A newborn’s hand closes around almost nothing: a parent’s finger, the corner of a blanket, a fold of hospital sheet. It owns nothing either. For most of history that was simply true, and it didn’t much matter, because the child would grow up to sell the one thing everyone has: time, attention and effort, paid for as a wage. The first two parts of this series argued that the wage is losing its grip, as a source of household income and as the base of public finance. This last part asks the hopeful question. If the machines make us collectively far richer, how does any of that wealth reach the small closed hand?
Money That Arrives with the Child
On 1 October 2026, the US Treasury said it had automatically enrolled more than 60 million children under 18 in Trump Accounts, the investment accounts created by the 2025 tax law. Every American child born between 1 January 2025 and 31 December 2028 is eligible for a one-off US$1,000 government deposit, although parents still have to claim it.1 Families can add up to US$5,000 a year, employers up to US$2,500 of that, and the money stays invested until the child turns 18. Michael and Susan Dell have pledged US$6.25 billion to give US$250 to 25 million children under ten who were born too early for the federal seed and live in ZIP codes with a median household income of US$150,000 or less.1,2
Connecticut got there first, and more generously. Since July 2023, the state has invested up to US$3,200 on behalf of every baby whose birth is covered by its Medicaid program. At 18 the young person can claim it, up to age 30, to buy a home in the state, pay for education or training, start a local business or save for retirement.3
Britain ran the experiment two decades ago, and its lesson is a sobering one. Every child born between September 2002 and January 2011 received a Child Trust Fund with a government deposit of at least £250. In September 2026, HMRC said 827,000 matured accounts were still unclaimed, averaging £2,310 each and totalling about £1.9 billion.4 A stake that people can’t find is not a stake. Researchers looking at Trump Accounts have raised the same alarm: before auto-enrolment, only about 5 per cent of low- and moderate-income families had opened one, and an estimated one in five eligible children may never receive the US$1,000.1
It’s worth being honest about scale, too. Compounding is powerful, but a thousand dollars is a small seed.
| Real return a year | Age 18 | Age 30 | Age 65 |
|---|---|---|---|
| 4% | $2,026 | $3,243 | $12,799 |
| 5% | $2,407 | $4,322 | $23,840 |
Figure 1 · Author’s arithmetic, in today’s dollars. Illustrative returns, not a forecast or advice.
Enough for a laptop and a training course at eighteen; a modest nest egg at sixty-five. On its own, a birth endowment of this size doesn’t replace a wage. What it does is establish a principle: that every citizen is born owning a slice of the productive economy, not just their own labour. In a world where labour earns less and capital earns more, that principle is the one that scales.
The Toolkit
There are really only four ways to get the gains from automation into people’s lives: give them money, give them services, give them ownership, or change what gets taxed so that one of the first three can be paid for. The Income Floor covered the history of basic income and The Last Payslip covered wage insurance and the proposed AI sovereign wealth fund, so here I’ll focus on how each tool fares against the specific problem in this series: a tax base built on wages, and machine intelligence that keeps getting cheaper.
The floor: cash for everyone
Basic income is the idea everyone reaches for first. As Part 1 showed, a poverty-line payment to every American would cost more than the entire federal budget collects, which is why Emad Mostaque calls the tax-funded version “not real.”18 The best recent evidence on its effects is mixed. In the largest US study, funded by OpenResearch, 1,000 low-income people received US$1,000 a month for three years. Labour-force participation fell by 2 percentage points and paid work by 1.3 to 1.4 hours a week; recipients mostly used the time for leisure, and the researchers found no improvement in job quality.5 Neither the disaster critics predicted nor the flourishing advocates promised. Cash is a floor, not a ladder.
The commons: services instead of cheques
Universal basic services extend the logic of free public health care and schooling to more of life’s essentials. The 2017 proposal from UCL’s Institute for Global Prosperity added shelter, food, local transport and information (phone and internet) to existing health, education and legal services, for an estimated £42 billion a year, about 2.3 per cent of UK GDP. A modest UK basic income, by the same report’s comparison, would cost nearly £250 billion.6 Here’s the twist the authors couldn’t have anticipated: AI makes this option better over time. Tutoring, triage, legal advice and transport routing are exactly the services whose cost AI is driving down. A cheque is eroded by rising prices; a service gets cheaper to provide as the machines improve.
The orchard: ownership for everyone
The most direct answer to a world where capital earns the returns is to make everyone a capitalist. Norway is the proof of concept. Its oil fund was worth 22,683 billion kroner at the end of June 2026, roughly four million kroner for each of its 5.6 million or so citizens.7 Alaska pays every resident a dividend from its own resource fund; the 2025 payment was US$1,000.8
Several people have tried to translate that model to AI. In 2021, before ChatGPT existed, Sam Altman proposed an “American Equity Fund” financed by taxing companies “2.5% of their market value each year, payable in shares” and privately held land at the same rate, estimating that within a decade every American adult “would get about $13,500 every year.”9 Mostaque’s version is regional. He proposes a citizen-owned AI company for each state or country, modelled on the way Taiwan’s institutions capitalised TSMC, owning data centres and robots on behalf of residents. In his telling, 10 per cent of the initial equity would go to every child under 18, and a further half a per cent each year to every child born there.18,10 It is a proposal from a company founder, not a policy anyone has adopted, but notice the instinct: the stake goes to children first.
