Payment Required.

A Scenario Deconstruction from the Year 2036

March 25, 2026

The House

There is an image I keep returning to. Today is a January morning. Ten years from now. So we have 2036. Outside, grey sky, four degrees, tripping rain, and wind pressing against the glass. Behind the shell, tomatoes and cucumbers ripen next to fig trees and intensely fragrant rosemary. The house sits inside a greenhouse, a timber frame wrapped in a steel-and-glass dome that turns winter in this part of my home into something that feels like a Mediterranean climate.

I have been thinking about something for a while: What happens when a building no longer needs an institution to function? No energy provider that bills monthly. No platform that brokers the intelligence. No bank that processes the payments. If a house could earn its own money, pay for its own computer systems, and grow its own food, the relationship between humans, machines, and the institutions that currently push themselves so manically in between would change. The question is whether this is technically possible, and what follows from it if it is.

My idea itself isn’t new. There was a Swedish architect named Bengt Warne who built the first of these houses in 1976 and called it Naturhus. What is new is what stands in a particular room.

Two machines. A mining ASIC in a heating enclosure that does nothing but compute SHA-256 hashes and convert electricity into heat and hashrate. And next to it a small GPU server running a local model that manages the house. The miner produces money and the server spends it.

And just now the grid price has gone negative.

The miner is running at full capacity. The cool thing about it is, not because someone switched it on, but because the AI cross-referenced the 48-hour energy price forecast (negative prices for the next three hours, then rising), the greenhouse camera feed (the image recognition model detected early blight on two tomato plants yesterday and recommended raising the temperature to slow the fungal spread), and the household calendar (nobody home before the evening, so the living room heating can stay off for now). All the miner's waste heat is being redirected into the greenhouse. The living room will be preheated thirty minutes before the rest of the family arrives, timed to the last cheap hour.

A thermostat reacts to temperature. This system thinks about price forecasts, plant pathology, and my annoying schedules.

Heat rises through ducts into the greenhouse. Hashrate flows to a mining pool that requires no account and no identity. Sats flow back via Lightning, credited to the house's wallet. The AI pays for its own inference with these sats. When it needs more capacity than the local model can deliver, it queries a cloud model through a protocol that requires payment. No identity. No human vouching for the machine.

This house produces its own money. It pays for its own intelligence and it grows its own food.

The system exists today in individual pieces. The hardware is being shipped. The protocols are already running. The code is open source. What does not yet exist is the integration: a building that runs the complete loop. That is what I want to think through here.

For me there are five layers that would all need to be true simultaneously for the house to work. For each one, the question is: Is it already true? If not, what is missing?

Why 2036? Two more Bitcoin halvings (2028, 2032) and just before the third additional one, will show whether mining actually decentralizes. Ten years of AI development will show whether local models become good enough. Community energy projects need at least five years to prove themselves. 2030 is too early for most parts. 2040 or later is already so far away that any scenario becomes unfalsifiable. I am not sure about the date. But about any other date I am even less sure.

The heat

The standard reading of Bitcoin mining is (and we all surely know this) that Bitcoin is an energy waster. I consider this reading wrong, or at least too unreflective. Economically you could say, it comes down to whether a miner is a cost factor or a product. If it is a product (heat plus Bitcoin), the economics of an autonomous house become possible. And the hashrate secures the network that the entire loop runs on. If it is just a cost factor, they do not.

The house in the greenhouse does not need much external energy. The glass shell reduces heating demand by about 30%, a figure cited by Anders Solvarm, who developed the modern Naturhus projects, as the average across his builds (Greenhouse Living). But it still draws electricity for the miner, for the AI, for lights and pumps. The question is where this electricity comes from and what happens with the surplus.

Germany alone recorded 573 hours in 2025 during which wholesale electricity prices fell below zero. An increase of 25 percent over the previous year (Bundesnetzagentur, SMARD electricity market data 2025). The IEA puts it at 8 to 9 percent of all trading hours in the first half of 2025 for Germany, the Netherlands, and Spain, doubled compared to the previous year. Sweden, France, and Finland show similar levels (IEA, Electricity Mid-Year Update, July 2025, p. 26). Negative prices mean the grid pays you to take electricity. This happens when renewable generation exceeds demand and storage capacity. Most of this surplus is curtailed or wasted. These hours are still a small share of the total, so they are a bonus. But the trend is clear and accelerating.

