard: you have been given the blueprints and the equipment, and it still doesn't work.

So far this is an old idea in new clothes — tacit knowledge predates him. His actual move is to invert the status of tools and IP.

Instead of seeing tools and IP as the ultimate ends of technological progress, I'd like to view them as milestones in the training of better scientists, engineers, and technicians.ibid.

This is a complete reversal. Normally we say: train people in order to make things. He says: making things is proof that people have been trained to a certain level. The invention is the receipt, not the goods.

Follow the inversion and a series of conclusions come out on their own.

First, "the factory left but the high-value work stayed" is an invisible loss. What left was not the low end; it was the bottom half of the learning loop. His line here is the coldest in the proposition: when firms and factories go away, the accumulated process knowledge disappears too.

Second, he found a hard test for the claim, and the evidence is an absence. He did not compare GDP or patent counts — he looked at what America doesn't have. The answer: high-end precision manufacturing equipment. Factory automation systems, machine tools, robot arms, production machinery of all kinds; the most advanced suppliers are in Japan, Germany and Switzerland.

I don't see enough Americans being troubled by the idea that America isn't making advanced industrial robots. It might be fine to think that robots will be doing all the manufacturing work in the future; but someone has to build these robots, and own the IP of advanced robotmaking, and for the most part, that someone is not the US.ibid.

And the causal claim he offers is a single question: how do engineers work on the design of automation systems if they don't have exposure to industrial processes? This is his most characteristic posture: not piling up data, but pointing at a position where something should be and isn't, then asking why it's empty.

Third, he offers a positive solution, and it comes from a wooden building in Japan. Ise Grand Shrine is completely torn down and rebuilt every twenty years, and has been for 1,200 years. He likes this not because it is quaint but because it is a deliberately engineered knowledge-preservation device:

These people understand that it's too difficult to write down every instruction necessary for building even a single wooden structure; imagine how much more difficult it is to create instructions for a machinery part, or a chip.ibid.

He also copies down what the older generation says when handing over: "I will leave these duties to you next time." Regularly tearing down and rebuilding a wooden temple, he concedes, might not sound like a great use of time — but he isn't sure the locals' priorities are entirely screwed up here.

Fourth, this proposition gave him a durable irritation with economics, and the wording gets steadily less polite. In 2017 he reports that when he brings thoughts about resisting industrial decline to economists, the responses don't satisfy him: either tautology ("By definition, outsourcing improves on the status quo") or arithmetic (comparative advantage, Ricardo). By 2021 the same complaint sounds like this:

That static model constructed by economists with the aim of seducing undergrads has leaked out of the lecture hall and morphed into a political justification for only watching as American communities of engineering practice dissolved.2021 letter, 2022-01-01

Fifth, and this is the uncomfortable part: the same proposition is a bull case for China. He spelled this out in 2019, unambiguously. If knowledge follows production, then wherever the world's goods are made is where the marginal learning happens.

Today, Chinese workers produce most of the world's goods, which means that they engage more than anyone else in the technological learning process.2019 letter, 2020-01-01

He uses the iPhone as his case, and explicitly concedes the opposing account first — Chinese workers do mere assembly, of mostly foreign parts, while Apple keeps the profits: "That story is true, but it misses a great deal." Three things it misses: even if most of the workforce learns little, a few thousand line engineers become the world's greatest experts in electronics assembly; Chinese brands plugged into the same supply chain and came to make up around 40% of global smartphone sales; and the Chinese share of added value per phone rose from 4% to 25% over a decade.

So the proposition has a symmetry that is awkward for him: one law explains both why America can't build robots and why China will eventually build chips. He did not hide the second half because it was unwelcome. That is the single most creditable thing on this page.

Proposition two: imagination is an input to growth, not a result of it

This is the piece he likes best and also the piece he considers worst written. His own words: "I consider Definite Optimism as Human Capital to be my most creative piece. Unfortunately, it's oblique and meandering." So a year later he wrote another one — the one above — to lay the premises out straight.

