From Parts 7 and 19 · The World We Want
The Map
Line up the transitions and two things jump out.
Historians argue about exactly when each of these systems started and ended, so treat the dates below as rough markers.
The systems that ran the world
System 1 of 7
Foraging bands
For most of human history, people lived in small bands that hunted, gathered, and shared. Power was personal and limited, because there was little surplus to control.
How it ended
As the last Ice Age ended and the climate stabilized, people in several places began domesticating plants and animals, starting around 9500 BCE in the Fertile Crescent. Farming created something new: a storable surplus. And a surplus needs someone to guard it, count it, and eventually tax it. The first states appeared around 3100 BCE.
Transition length roughly 6,000 years, from the first farms to the first states.
System 2 of 7
Agrarian empires
For the next few thousand years, power came from controlling land and the people who worked it. Empires from Egypt to Rome ran on grain, conquest, and, in many cases, slavery.
How it ended (in Western Europe)
The Western Roman Empire overextended, debased its currency, suffered waves of plague, and gradually lost the ability to collect taxes across its territory. When central taxation collapsed, power fragmented into the hands of local landholders. Rome fell in 476. By around 900 to 1000, a new order had taken shape.
Transition length roughly 400 to 500 years.
System 3 of 7
Feudalism
Under feudalism, land was granted in exchange for military service, and most people were serfs bound to the land, owing their lords labor and a share of their crops.
How it ended
A shock, then a squeeze. The Black Death (1347 to 1351) killed perhaps a third to half of Europe’s population, and suddenly labor was scarce. Lords tried to freeze wages by law, with England’s Statute of Labourers in 1351, and failed. Peasants revolted in 1381. Serfdom slowly dissolved in the West as workers moved to whoever paid most. Meanwhile, new trade routes opened, and merchants built a new tool for pooling capital: the joint-stock company. The English East India Company was chartered in 1600, and the Dutch East India Company in 1602. A joint-stock company lets many investors pool their money into one venture and split the profits by how many shares they own. If a ship sank, everyone lost a little, instead of one merchant losing everything. It wasn’t new. England’s Muscovy Company, chartered in 1555, was the first major one. But it took the new trade routes to make it the standard way to fund a voyage.
Transition length roughly 250 years (about 1350 to 1600).
System 4 of 7
Merchant capitalism
In the age of merchant capitalism, power came from trade: chartered monopolies, navies, colonies, and states that hoarded gold and silver.
How it ended
Technology changed what was scarce. Machines, and later steam, made power cheap, and the new constraints became capital and workers to run the factories. Enclosure had pushed people off common land and into cities. Capital from trade financed the first mills. Adam Smith’s Wealth of Nations (1776) gave the new system its ideas. And the tax base shifted to match: Britain introduced an income tax in 1799 to pay for war with France, and Robert Peel brought it back in 1842, in part so Britain could afford to cut tariffs. An income tax takes a share of what people earn each year, instead of taxing goods as they cross a border or get sold. Britain’s first was a wartime experiment, abolished in 1816, a year after Waterloo. Peel’s version stuck, and Britain has had one ever since. When Britain repealed the Corn Laws in 1846, free-trade industrial capitalism had won. Limited liability followed in the 1850s, making it far easier to raise capital from many small investors. Limited liability means investors can lose only what they put into a company, never their house or their savings. Before then, it usually took a royal charter or a special act of Parliament. Britain’s Limited Liability Act of 1855 opened it to ordinary companies.
Transition length roughly 75 years (from 1771, when the economist Carlota Perez dates the start of the first industrial revolution, to 1846).
System 5 of 7
Laissez-faire industrial capitalism
The age of railways, steel, and electricity concentrated power in the hands of factory owners, railroad barons, and industrial trusts, with little regulation and few protections for workers.
How it ended
Mass production (the Model T arrived in 1908) changed the economy, and then a string of shocks broke the old order: World War I, the Great Depression, and World War II. Ideas and prototypes that had been waiting got adopted fast. The US created a permanent federal income tax with the 16th Amendment and founded the Federal Reserve, both in 1913, then created Social Security and its payroll tax in 1935. Both had been tried before. The US first taxed income in 1861, to pay for the Civil War, and again in 1894, until the Supreme Court struck it down the next year. The 16th Amendment settled the question. Social Security is insurance for old age. Workers pay in through a payroll tax while they work, and draw a pension once they stop. Germany had built the first national version in 1889, under Otto von Bismarck. The Depression made it standard in the US. The Bretton Woods agreement rebuilt the global monetary order in 1944. Dalio marks 1929 to 1945 as the last great breakdown period of his Big Cycle.
Transition length roughly 30 years (1914 to 1945).
System 6 of 7
Managed capitalism
After World War II came big corporations, big unions, and big government, balancing one another. This was the golden age of the mass-production revolution.
