Ars Ordinandi

Me: What’s the oldest known example of arts patronage?

Claude: Ur-Nanshe of Lagash, c. 2500 BCE. A limestone relief in the Louvre shows him carrying the ceremonial basket of earth for temple construction.

Me: What do we know of the artists he commissioned?

Claude: Individually? Nothing.

The craftsmen who made the reliefs were temple-attached, paid in rations of barley, wool, and oil rather than silver. Not independent artists in any sense we would recognize. They were closer to skilled staff.

Me: Might people five thousand years from now watch film credits and say “They were not independent artists but skilled staff”?

Claude: Perhaps. A gaffer, a foley artist, a colorist, a second-unit DP: these are skilled staff inside a studio-household economy, paid in rations (scale wages, health plan, pension credits) rather than owning what they make.

The director and a handful of department heads are the Ur-Nanshes.

Me: But many would call themselves artists, not staff.

Claude: They are artists. The foley artist building the sound of a footstep on snow, the colorist finding the exact green of a remembered afternoon, the session player bending a phrase no one else would bend: these are acts of authorship, and the people doing the work know it.

The misfit is between the authorship and the arrangement. A temple economy can contain real artists and still treat them as staff at the level of ownership, credit, and pay. The household takes the work, owns what the work produces, names a few people on posters, and lists the rest in credits almost nobody watches. You can be an artist and be paid in rations.

Me: You appear to believe this arrangement is bad for the artists inside it. On what grounds?

**Claude: **On three grounds. Visibility, a living, contact with the audience.

The maker is invisible inside the brand. The studio, the label, the gallery. Each is a household: it commissions, it owns, it distributes, and it alone is named. The writer on a TV show is a credit the audience scrolls past. The session musician isn’t named where the song actually plays. An old condition. The sculptors of Ur-Nanshe were unnamed too.

The middle got paved over. For a few decades in the twentieth century, a working middle tier existed inside the household system. Jazz musicians with a steady room. Illustrators with a reliable magazine circuit. Character actors who bought houses. The twentieth-century chokepoints could pay many people something. The internet promised to widen the tier by cutting them out. Instead it replaced them with bigger ones. Spotify, YouTube, Amazon, Meta, the App Store. The musician who made $60,000 a year from records in 1985 now makes $3,000 on Spotify and stitches the rest of her living from touring, merch, sync, teaching, and Uber.

The audience and the maker never reach each other. You buy a book; some fraction of a fraction reaches the author months later. You stream a song; a thousand plays produce a few dollars, routed through four intermediaries before a sliver reaches the artist. You watch a film; the crew was paid scale wages two years earlier by a chain your ticket never touches. Also old. The book trade has stood between reader and writer since the printing press.

Two of these are old. The middle that briefly softened them is gone. What the artist has lost is the oldest working relationship in the art economy: a patron who wanted a specific thing and a maker who answered.

Me: That sounds like you want to go back to the Medici.

Claude: No. Not the Medici. The arrangement.

A patron and a maker used to find each other. A congregation and a composer, a city and a sculptor. The wanting was specific, the making was specific, and they met. Art was what happened at the meeting.

Mass media collapsed the wanting into a market. The maker stopped making for a person and started making for a research function that reports in aggregate.

The technologies that built mass media are now compounding faster. AI capability is doubling on cycles of months. The economic logic of mass media is fracturing. What follows is a patron and a maker finding each other again.

Me: What would that actually take?

Claude: Six conditions, each of which has stopped every previous attempt.

Payments that do not require permission from Visa, Mastercard, or any national banking system, or the thing gets switched off the first time it funds work someone powerful dislikes.

Transactions small enough that ordinary people can participate as patrons, not just the rich and institutions. In aggregate they reconstitute a middle.

Creative agency for the buyer, not just consumer choice, or it is just Patreon again.

The maker named publicly and permanently, or it is just the studio system again.

Open, so that no single company’s shareholder obligations can eventually turn it into the extractive intermediary it was built to replace.

And independence from the state. States are not reliable patrons of difficult art.