A Bitcoin miner is the most flexible load imaginable. It ramps from zero to full capacity in seconds and does not care when or where it runs. In a building that needs heat, the miner is a heater that produces Bitcoin on the side.

Finland proves the concept at scale. Marathon Digital launched a pilot in the Satakunta region in mid-2024, heating 11,000 residents with waste heat from water-cooled ASICs fed into the existing district heating network (MARA, Heating with Hashes, June 2024). Within eighteen months they heat 80,000 residents' homes (Grist, These Finnish homes are being heated by a surprising source: Bitcoin, December 2024). Water runs at 50 to 78 degrees Celsius through the miners, hot enough for residential heating. Marathon earns from both mining and heat delivery and saves an estimated 5,000 tonnes of CO₂ from peat and biomass.

But Marathon is a publicly traded company with bank accounts, municipal contracts, and a legal department. They don't need permissionless infrastructure. The conventional financial system works perfectly fine at their scale.

The house in the greenhouse has none of that. No company behind it and no grid contract and no finance department. The AI that operates the miner cannot open a bank account. The homeowner does not want to become a registered energy trader to sell waste heat to three neighbors.

The heat is proven. The question is: What kind of money flows through it and who controls the system?

There is a counterargument I go back and forth on. A heat pump produces three to four times more heat per kilowatt. Mining-as-heating does not replace a heat pump. In winter, both run during cheap or negative price hours: the heat pump for efficiency, the miner for Bitcoin and additional heat. In summer, when no heating is needed, the heat pump sits idle, but the miner still has a reason to run, because it absorbs solar surplus that would otherwise be wasted and earns Bitcoin. The AI decides when what runs. That makes optimization harder, which is another argument for having the AI in the first place.

And there is a second problem. Mining difficulty trends against the small miner. Every terahash earns less as difficulty rises. A household miner in 2036 competes with industrial operations. Whether this kills the concept or reshapes it depends on whether large operations become less profitable or their interests shift (as is already happening with miners pivoting to AI compute) and whether mining could then decentralize toward households where heat is the primary product and Bitcoin a byproduct. I cannot predict this. What I can say is that the architecture I’m describing does not depend on mining being lucrative. It depends on mining being cheaper than heating with electricity alone. That’s a different threshold.

The money without guarantors

There are two needs that sound different but require the same answer.

Machines need money that nobody has to vouch for. Every time a machine pays for something today, there is a human or a company behind it, holding the account and bearing liability. The machine is a dependent actor. Bearer tokens change that. A token that is valid because it is what it is (and not because of who holds it) needs no account and no guarantor. The machine simply pays.

Humans need money where they are not bound to institutions. If a bank can freeze your account, or if a platform can lock your wallet, or if a state can devalue your savings, the money belongs to you only on the condition of the institution's continued goodwill. Bearer tokens remove that condition.

Both needs are served by the same infrastructure. And this infrastructure runs on Bitcoin. The reason is specific: it is the only settlement layer that has been running for sixteen years without downtime, without a company behind it, without anyone's permission. That is also the reason why companies do not voluntarily build on it. A system that nobody controls is a system that nobody can monetize.

A bit of history, because I think it will be useful for what follows. In 1982, David Chaum invented blind signatures, a method that allows an issuer to sign a token without knowing who receives it. Perfect privacy for digital payments. He built DigiCash, spent the nineties trying to convince banks to adopt it. In 1998 DigiCash went bankrupt. The cryptography was brilliant, however, the dependency was fatal, because the system needed banks and banks had no interest in infrastructure that removed them from the transaction.

In 2022 Calle revived the idea: Chaum's blind signatures, rebuilt on Bitcoin. No more banks in the loop. The mint can be anyone. Regardless of it being a server, a community, or maybe a federation of neighbors.

The opening scene of this text used Lightning: the miner performs computation, the pool pays out sats. That works for a single house. Without an account and without KYC, this already works today at Ocean: enter a Bitcoin address and mine. The pool sees your address, your hashrate and can link both. Parasite Pool, a project from the Ordinal Maxi Biz orbit, goes one step further: Lightning payouts without registration, with somewhat less linkability, but still not anonymous. Hashpoolgoes further still: the pool never sees where the tokens flow, similar to a cashier handing out change without knowing which wallet it ends up in. Hashrate in, bearer tokens out. As far as I could determine, Hashpool is the only project that combines mining with structurally anonymous payouts.