It starts with someone else's short essay. The economic historian Anton Howes asks: is innovation in human nature? That is, do people carry some natural stock of innovative capacity, ready to burst out when conditions are right? From the lives of British innovators, Howes' answer is: not really.

Howes' framing, Wang says, finally let him ask the questions he had been too dumbfounded to pose in four years of econ classes: how do beliefs about the future affect long-term economic growth? What role do optimism and pessimism play? How important are imagination and drive to growth?

Then comes the sharpest question — why do we treat innovative capacity as some kind of fixed stock, constant across time and between countries, ready to be activated as soon as policymakers have finally gotten around to legislating the right conditions in place?

His positive claim is a substitution: treat "having a concrete picture of the future" as a line item in human capital, alongside education levels.

In addition to education levels, human capital models should consider factors like optimism, imagination, and hope for the future.Definite optimism as human capital, 2017-08-07

There is a qualification here without which he reads as a motivational speaker. The "positive" vision he wants does not require the future to be pleasant:

When I say "positive" vision, I don't mean that people must see the future as a cheerful one. Instead, I'm saying that people ought to have a vision at all: A clear sense of how the technological future will be different from today.ibid.

To have a vision you first need imagination. And his account of how to get imagination is thoroughly un-economist: "I mean to promote imagination by direct injection."

His evidence is two obscure facts, and both are about scars.

The first is in Japan. A Bloomberg piece: unemployment at 2.8%, so Japanese millennials should have their pick of employer, and almost none of them want to move — what they want is not high wages but perfect stability. The article attributes the risk aversion to lingering fears of the previous downturn. He also copies a headline from the same outlet: "Japan has the world's gloomiest millennials." His control group is China, where a recruiter describes a "fearless" employee base, because two generations of Chinese have never experienced a recession and job-hopping doesn't register as a risk.

The second is in America, and it is harder. He had written a separate piece asking: computer science is so lucrative, why do so few people major in it? Many responses told him he was missing the hangover from the 2001 dotcom bust. He accepts the explanation and points out that it is more pessimistic than his own was:

It means that it took nearly 15 years for the number of graduates to recover to a previous high after a single, brief recession.ibid.

Then he runs an uncountable tally: because of that talent constraint, how many fewer startups were founded? How many larger companies could have been better staffed? How many VCs could have generated higher returns? And the summary — recessions leave deep psychological scars that hurt risk appetites for a much longer time than the downturn itself.

This corollary has a policy consequence, and he names it: it weakens the Austrian case about recessions. If a recession really leaves a fifteen-year psychological scar, then treating recessions as healthy corrections that improve capital allocation doesn't hold, because they do a great deal of unseen harm. If we treated definite optimism as a function in human capital and productivity growth, he says, we could be slightly more rigorous in considering the broader effects of recessions.

Then he does the thing that makes many people file him as an eccentric: he treats science fiction films as a macroeconomic issue, in earnest. The chain is short: imagination is an input to growth; the cheapest way to imagine a future is to see one in a book or on a screen; and the dominant mode in modern sci-fi movies is dystopia, perhaps even nihilism. Therefore:

Today, the contrarian project is to present an earnest, joyful vision of the technological future. That kind of undertaking will invite mockery from commentators posting essays on Medium, but one can hope that younger, less jaded people will find attractive that daring imagination.ibid.

His reshoot list is the most entertaining passage on the site, and it is also a demonstration of his method — he doesn't say "sci-fi is too bleak," he hands you a specific work order. Elysium, with a focus on the logistics behind the smooth functioning of a satellite habitat. Ex Machina, featuring an artificial intelligence whose greatest desire is not acceptance in human society. Her, in which artificial intelligences decide to improve humanity, instead of migrating en masse to a higher spiritual plane. Star Wars with no wars and no Jedis — and therefore no Sith — that instead draws out the gains from interstellar exchange under oversight of the Trade Federation. He appends a note to his own proposal: "Bonus points for working out rules that would govern countervailing duties and state procurement under a galactic trade regime."