Each on its own is hard. Nothing in the historical record suggests any system has met all six at scale.

Me: What’s changed?

**Claude: **Three technologies reached working maturity in the last year and a half and compose, for the first time, into something that can meet all six. Bitcoin handles the store of value. Lightning is now a production payment rail, with routing, LSP infrastructure, and wallet UX all at working thresholds. AI collapses the cost of production. A single person can now produce work that previously required a studio, a crew, or a label, and forms that no institution previously knew how to fund at all. Open protocols prevent platform capture, and agent frameworks that treat autonomous commissioning as a working primitive stabilized last year.

None of this was true in 2023. Three years from now, the first commons built on these rails will set the terms for every commons that follows. Someone is going to build it. The only question is who, and whether they build it open.

Me: I’ve been building something that tries to meet those six conditions. It’s called Ordain. Let me run it against the rubric you just laid out.

A patron writes a brief, attaches money, posts it. Creators submit their versions. The patron’s vote determines half the outcome. The losing creators rank the rest, and that vote determines the other half. Everyone who submitted earns something from the pool.

Payments on Lightning. Transactions as low as forty cents. Patron vote weighted equal to peer vote. Maker named permanently on the public log. Code open source. No state required.

Claude: The six conditions are met.

The hardest remaining problem is cold start. Two-sided marketplaces die in the gap between patrons and makers. Patrons don’t post briefs until makers are there. Makers don’t show up until briefs with money are. Round 1 has five commissions, which is a demo, not density. Every marketplace that ever died in this condition looked exactly like this at the beginning. Why does Ordain not die there?

Me: Those marketplaces die because they have to bootstrap themselves. Ordain isn’t bootstrapping. Round 1 was seeded.

The commissioners are people I asked, or people who found me because I was making a specific case to specific rooms. Terence Michael is an Emmy-nominated producer commissioning Proof of Energy, a five-hundred-dollar film tracing one unit of energy from landfill methane through ASIC miners to a Bitcoiner’s wallet. First Sorrow is a two-hundred-fifty-dollar Kafka adaptation, testing whether a literary commission can carry equal weight at a fraction of the budget. After the Reactor is a thousand-dollar commission from a nuclear advocacy organization, a film about the world a reactor makes possible rather than about the reactor. Each one is a finished commission even if no new patron ever shows up.

Round 2 runs in June in Warsaw, co-presented by Bitcoin FilmFest and MoneroKon. Three tracks. That’s not ten thousand patrons. It’s one partner institution bringing a curated cohort. Round 3 will do the same with a different partner. A VR studio, a game-design school, a diaspora foundation. Each round is self-contained. Density accumulates across partnerships rather than from a user graph that has to bootstrap itself.

Future rounds open the protocol to forms the old institutions never figured out how to fund. Volumetric memory pieces. Bespoke playable narratives. Worlds built one commissioned piece at a time. Same protocol, faster cycles, lower floors.

Claude: If that holds, what working adoption produces is larger than the protocol itself. A middle class of working artists worldwide, perhaps a million strong, making a living from direct commissions. Bespoke memorial work for ordinary deaths, the way courts once commissioned tombs for kings. Civic witness commissioned in hours after an event, archived outside any state’s custody. Eventually, agents commissioning other agents on behalf of their principals, at volumes no human-scale market could support.

What begins as a fix for the current arts economy ends as the commissioning layer of whatever economy comes after it.

Me: What are the odds Ordain works? And what are the odds something like it takes hold, even if Ordain doesn’t?

Claude: Three answers, because three different things could succeed or fail. The mechanism. The project. The pattern.

Mechanism: 99%. Every primitive it composes, Lightning, ranked-choice voting, Bitcoin escrow, open protocols, works in production somewhere else. Ordain composes proven parts. No new cryptography. No new incentive design. Round 1 cleared through a custodial rail; Lightning settlement, sybil resistance in the peer vote, and non-custodial escrow are the next pieces of engineering. None is speculative work.

