Jaynit posted a clip of Marc Andreessen making an argument I keep coming back to (Jaynit's post). Technology collapses prices. TVs, phones, computers, anything electronic gets better and cheaper every year. But three things went the other way: housing, healthcare and education. Marc's explanation is that those are the sectors where government restricts supply and then subsidizes demand. You get fewer houses, fewer doctors, fewer accredited schools, and more money chasing them. Prices go up.

I think he's right. I also think the pieces that keep housing expensive are about to get hit from several directions at once. I wrote a short thread about it (here on X). This is the longer version.

How AI breaks the housing wall, three ways.

The way I see it, housing sits behind three walls. Labor, paperwork and distance. AI and robots go after all three.

Wall one: labor

A humanoid robot can use a hammer like I do.

That sounds obvious, but it's the whole point. We spent decades trying to automate construction by redesigning the house around the machine. Prefab panels, special factories, custom printers that pour concrete walls. Some of that works. But most houses are still built by people with ordinary tools on an ordinary job site, and that's not going to change overnight.

A humanoid doesn't need the job site redesigned. It's shaped like a person, so it picks up the same nail gun, climbs the same ladder, carries the same sheet of drywall. It fits into the process we already have instead of asking us to rebuild the process.

And it doesn't go home. It can frame at 2 AM. It doesn't get hurt, doesn't quit halfway through a project, doesn't have a bad week. Anyone who has tried to get a contractor to show up on time knows how much of the cost of building is really the cost of coordinating people. Waiting on the framer, then the electrician, then the drywall crew, each one on their own schedule.

I'm not going to pretend I know the exact year this gets good enough. Robots on real job sites are still early. But the direction is clear. When skilled labor stops being scarce, the shortage that drives up building costs goes away, and so does a big piece of the timeline.

Wall two: paperwork

Even if you could build a house in a few weeks, you can't start until someone says yes. Plans, engineering, code review, permits, inspections, comment letters from the city, revisions, resubmission. Most of that time isn't spent doing anything. It's spent waiting on paperwork sitting in a queue.

This is where AI agents are already useful today, not in some future. An agent can draft plans against the local code, check them before anyone at the city does, fill out the permit application, and answer the reviewer's comments overnight instead of two weeks later. The applicant side of bureaucracy is mostly reading long documents and filling out forms correctly. Models are very good at exactly that.

Here's the part I find interesting. Once every applicant shows up with a complete, correct, code-checked package, the approving side gets flooded. A permit office that was used to slow, messy submissions suddenly gets far more clean ones than it can handle. It can't hire its way out of that. It has to adopt the same tools. Then the review that sat on someone's desk for days takes minutes.

Paperwork friction was a rate limiter. Sometimes on purpose, sometimes just inefficiency. AI removes it on both sides of the counter.

Wall three: distance

This is the one I think people underrate.

Land isn't actually scarce. Land near jobs is scarce. Housing in and around cities is expensive because everyone needs to be close. Close to work, close to stores, close to the hospital, close to the airport. Living an hour and a half out means losing a big chunk of your day in a car and giving up a lot of convenience.

Driverless cars change that math. If the car drives itself, a long commute becomes time you can work, read or sleep. The distance is still there. It just stops costing you your attention.

Drone delivery does the same thing for errands. If the pharmacy, the hardware store and dinner can come to you in minutes, being far from town stops feeling like a sacrifice. You don't need to live near the city when the city comes to you.

Put those together and the supply of land that's actually livable gets much bigger, without changing a single zoning rule. That's the quiet way around Marc's supply restriction. You don't have to fight to build more in the places that are locked up if far more places become good places to live.

Housing goes the way of the TV

Stack the three together. Cheap land because distance stops mattering. Robot labor that works around the clock. AI paperwork that turns months of back and forth into days.

That's the same recipe that made electronics cheap. When the inputs get cheap and fast, the product follows. I think housing can go the way of the TV. Something that used to be a huge, once-in-a-lifetime expense becomes a product that gets better and cheaper over time.

The honest caveat

None of this removes the final gate. A city can still say no. Zoning can still cap how many homes go on a lot. Neighbors can still show up to the hearing. AI can make the application perfect and the review fast, but it can't make the decision for a government that doesn't want more housing. That part is political, and it's the hardest wall of all.

What I do think changes is the excuse. Right now "it's expensive and slow to build" is just true. When labor, paperwork and distance stop being the bottleneck, the only thing left standing between people and cheaper housing is the rule itself. That makes it a lot more visible, and a lot harder to defend.

I'm a builder, so I'm biased toward thinking the tools win. But I'd rather bet on that than on prices going up forever.