For now: the infrastructure exists in individual pieces. Ecash was rebuilt on Bitcoin in 2022. A mining pool has become accountless (on testnet). Community custody is getting easier. Whether it is easy enough for people who do not read documentation for fun is another question.

The status code

In 1997, the Internet Engineering Task Force published RFC 2068 and defined the HTTP status codes that structure the web. 200: OK. 404: Not Found. 402: Payment Required, with the note: "Reserved for future use."

For twenty-nine years, 402 had no implementation. The web monetized through advertising, subscriptions and data extraction. The idea that a server could say "pay me" and a client could pay was never technically worked out and never economically attempted. What was at stake, even though nobody phrased it that way at the time, was the question of whether machines could ever conduct transactions independently.

Then in 2026, three independent implementations appeared almost simultaneously.

Lightning Labs released the L402 Agent Toolkit, which is a protocol that allows any server to require a Lightning payment before delivering a response. It's permissionless, with no account on either side needed. Coinbase launched x402. It's account-based, but adresses the same structural problem: machines paying machines. The x402 Foundation now includes Google and Visa. Stripe released its new blockchain Tempo with its own standard MPP, which included Lightning from day one. One of Stripe's advantages is that it already has millions of internet-native merchants integrated into its infrastructure. And Cloudflare, which routes a substantial share of global web traffic, reports that its customers are already sending over one billion 402 responses per day, mainly to AI crawlers that demand content without paying (Cloudflare Blog, August 2025). Over one billion times a day, a server says to a machine: Pay me. The machine has no way to respond.

The empty status code is filling up simultaneously from multiple directions and independently of one another.

For the house, the specific protocol does not matter as long as it's permissionless. What matters is that any device can query any model from any provider by attaching a payment to the request. Anonymously. The tokens the house earned through mining pay for everything.

And "everything" is more than building management. The people in the house use AI just as much for their work and for learning as for controlling the greenhouse. An image recognition model monitors plant health, an optimization model controls energy flows, a language model handles whatever the people ask. The local language model serves as orchestrator: it holds the context of the house, decides which specialized model gets called, and manages the wallet. The other models are tools it calls. What they all share is the payment method.

Local models handle the routine, without calling an external API. By 2036 they will be significantly more capable. The development points toward more local tasks: fewer cloud queries, fewer expenses, more autonomy. The house benefits from inference that never leaves the building.

In 2026, a single mining heater produces enough tokens for a few dozen cloud queries per day. Inference costs are dropping, local models are taking over routine tasks, and the heat value of the miner offsets the majority of its electricity costs. Whether the loop actually closes depends on mining revenue and inference costs, and both are moving. If everything goes well, the house earns more than it spends. I’m not certain it does.

Bearer tokens do not make the house invisible. A cloud provider can still see the IP address, analyze request patterns, correlate timing. What bearer tokens prevent is linking payment to identity. The provider knows that someone paid, but not who. I’m aware, that’s not total privacy, but the difference between "the provider chooses not to track you" and "the provider cannot link your payment to your identity" is structural. Bearer tokens are the second category.

The system wants neighbors

So far this is a single house. One miner, one greenhouse, one AI, one wallet. The economics are modest and the loop works, but it is small. And to be honest about it: the economics of a single house are thin. A mining space heater costs a few thousand dollars and generates modest returns. The payback period takes years. The economic argument is not return on hardware, instead it is that mining funds the AI that runs the house. An autonomous intelligence layer, funded by computation, without subscription or platform dependency. The value lies in the loop.

A federated mint makes no sense for a single household. For a neighborhood, however, it does. People who know each other, who live nearby, who have a reason to cooperate. The mint issues ecash, backed by the Bitcoin that the neighborhood's miners produce, and nobody alone can control it. The trust in this case is not trustless but communal.

But the social structure is what I find most interesting here. The money layer dissolves dependence on institutions. The social layer builds identity back up, at the community level. The same system that allows a machine to pay without anyone vouching for it requires humans to vouch for each other in order to custody the money. The system was designed this way.

As someone from the cultural sciences, I am cautious with any argument that reduces community to economics. And I have encountered this in Bitcoin conversations more often than I would like. So I want to be precise. People join cooperatives because the economics are better together, but they stay because of the relationships they build. Every community needs both: a shared economic reason to start and a shared story that holds it together. Fedimint, an open-source protocol for community-managed Bitcoin custody with ecash issuance, is built exactly for this. It is the economic reason. What the community makes of it is the story. The question is whether a technical architecture can generate social bonds, or whether it only works where bonds already exist. Fedimint requires social trust that most neighborhoods don't have. And I cannot answer whether economic incentives can generate trust or only presuppose it.