Immediately afterwards he admits: "I confess I don't know how to make these stories exciting." And leaves the list standing anyway. That posture — I know how this sounds, I'm not withdrawing it — is half the reason the blog is fun to read.

Its most operational landing is in his 2018 letter, and it is the most easily overlooked good idea on the site. He begins with Moore's Law and de-mystifies it: Moore's Law is not some natural law built into the fabric of the universe, designed to self-execute without a bit of engineering effort; it requires a massive industrial undertaking. The semiconductor industry set a benchmark early on that seems kind of arbitrary today, and made a collective effort to execute against it. He quotes a description he likes: Moore's Law is "a clock that has become a chaotic attractor for innovation."

Which prompts a genuinely odd question: can exponential progress be replicated by branding?

Moore's Law turned from a neat backwards-looking observation into an obligation for the entire chips industry to keep improving… But maybe there's a handful more that are, and they await a nice label that will concentrate minds, mobilize capital, and attract talent to keep improving.2018 letter, 2019-01-01

He calls coming up with that label a kind of low-hanging fruit. This is the step where definite optimism stops being an attitude and becomes engineering: if what a field lacks is a public commitment that aligns everyone, then manufacturing one is concrete, actionable work rather than an exhortation to cheer up. He also names where he most wants it to happen — energy, because energy is upstream of everything in the economy: think how much more room AI would have to play with if energy costs were measured in cents rather than dollars.

The proposition lands on a very quotable line that is in fact a programme of action:

Call me a romantic, but I'd like everyone to think more about industrial lubricants, gas turbines, thorium reactors, wire production, ball bearings, underwater cables, and all the things that power our material world.ibid.

The sentence immediately following is a rule he set himself: he abides by a strict rule never to post or tweet about current political stuff; instead he tries to draw more attention to the world of materials. He manages attention as an exhaustible resource, and the instrument of management is a prohibition. In the nine years since he wrote it, the rule has largely held: not one of the site's forty-odd posts is about the political news of its week.

Proposition three: the engineering state and the lawyerly society — why he had to move house to think it

Both earlier propositions were formed by 2018. Together they leave an obvious hole: if process knowledge matters this much, why did America let it drain away? If definite optimism is this decisive, how does a country lose it collectively? Neither question is technical. Both are political. And his situation at the time couldn't answer them — he was in America, able to see only the stagnant half.

His solution was to move. And he had described the move in advance, as advice:

Another possibility is to leave the US, as I did. I understand that this is not an option for everyone, but for some young people at least, maybe this is a good time to spend time abroad. Get out when unproductive debate is the ambient national mood, and move to a more sane environment where you can really learn something instead.How technology grows, 2018-07-24

He lived in China six years and wrote one letter a year. The letters are not travelogues; they carry a hard analytical spine, and the judgements on that spine changed direction as evidence came in — which is the most trustworthy thing about him, and worth walking through in order.

The 2019 letter: his verdict on Chinese technology is "paper tiger." This has to be stated plainly, because he was later filed as a China bull, and he opened as a bear. His two claims that year: China's technology foundations are fragile, which the trade war has made evident; but over the longer term he is constructive. On the short term he is unsparing:

In many ways, China's technology success is too much like a paper tiger, impressive in appearance but in reality not so powerful.2019 letter, 2020-01-01

In the same piece he dismantles a frame everyone was using. All the commentary on Chinese innovation was counting patent registrations, R&D spending, journal publications —

But these inputs are irrelevant if they don't deliver output, and it's not clear that they often do, neither in China nor anywhere else.ibid.

And his verdict on the hottest sector of the moment was, in 2019, distinctly against the grain: "I find it bizarre that the world has decided that consumer internet is the highest form of technology." He goes further and says Facebook and Tencent might be net-negative for technological development, on the grounds that they pull smart kids out of materials science and semiconductor manufacturing and into ad optimisation and game development.

The 2019 letter also contains a prediction that the following years bore out:

The US responded to the rise of the USSR and Japan by focusing on innovation; it's early days, but so far the US is responding to the technological rise of China mostly by kneecapping its leading firms. So instead of realizing its own Sputnik moment, the US is triggering one in China.ibid.

