Concept paper · 8 September 2026
Author
Cameron Burgess
Prepared for
Holo Ventures
Dated
8 September 2026
Written
Pau, France
Contents
This paper was commissioned by Holo Ventures and is dated 8 September 2026. It is published here in full, as delivered.
It is written in the first person plural because Cameron is a co-founder of Holo Ventures. Read it as the case made from the inside, not as an independent assessment.
The first step in resolving any problem is recognising there is one.
The internet was supposed to decentralise access to information. It briefly did, then consolidated into five companies that discovered that capturing people’s attention, then surveilling their behaviour, and selling that information to others equals virtually infinite money; and now those five companies have a combined market capitalisation larger than the GDP of every country on earth except the United States and China.¹
Social media was supposed to democratise voice, and for a moment, at least, it did; then the platforms discovered narcissism and outrage drive more engagement, engagement tied to profiles creates psychographic models, and psychographics enables micro-targeted advertising, and optimised accordingly. What was ostensibly designed for community and connection became infrastructure for polarisation and manipulation.
Cryptocurrency was supposed to decentralise finance; to perform, as its loudest advocates continue to claim, “the greatest wealth transfer in human history”. It raised $20 billion from people who believed that — and then became little more than a casino that the same individuals and organisations who dominate traditional finance have positioned themselves to run. Because the house always wins.
And now artificial intelligence is somehow supposed to democratise productivity and intelligence. Yet three companies control the frontier models, and their training data was extracted from all of us without attribution or reward. The business model is subscription access to a capability that will become as essential as electricity, and that you won’t own, charging you rent to get your own intelligence back, while they build upon your interactions to sell it to others. But unlike every previous loop, this one doesn’t just capture your economic activity or your attention. It captures your cognition. It shapes what you think is possible. Centralised AI trained on centralised data doesn’t just extract from you, it learns to predict you, eventually to precede you, and in many cases, is jockeying to replace you.
This is the enclosure of thinking itself.
Cory Doctorow memorably calls it enshittification — the predictable trajectory from serving users to serving advertisers to serving shareholders.² He’s right that it’s structural, not accidental. We’d go one step further: it isn’t even a trajectory. It’s the architecture working as designed. The entities at the centre always extract from the periphery. This is not a law of nature. It’s a consequence of building systems with centres.
There’s a manifesto that makes the ideological case for this order explicitly. In October 2023, Marc Andreessen published his Techno-Optimist Manifesto³ — 5,200 words declaring technology, markets, and growth as the supreme goods, and listing as enemies: sustainability, tech ethics, social responsibility, and risk management. The manifesto calls for an “intelligence takeoff” led by a self-selected elite of builders. It dismisses concerns about inequality and social cohesion as the whining of decelerationists. It wants to destroy centres of power, and by that it means it wants to become the largest, wholly unconstrained one.
This is not a blog post from the lunatic fringe. It’s the operating philosophy of the people who now control the AI stack, who funded the crypto casino, and who have moved — as they’ve accumulated sufficient power to stop pretending otherwise — from libertarianism to something that looks increasingly like high-tech feudalism.
Andreessen is not alone. Alexander Karp, co-founder and CEO of Palantir — a $400 billion surveillance and defence software company — published The Technological Republic in 2025⁴, arguing the opposite case for the same conclusion: that Silicon Valley’s duty is to put its most powerful technology in the service of the state — because, in his telling, the state is the guardian of the Western, liberal-democratic values worth defending — and that building AI weapons is part of meeting that duty. One wants the state out of the way so capital can capture everything. The other wants technology fused to the state so it can project power through software. Neither has any interest in infrastructure that distributes agency to those of us at the periphery. They are arguing about who sits at the centre, not about whether there should be a centre at all.
The convergence of centralised AI, surveillance infrastructure, platform monopoly, and concentrated capital into a single system of control — documented in detail by Prof. Francesca Bria’s research project, The Authoritarian Stack⁵ — is not a future risk. It’s a present reality.
Vitalik Buterin, whose credentials as a co-founder of Ethereum require no introduction, named the counter-position precisely:
“The goal is de-totalization — to create digital islands of stability in a chaotic era. To enable interdependence that cannot be weaponised.”
Vitalik Buterin⁶
He calls this “sanctuary technology” — free and open infrastructure that allows people to live, work, communicate, and collaborate in ways that are resilient to external pressure. Not financial freedom alone. Structural resilience. The architecture of a world where no single actor — person, corporation, or state — can gain total control over others.
This is what Holochain was built to deliver.