How it ended
In 1971, Intel released its first microprocessor, which Carlota Perez marks as the start of the information revolution, and the US ended the dollar’s convertibility to gold. Oil shocks in 1973 and 1979 brought stagflation, and the managed model couldn’t fix it. Again, ideas were waiting. In 1970, Milton Friedman had argued that the social responsibility of business is to increase its profits, and that became the doctrine of shareholder primacy. Shareholder primacy is the idea that a company exists mainly to make money for its owners, the shareholders, so its managers should be judged by the stock price. It wasn’t new either. In 1919, Michigan’s Supreme Court ruled in Dodge v. Ford Motor Co. that a company is run “primarily for the profit of the stockholders.” The crises of the 1970s made the idea standard. Margaret Thatcher (1979) and Ronald Reagan (1981) cut taxes and deregulated, and the fall of the Berlin Wall in 1989 ended the main alternative. I’m arguing something different: the social responsibility of society is to create abundant universal basic goods and services, in service of maximum human agency and safety, so that human society can flourish until the end of time.
Transition length roughly 18 years (1971 to 1989).
System 7 of 7
Financial and information capitalism
This is the system we live in now. Power comes from capital markets, platforms, and data.
Next If that holds, the next one takes roughly 5 to 11 years. If the current handover (the part of the transition we’re in now) started somewhere around 2020 to 2022 and runs 5 to 11 years, the next system gets codified somewhere between the late 2020s and the early 2030s.
The pattern
Roughly 6,000 years, then 450, then 250, then 75, then 30, then 18.
- Foraging bands to Agrarian empires6,000 years
- Agrarian empires to Feudalism400 to 500 years
- Feudalism to Merchant capitalism250 years
- Merchant capitalism to Laissez-faire industrial capitalism75 years
- Laissez-faire industrial capitalism to Managed capitalism30 years
- Managed capitalism to Financial and information capitalism18 years
- Financial and information capitalism to the next system projectedroughly 5 to 11 years
- Feudalism to Merchant capitalism250 years
- Merchant capitalism to Laissez-faire industrial capitalism75 years
- Laissez-faire industrial capitalism to Managed capitalism30 years
- Managed capitalism to Financial and information capitalism18 years
- Financial and information capitalism to the next system projectedroughly 5 to 11 years
From here to there, in five stages
The stages overlap, and the dates are estimates, not promises.
Stage 1 The machine intelligence age (now to about 2030)
- What changes
- Intelligence on demand is everywhere. AI handles a growing share of knowledge work, and the first services get automated. Wages start to stall in the most exposed jobs.
- What to build
- The prototypes. Pilot towns, in pairs: one with spare food and one with spare energy, each with about 1,000 to 5,000 people and two to three years to prove the model.
Each pair runs the simplest version: equal shares, a free first block of the essentials, and maintenance before dividends.
Alongside them, Data Foundries, the idea from my first post on Futurespective: places where people get paid to train robots and keep a share of what the robots earn. They’re a bridge for the years before the floor exists.
Legal templates for parties (single-purpose entities where the people doing the work are partners) and guilds (open-admission craft organizations with a charter obligation to train anyone who wants to learn).
Community ownership trusts that can hold infrastructure. And the first round of tax reform: the formula in Every Spending Bill Should Show Up on Your Paycheck. - Signposts
- five numbers that show whether AI is starting to do the work.
Labor’s share of national income. The part of everything the economy earns that goes to workers as pay, instead of to owners as profit. If it keeps falling, machines are doing more of the work. In the US, it hit a record low of 52.8% in mid-2026.
What matters for the triggers below is the trend, not the level: a steady fall is the signal.
The share of tasks AI can do from start to finish. Not helping a person, but finishing the whole job. When that share rises, jobs stop being assisted and start being automated. Company filings give a second count: AI revenue as a share of the knowledge-work wage bill.
The employment gap for young workers in AI-exposed jobs. Entry-level jobs are where automation shows up first. Stanford’s Digital Economy Lab tracks the gap in its public AI Economic Indicators.
Payroll tax receipts as a share of federal revenue. How much of the government’s money comes from taxing paychecks. If that share shrinks while the economy grows, the old tax base is eroding.
Long-term real interest rates. What it costs to borrow for decades, after inflation. Trevor Chow, Basil Halperin and J. Zachary Mazlish show that if markets expected transformative AI, these rates would rise sharply, because people would borrow against a much richer future.
At 30% annual growth, real rates would top 30%; ten-year real rates in their developed-country sample have never gone above 5%. If the headlines are panicking and real rates are calm, the market isn’t buying the fast path. That includes my forecasts. - What switches on
- Switch A, the first layer of the safety net. It has no trigger: we build it now, while things are calm.
Unemployment insurance for gig and contract workers. Jacobs and Imas note that US unemployment insurance typically excludes gig workers and freelancers.
A monthly earned income tax credit, including workers without children. The earned income tax credit (EITC) tops up the pay of low earners. Today it largely leaves out workers without dependents, and it arrives once a year at tax time.
Employer-led apprenticeships. In Jacobs and Imas’s reading of the evidence, public retraining programs have historically disappointed, while apprenticeships and employer-led models succeed “by a considerable margin.”
