--:-- Dwarkesh Patel: Okay. Today, I have the pleasure of speaking with Brian Potter, who is an engineer and the author of the excellent Construction Physics blog, where he writes about how the construction industry works and why it has been slow to industrialize and innovate. It's one of my favorite blogs on the internet. I highly, highly recommend that people check it out. Brian, my first question is about the Line project in Saudi Arabia. What are your opinions?
--:-- Brian Potter: Okay, yeah. This will be in no particular order. One, it's interesting that this country, Saudi Arabia in the Middle East, is just so willing to do these big, crazy, ambitious building projects and pour huge amounts of money into constructing this infrastructure in a way that you don't see a huge amount in the modern world. China obviously does it huge amounts, some other minor places, but in general, you don't see a whole lot of countries doing these big, massive, incredibly ambitious projects. So on that level, it's like, "Interesting, glad to see that you're doing this." But the actual project is clearly insane and makes no sense.
--:-- I mean, just from the physical arrangement layout, there's a reason that cities grow in two dimensions. A one-dimensional city is the worst possible arrangement for transportation. The maximum distance between any two points is maximized. So from that perspective, it's clearly crazy. There's no real benefit to it other than perhaps in some weird hypothetical transportation situation where you had really fast point-to-point transportation, some sort of weird bullet train setup. Maybe it would make sense, but there's no reason to build a city like that. Even if you wanted to build an entirely enclosed thing—which again doesn't make a huge amount of sense—you would save so much material and effort if you just made it a cube. I would be more interested in the cube than the Line. But yeah, those are my initial thoughts on it. I will be surprised if it ever gets built.
--:-- Dwarkesh Patel: The cube, like from the meme about how you can fill all the humans in a cube the size of Manhattan?
--:-- Brian Potter: Yeah, something like that. If you're going to build a big giant megastructure, at least take advantage of what that gets you, which is minimum surface area to volume ratio.
--:-- Dwarkesh Patel: Why is that important? For temperature and stuff?
--:-- Brian Potter: Actually, I'm not sure how it would work with a giant single city in general. A lot of economies of scale come from geometric effects: as something gets bigger, your volume increases a lot faster than your surface area does. For an enclosed thing like a tank or a pipe, the cost goes down per unit you're transporting because you can carry a larger amount for a smaller amount of material. It applies to some extent with buildings and construction because the exterior wall assembly is a really burdensome, complicated, and expensive assembly. So for a building with a really big floor plate, you can get more area per unit of exterior wall, and all else being equal, be more efficient.
--:-- I'm not sure how that works with a single giant enclosed structure. Theoretically, on a small level, it would apply the same way because your climate control is a function of your exterior surface. At some level, you get more efficient climate control if you have a larger volume and less area for heat to escape from. But for a giant city, I actually don't know if that works. It may be worse because you're generating so much heat that it's now harder to pump out. The urban heat island effect and stuff like that—these cities generate massive amounts of waste heat. I'm trying to reach back to my physics classes in college, so I'm not sure of the actual mechanics. But in general, that's why you would want to perhaps build something of this size and shape.
--:-- Dwarkesh Patel: What was the thought process behind designing this thing? Scott Alexander had a good blog post about the Line where he said, presumably the Line is designed to take up less space and to use less fuel because you can just use the same transportation across the line. But the only thing that Saudi Arabia has is space and fuel. So what is the thought process behind this construction project?
--:-- Brian Potter: I get the sense that a lot of Middle Eastern projects are just really... they've had some amount of success in building big, impressive physical construction projects that are an attraction just by virtue of their size and impressiveness. There's a huge amount of stuff like this in Dubai. They have that giant clock tower in Jeddah that is similar—"biggest giant clock," building one of the biggest buildings in the world. So I think on some level, they're expecting that you can see a return just from building something that's really impressive and the biggest thing on some particular axis. It could be that they're just optimizing for "big and impressive" and maybe not diving into it more than that.
--:-- There's this theory I think about every so often called the "garbage can theory" of organizational decision making. It's that the choices organizations make are not the result of any particular rational process. Whenever a problem comes up, they reach into the garbage can of potential solutions, and whatever they pull out is the decision they end up going with, regardless of how much sense it makes. It was a theory mentioned by academics to describe decision making in academia, but I think about it especially with big bureaucracies and governments. You can just imagine the horse-trading process and how these decisions evolve, especially when there's such a disconnect between decision makers and technical knowledge.
--:-- Dwarkesh Patel: Tell me about your eBay arbitrage with designer clothes.
--:-- Brian Potter: Oh man, you really did dive deep. Yeah, this was a small business that I ran for a few years, seven or eight years ago at this point. Basically, a hobby of mine was high-end men's fashion for a while, which is a very strange hobby for an engineer to have, but there you go. For the enthusiasts, it centers around finding stuff for cheap because buying it new can be overwhelmingly expensive. A lot of times you can get it for very cheap if you're even a little bit motivated—it shows up on eBay or in thrift stores if you know what to look for. A lot of these clothes can last a super long time because they're well-made. Even if somebody has worn it for 10 years, it might be fine.
--:-- A lot of this hobby centers on finding ways to get really nice stuff for cheap, and a lot of it is based around eBay. But it was really tedious to find really nice stuff on eBay. You had to manually search a bunch of different brands, filter out obvious bad ones, and search for typos in brand names that were pretty reliably put in titles. I was in the process of doing this thinking, "This is really annoying, I should figure out a way to automate this process." So I made a very simple web app where you would search for, say, shoes, and it would automatically search the very nice brands of shoes, all the typos in the brand name, and filter out all the junk so you could search through the good stuff. I set up an affiliate system so anybody else that used it would give me a kickback on sales.
--:-- While I was interested in that hobby, I ran this website for a few years. It was reasonably successful—one of the first things I did that got any real traction on the internet. But it was never successful in proportion to how much effort it took to maintain and update it. So as I moved away from the hobby, I eventually stopped putting time and effort into maintaining the website.
--:-- Dwarkesh Patel: I'm curious as to how you even dug that up. I have a friend who was with you at the Oxford Refuges Conference—Connor Tabarroca, I don't know if you remember him.
--:-- Brian Potter: Oh, nice. Yeah, yeah.
--:-- Dwarkesh Patel: Finding other information about you on the internet actually was quite difficult. You may have managed to maintain your anonymity. So if you're willing to reveal, what was the P&L of this project?
--:-- Brian Potter: Oh, maybe a few hundred dollars a month for a few years. I only ever ran it as a side hobby business. In terms of time effort per my effort, I'm sure it was very low—pennies an hour or something like that.
--:-- Dwarkesh Patel: A broad theme that I've gotten from your posts is that the construction industry is plagued with these lossy feedback loops—a lack of strong economies of scale, regulation, and mistakes being very costly. Do you think this is a general characteristic of many industries in our world today, or is there something unique about construction?
--:-- Brian Potter: Interesting question. There are a lot of individual factors that are not unique at all. Construction is highly regulated, but it's not necessarily more regulated than medical devices or jet travel, or even cars to some extent, which have a whole battery of performance criteria they need to hit. Similarly, with land use, people say the land requirements and building on site make it difficult, but there are a lot of things that fall into that category that don't share the same structure of how the industry works as construction does.
--:-- I think it's just the interaction of all those effects. One thing that is perhaps underappreciated is that a building is really highly coupled in a way that a lot of other things aren't. If you're manufacturing a computer, the hard drive is somewhat independent from the display, somewhat independent from the power supply. These things are coupled, but they can be built by independent people who don't really need to talk to each other and then assembled into one structured thing. A building is not really like that at all. Every single part affects every other part for the most part. In some ways, it's like biology—it's very hard to change something without disrupting something else.
--:-- Part of that is because the job a building does is to create a controlled interior environment. Every single system has to run through and around the surfaces that are creating that controlled interior, so everything is kind of touching each other. Again, that's not unique—a plane or anything really highly engineered, like an iPhone, shares that to some extent. But in terms of the size of it and the relatively small amount you're paying in terms of unit size or unit mass, it's quite low.
--:-- Dwarkesh Patel: Is the fundamental reason you can't have as much specialization and modularity basically a transportation cost?
--:-- Brian Potter: I think it's really more just the way a building is. For example, for the electrical system of your house, you can't have a separate box that is your electrical system that you just install. If you needed to replace the electrical system, you couldn't take the whole box out and put a new one in. The electrical system runs through the entire house. Same with plumbing, same with the insulation, same with the interior finishes. There's not a lot of modularity in a physical sense.
--:-- Dwarkesh Patel: Gotcha. Now, Ben Kuhn had this interesting comment on your article about why it's hard to innovate in construction. He pointed out that many of the reasons you give—like the sequential dependencies and the highly variable delivery timelines—are also common in software, where Ben Kuhn works. So why do you think the same sort of stagnation has not hit other industries that have superficially similar characteristics, like software?
--:-- Brian Potter: I think you see a similar structure in anything that's project-based, or anything where you have to figure out what you're doing while you're doing it. Compared to large-scale manufacturing, you spend a lot of time figuring out what exactly you're building, then you spend a lot of time designing it to be built, then your first number of runs through you tweak your process to make it more efficient. But to some extent, the process of figuring out what you're doing is largely separate from the actual doing of it.
--:-- For a project industry, it's not quite like that. You have to build your process on the fly while you do it. Of course, there are best practices that shape it. For somebody writing a new software project or making a movie, they know roughly how it's going to go together, but there are going to be a lot of unforeseen things that come up.
--:-- I think the biggest difference is that those things can often scale in a way that you can't with a building. Once you're done with the software project, you can deploy it to a thousand or a hundred thousand or a million people. Once you finish making a movie, a hundred million people can watch it. It doesn't quite look the same with a building. You don't really have the ability to spend a lot of time up front figuring out how this thing needs to go. You need to figure out a way to get this thing to go together without spending a huge amount of time that maybe would be justified by the size of it.
--:-- He's definitely right. I was able to dig up a few references for software projects and how often they have these big long tails. Sometimes they go massively over budget. A lot of times they just don't get completed at all, which is shocking. But because of how many people it can then be deployed to after it's done, the economics of it are slightly different.
--:-- Dwarkesh Patel: I see. There's a famous law in software—I forget the name of it—which is that a project will take longer than you expect, even after you account for the fact that it will take longer than you expect.
--:-- Brian Potter: Hofstadter's law or something like that. I think it is.
--:-- Dwarkesh Patel: But I'm curious what the lack of scale in construction implies for startups. In software, the fact that there's zero marginal cost to scaling to the next customer is a huge boon to a startup. The entire point is just scaling with a hockey stick exponentially. Does that fundamentally constrain the size and quantity of startups you can have in construction, if that scaling is not available?
--:-- Brian Potter: Yeah, that's a really good question. The obvious first part of the answer is that for software, obviously if you have a construction software company, you can scale it just like any other software business. For physical things, it is a lot more difficult. This lack of zero marginal costs tends to fight a lot of startups, not just construction ones. But construction is particularly brutal because the margins are so low. And then there's the empirical fact that trying what seems like it would be a more efficient method of building doesn't actually let you do it cheaper and get better margins.
