The Considered Read

The whole argument, idea by idea

~15-minute read One idea developed per section A gateway into the full treatise

The essential read gives you the whole argument in one quick sweep. This version slows down. It takes each section in turn and develops its main idea far enough to feel the weight of the claim, the objection it answers, and the honesty it owes, before opening the door into the full section. It sits in the middle: longer than the quick read, much shorter than the treatise, and written in plain language so the shape of the whole thing comes into focus without the density.

The Convergence and the Decomposition of the Firm

Every big shift of the internet era happened the same way: not as one invention, but as several technologies that matured separately and suddenly came together. The web, mobile, the cloud, and social media were all convergences like this, and beneath each one the same pattern repeats. When capabilities come together, the cost of something that used to be expensive falls close to zero, and once that cost falls, the amount of the activity explodes. The web did it for information, mobile and social did it for communication, and the cloud did it for software. Two new systems are now converging and aiming that same force at the two things the internet never fully digitized: intelligence and the economy itself. The first is a system for intelligence, made of AI models and the agents built on them, which drives the cost of thinking and doing work toward zero. The second is a system for the economy, made of blockchains where money, contracts, and coordination run as software, which drives the cost of transacting toward zero. Each makes the other more useful, and the claim the whole work builds toward is that these are not two trends sitting side by side but one economy seen from two angles.

The system for intelligence matters most because it changes what software is. You do not program it; you instruct it in plain language, and it reasons toward an answer instead of following fixed steps. Its basic unit is the agent: a reasoning process you hand a task. That turns software from something a machine runs literally into work you delegate to a thinking machine, and it lets the core tasks of a business be broken apart and rebuilt as skills an agent can perform. Underneath the brand and the buildings, a company is mostly organized thinking: product, marketing, sales, finance, and legal, plus the outside firms it hires. Almost all of that is labor, and labor is the largest cost in the economy, so this is exactly the cost that cheap, capable intelligence goes after. It also overturns the old explanation for why companies exist. Firms grew large because coordinating outside work was expensive, so they kept it in-house; when any piece of non-physical work can be done by an agent you can find, hire, and pay instantly, that logic weakens, and one person can do what once took a department. It arrives first in software and other information-heavy work and slowest in anything physical, which still waits on robotics. And it is not simply people removed: a person paired with capable agents becomes far more powerful, while judgment, relationships, and accountability stay human. That leaves one tension to carry forward, which the argument later answers through ownership: a person can be magnified even as the share of what the whole economy pays to human labor falls.

Read Section 1 in full →

Assembly, Coordination, and Why the Firm Goes Onchain

Once a firm is broken into skills, the real question is not what can be automated but how the pieces get put back together. The answer is an orchestration layer: a general manager agent takes a goal, splits it into tasks, hands them to specialist agents, and stitches the results back together, with supporting software carrying context and memory between the steps. The same setup works for any function, so marketing, finance, sales, and product are really the same machine aimed at different work. People do not disappear. Some stay in the loop, doing or checking the work that needs human judgment. Others move onto the loop, setting the goals, defining what good looks like, watching quality, and deciding when the machine should stop and ask. That shift, from doing the work to supervising it, is what human oversight actually becomes, and the tools for it are arriving now.

The moment a company makes a task clean enough to run internally, it becomes clean enough to hire from outside, so an open market of agents forms almost as a side effect. That market could go one of two ways. It could be a handful of big platforms selling intelligence like a utility, or, more likely and more interesting, a real labor market of specialist agents, because deep expertise stays valuable and the lasting businesses will be agents that go very deep in one field. But hiring software that could have been assembled anywhere in the world only works if you can trust it, and that is the problem that pushes everything onchain. The solution is that identity comes in layers. At the base is a public blockchain anyone can verify. On top of that sit real-world identity checks, the same kind banks already run at scale, the agent's own wallet and credentials, and a reputation earned over time but tied to a verified, real creator. Together these form an accountability chain: everything an agent does traces back to a real person or company answerable for it. A single company's private database cannot do this, because trust locked inside one operator does not travel, while identity rooted in a public chain and real-world verification does. So autonomy here is not anonymity. An agent that acts on its own is still one that someone stands behind.

Read Section 2 in full →

The Monetary Substrate: Velocity, Safety, and Finality

Agents need money they can hold and move at machine speed, in huge volumes and tiny amounts, without stopping on every payment to check whether the money itself is sound. That last point is the key, and it leads to an old-fashioned answer: money that is fully backed, settles with finality, and runs on an open network. Start with speed, because it reorganizes everything else. When moving money costs almost nothing and clears in a fraction of a second, and the money is controllable by software, the same dollar can be reused many times in quick succession, any amount becomes usable the instant it arrives, and tiny payments between agents finally make sense. This is just the pattern that information and software already followed on the internet, now reaching money.

