Florence, 1450s. Us, now.

In 1436 a goldsmith with no architectural training closed the largest masonry dome ever built, over a cathedral begun a hundred and forty years earlier and still open to the sky at its crossing. Filippo Brunelleschi raised it without the wooden centering every master builder in Europe said was indispensable, because no forest in Tuscany held trees long enough to span the forty-five metres of the octagon. He invented the hoist that lifted the stone, the herringbone bond that let the bricks hold themselves up as the courses curved inward, and the double shell that made the whole thing light enough to stand. Four million bricks. Sixteen years. A city of forty thousand people watched it happen.

I have told this story on three pages of the Foundation's portals, each time in a paragraph, and each time the paragraph did the story an injustice. The dome is the usual illustration of Renaissance genius. It is a much better illustration of something else: what happens when a single person gets access to capital, to knowledge and to the permission to act, all at once, in a place that had never before put the three together. That is the argument I want to make at full length here. It is the argument underneath everything I build.

I. The parallel

Florence in the 1450s was not the richest city in Europe, nor the largest, nor the best governed. Venice had more ships, Milan more soldiers, Rome more authority. What Florence had was a coincidence, and the coincidence is the point.

The financial revolution. The Medici bank, founded in 1397, had by mid-century branches in Rome, Venice, Geneva, Bruges, London and Avignon. It was not the first bank and not the largest. What it did was make credit portable and abstract: a bill of exchange written in Florence could be settled in Bruges, and double-entry bookkeeping, which Florentine merchants had been refining for a century before Luca Pacioli codified it in 1494, turned a merchant's affairs into something that could be read, audited and financed by a stranger. Credit stopped being a favour between people who knew each other and became an instrument. A workshop could be funded on the strength of its books.

The knowledge revolution. Gutenberg's press was working in Mainz around 1450; the first Italian press opened at Subiaco in 1465 and Venice had one by 1469. Before the press, a manuscript copy of Euclid cost what a house cost, and a scholar's library was a life's inheritance. Within fifty years Europe held something like twenty million printed volumes. Knowledge stopped being scarce. Ideas travelled faster than the people who had them, and a workshop in Florence could read what a workshop in Nuremberg had learned.

The consciousness revolution. This is the one historians of technology tend to skip, and it is the one that made the other two matter. Civic humanism, the strand of thought that ran from Petrarch through Leonardo Bruni, the city's chancellor, made a claim that sounds ordinary today and was radical then: that a person's worth is in what they make and do, and that individuals, not only princes and popes, can shape the world they live in. Bruni wrote a history of Florence as the story of its citizens. That is a small thing to write and an enormous thing to believe. It gave the goldsmith permission to argue with the master builders.

Notice what the three did to a single workshop. A master could now borrow against his books, read Vitruvius in print (Rome, 1486) instead of by rumour, and describe himself, without blasphemy, as an author. Alberti wrote the first treatise on painting in 1435 for exactly such readers, and dedicated it to Brunelleschi. The workshop stopped being a place where a craft was inherited and became a place where a problem was solved, with money, method and the standing to try. That is the unit the Renaissance was made of.

Take any one of the three away and the dome does not happen. Without the bank, no one funds sixteen years of experiment. Without the press, the geometry stays in a monastery. Without humanism, Brunelleschi defers to his betters and the crossing stays open to the sky. The Renaissance was not caused by genius. Florence had no more geniuses per head than Bruges. It was caused by infrastructure meeting readiness: three independent systems reaching maturity in the same decades, in the same square kilometre, and reinforcing each other. Then it ran for a hundred and fifty years.

Nobody planned it. That is the part I keep coming back to. No council of Florence decided to have a Renaissance. The conditions were there, and the people who noticed the conditions built into them.

II. The three convergences

I think we are in that moment again, and I want to be precise about why, because the claim is easy to make and easy to dismiss. Every decade someone announces its own Renaissance. The test is structural: are there three independent systems, each reaching a threshold on its own, each of which makes the other two more powerful? I count three.

Intelligence: coordination gets cheap

For most of the twentieth century the firm existed because coordinating people inside a hierarchy was cheaper than contracting for the same work on the open market. That is Ronald Coase's 1937 argument, and the Manifesto spends a section on it; I will not repeat it here. What I want to add is what it looks like from inside a small organisation in 2026.

I run a foundation, its operations and its ventures' planning with a handful of people and a workforce of AI agents that plan, write, verify and report, under supervision. Five years ago each of those functions was a department. The agents replace the coordination cost that made the department necessary, the meetings, the handoffs, the translation of one specialist's output into another's input, and they leave the people. When that cost drops toward zero, the reason to be large drops with it. One person with fifty agents is a headcount I can point to.

The bank made creation fundable by making credit abstract. AI is making creation manageable by making coordination abstract. The parallel is exact.

Trust: verification gets cheap

Trust has always been the expensive part of any transaction. We pay for it in lawyers, auditors, escrow agents, notaries, brand premiums and the years it takes to build a reputation. Every intermediary in the modern economy exists because verifying a claim is costly and someone has to carry the cost.

Nick Szabo described smart contracts in 1994 as agreements that enforce themselves. Bitcoin has run without an operator since January 2009. What has changed in the last three years is the maturity of the tooling around them: zero-knowledge proofs that verify a claim without exposing the data behind it, decentralised identity that lets you prove one fact about yourself without handing over every fact, and formal verification that proves a contract behaves correctly under every input, rather than under the inputs a tester thought to try. My own training is in the last of these. I can say from inside the discipline that the gap between what the mathematics allows and what the industry ships is closing, fast.

When trust can be verified by a protocol instead of purchased from an institution, the institution's rent disappears. The press did this to the scribe. Programmable trust is doing it to the intermediary.