Here’s the thing: we’ve already done most of Switch A. Temporarily.
In March 2020, the CARES Act extended unemployment benefits to self-employed workers, independent contractors, freelancers and gig workers.
In 2021, the American Rescue Plan nearly tripled the maximum EITC for workers without qualifying children, from $543 to $1,502, for 2021 only. That same year, the IRS sent monthly child tax credit payments to families.
Each one was built in a crisis and allowed to lapse. Switch A is just making them permanent before the next one.
Stage 2 Machine labor arrives (about 2028 to 2035)
- What changes
- General-purpose robots start filling labor shortages in logistics, construction, care, and services. Prices for automated services fall. So does demand for the human labor that used to provide them.
- What to build
- A real transition safety net.
This is where I’ll concede something to the UBI camp: during the transition, a cash bridge makes sense. A check is the wrong destination, but it may be the right bridge while the goods themselves are still being built.
Pair it with in-kind provision wherever costs are already falling, starting with healthcare and housing. And start shifting taxes off human labor. - Signposts
- four numbers that show machine labor arriving, and who it’s paying off for.
Robot deployments per worker. How many robots are at work for every human worker. When it climbs, machine labor on demand is arriving. Track what a robot costs per hour of work, too.
The prices of services that machines now provide. If machines are really doing the work, those prices should fall. If they don’t, the savings are going somewhere else.
Watch the essentials against overall inflation, too: food at home, shelter and energy, from the Bureau of Labor Statistics’ Consumer Price Index. That comparison tells you which kind of trouble you’re in (more below).
The gap between productivity growth and wage growth. Productivity is how much the economy produces per hour of work. If output keeps rising while wages don’t, the gains are flowing to whoever owns the machines.
Long-term unemployment and median wages. How many people have been out of work for months, and what the typical worker earns. When both get worse together, wages are compressing. - What switches on
- Stage 2 itself, on data: when AI revenue reaches a set share of the knowledge-work wage bill, or the young-worker gap passes a set threshold. By then, Switch A should already be law.
Then Switch B, when wages compress. When long unemployment spells rise and median wages fall for a set period, the EITC converts into a negative income tax: a cash floor that pays out when your income falls below a threshold and phases out as it rises.
Milton and Rose Friedman put forward the first concrete proposal in Capitalism and Freedom (1962), though the British politician Juliet Rhys-Williams sketched the idea in the early 1940s. Friedman wanted one simple cash floor in place of a tangle of programs.
Unlike a universal basic income, it pays only people whose income has fallen, which, as Jacobs and Imas point out, makes it much cheaper.
Switch B is the bridge, and it switches on first. The cash path ends in the universal monthly cash floor in Every Spending Bill Should Show Up on Your Paycheck, which phases in. In year one, it pays what today’s consolidated benefit programs already cost, so the tax rate doesn’t jump.
From there, the floor grows with the tax base until it reaches its full size, about 5 points of the rate. Switch B keeps topping people up whenever wages fall.
That floor is the end state of the cash path, not of the series. It shrinks as each essential moves in kind. The destination is the lifeline block in Stage 4.
Stage 3 The build-out (about 2030 to 2040)
- What changes
- The scarcity flywheel spins up. AI-driven materials science, chemistry, and energy research start cutting the cost of the bottleneck goods.
Communities build microgrids for what they can’t go without: energy, water, and buffers of food and medicine. They trade for the rest, like chips and steel, which are far cheaper made at scale, for now.
The goal at this stage is resilience, not self-sufficiency: a town that can run its essentials through a disruption, without making itself poorer by making everything. - What to build
- Community nodes, built by parties (at first, AI-enabled startups, or “neo-labs”) as quests and financed by private capital and public investment, with the gift-back written into the charter from day one: when a node hits its sufficiency milestones for those critical needs, ownership transfers to the town’s trust, held for the people who live there, one equal share each.
Guilds document and teach the crafts the nodes need. - Signposts
- three numbers that show the build-out working where people live.
The share of a town’s basic needs produced locally. How much of its power and water a town makes for itself, and how many weeks of food and medicine it keeps on hand. When it rises, the nodes are doing their job.
Energy cost per kilowatt-hour. Energy is the input to everything else, so when it gets cheaper, almost everything else can too.
Housing cost per square foot. Housing is the biggest bill most families have. If the build-out works, this is where people will feel it. - What switches on
- The build-out at full speed, once two things hold together: general-purpose robots fall below a set cost per hour of work, and pilot nodes have passed their two-year tests. Cheap robots without a proven node design just build the wrong thing faster.
And Switch C, when labor decouples. If labor’s share of income keeps falling for several years while returns to capital rise, accelerate universal basic capital: speed up the node build-out and the gift-back.
Pay a national frontier dividend, the public’s share of frontier AI profits, into community investment funds that build nodes, instead of mailing it out.
Jacobs and Imas describe universal basic capital as something that could be “designed now but deployed” once the data shows that decoupling. That’s the right posture. On their panel’s agency scores, it ranked first of eleven policies, at 76.3 out of 100.