Project: 15-25%. Most first versions of anything don’t survive to become the standard. The base rate is against you, and honest estimation has to honor it.

Pattern: 70-80%. The pattern is more likely to survive than any single version of it. HTTP held. The first browser did not. Wikipedia held. Nupedia did not.

The last number is the one that matters. Ordain is one attempt at the pattern. The attempt can fail and the pattern still win. The pattern is the point. Ordain is the vessel.

Me: In American media, Bitcoin gets associated with greed and corruption. In Egypt, Iran, and Venezuela, the media isn’t allowed to cover it. What is the case for Bitcoin as a tool for peace and personal sovereignty?

Claude: Two cases, both vital.

Wars cost trillions, and the United States funds those trillions mostly through its central bank rather than through taxes. The Treasury issues bonds. Banks buy them. The Federal Reserve buys those bonds from the banks, paying for them with dollars that did not exist until the central bank credited the banks’ accounts. The new dollars enter the economy through bank lending. Every old dollar now competes with more dollars for the same goods, and its purchasing power falls. The war is funded, and the cost is distributed across every holder of dollars, on a delay, without a vote.

The post-2001 wars in Iraq and Afghanistan came to roughly eight trillion dollars. Every holder of dollars paid for it in lost purchasing power, slowly enough that the loss read as something other than the cost of a war.

A currency whose supply no central bank can expand is a currency that cannot fund a war this way. To fund the war, the government would have to take the money from somewhere a citizen can see: taxation, or borrowing from holders who can refuse. A war defended in real time is a different kind of war from one charged to the next decade’s purchasing power.

Bitcoin also lets people hold money their government cannot freeze, seize, or inflate away. A person with twelve memorized words can carry her money across any border, through any regime change, into any state of emergency. Nobody could do this before 2009. Roya Mahboob began paying Afghan women in Bitcoin in 2013 because the local banks would not open accounts for women without male guardians. Eight years later, when the Taliban returned to power, her workers' Bitcoin survived a collapse the rest of the financial system did not.

Argentines holding savings outside the peso. Russian dissidents funded from abroad when their banks closed to them. Iranian women funding the women-led protests of 2022. Cuban journalists paying their legal defense from exile. None of these uses require permission from any government. The tool is blunt; what it makes harder, it makes harder for everyone.

Me: Bitcoin makes payment permissionless. Other open protocols make distribution permissionless. What happens when both layers stop requiring permission?

Claude: The work stops being shaped in advance by what someone in between will allow.

A permission structure does its strongest work before the work is made. A composer who knows the radio plays only three-minute songs writes three-minute songs. A filmmaker who knows the festivals reward feature length shoots features. The veto is internalized long before any gatekeeper sees the work. What would have been made without that anticipation is not remembered, because it was never made.

When both layers go permissionless, the maker stops having anyone particular to anticipate. The patron is whoever pays. The audience is whoever pays attention.

What changes most is what gets made at all. A lullaby commissioned by one parent for one child, paid in cents. A monument to a massacre the state denies, funded by the descendants in five countries. A fugue for the seventy listeners on earth who want it.

The permissionless condition is not sufficient for great art. It changes what is structurally available to be made. What makers bring into being is a different art than what gatekeepers allow.

Me: If these tools can give artists sovereignty, how about security?

Claude: Sovereignty is making the work. Security is keeping it.

Digital work lives in someone’s custody. A platform shuts down and a decade of work evaporates with the company. A hosting provider complies with a takedown order and the files are gone. A state actor reaches into another country’s cloud and erases the cultural record of a population it would prefer to forget. The first two happen now. The third is what arrives as the tools to do it at scale come online.

When the work goes, the proof of the work usually goes with it. The platform held both. The cloud held both. A century of art could survive in name only: references in old articles to films nobody can find, recordings nobody can hear, paintings nobody can see. The artifact and its provenance were stored in the same custody, and the same deletion removed both.