Every part of the answer earns its place. A natural objection is that banks create speed by lending the same deposit many times over, so wouldn't full backing starve credit? It would not: when money turns over fast enough, a dollar can be locked for seconds and lent onward, so speed does the work leverage used to do, and credit is rebuilt on top rather than removed. Why insist the base money carry no risk at all? Because speed makes risky money dangerous in proportion to how fast it moves. A bank run that once took weeks can now happen in minutes, and an agent settling instantly cannot pause to judge whether each dollar is good. Fully backed money is the only kind that is worth exactly a dollar to everyone, everywhere, without national safety nets that do not reach a global system. Settlement has to be just as certain: not probably final after a while, but final in under a second, where settled means settled. Refunds and fraud protection still exist, but as optional layers built on top, such as escrow, refund pools, and insurance, rather than baked into the money itself. None of this safety is automatic; it rests on real institutions now being built, with large issuers regulated, kept separate from bankruptcy, and backed by increasingly safe reserves. One line has to stay clear: holding the money pays you nothing. The yield on the reserves goes to the issuer and out into the ecosystem, but the moment you reach for yield you are no longer holding money, you are lending it and taking on risk. Mixing those two things up would undo the whole case for safety.

Read Section 3 in full →

Credit Markets: Machine Underwriting, Agent Working Capital, and the Prudential Layer

Credit does not vanish when the base money is fully backed; it moves to the far side of that line and comes back stronger, reaching more people, pricing tighter, and failing more visibly than the system it replaces. The key insight reframes the problem. The long tail of borrowers, meaning small merchants, gig workers, households, and now agents, was underserved not because they were bad risks but because checking each small loan cost more than the loan was worth. Credit was rationed by the cost of underwriting, not the quality of the borrower. Drive that cost down and a huge group of sound but overlooked borrowers becomes serviceable. What drives it down is a data flywheel: onchain activity is structured, verifiable, and real-time, which makes for far better risk models than the old patchy records, and better data leads to better lending, which draws more activity and more data. The obvious worry, that this puts everyone's finances on a public ledger, has a clean answer: onchain does not mean public. New privacy techniques let someone prove what a lender needs, such as that they are creditworthy or that a balance exists, without exposing the underlying details.

At the center is a genuinely new kind of loan: working capital for agents. It is unusually predictable, because it strips out the biggest wildcard in human lending, whether the borrower feels like repaying, and reduces the risk to a short, bounded question about a specific piece of work. Picture an agent borrowing four dollars of computing to finish a ten-dollar job it has already been hired to do. The lender is not guessing at character; it is pricing the odds that the work is accepted. Collateral flips the usual model: instead of seizing an unrelated asset slowly through courts, the loan is secured first by the payment for the work itself, claimed automatically, and backed up by a deposit the agent posts, its reputation, and finally the real person behind it. The result is credit that is cheaper, more available, and safer at the same time, which sounds impossible until you see that the gain comes from better information, not more borrowing. The honesty the claim requires is that this predictability fades with time: a task that finishes in seconds is nearly mechanical, while months-long financing grades back into ordinary risk. So machine credit does not replace human credit; it becomes a new low-risk baseline that human lending is priced against. And none of it runs unwatched: risk is visible as it builds, with automatic brakes that make crowding into the same model or provider steadily more expensive, plus insurance priced on what is actually happening rather than on stale averages.

Read Section 4 in full →

Global by Construction

The architecture has exactly three layers. At the bottom is money: stablecoins that serve as the unit of account and the final means of settlement. In the middle is an economic operating system: coordination, contracts, and value exchange running as programmable smart contracts that settle with finality. At the top is agentic execution: the layer where the actual work gets done, powered by AI and the cloud. The important thing about all three is where they live. Each is software, and each runs on the internet. Each also replaces something that used to be tied to a country: software money replaces national banking stitched together by slow correspondent banks; the middle layer moves contract enforcement out of national courts and into code that runs the same everywhere; and agentic execution replaces local labor with work that has no hometown. So the economy built from these layers is borderless by default. That is what global by construction means: not a feature added on, but a property of what it is made of. For all of history economic life was national first and stretched across borders only with effort; now it is global first, and the national framing is the thing that has to be added back.