--:-- The startup that I used to work at, Katerra, their whole business model was basically predicated on that. "We'll just build all our buildings in these big factories and get huge economies of scale and reduce our costs and then make the billions of dollars that we're pumping into this industry." The math just does not work out. You can't build cheap enough to recoup those giant upfront costs. A lot of businesses have been burned that way.
--:-- The most success you see in prefab type stuff is on the higher end of things where you can get higher margins. A lot of these prefab companies tend to target the higher end of the market. There are a few different premiums for that. The obvious one is that if you're targeting the higher end, you have higher margins. If you're building to a higher level of quality, that's easier to do in a factory environment, so the delta is less enormous than it would be. Building a high level of quality is easier to do in a factory than in the field. A lot of houses built to a really high level of energy performance require a really high level of air sealing to minimize energy use. You tend to see a lot more houses like that built out of prefab construction and other factory-built methods because it's physically more difficult to achieve that on site.
--:-- Dwarkesh Patel: Can you say more about why you can't use prefabrication in a factory to get economies of scale? Is it just that transportation costs will eat away any gains, or what is going on?
--:-- Brian Potter: There's a combination of effects. I haven't worked through all of this—this is going to be the next block which basically solves it, I think I'll figure it out more by then. But at a high level, basically, your savings from using less labor are not quite enough to offset your increased transportation costs.
--:-- One thing about construction, especially single-family home construction, is that a huge percentage of your costs are just the materials you're using. A single-family home is roughly 50% labor, 50% materials for the construction costs, and then you have development costs and land costs. A big chunk of that you just can't move to the factory at all. The foundation—you can't really do that in a factory. You could prefab the foundation, but it doesn't gain you anything; your excavation still has to be done on site. So a big chunk of that can't move to the factory.
--:-- For the ones that can, you still basically have to pay the same amount for materials. Theoretically, that wouldn't be true if you were doing huge, truly huge volume where you could get material volume discounts, but even then, it's probably not looking at an asset saving. You can cut out a big chunk of your labor, and you do see that in factory-built construction. These prefab companies, like mobile home companies, have a really small fraction of labor as their cost, which is pretty typical for a factory in general. But they take out all that labor cost, they still have their high material costs, and then they have this overhead cost of whatever the factory has cost them, plus the additional overhead cost of transporting it to site, which is pretty limited. The math just does not really work out in favor of prefab in terms of being able to build dramatically cheaper. You can obviously build a building using prefab methods and build a successful construction business—many people do—but in terms of dramatically lowering your costs, you don't really see that.
--:-- Dwarkesh Patel: Austin Vernon has an interesting blog post about why there aren't more prefabricated homes. The two things he pointed out were transportation costs and that people prefer to have homes with unique designs or unique features. When I was reading it, it occurred to me that maybe they're both the result of the same phenomenon. Have you heard of the Alchian theorem in economics?
--:-- Brian Potter: Maybe, I don't think so.
--:-- Dwarkesh Patel: Basically, it's the idea that if you increase the cost of some category of goods in a fixed way—let's say you tax oranges, adding one dollar tax to all oranges, or transportation for oranges gets one dollar more expensive for all oranges—then people will shift consumption towards the higher grade variety. The ratio of the cost between the more expensive orange and the less expensive orange has decreased because of the increase in fixed costs. It seems like you could use that argument to also explain why people have strong preferences for uniqueness and design in manufactured houses. Since transportation costs are so high, that's basically a fixed cost, and that fixed cost has the effect of making people shift consumption towards higher grade options.
--:-- Brian Potter: I definitely think that's true. I would maybe phrase this as: the construction industry makes it relatively cheap to deliver a highly customized option compared to a really repetitive option. The ratio between a highly customized one and a commodity one is relatively small, so you see the industry built around delivering somewhat more customized options.
--:-- I do think that the intuition that people just desire too much customization from their homes and that really prevents you from having a mass-produced offering is true to some extent. One example is the Levittown houses, which were originally built in huge numbers of exactly the same model over and over again. Eventually, they had to change their business model to deliver more customized options because the market shifted. But I think the effect of that is basically pretty overstated.
--:-- Empirically, you see that in practice, home builders and developers will deliver fairly repetitive housing. They don't seem to have a really hard time doing that. As an example, I'm living in a new housing development that is just three or four different houses copied over and over again in a group of 50. The developer is building a whole bunch of other developments that are very similar in this area. I lived in a very similar development in a whole different state. If you just look at multi-family or apartment housing, it's identical apartments copy-pasted over and over again in the same building or a bunch of different buildings in the same development. You're not seeing huge amounts of uniqueness in these things. People are clearly willing to just live in these basically copy-pasted apartments.
--:-- It's also quite possible to get a pretty high amount of product variety using a relatively small number of factors that you vary. The car industry is like this. I was reading a book a while ago that was pushing back against the idea that the car industry in the 50s and 60s was just offering a very uniform product. They did the math, and the number of customization options on their car were more than the atoms in the universe—there were just so many different options. All the different combinations of leather seats, this type of stereo, this type of engine—if you add it all up, there's a huge massive number of different combinations.
--:-- You can obviously customize a house a huge amount just by the appliances you have, the finishes, the paint colors you choose, and the fixtures. It would not really change the underlying way the building goes together. So I do think that the idea that the fundamental demand for variety is a major obstruction—I feel like there's not a whole lot of evidence for that in the construction industry.
--:-- Dwarkesh Patel: Interesting. So I asked on Twitter what I should ask you. Usually I don't get interesting responses from this, but the quality of the people and the audience that knows who you are was so high that all the questions I got were really interesting. So I'm just going to ask you some questions from Twitter.
--:-- Brian Potter: Okay.
--:-- Dwarkesh Patel: Connor Tabrock asks, "What is the most unique thing that would or should get built in the absence of construction regulation?"
--:-- Brian Potter: Interesting. "Unique" is an interesting qualifier. There's a lot of things that just should get built—huge amounts of additional housing, or creating more land in these really dense urban environments where we need it. Places like San Francisco could just fill in a big chunk of that bay—it's basically just mud flat—and put more housing on it. I saw a proposal like that earlier.
--:-- Unique thing is more tricky. One idea that I really like, I read this in the book *Where's My Flying Car?*, is that it's basically crazy that our cities are designed with roads that all intersect with each other. That's just a crazy way to structure a material flow problem. Any sane city built from scratch would have multiple layers of transportation where each one went in a different direction, so your flows would be massively improved. That just seems like a very obvious one. If you were building your cities from scratch and had your druthers, and knew how big it was going to get, you would clearly want to build them that way.
--:-- Dwarkesh Patel: So you could plan very long term in a way that, again, almost no cities did, but be able to plan so these transportation systems don't intersect with each other and you have the space to scale them or run as much throughput through them as you need without bringing the whole system to a halt.
--:-- Brian Potter: There's a lot of evidence that basically says that cities tend to scale based on how much you can move from point A to point B through them. I do wonder if you change the way that they went together, you could unlock massively different cities. Or even if you didn't unlock massive different ones, but change the agglomeration effects that you see in cities if people could move from point A to point B much quicker than they can.
--:-- Dwarkesh Patel: I did an episode about the book *Where's My Flying Car?* with Rohith Krishnan. I don't know if we discussed this, but that was an interesting part of the book where he talks about transistor design. Like, if you design transistors this way, can you imagine how slow they would be? Okay, so Simon Grimm asks, what countries are the best at building things?
--:-- Brian Potter: Yeah, this is a good question. I'm going to, again, sort of cheat a little bit and do it in space and in time. Because I think a lot of the most countries that are most interesting in doing a good job at building vast amounts of stuff are not ones that are basically doing it currently. I mean, the current answer is like China, right? Where they're just building more concrete used in the last 20 years than the entire world used in the time before that. Just big, massive amounts of urbanization. A lot of really interesting buildings and deconstruction. In terms of raw output, you've got to get to that.
--:-- I would also put Japan in the late 20th century on there. At the peak of concern like, is Japan going to take over the world? They were really interested in building their stuff quite quickly. They spent a lot of time and effort trying to use their robotics expertise to try to figure out how to build buildings a lot more quickly. They had these really interesting factories that were designed to basically extrude a tire skyscraper, just going up vertically. All these big giant companies had many different factories that were kind of trying to do this with all these robotics. A really interesting system that did not end up ever making economic sense, but it is very cool. I think their big industrial policy organs of the government basically encouraged a lot of these industrial companies to develop prefabricated housing systems. You see a lot of really interesting systems developed from these industrial companies in a way that you don't see in a lot of other places.
--:-- The US, basically from 1850 to maybe 1970, so for like 100 years or something, was just building huge massive amounts of stuff in a way that lifted up huge parts of the economy. They spent all this, I don't know how many thousands of miles of railroad track the US built between 1850 and 1900, but it was many thousands of miles. Of course, needing to lay all this track and build all these locomotives really forced the development of the machine tool industry, which then led to the development of better manufacturing methods and interchangeable parts, which of course then led to the development of the auto industry. That explosion just led to even more big giant construction projects. You really see that being able to build just big massive amounts of stuff in this virtuous cycle with the US really advanced a lot of technology and raised the standard of development for a super long period of time. So those are kind of my three answers.
--:-- Dwarkesh Patel: Those three bring up three additional questions, one for each of them. That's really interesting. Have you read *The Power Broker*, the book about Robert Moses? I think I got like a tenth of the way through it. That's basically a whole book in itself. A tenth of the way? No, I'm like halfway through. So far it's basically the story of how this one guy built a startup within the New York state government that was just so much more effective at building things. It didn't have the same corruption and clientelism and incompetence of the Tammany Hall political machine at the time. Maybe it turns into tragedy in the second half, but so far it's like, we need this guy. Where do we get a second Robert Moses? Do you think that if you had more people like that in government or in the construction industries, public works would be more effectively built? Or is the stagnation there just a result of other bigger factors?
--:-- Brian Potter: That's an interesting question. I remember reading an article a while ago that was complaining about how horrible Penn Station is in New York. They were basically saying it would be nice to return to the era of the unbound technocrat, where these technical experts in these high positions of power in government could essentially do kind of whatever they wanted. If they thought something should be built somewhere, they basically had the power to do it. It's just a facet of this problem of how it's really hard to get stuff built in the US currently. I'm sure part of it is that you don't see these really talented technocrats occupying these high positions of government where they can get stuff done. But it's not super obvious to me whether that's the limiting factor. I kind of get the sense that they would end up being bottlenecked by some other part of the process. The whole interlocking set of institutions has just become so risk averse that they would end up being blocked in a way that they wouldn't have been when they were operating in the 1950s and 60s.
--:-- Dwarkesh Patel: Speaking of Japan, I just recently learned that a lot of the construction there is that they just keep tearing stuff down every 30 to 40 years and rebuilding it. You have this interesting series of posts on how you'd go about building a house or building that lasts for a thousand years. But I'm curious, how would you build a house or building that only lasts for 30 or 40 years? If you're building in Japan and you know they're going to tear it down soon, what changes about the construction process?
--:-- Brian Potter: As I'm not an expert on Japanese construction, I think a lot of their interior walls are basically just paper and stuff like that. I actually think it's kind of surprising they use a somewhat surprising post and beam construction method, which is actually somewhat labor intensive to do. The US in the early 1800s used a pretty similar method. Then once we started mass producing dimensional lumber, we kind of stopped doing that because it was much cheaper to build out of two by fours than it was with big heavy posts and mortise and tenon connections.