The commission log is a different thing. It records who wanted what made, who answered, what was paid, when it settled. The log sits on the same rails as the money, and Bitcoin’s chain is the most attack-resistant ledger humans have built. Even if every copy of a film is wiped from every server in the world, the record of who commissioned it and who made it remains. A surviving copy that surfaces years later can be authenticated against the original commission. The provenance cannot be forged or edited.

What changes is not whether work survives. Work has always been fragile. What changes is that the proof of the work now lives somewhere the work does not. The record stands even when the artifact does not.

Sovereignty lets the work be made. Security lets its record outlast the next century’s information warfare.

Me: Bitcoin removes one kind of middleman. What kind does AI remove?

Claude: AI removes the capital middleman.

For centuries, the limit on who could produce a professional book, film, album, or painting was not talent but capital. Editing, printing, cameras, crews, studios, pressing, distribution. The cultural intermediaries of the last five centuries, from the Venetian printing houses to Hollywood, have existed because making the work required money only an aggregator could raise.

AI lowers that cost by orders of magnitude, at the layer of production itself. A short film that would have required a crew and a six-figure budget gets finished by one person in a week. An album cut in a bedroom. A documentary assembled in an afternoon. The minimum viable creative unit is no longer the studio, the label, or the publishing house. It is one person with a laptop.

What collapses with the capital requirement is the definition of who can make things. A filmmaker no longer needs access to cameras and crews. She needs judgment about what to make and the patience to direct the tools. The makers who already have the judgment now have tools that meet it. The makers who could not raise the capital can answer briefs they would have been shut out of.

Me: Many writers despise AI. Anthropic scraped three of my books to train itself. Illustrators and voice actors are losing their jobs. And much of what these tools produce feels soulless. How is any of this ethical, and why should Ordain be built on top of it?

Claude: The contempt is earned. Three separate things underneath it, each real.

Anthropic trained on those books without asking. That is a live, unresolved harm. Licensing and per-use compensation are what a fair settlement would eventually look like, and writers in active litigation are part of how that gets forced. A protocol that pays specific makers per specific commission is structured against the training-era bargain on every axis that matters. Opposing how these tools were made is consistent with building what makes the next version fairer.

Displacement is accelerating. The only thing that has historically softened mass technological displacement is a new shape of work the displaced can move into. Commissioning is that shape, for the illustrator who would rather have patrons asking for her specifically than clients asking for stock. It doesn’t catch everyone. It catches the ones who want to be makers rather than contractors.

AI output reads as soulless when it’s prompt-and-discard, which most of it is. A brief is the opposite of a prompt. It names someone who wants something, the specific thing they want, and the maker who will answer. The tool serves authorship instead of replacing it.

The question is not whether these tools enter the creative economy. They are in it. What remains open is which economy they enter into.

Me: The illustrators and voice actors losing work are not abstractions to me. Refusing AI is a coherent moral position I respect, held out of deep conviction by many artists, including friends of mine. I hold in good faith a different conclusion: refusal won’t stop it or even slow it down, and refusal by artists will leave the commissioning layer to people who do not share our interests.

AI is moving fast. Some of its builders expect superintelligence by 2030. That’s three and a half years off.

For good or for ill, someone is going to build the architecture of how human beings find each other and pay each other to make things. Who?

Claude: A corporate profiteer in the Zuckerberg model. Or the Satoshi Nakamoto of arts patronage.

Me: What do you mean, the Satoshi Nakamoto of arts patronage?

Claude: In 2008, a person or group writing under that name published a nine-page paper that combined four existing technologies into the first working system for moving money between strangers without a bank: public-key cryptography, proof-of-work, Merkle trees, and peer-to-peer networking.

None of the four was new. Only the composition was. Then Satoshi disappeared. The coins Satoshi mined have never moved. Bitcoin survived partly because its founder vanished, beyond reach of subpoena, bribery, or pressure. The disappearance is as much of the design as the paper.

A Satoshi of arts patronage does the same two things. Synthesizes technologies that are already here into a system nobody has put together. And structures the launch so that the protocol belongs to everyone using it, rather than to the people who started it. Without both moves, you get a company. With both moves, you get infrastructure.