Consciousness: consent gets withdrawn

This is the convergence technologists miss, and they miss it for the same reason historians skip humanism: it is not a machine, so it does not look like infrastructure. It is.

The extraction economy runs on consent. People accept that a platform owns their audience, that an employer owns their output, that an intermediary takes a third of the value they create, because the alternative has not existed. Over the past decade that consent has started to go. Regulators in Europe have written it into law with the Digital Markets Act. Two hundred million people are attempting independent economic lives as creators, freelancers and one-person companies, most of them on rented land they know is rented. Workers who left their employers in the years after 2020 did not all come back. Cooperatives, community land trusts and open-source foundations are being founded by people who have read the terms of service and declined.

All of this is demand for a kind of infrastructure that does not yet exist at scale: infrastructure where the person who creates the value keeps the title to it. Bruni gave the goldsmith permission to argue. The same permission is being taken, this time, by hundreds of millions of people who have decided the current arrangement is not the only one possible.

Why three and not one

Each of these has been advancing on its own for a long time. AI research goes back seventy years. Crypto has been building since 2009. People have wanted control over their economic lives for as long as there have been economies. The singular thing about 2025 to 2030 is that all three are crossing thresholds in the same window, and each one needs the other two.

The reinforcement is concrete, and it runs in both directions. An agent that can negotiate, buy and deliver on its own is worthless the moment its counterparty cannot verify who it acts for and whether it will settle; the trust layer is what makes the coordination layer usable outside a sandbox. A verification protocol that requires its users to manage keys, gas and proofs by hand will be adopted by cryptographers and nobody else; the intelligence layer is what puts it in the hands of the two hundred million. And a demand that has nowhere to go stays a grievance. Each convergence is the missing piece of the other two.

Agents that coordinate at machine speed need trust that verifies at machine speed, or the first fraud ends the experiment. Cryptographic trust needs intelligence to become usable by people who will never read a private key. Both need human demand, or they are cathedrals, which is a problem I take up in the companion essay. Carlota Perez's model of technological revolutions and Brian Arthur's increasing returns explain why a window like this closes rather than stays open; the Manifesto's first section carries both, so here I will only say that the mechanism is real and has a name.

III. Why the window closes

The Florentine window did not close because the thesis expired. Banking, printing and humanism kept working for centuries. It closed because the infrastructure got captured. By the 1530s the Medici were dukes, the republic was over, and the institutions that had made a goldsmith's argument possible were owned by a family. The conditions for a Renaissance persisted. The access did not.

The capture was not dramatic. No one abolished credit or banned the press. The Medici simply became, over three generations, the only door: the bank that could fund you, the court that could commission you, the family whose approval made the argument admissible. Cosimo funded the Platonic Academy and Lorenzo chose which artists worked. The infrastructure was intact and it now had an owner. Cultural production continued, and it continued on the owner's terms.

I expect the same of this window, and I expect it to close faster, for three reasons that are visible today.

Intelligence is being concentrated. The frontier models are trained by a handful of companies with the capital to do it, and the agents most people will use are being built inside those companies' platforms, on their terms of service, with their memory of you. An agent economy assembled on that foundation is the extraction economy with better tooling. The coordination cost drops to zero and the saving accrues to the platform.

Trust is being regulated toward the incumbents. Every jurisdiction that writes rules for programmable trust writes them with the existing intermediaries in the room. Some of that is prudent. Much of it is capture: a compliance regime that only a bank can afford is a moat, whatever it is called. If the rails for verifiable trust end up licensed the way the rails for payments are licensed, the protocol will exist and the access will not.

Consent is being renewed by exhaustion. People who have declined the current arrangement need somewhere to go. If sovereign alternatives are not operational when they look, they go back. Withdrawal of consent is a window in itself, and it has a shelf life. A generation that tried to leave and found nowhere to stand will not try again.

The mechanism underneath all three is path dependence. Infrastructure built in a deployment window compounds: every user, every integration, every developer who learns the stack makes the next one more likely to choose the same stack, until the choice stops being a choice. Whoever builds the rails of the agent economy in this decade will own its defaults for the next several. The agents are coming regardless. The question is what infrastructure they find when they arrive, and by 2030 that question will have an answer.

That is what I mean when I say the window closes. The forces do not stop. The access to them does.

IV. What we're building instead of waiting

The Florentines did not wait for a Renaissance to be declared. They built into the conditions. That is the only response to a closing window I know of, and it is the reason Fucina Nexus Foundation exists.

The Foundation is a non-profit in Rome, and its programme is a set of ventures, thirteen of them at the time I write, each solving a problem for real users and each, by design, needing a piece of sovereign infrastructure to do it: verified identity, programmable trust, value exchange without a gatekeeper, governance that does not require a boardroom. What the Foundation does with those ventures, and what it commits to, is written on its own pages, in its own voice. I will not restate that here. This essay is my reasoning, and the reasoning is what I own.

What I will say is where the reasoning leads. The three convergences give a builder three instructions. Build for one person with fifty agents, because that is the unit of production this decade is producing. Build trust into the protocol and prove it, because the agent economy will move too fast for trust purchased after the fact. And build for people who have already withdrawn their consent, because they are the demand, and they will not wait long.

The Sovereignty Stack, published in March 2026, is the full statement of what those instructions add up to: the window, the diagnosis, the stack, the model for harvesting infrastructure out of ventures that needed it, and the alignment between the people who build and the people who fund them. This essay is its first section at the depth the first section deserved. The companion essay on cathedrals is its fifth. Read them in that order, then read the whole.

Brunelleschi did not know he was living in a Renaissance. He knew a dome needed closing, that the tools to close it had arrived, and that nobody had yet put them together. He put them together.

We are in that moment again. The forge is lit.

Ex Fucina, Nexus.