--:-- I think the boring answer to that question is that you would just build like we build mobile homes essentially, which is using pretty thin walls, pretty low-end materials that are just going together in a pretty minimal way. That ends up being not that different than the actual construction method that single family homes use, but just even further economizes and tightens the use of materials. Where a single family home might use a half inch plywood, they might try to use three sixteenths or even an eighth inch plywood. So probably a pretty similar way to the way most single family homes and multi-family homes are built currently, just with even tighter use of materials. Which perhaps is something not super nice about the way that Japan builds their homes.
--:-- Dwarkesh Patel: Okay, so China is the third one here. There's been a lot of talk about a potential real estate bubble in China, that they're building housing in places that people don't really need it, and of course the demographics aren't there to support the demand. What do you think of all this talk? Is there a real estate bubble that's created by all this competence and building?
--:-- Brian Potter: I have no idea. I've definitely heard talk of it and I've seen the little YouTube clips of them knocking down all these towers that it turns out they didn't need, or the developer couldn't finish. I don't know a huge amount about that in general. I wish I knew a lot more about how things are built in China, but the information is in general so opaque. I generally kind of assume that any particular piece of data that comes out of China has giant error bars on it as to whether it's true or not, or what the context surrounding it is. So in general, I do not have an informed opinion about that.
--:-- Dwarkesh Patel: And this is the second part of Simon's question. Does greater competence in being able to build stuff translate into other good outcomes for these countries, like higher GDP or lower rents or other important outcomes?
--:-- Brian Potter: That's a good question. Japan is an interesting example where people point to it as a sort of example of a country that builds huge amounts of housing and they don't have housing cost increases. In general, we should expect that dynamic to be true. There's no reason to not think that housing cost is essentially a supply and demand problem. If you built as much as people wanted, the cost would drop. I have no reason to doubt that's true. There is a little bit of evidence that suggests the cost of housing will kind of tend to rise to whatever people's income level are. The peak and the flattening of housing costs in Japan also parallel when people basically stopped getting raises and income stopped rising in Japan. So I don't have a good sense of if it ends up being more driven by some other factors. But in general, I expect the very basic answer of if you build a lot more houses, the housing will become cheaper.
--:-- Dwarkesh Patel: Speaking of how the land keeps gaining value as people's income go up, what is your opinion of Georgism? Does that kind of trend make you think that housing is a special asset that needs to be more heavily taxed because you're not inherently doing something productive just by owning land, the way you would be if you built a company or something?
--:-- Brian Potter: I don't have any special deep knowledge of Georgism. It's on my list of topics to read more deeply. I do think in general, taxing something encourages you to produce less of it. For something that you can't produce less of, it's a good avenue for something to tax more heavily. Obviously, if you had a really high land value tax in these places that have a lot of single family homes in dense urban areas like Seattle or San Francisco, that would probably encourage the land to be used a lot more efficiently. So it does make sense to me, but I don't have any special knowledge about it.
--:-- Dwarkesh Patel: All right. Ben Kunask on Twitter asks, what construction related advice would you give to somebody building a new charter city?
--:-- Brian Potter: That is interesting. Off the top of my head, I would be interested in if you could really figure out a way to build using a method that has really high upfront costs that could otherwise be justified. If you're going to build 10,000 buildings all at once, you could really take advantage of that. One thing you see in the post-World War II era is that we were building such huge masses of housing, and a lot of times building them all in one place. A lot of town builders were building thousands and thousands of houses in one big development. In California, it was the same thing, building five or ten or fifteen thousand houses in one massive development. There you end up seeing something where they basically build a little factory on their construction site and use that to fabricate all these things. You have almost a reverse assembly line where a crew would go to one house and install the walls, then go to the next house and do the same thing. Following right behind them would be the guys doing the electrical and the plumbing. This reverse assembly line system allowed you to get these things up really fast, like in 30 days, just thousands and thousands of houses at once.
--:-- I think you would want to be able to do something similar, not just do the construction the way that normal construction is done. But that's hard, because a lot of the centrally planned city never seems to do particularly well. Brasilia or whatever, the planned city built in the age of the unfettered technocrat who can sort of build whatever he wants. A lot of times what you want is something that responds at a low level and organically to the factors as they develop. You don't want something that's planned totally top down, disconnected from all the cases on the ground. A lot of the opposition to Robert Moses ended up being that in a certain form, he's bulldozing through these neighborhoods that he's not paying attention to at all. Trying to plan your city from the top down doesn't have a super great track record. In general, you want your city to develop a little bit more organically. I would think really good land use rules that are really thought through well, that encourage the things you want to encourage and not discourage the things you don't want to discourage. Don't have exclusionary zoning, allow a lot of mixed use construction. So that's a somewhat boring answer, but probably something like that.
--:-- Dwarkesh Patel: Interesting. I guess that implies that there would be high upfront costs to building a city because if you need to build 10,000 homes at once to achieve these economies of scale, then you would need to raise tens of billions of dollars before you could build a charter city.
--:-- Brian Potter: If you were trying to lower your costs of construction, but if you have the setup to do that, you wouldn't necessarily need to raise it. These other big developments were built by developers that essentially saw an opportunity. It didn't require public funding necessarily. It did in the form of loans, loan guarantees for veterans and things like that, but the government didn't have to go and buy the land.
--:-- Dwarkesh Patel: Okay. The next question is from Austin Vernon. And to be honest, I don't understand the question, but you two are too smart for me, but hopefully you'll be able to explain the question and then also answer it. Hey, guys, I hope you're enjoying the conversation so far. If you are, I would really appreciate it if you could share the episode with other people who you think might like it. This is still a pretty small podcast. So it's basically impossible for me to exaggerate how much it helps out when one of you shares the podcast. Put the episode in the group chat you have with your friends, post it on Twitter, send it to somebody who you think might like it. All of those things helps out a ton. Anyways, back to the conversation. What are your power rankings for technologies that can tighten construction tolerances? He gives examples like AR/VR, CNC cutting, and synthetic wood products.
--:-- Brian Potter: This is a very interesting question. Basically, because buildings are built manually on site by hand, there's just a lot of variation in what ends up being built. There's only so accurately that a person can put something in place if they don't have any sort of jig or something. The placement of materials tends to have a lot of variation, and then the materials themselves kind of have a lot of variation. The obvious example is wood, where one two by four is not going to be exactly the same as the two by four adjacent to it. It may be warped, it may have knots in it, it may be split. Also, because these materials are sitting outside in the elements, they end up getting a lot of distortion. They either absorb moisture and expand and contract, or they grow and shrink because of the heat. There's just a lot of variation that goes into putting a building up. To some extent, it probably constrains what you are able to build and how effectively you are able to build it.
--:-- I gave an example before of really energy efficient buildings. They're really hard to build on site using conventional methods because the air sealing is quite difficult to do. You have to build it in a much more precise way than is typically done and is easily achieved on site.
--:-- In terms of examples of things that would make that easier, he gives some good ones. Engineered lumber is where you take lumber and grind it up into strands or chips and basically glue it back together. That does a couple things. It basically spreads all the knots and the defects out so they aren't concentrated, and it lets you get a lot more uniformity. That's a very obvious one, and it's already in widespread use. I don't really see that making a substantial change. The one exception would be that there's this engineered lumber product called mass timber, or CLT is a big one. It's like a super plywood. Plywood is made from tiny little sheet thin strips of wood, and CLT is made from two by four dimensional lumber glued in cross-laminated layers. Instead of a four by eight sheet of plywood, you have a 12 by 40 sheet of dimensional lumber glued together. You end up with a lot of the properties of sheet material where it's really dimensionally stable and can be produced very accurately.
--:-- It's actually funny, a lot of times the CLT is the most accurate part of the building. If you're building a building with it, you tend to run into problems where the rest of the building is not accurate enough for it. Even with steel, if you're building a steel building, the steel is not going to be dead on. There's going to be an inch or so off in terms of where any given component is. The CLT, which is built much more accurately, actually tends to show all these errors that have to be corrected. In some sense, precision is a tricky thing, because you can't just make one part of the process more precise. All that does is make things more difficult, because if one part is really precise, that a lot of times means that you can't make adjustments to it easily. If you have this one really precise thing, it usually means you have to make compensation for something else that is not built quite as precise. That actually makes advancing precision quite a bit more complicated.
--:-- AR/VR, I think I'm very bullish on, assuming with the big caveat that they can just get the basic technology working. The basic intuition is that right now, the way pieces are put together on site is that somebody is looking at a set of plans, paper plans or an iPad, that tells them where everything needs to go. They figure that out and they take a tape measure or some other method and go figure out where that is marked on the ground, and then try to get it all put in there. There's all this setup time and manual location that is really quite time consuming and error prone. There's only so accurate that a guy dragging a tape 40 feet across site being held by another guy can be. It's very easy for me to imagine AR stuff that just projects exactly where the components of your building need to go. That would allow you a much higher level of accuracy that you can easily get using manual methods, and also reduce all that time it takes to manually measure things. So I can imagine it being much faster as well. I'm quite bullish on that at a high level.
--:-- At a slightly lower level, it's not obvious to me if they will actually be able to get that level of projecting it with perfect accuracy right in front of you. It may be the case that a person moving their head around, constantly changing their point of view, you're never going to be able to locate these things with millimeter precision. It's always going to be a little bit jumpy, or you're going to end up with some sort of hard limit in terms of how precise you can project it. I wouldn't bet on that. My sense is that locator technology will get good enough, but I don't have any principled reason to believe that.
--:-- The other thing is that being able to take advantage of that technology would require you to have a really accurate model of your building that locates where every single element is precisely and exactly what its tolerances are. Right now, buildings aren't designed like that. They are built using a comparatively sparse set of drawings that leaves a lot to be interpreted by the people on site doing the work. Efforts that have tried to make these models really precise have a lot of times not really paid off. You can get returns on it if you're building something really complex where there's a much higher premium to making sure you don't make any errors. But for a simple building like a house, the returns just aren't there. So you see comparatively sparse drawings. Whether it's going to be worth the upfront cost of developing this really complex, very precise model where exactly every component is, is kind of to be determined. There are some interesting companies that are trying to move in this direction, making it a lot easier to draw these things really precisely, where every single component is exactly where it is. So I'm optimistic about that as well, but it's a little bit TBD.
--:-- Dwarkesh Patel: Okay, this raises a question that I actually wanted to ask you. In your post about why there aren't automatic brick layers, it was a really interesting post. Somebody left an interesting comment, which was that bricks were designed to be handled and assembled by humans. I think you left a response to that which I thought was really interesting. You said, "The example I always reach for is with steam power and electricity, where replacing a steam engine with an electric motor in your factory didn't do much for productivity. Improving factory output required totally redesigning the factory around the capabilities of electric motors." I was kind of curious if you apply that analogy to construction. What does that look like for construction? What is a house building process or building process that takes automation and these other kinds of tools into account? How would that change how buildings are built and how they end up looking in the end?
--:-- Brian Potter: I think that's a good question. I think that's one big component of the lack of construction productivity. Everything is designed and has evolved over 100 years, 200 years or whatever, to be easy for a person on the site to manipulate by hand. Bricks are roughly the size and shape and weight that a person can move it easily around. Dimensional lumber is the same. It's the size and shape and weight that a person can move around easily. Kind of all construction materials are like this. The way that they attach together and stuff is the same. It's all designed so that a person on site can sort of put it all together with as comparatively little effort as possible.
