Tanks vs Trade, II: American lawyers are from Mars, Chinese engineers are from Venus

I've now read 4 or 5 reviews of Dan Wang's book Breakneck written by his friends, all of which start by enthusiastically recommending it before presenting the work as if Thomas Friedman was traveloguing on Substack.1 The short version of the argument is that the US is a country of lawyers blocking stuff, China is a country of engineers building stuff. Despite this, the summary goes, the two places have a lot in common (what distinguishes Americans and Chinese, from all others it seems, is pragmatic materialism).

I’ve read Wang’s letters over the years and found them long on interior monologue but short on ideas: the pay-off was never worth it. Henry Farrell’s positive review convinced me not to read the book for basically this reason, and it would be helpful to read his take before reading the below. (For “in the weeds” reviews, Noah Smith demonstrates that Wang’s empirical premise is incorrect at surface level; Jonathan Sine shows that it’s even less correct at deeper level. But it’s great anyway!)

I agree with Henry that there is significant utility in “process knowledge” — oversimplification: if you can turn a screw assembling an iPhone case you can also turn a screw assembling an EV — but it is an instrumental benefit, not a structural benefit. Process knowledge is specific to a system of production, it does not transform systems of production.

Henry has been interested in this topic since his dissertation research, during which he learned ways in which European industrial shops solved micro-level problems via relational knowledge-sharing. If you’ve read Henry’s work over the years you will know why these things matter to him, and I do not think he’s wrong to prioritize them.

These shops are mostly closed now because the problems they faced were structural. Solving micro-level problems through maintenance of process-knowledge did not lead to development, it spent resources on deprecated models.

What we learn from dead industries (e.g., of Northern Italy, or North Carolina) is that process-knowledge cannot “engineer the future”. Not under capitalism. Henry’s teabag-stapler in Bologna engineered a clever teabag-stapling machine, which was temporarily useful. But he did not invent the CNC machine, and once that technology was made accessible via the proliferation of computational power all of the process-knowledge involved in teabag-stapling lost its market value not just in the teabag sector but throughout mechanical industry.

All that's solid melts into algorithm.

The guy who invented the modern CNC machine (John Parsons) was called the “Father of the 2nd Industrial Revolution” by the Society of Manufacturing Engineers, a group which knew something about process knowledge. Parsons transformed the system of production not through creating a machine built for purpose — although he did do that — but through creating a technology which diffused through global industry (it is the domain of Chinese process knowledge today, and before them the Japanese and Germans), in the process redefining which forms of process-knowledge were valuable.

Henry summarizes the crux of the question:

If the U.S. and China ever did end up in a real confrontation, which would you prefer to dominate in: software like the U.S, or hardware like China? Dan’s answer is that algorithms on their own “don’t win battles.” I suspect that many people in national security would agree.

Many people in national security express confidence in their ability to win the last war, but do they know how to win the next one? There’s a reason why China doesn’t want the US’s hardware, and it is because of software. (That is a two-way street.)

Ukraine's answer is clear: hardware does not win battles on their own. The utility of Chinese artillery in a “real confrontation” is being directly tested, against a foe much less formidable than the United States, and the results are not particularly impressive. That is also the clear lesson of Gaza, where American bombs are inadvertently building a Palestinian state. And of Afghanistan, in which American bombs were no match for the combination of IEDs and patience. The US just dropped some of the biggest bombs ever engineered on Iran's most valuable process-knowledging facilities, but Israel did significantly more long-term damage with an algorithm and some cheap pagers. Not for the first time.

Strap in for Q-Day, in other words.

Being the best at teabag stapling or tank construction in the 1990s does not save you from CNC-powered scale economies nor algorithmically-guided drones in the 2020s. Turns out the tanks are deathtraps in the 2020s, they just get in the way, all of that engineering was a bridge to nowhere: that tank got blown up by garage technology, technology that can be purchased at retail. What will be the death traps of the 2040s? Fighter jets? Submarines? Will it be desirable to possess nuclear stockpiles if algorithmic worms can melt them down on your own territory?

Russia’s superior process-knowledge in tank construction relative to Ukraine gave it over-confidence that led to miscalculation. Pride cometh before the fall.

Similarly, the process-knowledge from Sputnik was not engineered into the World Wide Web; instead the USSR process-knowledged itself out of existence by never once creating any new product that was in demand (per Branko Milanovic). Process-knowledge is not creation, and it is thus a depreciating asset in the context of technological change. Process-knowledge can even be algorithmitized, to a significant degree, and then capital-for-labor substitution is only logical. Is it better to know how to work hard or to know how not to have to?

Lastly, the "rival great powers have things in common, actually" genre is well-worn, it usually obscures more than it enlightens. We can do better. Sean Starrs -- who barely escaped Hong Kong with his freedom, and thus has quite a different view of the potency of China's capitalist administrative legalism -- is more to the point: China's process-engineering has reinforced American power and boosted America's development, it has not reduced the US’s status but enhanced it. Structural power is structural.

The same is true in reverse: American structural power has reinforced the potency of Chinese process-knowledge so long as they operate within the structure. But their process-knowledge has little value outside of the system, and is thus subject to depreciation, as the Soviets learned.

I agree with Henry when he writes “There are aspects of context that simply cannot be represented well by quantitative data.” But his colleague John Yasuda, my former colleague, has tons of actual qualitative research — as opposed to memoir-as-theory — demonstrating China's inability to process-knowledge their way to high sophistication, because the needed feedback loops for that do not come only from “real life” implementation in an engineered environment.

That would not have surprised Susan Strange.2 It does not surprise Sean Starrs. If they are correct then China is more likely to assemble the future, not design it, unless they focus less on refining the wheel via the development of process-knowledge, and more on breaking the wheel to build a more powerful one using a new technology entirely. That is something they have yet to do. They have not yet transformed systems of production.

That probably requires… lawyers. Not only for the protection of intellectual property (including process-knowledge) against an extractive state, but to secure a rights-based society that encourages improvisation in the first place.

I have yet to read a review of Dan's memoir that engages with any of these themes. Instead, there's a Tocquevillean level of abstraction in these conversations that ignores the obvious fact that American engineers are currently destroying the world by operating under a "move fast break stuff" ethos. Some legalism!