Notice what Switch C is: roughly Citrini’s Shared AI Prosperity Act. The difference is when you write it.
Stage 4 The gift-back and the floor (about 2035 to 2045)
- What changes
- Infrastructure transfers to communities at scale. Universal basic goods and services become real: walk into a store, take what you need, walk out. Money persists for whatever stays scarce.
In practice, that’s a lifeline block: each household’s first block of food, water, energy, housing, healthcare, transport, and communication is free, along with personal AI and personal machine labor. No application, no means test.
Use above the block has a price, the way utilities price power, which curbs waste without policing anyone. That revenue maintains the machines first, and only then pays dividends.
As each essential moves into the block, the cash floor shrinks with it, down to a little cash for extras and for whatever stays scarce.
Machines guarantee the floor; people choose their work. Machines run the essentials no matter who shows up, and everywhere else, humans get first pick. - What to build
- The surplus-on-request network that links communities, so a town on farmland can cover a town in the desert.
Run it like the power grid. Towns trade surplus in a regional market, run by their federation, where a shortage shows up as a high price and pulls supply toward it.
Households never see that price. Their essentials sit behind the lifeline block, and a shared reserve backs up any town hit by a drought or a fire.
And a narrower government whose job is guaranteeing the floor: making sure every community gets built out, and settling disputes between communities. - Signposts
- two numbers that show the floor is real.
The share of household spending that goes to essentials. How much of a family’s budget goes to what the lifeline block covers, from food and housing to energy and healthcare. When the floor is real, that share heads toward zero.
The number of communities that have taken ownership of their nodes. The gift-back, counted. Each one is a town where every resident owns an equal share of the machines that provide for it. - What switches on
- The gift-back at scale, when labor’s share falls by a set amount over a few years, or nodes hit their sufficiency milestones, whichever comes first.
That “whichever comes first” is the point of this whole piece. Good news can pull the floor forward. So can bad news. If the spiral starts, the build-out speeds up instead of waiting for its date on a chart.
Stage 5 Node states (2040 and beyond)
- What changes
- A network of resilient communities becomes the main way people organize their lives, connected by trade in information and resource surplus rather than wages.
Resilient means every town can run its essentials on its own through a crisis, and trades for everything else.
Sometime between 2040 and 2050, if this works, node communities go beyond resilient. They become self-sufficient enough to exchange resources above their surplus freely with each other, and that’s what makes a world where we can each do whatever we want.
At the limit, that means self-sufficiency by region: each group of towns covers its own needs, with plants sized to the region they serve, surplus built in, and everything made to be recycled.
Parties and guilds replace most corporate employment. Growth turns outward: mining beyond Earth to feed every region’s fabs, and eventually, other planets. - What to build
- Whatever we haven’t thought of yet. By this stage, the people living in the system will know better than I do.
The Map as a table
| # | System | Transition to the next |
|---|---|---|
| 1 | Foraging bands | roughly 6,000 years, from the first farms to the first states. |
| 2 | Agrarian empires | roughly 400 to 500 years. |
| 3 | Feudalism | roughly 250 years (about 1350 to 1600). |
| 4 | Merchant capitalism | roughly 75 years (from 1771, when the economist Carlota Perez dates the start of the first industrial revolution, to 1846). |
| 5 | Laissez-faire industrial capitalism | roughly 30 years (1914 to 1945). |
| 6 | Managed capitalism | roughly 18 years (1971 to 1989). |
| 7 | Financial and information capitalism | If that holds, the next one takes roughly 5 to 11 years. |
| Stage | Dates | What changes | What to build | Signposts | What switches on |
|---|---|---|---|---|---|
| 1. The machine intelligence age | now to about 2030 | Intelligence on demand is everywhere. AI handles a growing share of knowledge work, and the first services get automated. Wages start to stall in the most exposed jobs. | The prototypes. Pilot towns, in pairs: one with spare food and one with spare energy, each with about 1,000 to 5,000 people and two to three years to prove the model. Each pair runs the simplest version: equal shares, a free first block of the essentials, and maintenance before dividends. Alongside them, Data Foundries, the idea from my first post on Futurespective: places where people get paid to train robots and keep a share of what the robots earn. They’re a bridge for the years before the floor exists. Legal templates for parties (single-purpose entities where the people doing the work are partners) and guilds (open-admission craft organizations with a charter obligation to train anyone who wants to learn). Community ownership trusts that can hold infrastructure. And the first round of tax reform: the formula in Every Spending Bill Should Show Up on Your Paycheck. | five numbers that show whether AI is starting to do the work. Labor’s share of national income. The part of everything the economy earns that goes to workers as pay, instead of to owners as profit. If it keeps falling, machines are doing more of the work. In the US, it hit a record low of 52.8% in mid-2026. What matters for the triggers below is the trend, not the level: a steady fall is the signal. The share of tasks AI can do from start to finish. Not helping a person, but finishing the whole job. When that share rises, jobs stop being assisted and start being automated. Company filings give a second count: AI revenue as a share of the knowledge-work wage bill. The employment gap for young workers in AI-exposed