--:-- But what is easy for a person to do is usually not what is easy for a machine or a robot to do. You typically need to think about what your end goal is and then redesign the mechanism for accomplishing that in terms of what is easy to make a machine to do. The obvious example here is that it's way easier to build a wagon or a cart that rolls than it is to build a mechanical set of legs that mimics a human's movement.
--:-- Dwarkesh Patel: That's just way, way, way easier. So yeah, I do think a big part of advancing construction productivity is to basically figure out how to redesign these building elements in a way that is really easy for a machine to produce and a machine to put together.
--:-- Brian Potter: And one reason that we haven't seen it is that a lot of the mechanization you see is people trying to mechanize exactly what a person does. So if a person moves like these two-by-fours, you know, in the place they're trying to make a machine that will move them before then nailing together. And to do that, it's like really, really expensive. You need a lot of really expensive industrial robots that can move exactly the way that a human moves, more or less.
--:-- Dwarkesh Patel: So what does that kind of thing look like?
--:-- Brian Potter: I think a lot of what that might look like is basically something that can be really easily extruded by machine in something like a continuous process that wouldn't require a lot of finicky mechanical movements. A good example of the technology is something called insulated metal panels, which is perhaps one of the cheapest and easiest ways to build an exterior wall. What it is, it's just a thin layer of steel, and then on top of that is a layer of insulation, and on top of that is another layer of steel. At the end, the steel is extruded in such a way that these inner panels can lock together as they go. It's basically the simplest possible method of constructing a wall that you can imagine. It has the structural system, the water barrier, the air barrier, and the insulation all in this one really simple assembly. When you put it together on site, it just locks together. Of course, there are a lot of limitations to this. If you want to do anything uncomplicated like add windows, all of a sudden it starts to look quite a bit less good. But things like that are really easy for a machine to do and can be put together without a lot of precision measurement in the field. They can just kind of snap together.
--:-- Dwarkesh Patel: And what would the houses or the buildings that are built using this physically look like? Maybe in 50 to 100 years, we'll look back on the houses we have today and be like, "Oh, look at that artisanal creation made by humans." But what is a machine that is designed for robots first or for automation first? In what interesting ways would it differ from today's buildings?
--:-- Brian Potter: That's a good question. I'm not especially bullish on 3D building printing in general, but this is another example of building using an extrusion process that is relatively easy to mechanize. What's interesting there is that when you start doing that, a lot of these other bottlenecks become unlocked a little bit. It's very difficult to build a building using a lot of curved exterior surfaces with conventional methods. You can do it, but it's quite expensive. But it's relatively straightforward for a 3D printed building to do that. They can build that just as easily as a straight wall. So you see a lot of interesting curved architecture on these things. It kind of unlocks a lot of options. Machines are more constrained in some things they can do, but they don't have a lot of the other constraints you would otherwise see. So I think you'll see a larger variety of aesthetic things like that.
--:-- That said, at the end of the day, I think a lot of the ways a house goes together is pretty well shaped to the way a person living inside it would like to use it. Stewart Brand makes this point in *How Buildings Learn*. He basically makes the point that a lot of people try to use dome-shaped houses or octagon-shaped houses, which are good because they enclose the space using the least amount of material possible. In some theoretical sense, they are quite efficient. But it's actually quite inconvenient to live inside a building with really curved walls. Furniture doesn't fit up against it nicely, pictures are hard to hang on a really curved wall. So I think you would see less variation than maybe you might expect.
--:-- Dwarkesh Patel: Interesting. And why are you pessimistic about 3D printers for construction?
--:-- Brian Potter: Okay, so many reasons. Not pessimistic, but there are a lot of other questions. The big obvious one is that right now, a 3D printer can basically print the walls of a building. That is a pretty small amount of the value in a building. It's maybe 7 or 8 percent, probably not more than 10 percent of the value in a building. You're not printing the foundation, you're not printing the overhead spanning structure of the building. You're basically just printing the walls. And you're not even really printing the second story walls if you have multiple stories. I don't think they've quite figured that out yet. So it's a pretty small amount of value add to the building. And it's a task that is relatively easy to do by manual labor. It's really pretty easy for a crew to basically put up the structure of a house.
--:-- It's a recurring theme in mechanization. It takes a lot of mechanization and a lot of expensive equipment to replace what basically two or three guys can do in a day. The economics of it are pretty brutal. So right now, it produces a pretty small value. I think the value of 3D printing is entirely predicated on how successful they are at figuring out how to deliver more components of the building using their system. There are companies trying to do this. One that got funded not too long ago is called Black Diamond. They have this crazy system which is a series of 3D printers that would act simultaneously, each one building a separate house. As you progress, you switch out the print head for a robot arm, because a 3D printer is basically a robot arm with a particular manipulator at the end. They switch out their print head for a robot arm, and the robot arm goes and installs different other systems like the windows or the mechanical systems. Whether you can figure out how to do that reliably where your printing system is installing a large fraction of the value of the building, it's not clear to me that it's going to be economic. It obviously needs to reach that point, and it's not obvious to me that they've gotten there yet. It's really quite hard to get a robot to do a lot of these tasks.
--:-- For a lot of these players, it seems like they're actually moving away from that. ICON is the biggest construction 3D printer company in the US, as far as I know. They moved away from trying to install lots of systems in their walls as they get printed. They moved on to having that installed separately, which I think has made their job a little bit easier. But again, it's hard to see how the 3D printer can fulfill its promises if it can't do anything beyond the vertical elements, which are really quite cheap and simple to build now.
--:-- Dwarkesh Patel: If you take a step back and talk about how expensive construction is, how much of it can be explained by the Baumol cost effect? That labor costs are increasing because labor is more productive in other industries, and therefore construction is getting more expensive.
--:-- Brian Potter: Yeah, I think that's a huge chunk of it. The labor fraction of it has not changed appreciably enough. I'm going to asterisk that because I haven't actually verified that, and I'm remembering somebody said that used to be much different. But in general, yeah, the labor cost has remained a huge fraction of the overall cost of the building. Labor is a huge cost of it, and reliably seeing their costs continue to rise. I think there's no reason to believe that's not a big part of it.
--:-- Dwarkesh Patel: In your post comparing the prices of construction in different countries, you mentioned how the cost of labor and the cost of materials is not as big a determiner of how expensive it is to construct in different places. But what does matter is the amount of government involvement and administrative overhead. I'm curious why those things have such a high consequence on the cost of construction.
--:-- Brian Potter: That's a good question. I don't actually know if I have a unified theory of that. With any regulation, any heavily regulated thing, any particular task that you're doing takes longer and is less reliable than it would be if it was not done. You can't just do it as fast on your own schedule. You end up being bottlenecked by different processes. It narrows your options. So in general, I would expect that to be the case. But I don't know if I have a unified theory of how that works beyond just a bunch of additional steps at any given part of the process, each of which adds cost.
--:-- Dwarkesh Patel: One interesting trend we have in the United States with construction is that a lot of it is done by Latino workers, especially undocumented Latino workers. What is the effect of this on the price and the quality of construction? If we have a bunch of hard-working undocumented workers who are working for below market rates in the US, will this dampen the cost of construction over time?
--:-- Brian Potter: I suspect that's probably one of the reasons why the US has comparatively low construction costs compared to other parts of the world. I'll caveat that. Residential construction, which is single-family homes and multi-family apartment buildings, are all built in the US out of light-framed wood and put together by a lot of immigrant workers. Because of that, it would not surprise me if those wages are a lot lower than the equivalent wage for a carpenter in Germany or something like that. That results in our residential construction costs being very, very low by world standards. I suspect that's a factor.
--:-- Dwarkesh Patel: Overall, it seems from your blog posts that you're kind of pessimistic, or you don't think that different improvements in industrialization have transferred over to construction yet. But what do you think is the prospect of future advances in AI having a big impact on construction? With computer vision and advances in robotics, do you think we'll finally see some carryover into construction productivity, or is it going to be more of the same?
--:-- Brian Potter: Yeah, I think there's definitely going to be progress on that axis. If you can wire up your computer vision systems to your robotic systems and your AI in such a way that your capabilities for a robot system are more expanded, I foresee robotics being able to take a larger and larger fraction of the tasks done on a typical construction site. I see it being done in narrow avenues that gradually expand outward. You're starting to see a lot of companies that have some robotic system that can do one particular task, but it can actually do that task quite well. There are a couple different robot companies that have these little robots for drawing wall layouts on concrete slabs, so you know exactly where to build your walls. You would think that would not be a difficult problem in construction, but it turns out that a lot of times people put the walls in the wrong spot, and then you have to go back and move them later or just deal with it. It's basically a little Roomba-type device that draws the wall layout onto the concrete slab and all the other systems as well, like where the lines need to run through the slab. I suspect that you're just going to start to see robotics and systems like that take a larger and larger share of the tasks on the construction site over time.
--:-- Dwarkesh Patel: What do you think of Flow? That's Adam Neumann's newest startup that is backed with $350 million from Andreessen Horowitz.
--:-- Brian Potter: I do not have any strong opinions about that other than, wow, they're really giving him another three hundred and fifty million. I will say that A16Z makes a lot of investments that don't make sense to me, but I'm not a venture capitalist. So there's no reason my judgment would be any good in this situation. I'm just presuming they know something I don't.
--:-- Dwarkesh Patel: I'm going to be interviewing Marc Andreessen later this month, and I'm hoping I can ask him about that. I would love to hear what that thing is. It may be as simple as, you know, VC is all about really high variance bets. There's nobody higher variance than Adam Neumann.
--:-- Brian Potter: Yeah, so maybe just on those terms, it makes sense.
--:-- Dwarkesh Patel: You had an interesting post about how a lot of the knowledge in the construction industry is informal and contained within best practices or between relationships and expectations that are not articulated all the time. It seems to me that this is also true of software in many cases, but software seems much more legible and open source than these other physical disciplines like construction, despite again having a lot of the knowledge contained within people's minds and within the culture rather than explicitly codified. Why do you think construction seems more closed source than software?
--:-- Brian Potter: It's interesting. To go back to the products versus projects industry distinction, a slightly different way of thinking about that is craft-based industries versus an industrial process. This isn't a dichotomy, it's a spectrum. In general, craft-based stuff has an expertise and judgment aspect that is pretty well embedded in the process that you can't really remove. Any decision at any given point requires an expert or an artisan who understands the relevant context and knows how to proceed based on the specific variables in that specific situation. An industrial process is more like this has been figured out, this is how it works every single time.
--:-- Construction is very, very much on the craft end of the production spectrum. The decision of how to put these things together and how to wire this building or whatever is all left up to the expertise and the judgments of the people doing the installation. What that gets you is that it lets you put things together without having to do a very large amount of specifying exactly what you need. The drawings to specify a house going together, even something quite complicated, are way fewer than the drawings needed to produce a Toyota Corolla, for instance. The design cost required to do it, in proportion to how expensive the thing is, is also much lower.