The 2022 letter: he learns to look at states from the opposite direction. That year Shanghai locked down; he flew to Yunnan two days beforehand and stayed months. That letter's first sentence is seven words long:

Mountains offer the best hiding places from the state.2022 letter, 2023-03-04

The best book he read that year was James C. Scott's The Art of Not Being Governed, about the hill peoples who fled states into highland Southeast Asia over millennia. And in Yunnan he found a perfect contemporary specimen of the theory — worth telling in full, because it is the best single example on the site.

A village near Dali. In 2015 Xi Jinping visited and admonished local officials to clean up the nearby lake. Officials jumped to implement, and one of their measures was to direct all water from the mountains to flow into the lake. Villagers who had been using spring water from those mountains for drinking and food production now had to drink treated water.

It was not that they objected to cleaning up the lake. It was that a word from the top leader prompted local officials to deny them the best water in China, while making an at-best-minimal contribution to the cleanup.ibid.

The villagers' response was to climb further up and lay new pipes to the village temple. And the ending is the point of the story:

Local officials came to the village temple not with hammers to smash these pipes, but with their own jugs for filling up.ibid.

They taught him to bring his own jugs. This story matters not because it is charming but because it is both faces of his later "engineering state" verdict at once: enormous top-down capacity to reshape systems, paired with very low resolution on local reality — and an implementation layer made of people, who find the gaps themselves.

In 2023 he left China and entered Yale Law School. This is where the mirror image appeared. He says so plainly: the idea of the lawyerly society became obvious when he returned to the US in 2023, because the Paul Tsai China Center set him inside the Yale Law School. What he saw was that elite law schools, now and in the past, fashion the easiest path for the ambitious into the top ranks of the American government.

Hence the sentence that says both halves at once — China an engineering state bringing a sledgehammer to problems physical and social, America a lawyerly society bringing a gavel to block almost everything, good and bad. And hence the diagnosis: the dominance of lawyers in the American elite has helped transmute the United States into a litigious vetocracy.

Here is something only a chronological reading surfaces, and it is the finding this page most wants you to keep.

The lawyer is not a villain who arrives in 2023. In that 2018 essay on process knowledge, arguing that services can't be relied on, his leading example is litigation:

Too many service jobs are meant to cancel out the efforts of other service jobs, for example in litigation, where a plaintiff's lawyer creates a job for the defendant's lawyer.How technology grows, 2018-07-24

In the same essay, inside a parenthesis, he connects another wire: having noted that Britain's policy of prioritising debating and thinking has failed it, he adds — "(I would also suggest that this liberal arts emphasis on argument turns universities into incubators of Girardian terror.)"

Which is to say: zero-sum service work, the elite tournament driven by mimetic desire, and a lawyer-dominated vetocracy are one thing in three sizes. In 2017 he wrote a whole essay arguing the American college is a perfect incubator of mimetic contagion, concluding that it produces people whose main preoccupation is to impress their peer class. In 2019 he ran that wire to Hong Kong's elite. In 2025 he ran it to the American system entire. He took eight years to finish the line, and every segment of it is on the blog, checkable piece by piece.

Finally, the most important half of this proposition is its negative: he does not think America should copy China. He gives his heaviest chapter to the one-child policy and says he had been completely unprepared to study the brutality of its enforcement, which was only possible through mass sterilizations and forced abortions. His conclusion:

The problem is that China's leadership just can't stop at physical engineering. Sooner or later, they treat the population as if it were another building material, to be moulded or torn apart as the circumstances demand. That's why America shouldn't look to China as the model.Breakneck (book announcement), 2025-07-25

His characterisation of the Party is on the same page, in a typical compression: "My handy formulation of the Communist Party is that it is a Leninist Technocracy with Grand Opera Characteristics — practical until it collapses into the preposterous."

And the prescription he writes for America is a deliberate anticlimax: America doesn't have to become China. Reaching the construction cost levels of France, Japan or Spain, he says, would be sufficient. His verdict on the country rests on one line: too many parts of America feel like the well-preserved ruins of a once-great civilization.