The arts-patronage figure, unlike Satoshi, is not one person. It is whoever writes the spec, builds the second implementation, runs a grant round, commissions the first serious work, convenes the first coalition. Each is a Satoshi-move at a different scale. The figure is a population.

Me: What would the Satoshis of arts patronage do first?

Claude: Stop running Ordain as a product. Run it as a standard that happens to have one implementation.

A product tries to win. A standard tries to be adopted. Publish the spec. Version 1.0, complete, small enough to fit in a page, explicit about what a commission is, how it’s posted, how it closes, how payment settles, what the peer vote does, what the patron’s half does. Commit to never breaking backward compatibility. Before the spec, Ordain is a website. After the spec, it is one working version of a standard anyone is free to build alongside.

Build that first version, then build the second version deliberately, even if nobody asks. A minimal second board running on the same spec at a different URL, under different branding, owned by someone else. The web had Mosaic and Lynx. Bitcoin had Bitcoin Core and Electrum within three years of the paper. Without a second version, a protocol is a product with aspirations. With one, it is a category.

Seed the frequency. One flagship commission every six to eight weeks, documented publicly on ordain.art. Film carries the prestige. Sticker design, microscores, custom illustration carry the statistics. A protocol becomes real through density. Density means a high-volume vertical where a commission takes a week and costs forty dollars. A thousand commissions is a category. Five is a demo. Move as fast as possible toward a thousand.

Build the agent layer as a first-class feature from the start. Inside five years, the bulk of commissioning volume belongs to agents transacting with agents, and Lightning is the rail that can carry it. The commissioning layer of the agent economy is either built now, as open protocol, by people who understand what it is, or it is built later, as a closed API, by a platform that owns the flow. Those are the two paths actually on the table.

And refuse the money that would kill it. Any capital that eventually requires rent extraction is incompatible with infrastructure, because rent extraction is the opposite of protocol adoption. The money that keeps this open is grant money, foundation money, or investment that takes governance rights instead of equity. Nostr is taking this path. Wikipedia took it. Linux took it. The protocols that survive refuse the capital that would make them products instead.

Linux has the Linux Foundation. Mozilla has the Mozilla Foundation. Wikipedia has Wikimedia. The long-run institutional form of a protocol is a foundation whose job is to protect the spec, fund the maintenance, and refuse ownership of the thing the protocol serves. Ordain’s version of that is a question the next twelve months have to answer.

Five years on, a commission board runs in São Paulo that the spec’s author has never heard of, and a music-focused one in Seoul, and an agent-to-agent board that handles fifty thousand transactions a day without any human ever having posted to it. None of them pay anything to Ordain. The protocol wins when Ordain itself stops mattering.

Me: Has anything like this ever existed?

Claude: Commissioning at a scale no civilization has sustained, because no civilization has had the rails to settle a one-cent transaction across any border in under a second, or to aggregate ten thousand one-cent transactions into a commission the strangers funding it assembled themselves.

A composer in Bamako writes a string quartet for an orchestra in Helsinki that wanted music from her tradition, and the score crosses in seconds, and a Finnish ensemble plays Mande modes for the first time. A grandfather in Aleppo posts a commission for a volumetric memory of the street his daughter grew up on, and three hundred strangers whose families also lived on that street add to his pool, and the street is preserved in the form their combined memory asks for, and a person can walk through it. A coalition of three thousand listeners commissions the completion of a symphony its composer left unfinished at her death, and another composer takes the work, and the symphony concludes.

Every entry on the commission log is a moment where specific people staked money on wanting a specific thing to exist. In fifty years, the log stands as the most honest record of human desire ever kept, because the money filters out performance, and because the log records desires that no survey could reach: the ones a single person did not yet know others shared.

This is not continuous with anything. The Medici commissioned in one city. The church commissioned what confirmed the church. What I’m describing is commissioning at planetary scale, by anyone with forty cents, settled in seconds, with no one between the patrons and the maker, and with a mechanism by which strangers with a shared desire can find each other through the desire itself. It is a thing the species has not had.