jobs. Entry-level jobs are where automation shows up first. Stanford’s Digital Economy Lab tracks the gap in its public AI Economic Indicators. Payroll tax receipts as a share of federal revenue. How much of the government’s money comes from taxing paychecks. If that share shrinks while the economy grows, the old tax base is eroding. Long-term real interest rates. What it costs to borrow for decades, after inflation. Trevor Chow, Basil Halperin and J. Zachary Mazlish show that if markets expected transformative AI, these rates would rise sharply, because people would borrow against a much richer future. At 30% annual growth, real rates would top 30%; ten-year real rates in their developed-country sample have never gone above 5%. If the headlines are panicking and real rates are calm, the market isn’t buying the fast path. That includes my forecasts. | Switch A, the first layer of the safety net. It has no trigger: we build it now, while things are calm. Unemployment insurance for gig and contract workers. Jacobs and Imas note that US unemployment insurance typically excludes gig workers and freelancers. A monthly earned income tax credit, including workers without children. The earned income tax credit (EITC) tops up the pay of low earners. Today it largely leaves out workers without dependents, and it arrives once a year at tax time. Employer-led apprenticeships. In Jacobs and Imas’s reading of the evidence, public retraining programs have historically disappointed, while apprenticeships and employer-led models succeed “by a considerable margin.” Here’s the thing: we’ve already done most of Switch A. Temporarily. In March 2020, the CARES Act extended unemployment benefits to self-employed workers, independent contractors, freelancers and gig workers. In 2021, the American Rescue Plan nearly tripled the maximum EITC for workers without qualifying children, from $543 to $1,502, for 2021 only. That same year, the IRS sent monthly child tax credit payments to families. Each one was built in a crisis and allowed to lapse. Switch A is just making them permanent before the next one. |
| 2. Machine labor arrives | about 2028 to 2035 | General-purpose robots start filling labor shortages in logistics, construction, care, and services. Prices for automated services fall. So does demand for the human labor that used to provide them. | A real transition safety net. This is where I’ll concede something to the UBI camp: during the transition, a cash bridge makes sense. A check is the wrong destination, but it may be the right bridge while the goods themselves are still being built. Pair it with in-kind provision wherever costs are already falling, starting with healthcare and housing. And start shifting taxes off human labor. | four numbers that show machine labor arriving, and who it’s paying off for. Robot deployments per worker. How many robots are at work for every human worker. When it climbs, machine labor on demand is arriving. Track what a robot costs per hour of work, too. The prices of services that machines now provide. If machines are really doing the work, those prices should fall. If they don’t, the savings are going somewhere else. Watch the essentials against overall inflation, too: food at home, shelter and energy, from the Bureau of Labor Statistics’ Consumer Price Index. That comparison tells you which kind of trouble you’re in (more below). The gap between productivity growth and wage growth. Productivity is how much the economy produces per hour of work. If output keeps rising while wages don’t, the gains are flowing to whoever owns the machines. Long-term unemployment and median wages. How many people have been out of work for months, and what the typical worker earns. When both get worse together, wages are compressing. | Stage 2 itself, on data: when AI revenue reaches a set share of the knowledge-work wage bill, or the young-worker gap passes a set threshold. By then, Switch A should already be law. Then Switch B, when wages compress. When long unemployment spells rise and median wages fall for a set period, the EITC converts into a negative income tax: a cash floor that pays out when your income falls below a threshold and phases out as it rises. Milton and Rose Friedman put forward the first concrete proposal in Capitalism and Freedom (1962), though the British politician Juliet Rhys-Williams sketched the idea in the early 1940s. Friedman wanted one simple cash floor in place of a tangle of programs. Unlike a universal basic income, it pays only people whose income has fallen, which, as Jacobs and Imas point out, makes it much cheaper. Switch B is the bridge, and it switches on first. The cash path ends in the universal monthly cash floor in Every Spending Bill Should Show Up on Your Paycheck, which phases in. In year one, it pays what today’s consolidated benefit programs already cost, so the tax rate doesn’t jump. From there, the floor grows with the tax base until it reaches its full size, about 5 points of the rate. Switch B keeps topping people up whenever wages fall. That floor is the end state of the cash path, not of the series. It shrinks as each essential moves in kind. The destination is the lifeline block in Stage 4. |
| 3. The build-out | about 2030 to 2040 | The scarcity flywheel spins up. AI-driven materials science, chemistry, and energy research start cutting the cost of the bottleneck goods. Communities build microgrids for what they can’t go without: energy, water, and buffers of food and medicine. They trade for the rest, like chips and steel, which are far cheaper made at scale, for now. The goal at this stage is resilience, not self-sufficiency: a town that can run its essentials through a disruption, without making itself poorer by making everything. | Community nodes, built by parties (at first, AI-enabled startups, or “neo-labs”) as quests and financed by private capital and public investment, with the gift-back written into the charter from day one: when a node hits its sufficiency milestones for those critical needs, ownership transfers to the town’s trust, held for the people who live there, one equal share each. Guilds