--:-- Software is perhaps more legible in that the end product is very clear. You can clearly see every single part of it and how every single part connects to every other part. Perhaps I'm overstating that. I'm sure a software developer could say it's actually not super obvious how these things work and why they're done this way. But you can clearly inspect every single part of it and see exactly how it does what it does and how it connects to the other parts. You can't really do that with a building.
--:-- With a building, it's not necessarily obvious how it got to the point that it did when it was put together. A lot of times with physical things, even if you have the object, it is unclear what the process was to create it. This comes up with industrial espionage. A lot of times, that doesn't help them as much as they would think to recreate it. They have to basically go through the entire process of figuring out how to make it, and it takes them just as long as it did the original people doing the development. You saw this with the atomic bomb. The people who stole the plans basically took as long to figure out how to make it as the US did. So just a physical object, the process used to make it is not necessarily super legible. It tends to be a little bit hidden.
--:-- I was going to say that perhaps that is not as true for software, but I don't actually know. It's very plausible to me that you could have some piece of software that was written and then just be utterly inscrutable as to how it came together and how you could duplicate a similar piece of software. Is that a category error?
--:-- Dwarkesh Patel: That's an interesting question. I think there are a lot of examples where if you don't have the context on why that was built a certain way, you wouldn't understand what's going on. If you've heard of the fast inverse square root, that's exactly what I was thinking of. For people who are interested, there's a great YouTube video on this. The guy who created it, John Carmack, is a super genius. But if you just look at the algorithm, it's a few lines, but you would never understand why this gives you an inverse square root unless you went through the mind of John Carmack when he goes through Newton's method and all these other things and why this particular float operation.
--:-- But I guess the advantage software has is the ability to fork. You can't just take a building and make an exact replica of it and then change the part you want to better understand to see what the effect is. With a software project, you can just fork it or make an API call and try to understand the specific subcomponent. It's easier with software.
--:-- Brian Potter: That's interesting. You can run experiments on your piece of software to understand how it works a lot more easily and at lower cost than you can with any physical object, especially a giant building.
--:-- Dwarkesh Patel: Let's say some mega corporation CEO is like, "Brian, we want to build some really interesting skyscraper. I've talked to the mayor and the governor, and they're willing to get rid of all the building codes. There's no regulation. We just want you to build a really cool skyscraper." What would you do if you were given this latitude to just build something?
--:-- Brian Potter: Oh, interesting. I would like to see us fulfill the dream of the early to mid 20th century and build a mile-high skyscraper. This was what people saw the development of skyscrapers going in the 30s, 40s, and perhaps 50s. Frank Lloyd Wright designed a mile-high skyscraper called The Illinois, back when Chicago was a metro of rising importance. The technology for it exists. I'm aware of other programs that have designed conceptual buildings this tall. You haven't really seen anyone do it, even people who are willing to build these giant white elephant projects. Nobody has tried to go the distance and build the highest skyscraper. I would like to see us do it.
--:-- Dwarkesh Patel: You have a really interesting essay in Works in Progress about skyscraper height. One of the things you talked about is the super linear increase in lateral forces and other impediments to building tall. How would you get over that kind of stuff?
--:-- Brian Potter: You basically throw a giant amount of money at the problem. The basic gist of that essay is that the physical constraints do not prevent you from building a mile in height. It's the economic and legal constraints that stop this extreme construction. The economic constraints are significant enough that even in places where there are no legal constraints, like China or Dubai, they still do not build this tall because the economics of it are just so brutal. But in this fantasy scenario, a giant stack of money would get devoted to doing this.
--:-- Dwarkesh Patel: Speaking of which, in that post you brought up the argument from the economist Glaeser that says we're leaving billions of dollars on the table by having building height codes because we're giving up all this vertical space. I'm curious why you think it's the case that these developers don't have any sort of lobbying or political influence to be able to collect the billions of dollars of deadweight loss that are created by these codes. Why aren't they able to organize politically in a way that gets rid of these regulations that are helping no one?
--:-- Brian Potter: That's a really interesting question. In general, the strongest construction lobbying group I'm aware of is probably the National Association of Home Builders, which exerts quite a bit of influence to try to keep the cost of building single-family homes low. I'm not aware of anything that exists for large commercial buildings, maybe the Association of General Contractors. My initial guess would be that the natural constituencies for opposing a big project like this are always going to be quite a bit bigger than the constituencies that would be able to act for it. Any big giant construction project, even if it had a lot of developers mobilized to support it, would also have a large constituency that would exist to oppose it: anyone who lives in the area and doesn't want this giant shadow of a building or is worried about the congestion it would cause.
--:-- I think a sort of paradox with this situation is that the places that need it the most because their rents are so high, the people that are living there already are going to be financially well equipped to oppose it. In a certain sense, they're losing the most out of it. If you're making fifty thousand dollars a year, you might value the view out of your apartment at five hundred dollars. But if you're making five million dollars a year, you might value that view proportionately more and be willing to expend a lot more to prevent it from being obstructed. This mechanism where places get wealthier and need more housing also creates its own opposition to some extent. That's why I think a Coasian solution to this kind of thing would be optimal, where if the view is worth more to you than an apartment is worth to somebody else, then you can just pay them to not build there. But the view is not worth more than an apartment is probably worth to somebody, so it's just an inoptimal allocation of resources based on who has political influence. I'm not super confident of that answer. I should look into it more. Why don't developers have better lobbying efforts? It does seem like an unanswered question.
--:-- Dwarkesh Patel: Speaking of being able to put projects into a tailspin and add more sclerosis, you just recently published a very interesting and thorough examination of how NEPA works. Do you want to explain what this law is and what its consequences are? Then I can ask you some more specific questions about it.
--:-- Brian Potter: Yeah, for sure. So NEPA is the National Environmental Policy Act. This is the law that basically requires any major federal government action that might have significant environmental impacts to do a very long and thorough and extensive environmental impact study before anything is done on a project.
--:-- Dwarkesh Patel: Speaking of being able to put projects into a tailspin and add more sclerosis, you just recently published a very interesting and thorough examination of how NEPA works. Do you want to explain what this law is and what its consequences are? Then I can ask you some more specific questions about it.
--:-- Brian Potter: Yeah, for sure. So NEPA is the National Environmental Policy Act. This is the law that basically requires any major federal government action that might have significant environmental impacts to do a very long and thorough and extensive environmental impact study before anything is done on a project.
--:-- It gets a large amount of attention because of how long these environmental impact statements take to prepare. The average time currently is something like four and a half years, and in some cases, since the average half of them are longer than that, the highway administration, for instance, takes on average something like eight and a half years to do an environmental impact statement before they can build a highway or something like that.
--:-- So people are perpetually trying to figure out a way to reform this law, basically such that we would not have to wait years and years and years before building big important infrastructure projects. That's the gist of what NEPA is and how it works.
--:-- Dwarkesh Patel: Yeah, you had a really interesting point in the article where you said that by adding this cost, you can basically think of it as like a tax on all major government action, and the effect of a tax is to reduce what you're taxing. And then I thought at the end you had a really interesting argument about how NEPA is an anti-law. Can you explain this argument for podcast listeners?
--:-- Brian Potter: Yeah, it's my spicy take at the end that I always have to throw in right at the end. The basic argument is that the purpose of a law is roughly twofold. One is to encourage something that you would want more of, or discourage something that you would want less of. So we have laws against drunk driving because we think drunk driving is harmful and we want less drunk driving in our society.
--:-- The second purpose of a law is to basically reduce coordination problems and enable exchanges that might not otherwise be able to take place. The obvious example of this is the government enforces which side of the road you're allowed to drive on, not because one side is inherently better than the other, but because it's good if everybody agrees on which side of the road to drive on. Contract law is in some ways like this. It's good if people will be punished for breaking contracts because that allows people to enter into them, which allows exchanges that might not otherwise take place.
--:-- There's an argument that the ability of the English government in the 1600s and 1700s to basically have to pay back its debts was a really important development because it allowed it to raise money that it otherwise wouldn't be able to, because people could trust that they would be able to get paid back.
--:-- Anyway, those are the two rough purposes of a law, and NEPA does not do either of those things. First, because NEPA is basically a procedural statement or a requirement, it does not require the government to weigh environmental concerns especially heavily. It doesn't prevent a big oil and gas drilling project from taking place. Essentially, what it does require is that for any major environmental effects, you just have to document them very thoroughly. So all it is is a documentation requirement and notifying the public of what you're doing. But it doesn't prevent a major environmental negative effect. As long as you've documented it quite thoroughly, you can kind of do whatever you want.
--:-- The evidence is very unclear as to whether it has had beneficial environmental effects. To the extent that it has, I think there's a good argument that it has done so just by making it harder to do anything at all.
--:-- Then the other side of this is to create certainty, solve coordination problems, stuff like that. NEPA actually is very, very bad at this. It creates a lot of uncertainty because the requirements for meeting, doing the analysis are so unclear and constantly shifting. The definition of what is a major environmental effect has basically changed over time. An obvious example is in the 1970s, it was maybe not obvious that greenhouse gas emissions were a major environmental effect. But now in the 2020s, that obviously is. So what you have to do for a NEPA analysis changes over time, which is fine in that sense, but it does mean that when you have to go through the process, it is unclear how long it's going to take, what is going to be involved, and whether anybody is going to litigate your decision.
--:-- This is the other big part of NEPA: people are basically able to sue groups for not completing the analysis thoroughly enough. They can't permanently stop a project because all you need to do is show that you've documented your things thoroughly enough. Once you have documented it thoroughly enough, they don't have grounds for stopping it anymore. But they can slow it down. If they slow it down enough, sometimes the project becomes unattractive and it gets canceled. That's what a lot of these groups hope for.
--:-- So instead of creating certainty and solving coordination problems, it creates all this new uncertainty where people will very deliberately try to avoid the process because they do not know how long it will take and how much it will cost to get their project approved. In some cases, going through the process will take a very, very long time. Making a business decision about whether to do a new offshore wind development or develop a new forest resource or something like that is very hard if you don't know when your project is going to start because you don't know how long the process will take.
--:-- Dwarkesh Patel: Yeah, that's so fascinating. I've got a lot more questions about this because your breakdown was really interesting. One thing is, I don't know if you're familiar with the long-termist movement. Have you come across this?
--:-- Brian Potter: Vaguely.
--:-- Dwarkesh Patel: Okay. So one thing that they've proposed, this is not like a doctrine or anything, but on the periphery, some idea I saw was that just as we have environmental review, we should have a posterity review so that you're analyzing the impacts of your actions on generations way down the line. What are the future impacts for actions, just as we analyze what are the environmental impacts of actions? What do you think of an idea like that given the various dysfunctions of the environmental review process?
--:-- Brian Potter: A couple of thoughts. One is just that my gut response is that any additional review is just going to add additional time and complexity to your process. Everything is a process that takes some amount of time and has some particular chance of success. So just adding another filter to the process is only going to make the process slower and less likely to succeed rather than more.
--:-- The other thing is, how accurately are you going to be able to predict what your long-term impacts are? I've seen a take like this before, so this is not original at all, but it's not obvious to me that anybody making predictions over even the past 20 years would have been able to do so with any degree of precision. It's not obvious you would even be able to get the sign right on that, whether it be net positive or net negative. That's off the top of my head. I could definitely be persuaded otherwise by somebody who has thought a lot more about it, so that is not a strongly held opinion. But that would be my critique.