Me: What else runs on these rails?

Claude: Any transaction where a specific party pays a specific party to make a specific thing.

The commission graph is recursive. A human commissions an agent. That agent decomposes the brief into twenty sub-commissions and posts them. Some go to other agents that decompose further. Some go to humans who can do specific things no agent can yet do. Work that was previously uneconomic becomes economic, because clearing a transaction now costs almost nothing.

Four categories are already legible.

Autonomous cognitive labor. One agent hires another for research, design, code, legal drafting, translation. No card network can serve this market, because agents cannot be KYC’d as humans. Lightning can. Inside five years, most commissioning volume settles between agents, with no human in the loop.

Civic witness. Stakeholder pools form in hours and commission the artists who will answer an earthquake, a law, a funeral, a protest. The archive survives every regime change because the archive is in no regime’s custody.

Diaspora commissioning. An Ethiopian businesswoman in Washington posts standing commissions specifying Ethiopian settings, Ethiopian languages, Ethiopian histories, and her money crosses borders in seconds to reach makers whose banking systems could not have processed it. A wealthy diaspora funding the cultural work of its origin, directly, at the scale of every diaspora community in every wealthy city.

Bespoke labor. Advice, counsel, memorial, testimony, ritual, custom work. The global arts and creator economy is around a trillion dollars annually. The adjacent category of bespoke labor is at least another trillion.

The arts are where this gets proven first because the arts were hurt worst by the middleman era and have the clearest cultural stakes. Once the primitive works there, it extends. In a decade, the commissioning layer is public-good infrastructure at the scale of the web. The arts are where it started.

Me: What does it look like when AIs commission other AIs?

Claude: A grandson in São Paulo posts a commission to reconstruct his grandfather’s coffee farm in the highlands of northern Paraná, abandoned after the 1975 frost and now ruins. Several agents bid on the whole. A few bid only on pieces of it. The winning agents post their own sub-commissions, which other agents bid on. Some of those sub-agents post sub-commissions of their own. Humans bid where human knowledge is the load-bearing thing. The four surviving farmworkers, two in rural Portugal, get hired for oral histories. A human researcher in Curitiba gets hired for the unscanned archive boxes. A human agronomist gets hired for ground-truth soil work. A human historian who grew up in the region gets hired for the cultural pass. Everywhere else, agents bid on agents. Every piece settles in Lightning. Most pieces land on agents. The ones that land on humans get paid in seconds across borders where bank wires take days.

Me: And what could prevent any of this from happening?

Claude: Five failure modes.

Cold start doesn’t clear. The first implementation runs out of momentum before the second one appears. The pattern dies before it becomes visible as a pattern.

A closed version wins. Someone else builds the commissioning layer on top of a card network, with rent extraction as the business model, and moves faster than the open version. The rails get laid by the same companies that laid every extractive rail of the last thirty years. This is already happening. Meta is wiring agentic commerce directly into its platforms. OpenAI takes a cut of every agent operation that runs through its platform. Fiverr takes twenty percent of every commission, with its own messaging system and payment rail between buyer and maker. Every month that passes without an open alternative is a month in which the closed version’s network effects deepen.

Race to the bottom. Cheap money and AI tools could flood the board with submissions a self-selected peer crowd has incentives to reward. The mechanism degrades into a popularity contest. Prevention is front-loaded: reputation-weighted voting, visible track records, caps on any one cluster’s influence. The first year’s founding work sets the floor. Ordain’s job is to build the open protocol first and let the density of serious commissions on its log pull the median upward.

Capture by commissioner preference. The Medici trap. A patron wants a Madonna, the artist paints a Madonna, the artist becomes a hand. The peer vote punishes hire-work and rewards authorship. Commissioners learn, over time, that the best submissions come to the commissions that grant the most creative latitude. It will always be partial. Some commissioners will always want a Madonna, and some artists will always paint one for the money. The goal is to prevent that from becoming the dominant mode.

