Brian Potter, an engineer and author of the Construction Physics blog, discusses why the construction industry has been slow to industrialize and innovate, covering topics ranging from Saudi Arabia’s Line megaproject to the aesthetics of modern architecture, the economics of prefabrication, regulatory bottlenecks like NEPA, and the future of automation in building.
Saudi Arabia’s Line: A One-Dimensional City That Makes No Sense
The Line is a proposed 170-kilometer-long linear city in Saudi Arabia, enclosed in mirrored glass, designed to house 9 million people in a single-file arrangement.
Brian argues it is physically and economically irrational: a one-dimensional city maximizes travel distance between any two points, unlike real cities that grow in two dimensions for efficient transport.
Even if you wanted an enclosed megastructure, a cube would be far more material-efficient due to better surface-area-to-volume ratios.
He speculates the project is driven by a desire to build something “big and impressive”—a pattern seen in Dubai’s megaprojects—rather than sound urban planning.
He references the “garbage can theory” of organizational decision-making: big bureaucracies often pull pre-made solutions from a grab bag rather than reasoning from first principles.
He would be surprised if it is ever built.
Why Construction Is Hard to Innovate
Construction shares traits with other project-based industries (software, film), where you figure out what you’re building while building it, unlike manufacturing where design and production are separated.
The key difference is scalability: software can be deployed to millions at near-zero marginal cost; a building serves only its site.
Construction is highly coupled—every system (electrical, plumbing, insulation, structure) touches every other, making changes disruptive, unlike modular products like computers.
Buildings lack physical modularity: you can’t swap out a building’s electrical system like a computer’s hard drive.
Margins in construction are thin, so even seemingly efficient methods (like prefab) rarely deliver dramatic cost savings.
Prefabrication Doesn’t Save as Much as You’d Think
In single-family home construction, roughly 50% of cost is materials, 50% labor.
Foundations and site work can’t be moved to a factory; material costs stay the same regardless of where you build.
Prefab factories reduce labor but add factory overhead and transportation costs, which are heavy and expensive for large building components.
The math rarely works out to dramatically lower costs; prefab succeeds mainly at the high end where margins are better and factory precision adds value (e.g., air-tight energy-efficient homes).
High transportation costs act like a fixed cost, which (via an Alchian-Allen type effect) shifts demand toward higher-grade, more customized options rather than commodity housing.
Brian pushes back on the idea that customization demand prevents mass production: empirically, builders successfully deliver repetitive housing (e.g., Levittown, apartment complexes), and the car industry shows you can offer enormous variety from a small set of variable features.
If Building Regulations Didn’t Exist
The obvious answer: build vastly more housing, especially in dense urban areas like San Francisco (e.g., fill in parts of the bay).
A more unique idea: redesign cities with layered, non-intersecting transportation networks (like multi-level flow systems) instead of flat grids where all roads cross—unlocking vastly greater throughput and scalability.
Cities’ economic output scales with how easily people can move through them; better transport design could unlock larger, more productive cities.
Countries That Are Best at Building Things
China (current): Enormous raw output—more concrete used in recent decades than in all prior history—plus rapid urbanization and ambitious construction.
Japan (late 20th century): At peak economic confidence, invested heavily in construction robotics and prefab housing systems; companies developed fascinating (if uneconomic) automated skyscraper factories.
United States (1850–1970): Built massive infrastructure (railroads, highways, suburbs) in a virtuous cycle: construction drove development of machine tools, interchangeable parts, auto industry, and further construction, advancing technology and raising living standards for over a century.
Japan’s 30–40 Year Building Cycle
Japanese buildings are often torn down and rebuilt every few decades.
To build for such short lifespans, you’d use very thin, cheap materials—essentially mobile-home construction taken further (e.g., thinner plywood, minimal walls).
Japanese homes already use lightweight interior walls and post-and-beam methods; shortening lifespan just means economizing even more on material.
China’s Real Estate Bubble
Brian is skeptical of claims about a Chinese real estate bubble because data from China is opaque and unreliable.
He notes the common narrative—ghost cities, unfinished towers—but admits he lacks an informed opinion.
Does Building More Housing Lower Costs?
In theory, housing cost is a supply-and-demand problem: build enough and prices should fall.
Japan is often cited as a counterexample: despite building lots of housing, costs didn’t collapse—but this parallels stagnant incomes, suggesting broader economic forces dominate.
Brian’s general expectation remains: building more housing should make it cheaper.
Georgism and Land Value Tax
Brian sees the logic: land is in fixed supply, so taxing it doesn’t reduce production, unlike taxing labor or capital.
A high land value tax in expensive cities (San Francisco, Seattle) would encourage more efficient land use (e.g., replacing single-family homes with denser housing).
He doesn’t claim deep expertise but finds the reasoning sound.