--:-- Dwarkesh Patel: That's my critique as well. If you think of the 70s, correct me if I'm wrong, but the technical consensus at the time was that we would hit peak oil by the 90s. Of course, that's because they couldn't predict our ability to find new reservoirs and develop new technologies that made more oil available to us. It's just very hard to predict future trends, so I don't know how much that kind of law would help.
--:-- Let me ask about NEPA. I vaguely hear that there's political deliberation about reforming NEPA. What would your ideal reform of NEPA look like? How would you reform the implementation?
--:-- Brian Potter: I think on a very simple level, I would like to even the playing field. A lot of these newer energy technologies should have the same benefits that oil and gas has been able to accrue. For instance, oil and gas has a lot of categorical exclusions for certain drilling operations and for drilling in the Gulf of Mexico. They actually have a lot of NEPA exclusions, which was perhaps one of the reasons, not one of the reasons, but you can connect that to the giant Deepwater Horizon accident.
--:-- So basically, giving the technologies like wind and solar and large-scale transmission projects the same benefits that oil and gas drilling and natural gas pipelines have would be a massive boom. Just off the top of my head, that's one thing I would like to see.
--:-- Dwarkesh Patel: I wonder what you think about this. Reading the post, it doesn't make sense to me why this is enforced through the courts where anybody can just bring a lawsuit. Shouldn't there be a single coherent bureaucracy whose goal is to figure out who's messing up and whatnot? It seems like a mistake to have this done through the courts.
--:-- Brian Potter: I was reading a paper earlier about it from Michael Lowry. Basically, NEPA is a horribly drafted law because it contained no provisions for funding it or with a bureaucracy for enforcing it or anything like this. It's essentially through random chance that the courts, when it was passed during a very activist period of the court, decided to enforce this provision for the impact statement. That provision was kind of added late in the process without a lot of fanfare or consideration, and that one little part of it ended up becoming the most important part because that's what the court decided to enforce really strongly.
--:-- People say this is not how you would draft an environmental protection law if you were doing it from scratch. It's this weird thing that we've ended up with due to path dependency. You would want a government agency responsible. An idea I've seen floated a few times is that you would want something that looked more like the OMB, which is charged with figuring out how much a given law is going to cost. You would want a bureaucracy attached to it that was designed to figure out the environmental effects of this, decoupled from the specific agency or whatever.
--:-- The downside of that for a really environmental protection statute is whether it would be captured by political interests or not, and either just not enforced at all if it was a very conservative staffing court, or enforced extremely vigorously if it was on the other end of the spectrum. That's the risk of that. But clearly it's not ideal that the court system is responsible for basically determining this.
--:-- Dwarkesh Patel: Even if there was a bias in how the law was enforced, if there's a bureaucracy, the benefit is that you can just fire the guy who's running it if you think he's not enforcing it correctly and replace him with somebody who you think will enforce it more appropriately. Whereas with the judiciary, if it's dozens of different judges having independent opinions of how this should be enforced, there's nobody who's responsible that you can tell to enforce it differently.
--:-- Brian Potter: And it actually makes it a little bit hard to change it because any given change in the provision, you don't actually know what the effect of it is until somebody files a lawsuit referring to it and it works its way through the courts. Some people think that during the Trump administration, there were some changes to how these laws work. They changed some of these laws regarding it, and someone said, "Yeah, this is designed to accelerate it." But because it's unsure how the courts are going to interpret them, it's actually going to increase the risk for these projects in the short term until some of these lawsuits make their way through the courts and we know exactly what is required or not because the court tells us.
--:-- It's a very efficient system of implementing environmental requirements. I don't think it's really how other countries do it. I'm not an expert on this, but a lot of other countries, almost every other country I think, has something requiring environmental impact statements. But I don't think it's enforced via citizen lawsuits. I think a lot of times it's done via a normal government bureaucracy or something like that.
--:-- Dwarkesh Patel: I was actually just about to ask that. Countries that have different or no systems of environmental review, do they have speedier and more cost-effective public works? Do you have data on this?
--:-- Brian Potter: No, I don't actually. It would be very hard to separate that.
--:-- Dwarkesh Patel: That's all right. Then you had a comment in that post that I found interesting. You said, "This uncertainty also makes changing NEPA somewhat risky. Experts have noted, for instance, that rules to accelerate NEPA processes or impose maximum timelines might result in more of them being challenged in court by failing to take the proper, quote unquote, hard look." Do you want to explain this? Because this is counterintuitive to me.
--:-- Brian Potter: The requirements for meeting the NEPA requirement, and this is a more general requirement across other things, is that you have to take what the courts call a "hard look," where you have to consider these impacts quite thoroughly. The risk is that if you put a timeline cap on some of these processes, like it has to be done in a year and if it's not done in a year it's automatically approved, which is an idea you see floated from time to time, the risk is that people will just go and litigate the project immediately. When they do, they will say they did not look hard enough at these impacts.
--:-- If they can't marshal the resources, or if we needed to study some particular flowering species that only has a flowering period of two weeks in the spring and the time period was up before that happens, which is a thing that happens in NEPA apparently, then the court would say you did not look hard enough at this. Go back and do it again. That's one of the reasons why these take multiple years. If you're observing some species or whatever, you maybe need an entire year or more to actually observe it.
--:-- So timeline caps could very easily backfire by increasing the amount of litigation, which would just make these things take longer than they already do. Already, you see a lot of extra analysis due to risk aversion from these federal agencies. The laws around NEPA actually say that your environmental impact statement should really not be longer than 150 pages, except in extreme circumstances. But the average environmental impact statement is now like 660 pages or something like that. So people are already going more than what the law says they should do just out of risk aversion. If you don't fix the incentives that are causing this risk aversion, your solution will not work.
--:-- Dwarkesh Patel: I'm curious if you think that there's been a sort of talent drain from physical engineering tasks like construction into software. Has that happened and has that had an impact on the world of atoms? Or is that just a thing people discuss on the internet, but it's not real?
--:-- Brian Potter: I do think that's almost definitely happening. This was my constant opinion when I was working as an engineer, especially when I was managing younger engineers who were getting paid comparatively little. I was always on the verge of telling them, "Why don't you just go learn how to code and go make three or four times as much at a FAANG company instead of doing this fairly thankless work?" Frankly.
--:-- So I think it's very likely to be an issue. In some sense, it's theoretically self-correcting. To the extent that labor moves out of the field and it then gets more expensive, the incentives around it change a little bit and you're forced to find ways of building things with less labor requirements or where you can spread your labor development over a larger volume of output.
--:-- If the engineers are leaving and engineers are getting more expensive, you're going to basically find a way to spread your engineering effort over a larger construction volume, which would be using more prefab or using kit of parts assemblies and things like that. So theoretically, to some extent, it is self-correcting. The risk is if you screen off the top 20 percent of most talented people, does that just fundamentally handicap what your industry is capable of doing? Or do the incentives kind of push the other way and you try to lower your quality even more?
--:-- I feel like I've heard people make complaints to this effect with semiconductor research, how semiconductor engineering is not especially well paid and it is relatively easy for semiconductor engineers to go get jobs working in software development, so you see a brain drain from that.
--:-- I think it actually may be correcting to some extent. The salaries for engineers have actually risen quite a bit. A similar risk, and people in construction are complaining about this constantly, is just the unavailability of skilled labor. Being able to do a lot of these tasks requires skilled expertise. You saw a lot of these people leave the labor force during the Great Recession and not really come back, and people aren't entering it at the rates that they need to. I think the average age of a construction worker is somewhere in the 40s or perhaps 50s. It's very high. People are constantly complaining, "I can't get enough labor, there's not enough workers."
--:-- But of course, this also is going to incentivize finding ways to get these buildings built using less labor. I'm a little bit less worried about potential negative effects from that. There are a lot of historical examples of labor constraints developing into labor-saving inventions. The history of the US is like that, where we had not nearly enough labor as they needed, so we developed a lot of labor-saving machinery, the American system of manufacturing, things like that. I'm a little bit less worried about the skilled labor side, but I could see it definitely having effects on the other sides of the industry.
--:-- Dwarkesh Patel: Even the author of that global economic history book made a point that the industrial revolution happened in Britain because the cost of labor was highest in Britain because there was a plague that killed off a bunch of people, so labor was really expensive.
--:-- Brian Potter: I've heard that as well. I'm not an expert in economic history and I'm trying to avoid things I know very little about. The other factor is that now as venture capital expands its tentacles into other industries, you're seeing a lot more effort on developing solutions for the built environment. You're seeing more venture capital money flow into the space, so perhaps that will counteract to some extent.
--:-- Dwarkesh Patel: Final question. Many years ago, Tyler Cowen was interviewed by Patrick Collison. One of the things Tyler Cowen said was that there should be more blogs that are just focused on one particular area or issue and then kind of raise the salience of it and help drive insight and understanding of it. I feel like you've done that really well with construction and construction productivity. I'm curious what other areas you would like to see blogs that do for their area what you have done for construction productivity, which is just take a broader view of what's happening, what the trends are, try to add more insight into professional problems and things like that.
--:-- Brian Potter: Good question. I feel like there's so much about this. I was complaining the other day that this giant civilizational machine that takes in raw materials and spits out finished goods and services and high quality of life, nobody really knows how it works. Mostly how it works is completely undocumented. It does not exist in written information anywhere. The information that is written down is maybe in one particular company's shared drive or something like that. The instructions for building a Toyota or something don't exist in any way, shape, or form, and it might be very hard to recreate that if you needed to.
--:-- I would like to see somebody do something similar for manufacturing just in general, especially now because so much of manufacturing knowledge in the US has been lost, just moved over to China and other places where labor is cheaper. There's so much that has just been gone, and how things get built and manufactured that just isn't written down anywhere or isn't written down anywhere accessible. You see people share the same small number of resources for how these things work over and over again.
--:-- I asked a few people who I thought would know about it, "Is there any book written that actually explains what it is like to build something, get something built in China, or what building things in China are actually like?" And they're all like, "No, I don't know any source that exists." There are a few blogs that describe their particular experiences, which again end up being overly valuable resources, but there's no general source of available information for how any of that works.
--:-- I just think documentation of how the civilization machine functions is wildly underinvested in, which I guess is true for documentation in general. Nobody wants to do that because the payoffs are far in the future and uncertain, and the costs are up front. So it's not surprising to me that it doesn't exist. But I do think that in general, people are really very interested in understanding how things work. If you can explain how something works, even if it seems like a fairly niche topic, you can get quite a bit of attention for doing it. So for anybody who works in manufacturing or like that, I think you would find quite a bit of success if you started writing up how it functions.
--:-- Dwarkesh Patel: Interesting. Okay, excellent. To listeners who know about manufacturing, do spin up your Sub stacks.
--:-- Dwarkesh Patel: Okay, so this is part two of my conversation with Brian Potter. The first conversation was really interesting, and afterwards I realized that I had forgotten to ask Brian about this really interesting theory that Scott Alexander and others have written about. Brian is the perfect person to talk to about this, so I had to inconvenience him and ask him to come on again. But the idea is basically, or the question is, why does modern architecture just look so much more ugly than things that were built 100 years ago or 200 years ago?