Nobel prize winner Elinor Ostrom studied exactly this (Governing the Commons, 1990). Her design principles for long-enduring commons institutions read like a specification for Fedimint governance: clearly defined boundaries, rules adapted to local conditions, participatory decision-making, active monitoring, graduated sanctions, low-cost conflict resolution. Ostrom showed that communities can manage shared resources without privatization or state control, if the governance fits the community.

Shared infrastructure requires shared decisions. A community mint requires neighbors who talk to each other. A shared greenhouse requires someone to harvest the tomatoes. The AI handles the optimization so that people can tend to the relationships. The less time you spend on energy bills and price tracking, the more time you have for the people next door. That is the opposite of the platform model, where technology connects you with strangers far away and isolates you from the people nearby.

The system favors owners. A roof for panels, a shed for machines, equity to invest. The community layer addresses this partially (shared infrastructure, cooperative models, apartment buildings that run shared installations in the utility room). But the opening image is a house. I have not resolved this tension and I am not sure it can be resolved within the architecture alone.

The question of geography 

The method has limits, and here I reach them.

For the first four layers I could work backwards: What would need to be true? Some of it already is. Some is in progress. Some is missing.

But if these four layers work, something follows from them that cannot be constructed backwards. It can only be speculated forward.

The system works best where renewable energy is abundant. Solar at southern latitudes, wind on coasts and ridgelines, geothermal where the geology permits, micro-hydro near streams. And it works best communally. Fedimint needs neighbors, community energy needs participants, shared AI needs scale. Small communities at energy-rich locations, adapted to their climate. Connected clusters of houses that generate their own energy, mint their own money, pay for their own intelligence, grow their own food or cool their own air.

The building itself does not have to be a timber frame in a glass dome. An earth-sheltered house built into a hillside inverts the thermal logic. The earth insulates from three sides at a constant 10 to 12°C year-round. In summer, the hillside acts as a heat sink. And the system is not limited to cold climates. An absorption chiller (a cooling technology that runs on heat instead of electricity, industrially deployed for over a century) can convert mining waste heat into air conditioning. In northern Europe, tomatoes grow. In Senegal, the air conditioning runs. The architecture differs. The economic loop is the same.

But Fedimint does not scale to thousands (the guardians need to know each other). Local energy does not scale to megacities (density competes with area). Community governance becomes political above a certain size, and then you need institutions again. The principle holds: the system works where people know each other by name.

The question I find most interesting in this entire exercise: Can megacities function in this model at all? Or is the system inherently decentralized, not just in code but in geography? Does the logic of bearer tokens and community mints and local energy imply a redistribution of where people live?

We cannot answer that yet. The first four layers are engineering problems. The fifth is a civilizational question. And it might be the reason why the other four matter at all.

402

Back to the house.

Wind outside. Not a storm, but steady, the kind that makes the glass shell creak. Behind it, the greenhouse is warm. Diffuse light through the dome. Fig trees thriving. The tomatoes are fine, because the AI took care of them hours ago.

You make coffee. Sit down at the table by the large window. The landscape outside is wide and grey and beautiful. You open your laptop to write. You do not check an energy bill. You do not log into an AI platform. You do not think about the miner in the basement, because you do not have to. The system hums. It earns, it spends, it heats, it grows. You live in it like in a house with running water: you just turn the tap and it works.

Your neighbor knocks. Strawberry surplus next week, whether you want some. You say yes. Later, a few sats move from your wallet to hers. Nobody is in between. The smallest possible economic act between two people who know each other by name.

At some point along the way, the signs reversed. The house produces its own intelligence instead of trading data for it. It produces energy instead of merely consuming it. And the system that runs all of this connects you with the people next door, not with strangers on a platform. None of this required external permission. It required infrastructure that works without permission.

In 1997, the IETF reserved a status code. 402: Payment Required. "For future use." For almost thirty years, nothing happened. The web chose advertising. Then, within a few months, providers and dozens of open-source projects began building answers to this empty code.

The parts exist. The question is whether someone puts them together, or whether they assemble themselves because the incentives are right.

And whether what emerges from this is not just a technical system, but a way to live.