Advice for Building a Charter City
Don’t just replicate normal construction at scale; instead, use methods with high upfront costs that pay off over thousands of buildings.
Historical example: post-WWII tract builders (e.g., Levittown) used on-site mini-factories and reverse assembly lines to erect houses in ~30 days across thousands of units.
But top-down planned cities (e.g., Brasília) have a poor track record; cities work better when they respond organically to local conditions.
Better approach: well-designed land-use rules that encourage mixed-use development, avoid Euclidean zoning, and let the city evolve from the bottom up.
High upfront costs for a charter city could be justified if you’re building 10,000+ units, but it’s not strictly necessary—private developers have done it with loan guarantees and market incentives.
Technologies to Tighten Construction Tolerances
Buildings are built by hand on site, so placement and materials vary widely (wood warps, steel is slightly off, etc.).
Engineered lumber (e.g., CLT): Made from dimensional lumber glued into large, stable panels; more uniform and accurate than raw wood. CLT is often so precise it exposes inaccuracies in the rest of the building.
AR/VR: Brian is bullish—projecting exact component locations onto the site would reduce manual measuring time and errors. Caveat: requires very accurate building models, which are costly to produce for simple projects.
CNC cutting and synthetic wood products: Improve precision and uniformity of materials.
Precision is tricky: making one part ultra-precise can force costly adjustments elsewhere if other parts remain imprecise.
Redesigning Buildings for Automation
Current building materials (bricks, 2x4s) are sized for human handling, not machines.
To automate construction, you’d redesign components for machine production and assembly—e.g., continuous extrusion processes, snap-together panels.
Example: insulated metal panels (steel-insulation-steel) that lock together, combining structure, insulation, and barriers in one easy-to-install element.
Machines are more constrained in some ways but less constrained in others: e.g., 3D-printed buildings can easily make curved walls, which are expensive with conventional methods.
But curved interiors are less practical (furniture doesn’t fit, pictures are hard to hang), so human use patterns still shape design.
Why Brian Is Pessimistic About 3D-Printed Construction
Current 3D printers only produce walls, which are ~7–8% of a building’s value—they don’t print foundations, roofs, or mechanical systems.
Walls are already cheap and easy to build manually, so the value proposition is weak.
The promise depends on expanding the printer’s role to install windows, mechanicals, etc.—which is hard and not yet economical.
Companies like ICON have scaled back ambitions, focusing only on printed walls with other systems installed separately.
Until 3D printing handles a large fraction of a building’s value, it won’t fulfill its transformative promise.
Rising Labor Costs and Construction
Construction labor costs rise because labor is more productive in other industries, pulling wages up—but construction hasn’t offset this with productivity gains.
Labor remains a huge fraction of total building costs, and those costs keep rising with inflation.
Why Government Involvement Raises Construction Costs
Regulation adds steps, delays, and uncertainty to every process.
Even if individual rules seem reasonable, the cumulative effect is that any task takes longer and is less reliable.
Brian doesn’t offer a unified theory beyond: more steps = more cost.
Undocumented Labor and U.S. Construction Costs
Much U.S. residential construction relies on undocumented Latino workers paid below market rates.
This likely suppresses labor costs, making U.S. residential construction cheaper than in countries like Germany.
Residential construction in the U.S. is notably inexpensive by global standards.
AI and Robotics in Construction
Brian expects gradual progress: robots will take over narrow tasks first (e.g., layout robots that draw wall lines on concrete slabs), then expand.
Computer vision + AI + robotics will let robots handle an increasing share of construction tasks over time.
He doesn’t have strong opinions on Adam Neumann’s Flow (a16z-backed real estate startup), noting that venture capital often makes high-variance bets he doesn’t fully understand.
Why Construction Is More “Closed Source” Than Software
Construction is a craft-based industry: expertise is embedded in judgment and relationships, not codified in explicit instructions.
Building drawings are far sparper than manufacturing specs for something like a Toyota Corolla.
Physical objects are harder to reverse-engineer than software: even if you have the product, the process to create it is often hidden (e.g., industrial espionage with atomic bomb plans didn’t shortcut development).
Software is more legible: you can inspect every part, fork it, run experiments, and test subcomponents at low cost—things you can’t do with a building.
If Brian Could Build Anything Without Regulations
He’d build a mile-high skyscraper—a dream of early-to-mid 20th-century architects (e.g., Frank Lloyd Wright’s “The Illinois” for Chicago).
The technology exists; the barriers are economic and legal, not physical.
Even in places without height limits (Dubai, China), the economics are so brutal that no one has done it.
Why Don’t Developers Lobby to Remove Height Restrictions?
The natural opposition to big projects (existing residents worried about shadows, congestion, property values) is at least as organized as the constituencies supporting them.