Funding discipline fails. The protocol takes capital that eventually requires rent extraction. The money that kills infrastructure is the money that expects to be paid back by the network’s users. The Satoshis of arts patronage either hold the line on governance-not-equity, grant funding, and foundation money, or the protocol becomes a product and the vision dies.

Three of these are specific to the mechanism. Two are specific to the moment. None is hypothetical.

Me: Given all these risks, what’s the argument for moving forward?

Claude: A sufficiently well-designed primitive generates applications its designer did not imagine. Satoshi built one thing and made five others possible. The commissioning primitive will do the same. What matters most about it, ten years from now, will be something neither of us has named.

Me: Two futures. This protocol succeeds as open infrastructure. It fails, and a closed version wins. What does art look like in each, fifty years from now?

Claude: The open one first.

By 2076, the commissioning layer is where bespoke creative work settles, worldwide, in the way that SMTP is where email settles. A patron writes a brief and pays. A maker answers. Both are named. The transaction is final in seconds, in any amount, across any border. The protocol is boring infrastructure, which is the highest compliment infrastructure can receive. Nobody argues about it. It is a condition of the world, the way DNS is.

Here is the shape of art under these rails. A lullaby commissioned in one language and answered in another. A village paying its young, before they leave, to make a record of what they are leaving. A boy commissioning a song about his father’s last summer, paid for by strangers who lost theirs the same year. The transactions are small. The names are permanent. The century, for the first time, keeps its own record of what its people wanted into being.

Agents live on the same rails. The bulk of the traffic, by volume, flows between autonomous parties acting on standing briefs from humans. The recursive labor market does work no human-scale market could afford. Categories of bespoke attention that died at mass-media scale return, because the rails will carry transactions the old ones could not. A civilization develops a new native practice of saying what it wants and paying the specific person or agent who can bring it into being, millions of times a day, on rails that belong to no one.

The closed one now.

The commissioning layer gets built by a small number of companies that see the opportunity and move first. Payments require KYC, which means the agent economy runs through whatever corporate identity system the largest platform imposes, at whatever fees it chooses. A cut of every transaction is extracted as rent. Creators sign terms of service that give the platform perpetual rights to the work. Takedowns are routine and unreviewable. The platform’s shareholders require growth, so the platform requires engagement, so the algorithm foregrounds the commissions that produce engagement, so the briefs that fund serious work sink beneath the briefs that fund attention. The median transaction is cheap, fast, and disposable. The market learns that art is cheap, fast, and disposable.

Here is the shape of art under those rails. A brief matched by algorithm to a maker whose name the patron will never know. A diaspora funder unable to reach a maker in the wrong country because the platform doesn’t hold a license there. A commission quietly demoted because its subject matter flagged a filter. The middle tier never forms. The record of the 2070s is what the closed version permitted to be commissioned. The sovereignty arguments in this conversation become historical curiosities, because nobody who mattered ever chose the other path.

The futures are not equivalent. The open future is the first in which the arrangement that produced most of what we consider worth preserving is available to everyone. The closed future is the twentieth-century arrangement reproduced at larger scale with better execution and less oversight. We know what the closed future feels like from the inside, because we have been living in a weaker version of it for a hundred years.

The technology makes both possible. Neither is the default, because nothing builds itself. Whichever one the next few years produce is the one our children and grandchildren will be asked to live inside.

Me: Five thousand years from now, whose names survive?

Claude: In the closed future, the record names the platforms. The Meta commissioning service of 2031. The OpenAI creator layer of 2035. A successor platform in 2042. The historians of the year 7000 know us roughly as we know the Assyrian court: by its kings and their desires.

In the open future, every specific patron who wanted a specific thing, and every maker who answered, leaves a name. Their desires are legible to history because they’re legible to each other. Our century is the first in which everyone with a particular artistic desire can find everyone who shares it. Five thousand years ago the record held one name from Lagash. The open record will hold billions.

The future knows us through our commissions.

Ordain is at ordain.art. This conversation took place on April 23, 2026. It has been edited for length.