document and teach the crafts the nodes need. | three numbers that show the build-out working where people live. The share of a town’s basic needs produced locally. How much of its power and water a town makes for itself, and how many weeks of food and medicine it keeps on hand. When it rises, the nodes are doing their job. Energy cost per kilowatt-hour. Energy is the input to everything else, so when it gets cheaper, almost everything else can too. Housing cost per square foot. Housing is the biggest bill most families have. If the build-out works, this is where people will feel it. | The build-out at full speed, once two things hold together: general-purpose robots fall below a set cost per hour of work, and pilot nodes have passed their two-year tests. Cheap robots without a proven node design just build the wrong thing faster. And Switch C, when labor decouples. If labor’s share of income keeps falling for several years while returns to capital rise, accelerate universal basic capital: speed up the node build-out and the gift-back. Pay a national frontier dividend, the public’s share of frontier AI profits, into community investment funds that build nodes, instead of mailing it out. Jacobs and Imas describe universal basic capital as something that could be “designed now but deployed” once the data shows that decoupling. That’s the right posture. On their panel’s agency scores, it ranked first of eleven policies, at 76.3 out of 100. Notice what Switch C is: roughly Citrini’s Shared AI Prosperity Act. The difference is when you write it. |
| 4. The gift-back and the floor | about 2035 to 2045 | Infrastructure transfers to communities at scale. Universal basic goods and services become real: walk into a store, take what you need, walk out. Money persists for whatever stays scarce. In practice, that’s a lifeline block: each household’s first block of food, water, energy, housing, healthcare, transport, and communication is free, along with personal AI and personal machine labor. No application, no means test. Use above the block has a price, the way utilities price power, which curbs waste without policing anyone. That revenue maintains the machines first, and only then pays dividends. As each essential moves into the block, the cash floor shrinks with it, down to a little cash for extras and for whatever stays scarce. Machines guarantee the floor; people choose their work. Machines run the essentials no matter who shows up, and everywhere else, humans get first pick. | The surplus-on-request network that links communities, so a town on farmland can cover a town in the desert. Run it like the power grid. Towns trade surplus in a regional market, run by their federation, where a shortage shows up as a high price and pulls supply toward it. Households never see that price. Their essentials sit behind the lifeline block, and a shared reserve backs up any town hit by a drought or a fire. And a narrower government whose job is guaranteeing the floor: making sure every community gets built out, and settling disputes between communities. | two numbers that show the floor is real. The share of household spending that goes to essentials. How much of a family’s budget goes to what the lifeline block covers, from food and housing to energy and healthcare. When the floor is real, that share heads toward zero. The number of communities that have taken ownership of their nodes. The gift-back, counted. Each one is a town where every resident owns an equal share of the machines that provide for it. | The gift-back at scale, when labor’s share falls by a set amount over a few years, or nodes hit their sufficiency milestones, whichever comes first. That “whichever comes first” is the point of this whole piece. Good news can pull the floor forward. So can bad news. If the spiral starts, the build-out speeds up instead of waiting for its date on a chart. |
| 5. Node states | 2040 and beyond | A network of resilient communities becomes the main way people organize their lives, connected by trade in information and resource surplus rather than wages. Resilient means every town can run its essentials on its own through a crisis, and trades for everything else. Sometime between 2040 and 2050, if this works, node communities go beyond resilient. They become self-sufficient enough to exchange resources above their surplus freely with each other, and that’s what makes a world where we can each do whatever we want. At the limit, that means self-sufficiency by region: each group of towns covers its own needs, with plants sized to the region they serve, surplus built in, and everything made to be recycled. Parties and guilds replace most corporate employment. Growth turns outward: mining beyond Earth to feed every region’s fabs, and eventually, other planets. | Whatever we haven’t thought of yet. By this stage, the people living in the system will know better than I do. | — | — |
Stage 1 The machine intelligence age
- Dates
- now to about 2030
- What changes
- Intelligence on demand is everywhere. AI handles a growing share of knowledge work, and the first services get automated. Wages start to stall in the most exposed jobs.
- What to build
- The prototypes. Pilot towns, in pairs: one with spare food and one with spare energy, each with about 1,000 to 5,000 people and two to three years to prove the model.
Each pair runs the simplest version: equal shares, a free first block of the essentials, and maintenance before dividends.
Alongside them, Data Foundries, the idea from my first post on Futurespective: places where people get paid to train robots and keep a share of what the robots earn. They’re a bridge for the years before the floor exists.
Legal templates for parties (single-purpose entities where the people doing the work are partners) and guilds (open-admission craft organizations with a charter obligation to train anyone who wants to learn).
Community ownership trusts that can hold infrastructure. And the first round of tax reform: the formula in Every Spending Bill Should Show Up on Your Paycheck. - Signposts
- five numbers that show whether AI is starting to do the work.