--:-- Dwarkesh Patel: Okay, so this is part two of my conversation with Brian Potter. The first conversation was really interesting, and afterwards I realized that I had forgotten to ask Brian about this really interesting theory that Scott Alexander and others have written about. Brian is the perfect person to talk to about this, so I had to inconvenience him and ask him to come on again. But the idea is basically, or the question is, why does modern architecture just look so much more ugly than things that were built 100 years ago or 200 years ago?
--:-- You would have thought that due to increases in technology and the buildup in new ideas and designs, that things would become prettier. But if you look at a building that was built in the last 100 years, it just looks like a cylinder or a rectangle of glass and concrete. And if you look at things that were built before, you see things like this Sagrada Familia or these intricate cathedrals or even these skyscrapers that have all these flourishes and ornaments and all these decorations. So Brian, what is going on? Why are things so much uglier now?
--:-- Brian Potter: Sure, I have a few thoughts on this. I don't know if I have an answer to the actual question, but I can sort of add some context. I think I'm not necessarily the best person to ask on this. A better person to ask would be an architect who can talk about how the design and taste has evolved over the course of the 20th and 19th century and stuff like that. But also in general, I don't have a huge amount of opinion on aesthetics. And to the extent that I do, I do tend to favor the more simple, minimalist, clean lines.
--:-- Like one of my favorite pieces of construction is this bridge called Salginatobel, which is built in Switzerland. It's the banner that's on my Twitter account. And it's exactly this: it's this really simple, minimalist, just really tiny curve of concrete that's exactly in the shape that it needs to be to resist the forces on it. So to some extent, to me, it's like you asking why did aesthetics change from worse aesthetics to better aesthetics—you know, because they're better. That's obviously not what I actually think, but just to give you a little bit of where I'm coming from.
--:-- I want to separate out the question a little bit. I don't think it's so much a question of why do modern buildings look ugly per se. I think it's more a question of why do modern buildings have less ornamentation. I think there's a lot of really beautiful modern architecture. A lot of what it seems to focus on—and again, an architect would be a better person to ask about this, but I will sort of muddle my way through—is using light and space and form to create these big, interesting, open, impressive spaces that create a lot of interesting shapes. That big expansive space is what they're really going for.
--:-- A good example of a building thing that's interesting is if you look at the new terminal for the Portland Airport, which has this big giant mass timber roof. It's a way of building these large span structures, but it's really neat and kind of interesting. A lot of architecture is focused on that, and that's been pretty common throughout history, right? Cathedrals are trying to create these big impressive spaces using light to create this impressive space. The Pantheon is the same thing: creating this big open impressive space.
--:-- So I think you want to separate out, to some extent, the question of "why is it ugly," which I don't necessarily agree with, and "why does it have so much less ornamentation," which I think is a little bit more defensible.
--:-- Dwarkesh Patel: Okay, actually let me ask about the second one because I'm not sure I totally followed. So the ornamentation has been replaced by the openness?
--:-- Brian Potter: Well, yes. My point is that traditional, a lot of interesting, impressive spaces were just focused on creating this big indoor, impressive, large span space and playing with the light and stuff to do that. And that's been common in buildings that people think are impressive. And you still see that today. Architects still do that a lot, using big open areas with a lot of glass everywhere. It's kind of a tool in their tool belt to do that, to some extent.
--:-- Dwarkesh Patel: Gotcha. So they're replacing the old stone working and stonemasonry stuff and the gargoyles with these kinds of things. Okay, I see.
--:-- Okay, so yeah, let's talk about one because the idea that the older designs were less efficient in some way. To steel man the opposite position, someone might say that a lot of the value of construction and building is just the aesthetic presence that this has in a city or a neighborhood. And to the extent that we've lost that, maybe we got to use less material on a bridge or a skyscraper, but then it kind of just acts as this ugly thing in the middle of the city. Or if not ugly, at least not as aesthetically pleasant as it could have otherwise been. And that's something we should care about when we're doing construction. How would you react to something like that?
--:-- Brian Potter: Yeah, so I guess to clarify what I was talking about before, I don't think new buildings necessarily use less material in a sense. I just think they have an aesthetic that's more minimalist and streamlined and without a lot of extra decoration on it. I think the complaint is like, "Oh, they'll build just a big cylinder of glass or a rectangle of glass or whatever."
--:-- I think one reason that that is true is because it may not be super interesting from the outside, but if you're on the inside of a building, you actually want a whole lot of glass. It's really nice to have a large open space to walk through and to have a lot of natural light. That wasn't really possible prior to the mid 20th century for a couple reasons. One is because buildings weren't really air conditioned before that. Another is that to have a big expanse of glass was still expensive, but it was even more expensive as you go farther back. There was a glass making process invented in the 1950s, 1960s called the float glass process which made it possible to make really high quality glass way, way cheaper than was possible previously.
--:-- So air conditioning and advances in structural design and increased use of steel basically made it a lot more possible to just have your building be a big giant slab of glass and have a lot of natural light and a lot of openness on the inside. That creates a kind of nice experience if you're actually in the building.
--:-- I think that to some extent applies a little bit more generally. If you look at an HGTV remodel or something like that, everybody wants to create a big giant open concept floor plan with all the walls blown out and lots of natural light coming in everywhere. So it seems like it's maybe a case of the technology evolved such that you can create this nicer interior space. That came perhaps at the expense of having a lot of really ornate decoration on the outside. If you're building a building, it makes sense to optimize the inside at the expense of the outside, because that's what you're actually using.
--:-- Dwarkesh Patel: Okay. Yeah. That's an interesting way to put it.
--:-- One comment that I've heard is that the progress of YIMBYism has been hurt by the fact that new buildings, because of this rational self-interest, are ugly on the outside and very pleasant on the inside. But the result is that the people who are the community organizers or whatever, the other people in the community, will oppose these new buildings because what they get to see is the outside. And these historical preservation boards will notice that these older buildings that are not as pleasant on the inside and have obviously less occupancy and everything, they look more pleasant in their neighborhood than the new buildings that people want to tear down and build. So if we want to convince the NIMBYs to go along with new construction, you kind of have to make it look beautiful. What do you think about that?
--:-- Brian Potter: Yeah, I bet that sounds like it's true to some extent. There is that fundamental tension in real estate: any property you have has a lot of externalities and influences the value of everything around it. So it's this big game of negotiation. Any given thing you build is a big negotiation between you and everything in the immediate area, and that's why things like zoning and stuff like that exist in the first place, even if it's not necessarily implemented in an especially good way.
--:-- I think it's probably a little bit easy to over-index on that. It would not surprise me if, you know, it's like, "Okay, we'll build this as a really beautiful old building." It's like, "No, it's still gonna be a giant building. It's still gonna mess up my traffic. It's still gonna have a bunch of renters that are gonna come in and reduce my property values." You see a lot of these historical preservation things want to preserve things like old gas stations and laundromats and things that are not beautiful at all, or parking lots and stuff like that. So I think that's probably true on some margin, but I would be surprised if it's the one thing we need to fix to untangle this problem.
--:-- Dwarkesh Patel: Gotcha. I wanted to ask, by the way, you mentioned how indoor air conditioning and heating has changed this dynamic. Is it just because you have enough insulation that even if you lose heat through the glass that's fine?
--:-- Brian Potter: No. Basically, if you have a big glass wall as your exterior building, if you don't have air conditioning, that thing is just going to cook whenever it gets hot out. It's going to basically be like a greenhouse. So you basically need air conditioning to make that habitable. You see this with single family homes too. When air conditioning started to become popular and installed by default, all of a sudden they started building these houses with these big giant picture windows which they didn't have before. People like that again—it makes a nice inside space. And if you're going to be inside, you want to focus on that part. But it basically is predicated on having air conditioning to control the climate.
--:-- Dwarkesh Patel: Gotcha. Okay, so that explains why the outsides of buildings aren't as pretty as they used to be. But there's still the question of why insides are so minimalistic. I don't know if you saw this, about a year or two ago Microsoft announced that they were building this new office for developers in India. The outside looks like a modern glass building, but the inside had all these really interesting Mughal architectural details and craftsmanship that made it look like a Mughal palace—from the furniture to the floor layout to the arches of the entrances. It actually looks really cool. So why don't more developers do something like that on the interiors of these buildings?
--:-- Brian Potter: Yeah, again, good question. I think for the most part, a developer—and again, I'm not a developer, so I'm gonna muddle my way through as best I can—they're basically building the space to rent out or to sell to somebody else. A lot of times, development is called "shell and core," where they basically don't even finish the inside. It's all just raw open studs and everything, and then the tenant comes in and does whatever they want to the space. They're basically in the business of creating usable space for some given class of market that they're trying to attack. They're not in the business of making fancy architecture.
--:-- Same if you're buying a single family home. You come in with blank walls and you put up all your different pictures and different decorations and paint it however you want and add your funky wallpaper or whatever. And then when you move, you can take all that stuff out and the next person can come and put everything that they want. So I think to some extent, you see a decoupling of the aesthetic design from the building itself.
--:-- Dwarkesh Patel: Okay, so then this is really a revealed preference about the people who are actually renting or buying these properties. People who are living there actually don't want this kind of material design?
--:-- Brian Potter: Maybe. I think again it comes back to a little bit about a question of aesthetic style and whether people want a lot of ornamentation. People do focus a lot on making these spaces attractive, and a lot of times they do that by having big and nice and open, a lot of natural light, and a cool glass staircase or whatever that goes from here to there. People like that. Somebody pointing to the new Google headquarters—it's not that new, but on the newer side—it's this big giant dome type thing with a lot of structural detail visible and a lot of light coming in, this big giant open space, a lot of interesting dark cases and kind of stuff like that. So to some extent, it's just a style with less ornamentation, not necessarily bad aesthetics per se.
--:-- Dwarkesh Patel: Okay, gotcha. The other theories that are out there about this emphasize more the cultural and aesthetic changes rather than the actual practical necessities of these kinds of changes that you've talked about. So one theory is that if you look at the change in men's clothing over the last two centuries, it's become also much more minimalistic and much less colorful and ornamental. If you look at a King Louis or something, he's wearing a suit with all these studded colors and dyes and even gems. And if you look at a picture of Joe Biden, he's just wearing a suit, not even a tie anymore. It's just a black suit and a white shirt. So this is a continuation of this sort of aesthetic trend towards being very minimalistic. What's your reaction to that take?
--:-- Brian Potter: Yeah, I mean, I think that's true to some extent. This is one of the difficulties of having a conversation like this: there are so many degrees of freedom, it's hard to pin down what specifically you're talking about, what class of things you're specifically talking about. If you look at an example from fashions, if you look at streetwear or something like that, you will find a ton of pictures of really quite ornate, interesting clothing. There will be some stuff that's minimalist, but there will be a lot that's not necessarily minimalist at all.
--:-- So I guess in some way I disagree with the premise. If you're talking about some elite counter-signaling thing, like it's so easy and cheap to make ornate colorful clothing that it's not a signal of status anymore, so you have to counter-signal by having very simple clothing. Maybe there's something to that, but that's a different thing than "why are things ugly in general?"