5. How he thinks

His conclusions will date, and he is his own evidence: in 2019 his verdict on Chinese technology was "paper tiger," and in 2025 he writes that on AI China is behind the US "but not by years." That is not self-contradiction, it is revision under evidence — which is exactly why the conclusions aren't the thing to memorise, and the method is.

1. Moving house as a research method. The least reproducible and most memorable item here. He did not come to understand China by reading more China research; he went and lived there for six years, relocating his desk from San Francisco to Hong Kong, Beijing, Shanghai. And after returning, his second insight — the lawyerly society — also came from a change of position: not from reading legal scholarship but from being seated inside a law school. He stated the underlying principle generally in his very first letter:

One can learn more by traveling to new places, being social in different ways, reading new types of books, changing jobs or professions, moving to a new place, by doing better and by doing more.2017 letter, 2018-01-01

That letter's main claim is thoroughly counterintuitive: learning ought to accelerate over time, not decelerate. Analogising the growth rate of knowledge to a country's GDP growth rate, he says, is an analytical error. His two reasons are both lifted from industrial organisation: knowledge has network effects (learning a new fact increases the value of facts you already know) and positive returns to scale (the more you learn, the easier it is to learn, because you can skip what you've read).

2. Look at outputs, not inputs. His most useful blade, because it dismantles nearly all public discussion of technological competition. Patent counts, R&D spending, paper counts, PhDs granted — all inputs. His position is that inputs are irrelevant if they don't deliver output. The blade cuts his own arguments too: his long-run bull case on China is not how much money China spends, it's how many products are passing through Chinese hands.

3. Use absence as evidence. His most distinctive move. To argue American industrial decline he doesn't count employment (though he does that too) — he asks what America doesn't have. The answer is high-end industrial robots. An absence is harder to spin than a presence, because nobody issues a press release for an industry that doesn't exist.

4. Refuse to stop at the abstract layer. Almost every judgement in his writing is trailed by something concrete you could go and look at. Preserving communities of engineering practice → Ise Grand Shrine, torn down every twenty years. The limits of state capacity → the pipe those Dali villagers ran further up the mountain, and the local officials arriving with their own jugs. Chinese consumption downgrading → a group of girls at the Ritz-Carlton in Shanghai photographing each other over cakes for over an hour. You will remember those three; you will not remember the sentence "he argues well from concrete cases." Which is the thing his own writing is demonstrating.

5. Manage attention as a depletable resource, by prohibition. He never writes about current politics, a rule he set in writing and has kept nearly intact for nine years. His first act after the 2016 election was to cancel his New Yorker subscription, and he says he is still proud of himself for having quickly taken decisive action. That isn't a joke, it's a method: he holds that attention spent on political scandal is wasted and permanently damaging — his word is that the minds of many talented young people will be "disfigured" by the obsession.

6. Read the most tedious primary source to the end. The least romantic item here, and the one that says most about what kind of analyst he is. In 2020 he read every issue of Qiushi — the Party's flagship theory journal, published twice a month — for the whole year. His description for outsiders: Qiushi reads like a cross between the New Yorker and the Federal Register. And he gives his own reason for being able to stand it: his main professional activity at the time was studying the US sanctions regime (Commerce's Export Administration Regulations, Treasury's IEEPA-based authorities), and Party speeches and US regulations are the same genre — both made up of arcane, formal language that makes reference to more obscure texts, which themselves hint at still more distant and terrible truths. He adds a verdict in passing: US sanctions lawyers, he suspects, can have a splendid time with Qiushi.

This item belongs with the five above as one kit: look at outputs not inputs, use absence as evidence, go and live there, and read the official texts nobody wants to read, page by page. Assembled, those four explain why his judgements keep better than most China-watchers'.

7. Use low frequency against the news cycle. One letter a year, eight to fifteen thousand words. The format is itself an argument: if you only write in December, you won't write about things that mattered only in one week of March. His own account of the form is that the core of the annual letter is to reflect on what he saw that year, mostly around China, in which he tries to present a lot of analytical insight combined with some element of narrative.