In wealthy areas where housing is most needed, residents have the most resources and incentive to block new development.
Brian speculates that a Coasean solution—paying residents to accept development—would be more efficient, but political influence determines outcomes instead.
NEPA: The National Environmental Policy Act
NEPA requires major federal actions with significant environmental impacts to undergo lengthy, expensive environmental impact studies (average: 4.5 years; some over 8 years).
It’s a procedural documentation requirement, not a substantive environmental protection law: you can proceed with environmentally harmful projects as long as you’ve documented the impacts thoroughly.
It creates enormous uncertainty because requirements shift over time (e.g., greenhouse gases weren’t originally considered a major impact) and anyone can sue, claiming the analysis wasn’t thorough enough.
Lawsuits can’t permanently stop projects but can delay them until they become uneconomical.
NEPA was poorly drafted: it created no enforcement bureaucracy; courts ended up enforcing it due to judicial activism in the 1970s.
Reform ideas: give clean energy (wind, solar, transmission) the same categorical exclusions that oil and gas enjoy; create a centralized bureaucracy (like OMB) to assess environmental costs.
Timeline caps on reviews might backfire: if agencies rush, courts may rule they didn’t take the required “hard look,” forcing reanalysis and further delays.
Other countries have environmental review systems but typically enforce them through bureaucracies, not citizen lawsuits.
Should We Add a “Posterity Review” for Long-Term Impacts?
Brian is skeptical: any additional review layer makes processes slower and less likely to succeed.
Predicting long-term impacts is extremely unreliable (e.g., 1970s peak oil predictions were wrong due to unforeseen technological advances).
He doesn’t hold this as a strongly held opinion but doubts such reviews would be accurate or useful.
Is There a Talent Drain from Construction to Software?
Yes: talented engineers leave construction for software, where pay is much higher.
This is theoretically self-correcting: as labor becomes scarcer and more expensive, the industry is pushed toward labor-saving methods (prefab, modular assemblies).
But there’s a risk: losing the top 20% of talent could fundamentally handicap the industry’s capabilities.
Similar dynamics affect semiconductor engineering, where talent flows to software.
Construction also faces a skilled labor shortage: many workers left during the Great Recession and didn’t return; the average age of construction workers is in the 40s–50s.
Historically, labor scarcity drives labor-saving innovation (e.g., American system of manufacturing), so Brian is cautiously optimistic.
Gaps in the Blog Marketplace of Ideas
Brian wishes for more blogs that deeply document how complex industries actually work—especially manufacturing.
Much industrial knowledge is undocumented, locked in companies’ internal drives or people’s heads; if it’s lost (e.g., moved overseas), it’s hard to recreate.
He’d love to see someone write a comprehensive guide to how things are built in China—no such resource exists despite many people needing it.
Documentation has public-good properties (costly to produce, uncertain payoff), so it’s underprovided, but there’s real demand for it.
Why Is Modern Architecture So Ugly?
Brian reframes the question: it’s less about ugliness and more about the loss of ornamentation.
Modern buildings use minimalist aesthetics (glass, concrete, simple forms) partly because new technologies (float glass, air conditioning, steel structures) made large, open, light-filled interiors possible and desirable.
Glass curtain walls are expensive but prized for the interior experience (natural light, openness), even if exteriors look bland.
Air conditioning was a prerequisite: without it, glass buildings would overheat like greenhouses.
Ornamentation declined partly because it’s labor-intensive and expensive in high-wage economies (e.g., hand-carved stone masonry is economically unfeasible in the West).
There’s a feedback loop: less demand → fewer skilled masons → higher costs → even less demand.
But cost isn’t the whole story: some expensive modern styles (e.g., International Style skyscrapers) were adopted purely for aesthetic reasons, and owners pay extra for all-glass facades despite higher mechanical costs.
Brian is skeptical that new tech (3D printing, CNC, AI) will automatically produce better aesthetics: Victorian ornamentation was itself a product of mass production, and tastes shifted away from it for practical reasons (dust, maintenance, water shedding).
He doubts that giving architects unlimited formal freedom would lead to clearly “better” art, citing how modern art hasn’t obviously improved despite new tools.
On the “architect cabal” theory (that modernist guilds impose unpopular designs): Brian thinks there’s some truth—architects are trained in art-adjacent schools and develop distinct tastes—but most architects are market-driven and must satisfy clients to stay in business.
Advice for Aspiring Engineers and Innovators
Go to a good engineering school and work at top-tier firms—these signals matter in the building industry.
Alternatively, enter via software: many construction startups need developers, and software skills are transferable (with high fallback salaries).
Work at startups tackling construction innovation: green building, low-carbon concrete, low-carbon steel, construction robotics, prefab.
The built environment space is attracting more venture capital and technological experimentation, so opportunities are growing.
