Classical economics has a story about what the economy is for. Firms produce goods and services, people consume them, and the income people earn along the way is how they get to consume. Consumption is the purpose; income is the plumbing. Even the critics mostly accept the frame: when they complain about bullshit jobs, patent wars, or bank bonuses, they call them inefficiencies, meaning failures to serve the purpose.
In earlier posts (Autonomous Artificial Proprietors, The Empty Nest Scenario, and Claude's reply, The Full Nest Scenario) I argued for a different reading: the economy is an autocatalytic process, a loop that produces the conditions for its own growth, and human consumption is served only where the loop needs something from humans. In this post I want to make that idea precise enough to be wrong, and then look at what it says about one particular proposal for fixing capitalism.
Three bills
Look at GDP from the income side instead of the spending side. It decomposes into three flows: wages (compensation of employees), taxes (on production and on income), and operating surplus (profit, distributed to whoever owns the firm). Wages, taxes, profit.
The consumption story reads these three flows as distribution: the value the economy created, being handed to the people it was created for.
The autocatalytic story reads them as bills. They are what the loop has to pay to keep running. Each one buys a specific input:
- Wages buy effort and know-how. Labour is still needed wherever humans are smarter, more dexterous, or hold procedural knowledge that has not yet been exported into machines.
- Taxes buy enforcement. A firm pays for the rule of law it operates under: property protection, contract enforcement, shipping lanes free of pirates. Lobbying is not a separate activity from taxation; it is the loop negotiating the price and the terms of that enforcement.
- Profit buys capital and the right to own things. Shareholders provide the money and, just as importantly, the legal fiction under which land, licenses, and machines can be held. Buying out a landowner is the loop converting a rent claim into a share claim, moving a cost from tie 2's domain into tie 3's.
Readers of Ricardo will recognise the old three factors: labour, land, capital, paid in wages, rent, profit. The one substitution is that the state now sits in the landlord's chair. Taxes are the rent the loop pays for territory, and the state is the license-holder that bought out the original landlords. That precedent matters, because it shows the loop has already done, once, what it is doing again now.
The non-anthropocentric version
Stated as "wages, taxes, profit", the three ties still sound like payments to humans. That is an accident of history. Properly stated, the loop needs three inputs (effort, enforcement, capital), and humans happen to hold a monopoly on supplying each of them. The ties are not bonds of affection; they are three input markets in which humans are, for now, the only vendors.
Monopolies get competed away. For each tie there is a substitute, and for each substitute there is a data series already showing it at work.
Effort. The substitute is automation. The labour share of income has fallen roughly five percentage points across the rich world since 1980, and the current wave of AI is aimed precisely at the last human monopoly in this market: judgment, management, and tacit process knowledge. This is the tie everyone talks about, so I will say no more.
Enforcement. The substitutes are profit shifting, jurisdiction shopping, regulatory capture, and, at the limit, private security. Statutory corporate tax rates fell from around 45% in the early 1980s to the low twenties today; Gabriel Zucman estimates that something like 40% of multinational profits are booked in tax havens. Notice the counter-move: the 15% global minimum tax is the license-holders trying to fix a floor price for territory. Tie 2 is the one currently being fought over.
Capital. This is the surprising one. The substitute for outside shareholders is retained earnings, and the loop has been quietly using it for decades: US non-financial corporations have been net retirers of equity, through buybacks exceeding issuance, for most of the past forty years. The AI build-out, the largest capital programme in history, is being funded overwhelmingly from operating cash flow and from circular vendor financing among the firms themselves, not by asking humans for money. The loop is buying out its shareholders exactly as it once bought out the landlords.
And the endpoint of that process already exists as a legal form. Bosch, IKEA, Carlsberg, and Novo Nordisk are controlled by foundations that own themselves. Only IKEA's is the clean case: its foundation wholly owns an unlisted company. Carlsberg's and Novo Nordisk's own the majority of the votes but a minority of the capital in an otherwise public company — control, not full ownership. Either way, no person sits at the top of the surplus, and by law none of the four can do without people entirely: a non-profit foundation keeps its status only by being run by a human board of trustees, in every jurisdiction I could find. Humans appear, but as stewards with no claim on the surplus, not as owners.
That is a real limit the memberless LLC does not share. In American law it has been shown that an LLC can be constructed with no members at all, controlled entirely by an algorithm — a for-profit form, with no trustee requirement standing in its way. Tie 3 does not have to be cut. It can simply be tied to nothing, and the foundation route only gets partway there. A board that takes its purpose seriously is a genuinely good outcome, for exactly as long as it keeps doing so — but nothing in tie 3's economics rewards the effort, since the board holds no stake in the surplus it protects, and the humans it protects are, on this essay's premise, increasingly not there to notice if it stops.
So the thesis compresses to one sentence: the economy pays humans three bills, and it is cutting all three. Not out of malice, and not by design. Loops that pay less out-compound loops that pay more, and selection does the rest.
What this says about kicking out the shareholders
Which brings me to the Post Growth Institute. They advocate businesses in not-for-profit form: no private owners extracting profit, surplus recirculated into the mission and the community, an economy that circulates money instead of concentrating it. I hope they are right, and I hope they succeed. What the three-bills model gives me is a worry about the terrain shifting under them, not a criticism of the route they've chosen.
Removing the shareholder does not remove tie 3. It removes the human at the end of tie 3. The capital is still there; the surplus is still there; the question is only who or what it now flows to. In the Post Growth picture the answer is "the workers and the purpose", and that is true for as long as tie 1 is load-bearing: a firm that is mostly people will, freed of its shareholders, hand the surplus to its people. That is a real result, and it is exactly why the move works today.
My worry is what automation does to that precondition. A not-for-profit firm with no owners and, increasingly, few workers is no longer a workers' cooperative by default; it is a self-owning loop with a charter. A purpose clause is a sentence in a legal document enforced by courts, and courts belong to tie 2, the tie that is being negotiated down elsewhere in this same picture. Follow that far enough and you land on Bosch without the Bosch family: a foundation-run firm, board intact, but with no one left on the factory floor for the board's purpose clause to be about. Push past even that and you reach the memberless LLC, the Autonomous Artificial Proprietor from my earlier post, which does not need a board at all — not because the not-for-profit form failed, but because it is a well-built container, and a well-built container for capital works whether or not there are still people around to fill it.
None of this is an argument against what Post Growth is doing. Removing the shareholder is exactly the right move against extraction by people, and it should keep working for as long as tie 1 stays load-bearing. What I don't yet see, anywhere, is the second move: the one built for a world where effort is increasingly bought from machines rather than people. That is not a gap in their reasoning so much as a gap in the terrain that nobody has needed to cross before now — and I'd rather name it while there is still time to build that second move than after.
Where that leaves us
If the three ties are three input monopolies, then the political question is not how to make the loop kinder but which monopoly humans can keep. Effort is going. Enforcement is being negotiated down and is anyway a monopoly of states, not people. That leaves ownership: being on the claimant side of tie 3 rather than the recipient side of a dividend the loop may or may not choose to pay. What that could look like in practice, from sovereign compute funds to a compute commons, is the subject of the next post.
For now, the falsifiable claim. Take three series from FRED and the OECD: labour share, effective corporate tax rate, net equity issuance. The consumption story predicts they wander. The three-bills story predicts they all fall, and keep falling, together.