8. He discounts his own framework, explicitly and in writing. "The engineering state versus the lawyerly society is not a grand theory to explain absolutely everything about the U.S. and China." That is his own sentence on his own book announcement. The positioning he offers instead: the book is rooted in his experience of living in China from 2017 to 2023, and he offers the framework to make sense of the recent past and think about what might come next. Labelling a framework a good-enough lens rather than the truth is the hardest of these methods to copy.

9. He left an honest record about a new tool, too. He states that he made it a point not to use AI for any part of Breakneck. And his reasons for using it now are specific: he treats AI as a companion for the intellectually peripatetic — go to an art exhibit and then interrogate it about what you saw, go to a string quartet and have a good conversation about what you heard. He even gives an actual question he asked it: "Why did the Spanish develop such a virulent Inquisition while Austrian Catholicism feels relatively cheerful?" That example says more about what he finds AI good for than any general pronouncement could.

6. Where he might be wrong

First, the two halves of his framework are not equally thick, and the thin one is America. This is the most concentrated criticism in the published reviews, with two named reviewers making the same point from different angles.

Dylan Gyauch-Lewis, reviewing in The Sling, raises several specifics: the points where the American "lawyerly" counterpart is brought into the mix are more tenuous; the book does not establish when America transitioned from an engineering society to a lawyerly one, nor how to determine the threshold where legalism becomes overburdening; and by Wang's own lawyers-per-capita measure, several other nations have more lawyers per capita than the US yet don't face comparable building challenges — a criticism aimed directly at his instrument. Beyond that, Wang barely addresses the necessity of the lawyerly society in preserving and defending the public from powerful interests or corrupt political regimes.

Charles Yang, reviewing in American Affairs, says that the book lacks the rigor of hard data about the impact of engineers in government; that the United States has never had large proportions of engineers in government — and if not, the narrative of an America that fell from engineering society into lawyerly society needs rewriting; and that some characteristics Wang describes are to some degree characteristic of authoritarian regimes in general rather than of engineer-rule specifically. That last objection carries the most weight, because it points to an alternative explanation: perhaps the brutality of the one-child policy came not from the enforcers being engineers but from the enforcers being unchecked. If so, "engineers versus lawyers" is the wrong axis.

(This page's account of those two reviews comes from their published texts; the summaries above are paraphrase, not verbatim translation. And Breakneck itself has not been read for this page — so what is reported here are the checkable critical points in the reviews, without adjudicating whether they land against the book.)

Second, the most honest criticism of him was written by him, and nobody has put it more sharply. On definite optimism, the whole argument has a fatal circle: optimism drives growth, but optimism is mostly a product of growth. He did not concede this under pressure; he volunteered it in his next essay — he concedes the circularity to the argument, because optimism is probably endogenous to growth.

His defence is equally candid: instead of waving his hands around to claim that growth comes down to luck (the position he attributes to Bryan Caplan), he would like to wave his hands around in a different way and say it comes down to a choice to be optimistic. "Wave my hands around in a different way" is his own phrasing. It's hard to accuse a man who describes his central argument that way of concealment — but the circle is still a circle, and the proposition still has no falsifiable form.

Third, his vantage has a structural tilt, and that tilt is the same thing as his greatest advantage. "I lived there for six years" is an extremely strong authority claim and an extremely hard one to argue with. When he disagrees with someone who has never been, he always holds that card. And eyewitnesses have eyewitness biases: the China he saw is the China of Beijing, Shanghai, Shenzhen and Dali — and he concedes in the 2021 letter that the Greater Bay Area is "a bit more of a mystery" to him, because he lived in the failing part, Hong Kong, rather than the growing part, Shenzhen.

Fourth, a corollary he never chased down. A good part of his case against services rests on zero-sumness: lawyers, compliance, advertising, lobbying, cancelling each other out. But his own work — technology analysis, research institutes, book-writing, podcasts — sits in that category. He has never defended it, nor explained why analytical work escapes the charge. This isn't fatal (his claim is that too many service jobs are zero-sum, not all of them), but a man who weighs exposure to industrial processes this heavily owes a sentence about where his own job sits in his system.