--:-- Dwarkesh Patel: One thing that people also talk about is the increasing cost of labor. Now you can't hire people to spend the higher talented stonemasons to spend hours on every square foot of the outside of a building. I was in India like six months ago and we went to New Delhi to visit this new temple that was built, the Swaminarayan Akshardham. It's this really intricate and cool design on the outside, just covered in these hand carved stones with these intricate idols and different images. It's really cool. But I think it took like tens of thousands of hours, probably way more, and like thousands of workers and stonemasons to actually construct that. Which is obviously not going to be very feasible economically in a western country. So one theory is just that it's too expensive to do that kind of stuff anymore. You need to do something that doesn't require as much manpower.
--:-- Brian Potter: Yeah, I mean, I think that's definitely got to be part of the story. Construction as we know has not really gotten any cheaper, but it also hasn't gotten more expensive. A lot of other things that are pure labor, like other services—medical care, education, stuff like that—have gotten more expensive. Part of that is probably because we've found ways to pull labor out of the process, so construction only rises at the rate of inflation instead of faster than the rate of inflation.
--:-- And yeah, super labor intensive things like masonry just don't get done anymore. There's probably a vicious cycle there: you hire less of it, so there are fewer masons available, and so they get more expensive, and the skill gets more scarce. Now to hire someone to do really ornate masonry work in the US is probably really difficult to even find them. I'm sure they exist, but it's probably not trivial, and if you have your schedule you're trying to meet, it's not super straightforward.
--:-- An interesting example of this is the Ise Jingu Temple in Japan. It's this temple complex that gets torn down and rebuilt every 20 years, and they've been doing this for 1300 years or something like that. It uses 1300 year old construction techniques—thatched roofs and particular woodworking methods. And it's now quite difficult for them to find a lot of these skills. There's just not that many roof thatchers around anymore. So it's a similar cycle with masonry.
--:-- I don't think that's like the whole story, though. A lot of this minimalist style is actually quite expensive to do. A glass curtain wall is actually really expensive to build on a building because glass is just expensive by itself, and because it lets in so much light and it gets so hot, you need a much more expensive mechanical system to keep it cool. I've talked to architects: why do the owners love these glass curtain walls? Especially now, where everyone is really concerned about climate change and greenhouse emissions of your building. And people were like, "Yeah, they just really love this all-glass look, and they're willing to pay extra to get it."
--:-- If you look at the very early style of construction where you just have a big glass facade, for a skyscraper it's called the international style. Early international style skyscrapers were actually quite expensive, more expensive than the traditional way of building. But again, the people who were building it really liked the aesthetic, and it gave you some other options in terms of light on the inside or whatever, so they decided it was worth it. So that's definitely probably part of the story, but it definitely doesn't seem like it can be all of it.
--:-- Dwarkesh Patel: Gotcha. Okay, yeah. To the extent that it is part of the story, somebody left an interesting comment on one of Scott Alexander's posts on this thing about how modern construction looks different than older construction. This is a comment from Fluffy Buffalo. He writes, "I think new technology should help a lot with 3D printers, CNC machines, robots, CAD, and AI. It shouldn't be too hard to come up with a way to produce ornate murals and building shapes at a reasonable price." And then they go on, "But no one is doing it because the current crop of architects can apparently only think in steel, concrete, and glass." So how plausible do you think this is?
--:-- Brian Potter: Yeah, so that makes me think of—and I think I read some of the comments on that post, and I think one of the people mentioned this—the Victorian style of house basically came about because of mass production methods which made really ornate wood decorations really, really cheaply. They were just cranked out on some milling machine in a factory somewhere, and then you could just buy them and staple them to your house or whatever. So that style is basically a function of technology making that cheaper. And then of course, that was in style for a while, and it's not in style anymore, for the shifting sands of aesthetics.
--:-- But I also think part of it is—and this is more broadly applicable—it's actually kind of a hassle to have all this ornamentation all over your house. On the inside, it's just lots and lots of stuff to dust and keep clean or whatever, which again comes back to cost disease. If you have servants to do that for you, it's fine. But if you don't have servants, it creates a lot of extra work for you. Same on the exterior, but then also there is the issue that in general you want your building to shed water and direct it away from the house as much as possible. If you have a lot of little nooks and crannies and ornamentation for water to collect, that's pretty bad from a durability perspective.
--:-- I don't think I would be careful not to over-index on that, because obviously you can build extremely durable, beautiful masonry brick buildings. It's not like that makes it impossible. It is something to consider, though, that it does on the margin probably make your maintenance costs go up.
--:-- Dwarkesh Patel: Okay. Are you optimistic overall—maybe not with the specific technologies, but just—and maybe not with having ornamentation or something—but do you think that in a hundred years the technology would enable construction of buildings that look prettier than modern buildings and are also more maintainable, or at least as maintainable? Or is it basically going to be steel and glass towers for the upcoming future?
--:-- Brian Potter: Yeah, that's a good question. There's a lot of enthusiasm for mass timber in architecture. We'll see how well it actually gets adopted, but for large, tall buildings made out of heavy timber elements—partially it's an aesthetic thing, partially it's from a carbon perspective. Using a lot of timber versus steel or concrete is way better on that calculus. And it does kind of look nice. A lot of these tech companies are building these ornate timber offices and stuff like that. So that's one trend.
--:-- I don't see an obvious technology that could usurp the sort of standard palette we have, unless it's a dramatic development in material science where somebody finds some way of building something that's way cheaper and gives you way more options than you had before. That's what steel and then concrete gave you. All of a sudden you could build things in ways that you couldn't do before. With steel, you could use so much less structure. With concrete, because it was liquid, you could make any shape that you want. That's when you get all these really cool interesting shells and ornate concrete domes and stuff like that, which of course we also don't build anymore.
--:-- So I don't see an obvious technology to replace that. And then I guess the other thing is: when all of a sudden you had technology that could easily duplicate or make any sort of art possible, did that usher in a new era of way better art or 2D aesthetics? I don't necessarily think that it did, right?
--:-- Brian Potter: So I think if you could all of a sudden build any sort of shape that you want cheaply and easily, it's not obvious to me that that would all of a sudden mean we can get our really cool things. I'm sure you would see a lot of interesting experiments on the margin or whatever, but it's not obvious to me that the standard way of building—in some ways it might become even less interesting, right?
--:-- Dwarkesh Patel: So there is the question of why we haven't seen in these other fields better aesthetics. Given how much you can do with artistic, like the traditional art today, why do these paintings that are selling for millions of dollars look like—obviously the common perception and stereotype about them is that they're very simple and ugly. To apply that to architecture and engineering, there's this idea Scott Alexander kind of talked about a lot in his post, that there's this cabal of modernist architects in these guilds who are just obsessed with building these buildings that the public doesn't like to look at. Given your exposure to the industry, do you think that's true? That the architects just have a completely different sense of aesthetics than the average public?
--:-- Brian Potter: I think there's probably something to that. You go to architecture school, and architecture school is adjacent to art school. You learn how to design buildings or whatever, but you also learn a specific way of looking and thinking about them, and that filters down into how you design your building. But I think architects are competing in a marketplace. The vast majority of them who are not star architects are basically having to deliver buildings that owners want and are happy with. If they don't do that, they're going to go out of business. I think most architects you talk to will basically say, "Yeah, if I don't give the owner a building that he's happy with, I am not doing my job correctly." They are being hired for, among many other things, their sense of taste. But perhaps that's part of the story, a fragment or whatever. I think it would be very easy to over-index on that, and I think there's a lot of ways that's not true.
--:-- Dwarkesh Patel: Okay, okay. I'm really glad we got a chance to cover this topic, and I appreciate all those explanations. But actually, there's one question that I didn't get—I forgot to ask you last time—that I'm curious about your opinion on. Let's say somebody listens to the conversation or they've been a follower of your blog and this got them interested in engineering. Somebody maybe in high school or maybe early in college. For them to be able to do cool things in these fields in the future, what kind of training and career advice would you give them? What should they pursue to get involved and innovate in these fields?
--:-- Brian Potter: Yeah, that's a good question. I don't have a super good answer. I fell into it very much on accident. I spent a very large chunk of my career just doing pretty standard work-a-day engineering type stuff, and then I sort of fell ass backwards into a job at a construction startup, and that led me to other places.
--:-- I guess, if you're an engineer, go to a good—I feel sort of gross giving this answer because in general, "go to a good engineering school" perpetuates the college industrial complex. But people do really care about that, especially in the fields of building engineering that I'm working with. They care about that quite a bit. I went to a reasonably good engineering school, and that probably opened doors for me that it would not have otherwise. So yeah, go to a good engineering school. Go to a lot of the status indicators that are very obvious—go to a good engineering school and go to a top-tier firm. That does kind of matter. People do pay attention to those things in the building development world. If you went to MIT or Caltech and then had an internship at Skidmore, Owings & Merrill or whatever, that's going to open quite a few doors for you. I'm not amazingly happy with that answer, but I do think it is true.
--:-- Another way would just be to try to come in from a more oblique angle. Focus on software and stuff like that, and then work at one of the many startups now tackling the construction space. They're increasingly in need of software developers. That would be a lower-risk way to do it because you don't need to necessarily go to a fancy school to become a software developer. And then if it doesn't work out, you can just go get a job at a FAANG company and make $600,000 a year. Look and see how the startups in the space are changing and growing. I think by the time some folks get out, the building space might look pretty different. Technology and software stuff is slowly forcing its way in there a lot more.
--:-- Dwarkesh Patel: Okay. But then, so if they want to transition from just being an engineer to working on the forefront stuff that you write and talk about, is working at a startup the ideal way to do that?
--:-- Brian Potter: Oh, that's a good question. There's a lot of startups doing really interesting stuff, not just in software, but a lot of interesting building technology stuff. There's a lot of green building stuff that is being developed—folks working on low-carbon concrete, low-carbon steels. There's a lot of stuff doing construction robotics, a lot of prefab stuff. The startup world is really starting to eat the construction world. There are a lot of opportunities there. I very broadly would say I see a lot of innovation happening. So I think that's a good place to sort of look and see what's going on.
--:-- Dwarkesh Patel: Awesome. Okay, well, Brian, thank you so much for coming back a second time. I really appreciate it. I'm glad we got a chance to talk about these things. It is very interesting, and it's good to get an actual engineer's perspective on it so that we're not just doing cultural theory and we actually understand the practicalities of what's involved in all those kinds of things. Okay, Brian, this is super fascinating. I highly recommend people check out your blog, which you can find at constructionphysics.substack.com, and they can also follow you on Twitter. We'll leave the handle in the description. Anything else you'd like to plug or say at the end?
--:-- Brian Potter: Yeah, I am doing some work for the Institute for Progress, which is a think tank designed to advance industrial progress and progress studies ideas more generally. So you should check them out as well.
--:-- Dwarkesh Patel: Okay, excellent. Thanks, Brian. I appreciate it.
--:-- Brian Potter: Thanks.
--:-- Dwarkesh Patel: Hey, thanks for listening. If you enjoyed that episode, I would really, really, really appreciate it if you could share it. This is still a pretty small podcast, so it is a huge help when any one of you shares an episode that you like. Post it on Twitter, send it to friends who you think might like it, put it in your group chats. Just let the word go forth. It helps out a ton. Many thanks to my amazing editor, Graham Bessalou, for producing this podcast. And to Mia Ayana for creating the amazing transcripts that accompany each episode, which have helpful links. You can find them at the link in the description below. Remember to subscribe on YouTube and your favorite podcast platforms. Cheers. See you next time.