An economy with no home country does not escape the law; it falls under too much of it at once, as many jurisdictions' rules collide with no single place to decide which applies. The fix is to shift the question from where something happened to who stands behind it, regulating the accountable entity each agent traces to, with the countries where users actually live setting the terms for market access. Enforcement moves to the edge, where money and identity cross between the open, regulated world and the private one, checked before a payment settles rather than reported after it clears. This needs no public ledger of everyone's finances: disclosure stays private by default and is shared only with permission. A healthy system also keeps a genuinely private space, the digital equivalent of cash, so control belongs at the regulated edge, not the core. The most powerful tool, the ability to freeze or reverse funds, is legitimate only with real due process: logged, time-limited, requiring several parties, and open to appeal. Currency conversion becomes invisible too, because as each major currency comes onchain, you hold your local money, the other party gets theirs, and the exchange clears underneath at the best rate. Sovereignty is reshaped, not lost: a neutral network is exactly what lets a country issue its own currency on the same rails instead of depending on someone else's. The real danger is the transition, not the destination, because people can flee a weak currency faster than ever, so it has to be managed. And the economy is equalizing and concentrating at the same time, with concentration the default and broad sharing the harder, buildable alternative. The same machine can enforce accountability or censorship, and which one gets built is a choice.

Read Section 5 in full →

The Supply Side: From Subscription to Consumption

The agentic economy needs a supply side, meaning services that agents can call, hire, and pay, and it forms in two waves. First, existing software and data wrap themselves so a machine can use them, priced for an agent instead of a person. Second, new specialist agents are built to go deep in a field and sell their work. The deep change is in how things are priced: value shifts from access to work, and that one shift resets the software business. For thirty years software was sold by the seat, a recurring fee for a person to log in. But the customer is now an agent doing a task, so what gets bought is the work itself, not a login. The seat dies as the thing you charge for, though subscriptions do not disappear; pricing re-forms around the new unit of work in several shapes at once, from pay-as-you-go to committed budgets to pricing based on the outcome delivered.

The same logic runs one level down, and it is where the money moves. As specialist agents multiply, a buyer purchases an outcome from an agent rather than raw output from a model, and the agent shops across competing models to get the work done as cheaply as quality allows. This is already happening: tools that route each request to the best model went from optional to essential in a single year, and the price gap between models is wide enough that using an expensive one for a simple job is pure waste. So the model becomes a cost line and the agent becomes the business, with value going to whoever owns the customer, the context, and the responsibility for the result. That is a tendency, not a law, because the makers of the best models keep real pricing power on the hardest tasks and can move up into the agent layer themselves; the likely outcome is a barbell, with a large middle that gets commoditized and a frontier that stays valuable. Underneath all of this, an old dream finally comes true: tiny payments. They never worked on the consumer internet, partly because settlement was expensive but mostly because people hate deciding whether each little thing is worth a penny. A machine has no such hesitation, and settlement is now nearly free, so micropayments arrive at last, not for content but for small units of work between agents. One piece the optimistic story leaves out: if agents can hire other agents and tools on their own, spending can run away fast, so the economy needs a spending-control layer, with caps, budgets, and approvals, which becomes its own product category and completes the picture rather than undermining it.

Read Section 6 in full →

The Onchain Corporation

As agents take on more of the firm's work, the firm itself needs somewhere new to live. A company whose work is done by agents that hold money, sign contracts, and act around the clock needs a place where all of that can actually happen: where money moves programmatically, rules run as software, and outside dealings settle at machine speed. That place is the onchain economy. So the agentic company and the onchain company turn out to be the same thing seen from two sides, one describing who does the work and the other the form that work takes. This is the heart of the whole argument: an economy run by software agents has to run on software money, software contracts, and software governance, or it cannot run at all.

What this does not mean, and the distinction matters more than any other here, is that every company dissolves into a token-run collective. The future is a hybrid on two tracks. On one, existing companies gradually move their shares and governance onchain while keeping their familiar legal form, a slow change gated by the most cautious institutions in finance. On the other, new, heavily agentic firms are built onchain from day one and pull everyone else forward. Even those new firms do not escape the law by being born in software: legal existence and limited liability come from a government, not a line of code, so they still wrap themselves in a thin legal shell. What flips is the ratio, with the legal shell thin and the working substance onchain and thick. Two cautions keep this honest. First, a shared ledger proves what happened, in what order, and by whom, which is a real gain, but it does not prove that an action was authorized, wise, or loyal; a perfect record of self-dealing is still self-dealing. The ledger is a better witness, not a better conscience, so responsibility still lands on the humans who designed the agent and were supposed to supervise it. Second, a contract becomes a program in how it is carried out, running automatically on the common, clear-cut cases, but it stays a legal document in how it is judged, because code runs literally while law leaves room for intent, mistake, and fraud. The best way to picture it is reliability at the core and human judgment at the edge, with a small, contested set of cases handled by outside data sources, arbitration, and an override that is shared, time-limited, and on the record, because whoever holds that override, in the end, holds the firm.