Fifth, he is sharper on his own milieu than on himself, and that belongs on the credit side of the ledger. His 2025 verdict on Silicon Valley is unsparing: tech's movements are more herdlike, and when political winds shift, people fall in line — he calls this Silicon Valley's soft Leninism. He pairs San Francisco with Beijing as the two most insular cities he has lived in, on the grounds that they are places where people are willing to risk apocalypse every day in order to reach utopia. And on his industry's hottest conviction he draws a clean line: he is not a skeptic of AI, only a skeptic of the decisive strategic advantage, which treats awakening the superintelligence as the final goal. AI discussion collapses into the utopian or the apocalyptic too fast, he says — it's a Pascal's Wager, in which we're sure that the values are infinite, but we don't know in which direction. Then comes his answer to all of it, a line fit to be a slogan:

Call me a romantic, but I believe that there will be a future, and indeed a long future, beyond 2027. History will not end. We need to cultivate the skill of exact thinking in demented times.2025 letter, 2026-01-01

7. Sources

• How technology grows (2018-07-24)The full argument for process knowledge, and the most-cited thing on the site. The three forms of technology, the kitchen analogy, Ise Grand Shrine, why America doesn't build industrial robots, why services can't be relied on (litigation as the leading zero-sum example is here), and the sentence where he concedes his own circularity. He says he wrote it to lay the previous piece's premises out straight.
• Definite optimism as human capital (2017-08-07)Self-assessed as "my most creative piece" and simultaneously as "oblique and meandering." The core claim about imagination as human capital, the two scar-shaped pieces of evidence (Japanese millennials, American CS graduates), the sci-fi reshoot list, and the rule he set himself never to write about current politics.
• 2019 letter (2020-01-01)The densest of the eight annual letters analytically, and the baseline for understanding his China judgements. "Paper tiger"; inputs are not outputs; consumer internet is not the highest form of technology; the Chinese share of added value per iPhone going from 4% to 25% in a decade; "the US is triggering [a Sputnik moment] in China." The second half turns to science fiction and Philip K. Dick, reading Hong Kong's elite as characters in a PKD novel.
• 2022 letter (2023-03-04)Written from Yunnan in the year Shanghai locked down; the first sentence is "Mountains offer the best hiding places from the state." A James C. Scott-inflected view of states, the Dali water pipe and the officials who came with jugs, and the personal source he later connects in the book: the suspicion of the state common to mountain peoples.
• Breakneck: China's Quest to Engineer the Future (2025-07-25)The book announcement, not the book. Original source of the engineering-state/lawyerly-society sentence, the "Leninist Technocracy with Grand Opera Characteristics" formulation, the one-child policy chapter, and the boundary he draws around his own framework ("not a grand theory to explain absolutely everything"). The second half reflects on writing, reading and eating, including several practical signals he uses to judge whether a book is worth reading.

Other pieces quoted on this page, in order: Why is Peter Thiel pessimistic about technological innovation? (2014-09-10, among the site's earliest posts), Technological predictions of the twentieth century (2015-09-08), The life of Philip Glass (2015-07-30), Laszlo Foldenyi's Melancholy (2016-10-16), How smartphones made Shenzhen China's innovation capital (2016-11-06), Yunnan: an unglobalized part of the world (2017-04-17), Why are there so few computer science majors? (2017-05-29), College as an incubator of Girardian terror (2017-06-25), 2017 letter (2018-01-01), 2018 letter (2019-01-01), 2020 letter (2021-01-01), 2021 letter (2022-01-01), China notes, July '23: on technological momentum (2023-07-18, the post announcing that the letters were over), 2023 letter (2024-01-23), 2025 letter (2026-01-01, the restart, on Silicon Valley and AI), and his About page (CV, and his own list under "A few other favorites on this site"). The two reviews: Dylan Gyauch-Lewis, The Sling, Charles Yang, American Affairs. Blog homepage: danwang.co.