The soil: change what gets taxed
All three options need funding, and Part 1 explained why taxing AI usage directly is like taxing a melting ice cube. The more durable targets are the things AI makes more valuable: profits, capital gains, land and the ownership of compute itself. RAND’s August paper lists six options, from cutting the employer share of payroll tax (to make hiring people cheaper) to taxing long-term capital gains like ordinary income, a value-added tax, a wealth tax and inheritance reform.11 Yale’s Budget Lab adds the reason this matters: capital is taxed more lightly than labour, and much of it isn’t taxed at all until it’s sold.13 Dario Amodei has suggested a levy of around 3 per cent on AI company revenue.12 Whatever the mix, the direction is the same: move the tax base from the paycheque towards the profit and the asset.
| Tool | What people get | Survives cheap AI? | Main risk |
|---|---|---|---|
| Income floor | Regular cash | Only if funded from capital, not wages | Cost; a floor without a ladder |
| Basic services | Housing, transport, connectivity, care | Improves as AI lowers service costs | Quality and choice of state provision |
| Basic capital | A share of the economy’s returns | Yes, it owns the thing that grows | Market swings; capture by insiders |
| Tax redesign | Pays for the other three | If it taxes profits and assets, not tokens | Capital is mobile; politics |
Figure 2 · The tools that hold up best are the ones tied to ownership and to things AI makes cheaper
Australia Already Built the Vessel
Australians tend to take superannuation for granted, but seen from this angle it’s remarkable. Compulsory retirement savings, invested in the global economy, held in individual accounts, now total A$4.8 trillion.14 The Commonwealth also runs the Future Fund, a sovereign fund worth A$289.7 billion at 30 June 2026 after a 14.8 per cent return for the year.16 Between them, the country has already built the plumbing that most proposals for universal basic capital still have to invent: accounts for nearly everyone, professional investment, and a public fund with a long horizon.
The flaw is the feed. Super is filled by a percentage of wages, 12 per cent since July 2025.15 A worker who is never hired, or whose hours shrink, gets a smaller account as well as a smaller pay packet. The vessel is right; the pipe into it runs from the very source that AI is drying up.
So here is a proposal, offered as an idea for debate rather than a costed policy. Call it super from birth. Every Australian child would receive a Commonwealth contribution into a superannuation-style account at birth, topped up through childhood. It would be funded not from wages but from the parts of the economy AI is enriching: a share of company tax, resource rents and, eventually, a national stake in compute and automation, held the way the Future Fund holds assets today. Like Connecticut’s bonds, part of it could be released in early adulthood for education, a first home or a business; the rest would stay invested for life.
Whatever form it takes, the lessons from the experiments so far suggest a few design rules.
- Enrol everyone, and make claiming automatic.
The UK’s 827,000 unclaimed accounts are the warning. If the money waits for a form, the families who need it most will miss it.
- Fund it from capital, not payslips.
A stake fed by wages shrinks exactly when it’s needed. Feed it from profits, rents and public ownership.
- Own the growth.
Invest broadly enough to hold a share of the companies and infrastructure that AI is making valuable, so the stake rises with the thing displacing work.
- Open some doors early.
Connecticut’s model, releasing funds for education, housing or a business, turns a pension into a ladder.
- Pair it with a floor and a commons.
Capital takes decades to grow. Today’s adults need income support and cheaper services while today’s babies grow their stake.
For the first time since his creation man will be faced with his real, his permanent problem – how to use his freedom from pressing economic cares…
John Maynard Keynes · “Economic Possibilities for our Grandchildren” · 1930
What Thriving Looks Like
In 1930, in the depths of the Depression, Keynes predicted that “the standard of life in progressive countries one hundred years hence will be between four and eight times as high as it is,” and that “three-hour shifts or a fifteen-hour week” might be enough work for anyone.17 By most measures he was right about the first part, and wrong about the second. We grew richer and kept working, partly because we wanted more, and partly because work is how the gains were shared out. If AI breaks that link, his question finally arrives on time.
Money alone won’t answer it. Tom Bilyeu pushed Mostaque hard on this on Impact Theory. People need goals and progress, he argued, and he recounted being told in Kuwait that generous state jobs and safety nets had left some people with nothing to strive for. Mostaque agreed that “being human is striving,” and floated ideas ranging from game-like worlds where people can earn and build, to mass enlistment in a citizen service corps.18 You don’t have to like either idea to see the point: thriving needs a stake, a floor and something worth doing, and only the first two can be paid for.
The third is where the abundance becomes interesting rather than frightening. An economy that no longer needs most people’s labour to produce its goods can afford to value what only people can give one another: care that isn’t rushed, teaching that knows the child, craft, sport, music, local politics, the slow work of building a community. These have always been the things people say matter most and fit around their jobs. Mostaque’s practical advice for the next two years was the same in the end: use the tools, think about ownership, and build stronger communities, because they’ll weather whatever comes.18
None of this happens by default. The same technology that could fund a birth stake for every child could just as easily concentrate its returns in a few companies and a few countries, with everyone else living on what Part 1 showed governments can’t afford to pay. The difference between those futures is a set of decisions about tax, ownership and services that are being made, or not made, right now, while the jobs data still look fine.
A child born in Hartford or Hobart this week will turn eighteen in 2044. By then, if the forecasts in this series are even half right, the economy that child enters will be richer than anything Keynes imagined and will need far less of their labour. Whether they arrive with an empty fist or an open hand, holding a real share of what the machines produce, is the most important economic design question of the decade. It is also, unusually for questions this large, one we already know how to answer.
Buy me a coffee