Labor’s share of national income. The part of everything the economy earns that goes to workers as pay, instead of to owners as profit. If it keeps falling, machines are doing more of the work. In the US, it hit a record low of 52.8% in mid-2026.
What matters for the triggers below is the trend, not the level: a steady fall is the signal.
The share of tasks AI can do from start to finish. Not helping a person, but finishing the whole job. When that share rises, jobs stop being assisted and start being automated. Company filings give a second count: AI revenue as a share of the knowledge-work wage bill.
The employment gap for young workers in AI-exposed jobs. Entry-level jobs are where automation shows up first. Stanford’s Digital Economy Lab tracks the gap in its public AI Economic Indicators.
Payroll tax receipts as a share of federal revenue. How much of the government’s money comes from taxing paychecks. If that share shrinks while the economy grows, the old tax base is eroding.
Long-term real interest rates. What it costs to borrow for decades, after inflation. Trevor Chow, Basil Halperin and J. Zachary Mazlish show that if markets expected transformative AI, these rates would rise sharply, because people would borrow against a much richer future.
At 30% annual growth, real rates would top 30%; ten-year real rates in their developed-country sample have never gone above 5%. If the headlines are panicking and real rates are calm, the market isn’t buying the fast path. That includes my forecasts. - What switches on
- Switch A, the first layer of the safety net. It has no trigger: we build it now, while things are calm.
Unemployment insurance for gig and contract workers. Jacobs and Imas note that US unemployment insurance typically excludes gig workers and freelancers.
A monthly earned income tax credit, including workers without children. The earned income tax credit (EITC) tops up the pay of low earners. Today it largely leaves out workers without dependents, and it arrives once a year at tax time.
Employer-led apprenticeships. In Jacobs and Imas’s reading of the evidence, public retraining programs have historically disappointed, while apprenticeships and employer-led models succeed “by a considerable margin.”
Here’s the thing: we’ve already done most of Switch A. Temporarily.
In March 2020, the CARES Act extended unemployment benefits to self-employed workers, independent contractors, freelancers and gig workers.
In 2021, the American Rescue Plan nearly tripled the maximum EITC for workers without qualifying children, from $543 to $1,502, for 2021 only. That same year, the IRS sent monthly child tax credit payments to families.
Each one was built in a crisis and allowed to lapse. Switch A is just making them permanent before the next one.
Stage 2 Machine labor arrives
- Dates
- about 2028 to 2035
- What changes
- General-purpose robots start filling labor shortages in logistics, construction, care, and services. Prices for automated services fall. So does demand for the human labor that used to provide them.
- What to build
- A real transition safety net.
This is where I’ll concede something to the UBI camp: during the transition, a cash bridge makes sense. A check is the wrong destination, but it may be the right bridge while the goods themselves are still being built.
Pair it with in-kind provision wherever costs are already falling, starting with healthcare and housing. And start shifting taxes off human labor. - Signposts
- four numbers that show machine labor arriving, and who it’s paying off for.
Robot deployments per worker. How many robots are at work for every human worker. When it climbs, machine labor on demand is arriving. Track what a robot costs per hour of work, too.
The prices of services that machines now provide. If machines are really doing the work, those prices should fall. If they don’t, the savings are going somewhere else.
Watch the essentials against overall inflation, too: food at home, shelter and energy, from the Bureau of Labor Statistics’ Consumer Price Index. That comparison tells you which kind of trouble you’re in (more below).
The gap between productivity growth and wage growth. Productivity is how much the economy produces per hour of work. If output keeps rising while wages don’t, the gains are flowing to whoever owns the machines.
Long-term unemployment and median wages. How many people have been out of work for months, and what the typical worker earns. When both get worse together, wages are compressing. - What switches on
- Stage 2 itself, on data: when AI revenue reaches a set share of the knowledge-work wage bill, or the young-worker gap passes a set threshold. By then, Switch A should already be law.
Then Switch B, when wages compress. When long unemployment spells rise and median wages fall for a set period, the EITC converts into a negative income tax: a cash floor that pays out when your income falls below a threshold and phases out as it rises.
Milton and Rose Friedman put forward the first concrete proposal in Capitalism and Freedom (1962), though the British politician Juliet Rhys-Williams sketched the idea in the early 1940s. Friedman wanted one simple cash floor in place of a tangle of programs.
Unlike a universal basic income, it pays only people whose income has fallen, which, as Jacobs and Imas point out, makes it much cheaper.
Switch B is the bridge, and it switches on first. The cash path ends in the universal monthly cash floor in Every Spending Bill Should Show Up on Your Paycheck, which phases in. In year one, it pays what today’s consolidated benefit programs already cost, so the tax rate doesn’t jump.
From there, the floor grows with the tax base until it reaches its full size, about 5 points of the rate. Switch B keeps topping people up whenever wages fall.
That floor is the end state of the cash path, not of the series. It shrinks as each essential moves in kind. The destination is the lifeline block in Stage 4.