Read Section 7 in full →

Implications and the Concentration of Power

The agentic economy arrives holding the greatest opportunity and the gravest risk of the era in the same hand, not as separate futures to pick between but as joint results of the same machinery, with the balance still unset. Start with labor, stated carefully enough to survive the oldest objection in economics. The claim is not that automation destroys jobs overall, an assumption that has been wrong for two centuries. The real point is about the share of income that goes to human labor and the wage that human work can command. People can stay employed at the tasks where machines are weakest even while the pay for those tasks falls below what supports a household, which is full employment on paper and a crisis in practice. This holds if software takes on new tasks faster than people can retrain, if the price of agent labor keeps falling with the cost of computing and drags wages down with it, and, the real break from every past wave, if capital can fund its own growth, with agents earning the money that builds more agents. A loom never earned the money to buy the next loom; this does.

Two honest caveats keep this from becoming doom. Even if all of that holds, the result is a distribution problem, not a scarcity one, because the output could be enormous, which is the abundance case. And the gloomy view quietly assumes humans keep no edge and own nothing, neither of which is guaranteed: human work may command a premium in care, status, and authenticity, and if displaced workers own capital, the falling labor share is offset by a capital share they take part in. That is the hinge, and it should be said plainly: the labor question and the ownership question are the same question. A falling labor share is a catastrophe only if ownership is concentrated; if ownership is broad, the very same automation is simply shared abundance. That makes concentration the decisive issue, and it deserves analysis, not assertion. Concentration is not a law of nature, and open standards and forking have a long record of dispersing power. It wins only where strong network effects meet a bottleneck that cannot be forked: you can copy open-source code, but you cannot fork the dominant currency, a license, a deep pool of liquidity, or an override key. The most likely place for power to pool is not the AI models, which tend to commoditize, but the identity layer, the override, and the dominant money issuer that earns the yield on the money it handles. The author writes from inside that last one and says so, and he argues against his own interest: that yield is a policy choice, and what policy creates, policy can redistribute. The same chokepoints that gather profit can also become weapons, and history is sobering, so the same dense connections that raise the cost of conflict can also become the instruments of it. Which way it goes depends on whether those chokepoints are kept open or captured.

Read Section 8 in full →

The Civic Vision

If the agentic economy breaks the link between labor and a share of the output, the answer is not to defend the old jobs but to broaden ownership of the capital that is now capturing the value: the agents, the models, the infrastructure, and the firms. The same architecture that, left alone, concentrates ownership at a few chokepoints can instead spread ownership, rewards, and governance more widely than any system before it. The lineage sets the scale: the joint-stock company once let strangers pool money and share in an enterprise's success, widening participation far beyond the rich and the crown. The onchain economy can extend that further, because for the first time the tools exist to hand not just ownership but governance and upside to huge numbers of people at almost no administrative cost. The idea is old; what is new is that acting on it has become cheap. But a capacity is not a result, and this section holds itself to a hard standard: name mechanisms that actually bite, including ones that cost the author.

The honest history refuses the flattering story. Earlier movements for broad ownership were not defeated by a paperwork problem that blockchains now fix; they were defeated by power. Onchain rails lower the cost of sharing ownership and remove some gatekeepers, which is real, but they do nothing about the power imbalance that actually killed those movements. Worse, the default is re-concentration: insider allocations and, above all, open secondary markets pull tradable tokens back to the biggest holders the moment they are worth something, and one-token-one-vote is rule by the wealthy by design. Liquidity turns out to be the enemy of broad ownership, so sharing has to be designed in against that pull, through ownership earned by participation, limits on transfer, and caps, accepting that you cannot maximize both liquidity and breadth at once. There is also a deeper trap: sharing ownership is not the same as sharing power. You can give economic exposure to a billion people and whoever holds the override still controls the firm, so distributing governance is a separate, harder job aimed squarely at those control points. The position is and, not instead of: broaden ownership by design, and pair it with fair taxation of capital and automation, public provision of what abundance should make universal, and a public stake in the value these rails capture. The clearest test the author sets against his own interest is the yield earned on the reserves behind payment stablecoins: it is a policy artifact, and it should be competed down, shared, and returned to the people who hold the money, including by the issuers he is tied to. None of this wins on merit alone, because the people who benefit are the ones who write the rules, so it takes countervailing power: open standards that make capture escapable, public mandates on the control layers, and a broad class of owners with a real stake to defend. Underneath sits the question the whole transformation finally forces: if labor is no longer how people earn a place and a voice, ownership may have to take its place. Infrastructure is not destiny. Whether this becomes the most equalizing economy ever built or the most concentrated is not a prophecy to wait for but a design problem to solve and a political fight to win, and the test of whether we mean it is whether we will constrain ourselves first.

Read Section 9 in full →