Sam Seitz

Dan Wang posts extremely infrequently, but his commentary is always of an exceptionally high caliber. Here are a few excerpts from his most recent post that I found particularly interesting.

The internet is important, and we’re likely still underrating its effects. But I don’t think that we should let innovation be confined entirely to the digital world, because there’s still too much left to build. The world isn’t yet developed enough that everyone has access to shelter, food, water, and energy at a low share of income. Hundreds of millions still live in extreme poverty, which means that manufacturing and logistics haven’t overcome the obstacles of delivering cheap material comfort to all.

And I submit we can’t bring ourselves to calling it the “developed” world until we’ve built so many other things. We go to work in subways built in the ‘70s, guided by signals equipment put in place in the ‘20s. We’ve been moving more slowly across the planet ever since we decommissioned the Concorde, at a time when global travelers want faster access to major hubs. Are we sure that the developed world is not undergoing its own premature deindustrialization? When people bring up that the fact that the digital world has become very fun, I tend to think that the response smacks of “Let them eat iPhones.”


I’d like more people to consider that there are major positive consequences for high economic growth. When people have experienced a few years of high growth, they’re conditioned to expect more of it. That expectation increases risk appetites for both companies and individuals: people have seen their lives getting better in a hundred different ways in the last four decades, and they can be optimistic that more things will improve. They’d be more comfortable starting new businesses or trying on new careers, and these activities won’t even feel like risky events, because new opportunities have always been coming along.

All of this is on top of the fact that higher growth improves our capabilities to deal with various problems. If we’re accustomed to low growth, it’s more difficult to imagine a big improvement to our lives unless we can force a redistribution of wealth. But if we’ve experienced high income growth for a few decades, then it’s easier to imagine that we can make our problems go away because we keep accruing greater resources to deal with them.


Does better capital allocation always lead to technological acceleration? I don’t think so. In fact, I’d like to argue that some of the responsibility for the loss of process knowledge can be attributed to the US financial sector, consisting of both investors and financial analysts, with its emphasis on return on capital. (This is also something that Andy Grove brought up.)

My fundamental argument is that technology ultimately progresses because of people, and in particular the amount of process knowledge they’ve managed to accumulate. In my opinion, the US financial sector has underappreciated how important it is to have a deep pool of technically-experienced workers. It’s certainly much easier to identify and measure the stock of tools and IP instead of process knowledge, which exists in people’s heads. As a result, investors and financial analysts have systematically rewarded the firms that are most eager to reduce headcount, which they see as a cost. But just because we can’t straightforwardly measure process knowledge doesn’t mean that we should dismiss its existence.


I believe that technological progress is not inevitable, and that we have agency on how hard to push it forward. A doubling of transistor density every 24 months is not some foreordained natural law, gifted to us through heavenly benevolence. That sort of progress doesn’t happen unless we put a bit of thought into it. Moore’s Law is not a promise, but a challenge, and we’ve met it pretty well so far.

One day, we can throw up our hands and declare that we’ve had enough innovation in semiconductors. “The future is in services instead, not in this kind of toxic manufacturing work.” We can fire all the nerds, throw out all their books, and shut down all these fabs. Let’s say it takes a few years for us to come back to our senses. When we subsequently want to revive the industry, it may not be as simple as plugging in the machines, blowing the dust off of the blueprints, and then happily expect production to resume at prior levels. The hard-won process knowledge held by these engineers will have decayed, and the workers will have to relearn a bunch of things.

It probably seems like I’ve just excerpted the entire essay, but that is not the case! I therefore encourage you to read the entire piece (and all his other posts).