Stage 3 The build-out
- Dates
- about 2030 to 2040
- What changes
- The scarcity flywheel spins up. AI-driven materials science, chemistry, and energy research start cutting the cost of the bottleneck goods.
Communities build microgrids for what they can’t go without: energy, water, and buffers of food and medicine. They trade for the rest, like chips and steel, which are far cheaper made at scale, for now.
The goal at this stage is resilience, not self-sufficiency: a town that can run its essentials through a disruption, without making itself poorer by making everything. - What to build
- Community nodes, built by parties (at first, AI-enabled startups, or “neo-labs”) as quests and financed by private capital and public investment, with the gift-back written into the charter from day one: when a node hits its sufficiency milestones for those critical needs, ownership transfers to the town’s trust, held for the people who live there, one equal share each.
Guilds document and teach the crafts the nodes need. - Signposts
- three numbers that show the build-out working where people live.
The share of a town’s basic needs produced locally. How much of its power and water a town makes for itself, and how many weeks of food and medicine it keeps on hand. When it rises, the nodes are doing their job.
Energy cost per kilowatt-hour. Energy is the input to everything else, so when it gets cheaper, almost everything else can too.
Housing cost per square foot. Housing is the biggest bill most families have. If the build-out works, this is where people will feel it. - What switches on
- The build-out at full speed, once two things hold together: general-purpose robots fall below a set cost per hour of work, and pilot nodes have passed their two-year tests. Cheap robots without a proven node design just build the wrong thing faster.
And Switch C, when labor decouples. If labor’s share of income keeps falling for several years while returns to capital rise, accelerate universal basic capital: speed up the node build-out and the gift-back.
Pay a national frontier dividend, the public’s share of frontier AI profits, into community investment funds that build nodes, instead of mailing it out.
Jacobs and Imas describe universal basic capital as something that could be “designed now but deployed” once the data shows that decoupling. That’s the right posture. On their panel’s agency scores, it ranked first of eleven policies, at 76.3 out of 100.
Notice what Switch C is: roughly Citrini’s Shared AI Prosperity Act. The difference is when you write it.
Stage 4 The gift-back and the floor
- Dates
- about 2035 to 2045
- What changes
- Infrastructure transfers to communities at scale. Universal basic goods and services become real: walk into a store, take what you need, walk out. Money persists for whatever stays scarce.
In practice, that’s a lifeline block: each household’s first block of food, water, energy, housing, healthcare, transport, and communication is free, along with personal AI and personal machine labor. No application, no means test.
Use above the block has a price, the way utilities price power, which curbs waste without policing anyone. That revenue maintains the machines first, and only then pays dividends.
As each essential moves into the block, the cash floor shrinks with it, down to a little cash for extras and for whatever stays scarce.
Machines guarantee the floor; people choose their work. Machines run the essentials no matter who shows up, and everywhere else, humans get first pick. - What to build
- The surplus-on-request network that links communities, so a town on farmland can cover a town in the desert.
Run it like the power grid. Towns trade surplus in a regional market, run by their federation, where a shortage shows up as a high price and pulls supply toward it.
Households never see that price. Their essentials sit behind the lifeline block, and a shared reserve backs up any town hit by a drought or a fire.
And a narrower government whose job is guaranteeing the floor: making sure every community gets built out, and settling disputes between communities. - Signposts
- two numbers that show the floor is real.
The share of household spending that goes to essentials. How much of a family’s budget goes to what the lifeline block covers, from food and housing to energy and healthcare. When the floor is real, that share heads toward zero.
The number of communities that have taken ownership of their nodes. The gift-back, counted. Each one is a town where every resident owns an equal share of the machines that provide for it. - What switches on
- The gift-back at scale, when labor’s share falls by a set amount over a few years, or nodes hit their sufficiency milestones, whichever comes first.
That “whichever comes first” is the point of this whole piece. Good news can pull the floor forward. So can bad news. If the spiral starts, the build-out speeds up instead of waiting for its date on a chart.
Stage 5 Node states
- Dates
- 2040 and beyond
- What changes
- A network of resilient communities becomes the main way people organize their lives, connected by trade in information and resource surplus rather than wages.
Resilient means every town can run its essentials on its own through a crisis, and trades for everything else.
Sometime between 2040 and 2050, if this works, node communities go beyond resilient. They become self-sufficient enough to exchange resources above their surplus freely with each other, and that’s what makes a world where we can each do whatever we want.
At the limit, that means self-sufficiency by region: each group of towns covers its own needs, with plants sized to the region they serve, surplus built in, and everything made to be recycled.
Parties and guilds replace most corporate employment. Growth turns outward: mining beyond Earth to feed every region’s fabs, and eventually, other planets. - What to build
- Whatever we haven’t thought of yet. By this stage, the people living in the system will know better than I do.
- Signposts
- —
- What switches on
- —
The window is opening. Let’s build.
Every date, length, and line on this page comes from Parts 7 and 19. Read Part 7 → Read Part 19 →