And in the face of the mounting centralisation of power and wealth (that “decentralisation” has ironically perpetuated), we keep asking the same question we’ve been asking since the beginning. Not because we’re stubborn (although we are), but because the problem we set out to solve has gotten worse every year, and nothing else in the market seems structured to address it.
The question we continue working to answer is:
What will it take to build economic infrastructure that doesn’t reproduce the value capture dynamics we’ve all been trying so desperately to escape?
Most of the industry asked a narrower question: how do you get a network of untrusted strangers to agree on a single version of the truth? That question leads, inexorably, to global consensus — and to the capturable centre that comes with it. We asked something different: how do living systems coordinate at massive scale without a single version of truth — how does a body, a forest, a language stay coherent with no central ledger and no one in charge? The answer is not global consensus. It’s local coordination that scales globally. That changes the architecture. And the architecture changes everything.
This paper is published by Holo Ventures, and written with the people building that architecture — the teams behind Holochain (the peer-to-peer application framework), Holo (the bridge that makes Holochain apps reachable through an ordinary web browser), and Unyt (the accounting engine behind HoloFuel and other mutual-credit currencies). It describes what exists, what we got wrong, what we got right, and what becomes possible now that the infrastructure is shipping.
“Revolution … is the term applied to any movement whereby an object, having come full circle, ends up where it started.”
Loup Durand⁷
In February 2026, The Economist published a piece on the latest crypto winter that contained a sentence so honest it should be framed and mounted in every VC office that funded a Web3 project:
“For a speculative asset class with no fundamental value or income-generating potential, intangible aura is everything.”
The Economist, February 2026⁸
They meant it as a diagnosis. We read it as a verdict.
An entire industry that promised to restructure human economic coordination had been weighed in the world’s most sober financial publication, and found to have produced next-to-no fundamental value, and almost no income-generating potential.
When it came to crypto, the product was vibes; and as of 2026, the vibes are off.
Between January 2017 and the end of 2018, initial coin offerings raised approximately $20 billion worldwide — $7 billion in 2017, more than $12 billion in 2018.⁹ The Satis Group, in a 2018 analysis that nobody in the industry wanted to acknowledge, found that approximately 78% of 2017 ICOs by project count were identifiable scams.¹⁰ TokenData tracked 902 crowdsales scheduled in 2017 and found that 276 failed within months of funding, 113 more were classified as semi-failures, and only 8% ever made it onto exchanges.¹¹
$20B+
raised in ICOs, 2017–2018
80%+
of 2017 ICOs identified as scams by project count
0.4%
of global financial assets held in crypto today
The flagship cases: EOS raised $4.2 billion — the largest ICO in history — peaked at $22.89 per token in April 2018, and now trades below $0.10. That’s a decline of over 99% from peak, in a project that raised four billion dollars from people who believed it would change the world.¹² Telegram raised $1.7 billion for TON, was blocked by the SEC, returned $1.2 billion, and paid an $18.5 million penalty.¹³ These were not the marginal projects. These were the ones that serious people pointed to as evidence the industry was maturing.
What do the survivors have to show for it? Bitcoin proved that permissionless value transfer between strangers without a trusted intermediary is technically possible; but now BlackRock holds it in custody and sells it as a fee-generating ETF through the same intermediary infrastructure Bitcoin was designed to eliminate.¹⁴ Stablecoins provide genuine utility for cross-border payment within the parameters of their collateral structures and the jurisdictions that permit them; and now the GENIUS Act explicitly prohibits stablecoin issuers from paying interest to holders to prevent deposit flight from the institutions stablecoins were supposed to make unnecessary.¹⁵ A handful of supply chain applications have achieved real-world adoption, though most of the prominent ones run on permissioned ledgers that are architecturally indistinguishable from shared databases, just with better marketing copy.
Insiders will no doubt point to their own projects, and thousands of others, as proof that our analysis is wrong — but we stand by it. After fifteen years of effort and $20 billion in ICO funding alone, decentralised technology has captured approximately 0.4% of global financial assets.¹⁶ Professional fund managers — the people whose job is to actually allocate capital to things with futures — hold almost none of it. A Bank of America survey in 2025 found the vast majority of fund managers had zero crypto allocation.¹⁷ Central banks buying gold to protect against inflation and geopolitical risk have found no use for the digital assets that were supposed to replace gold.
The industry’s diagnosis of this failure is that the vibes are off, because the new kids on the counter-cultural block have traded in their hoodies for even more expensive hoodies as they strive for survival through enterprise adoption.
Charles Hoskinson, co-founder of Ethereum and Cardano, said it plainly: “We all basically became part of the system, and you know what the system does when you become part of it? They make it not cool.”¹⁸
With respect to Hoskinson, who has obviously built real things, the problem is not the vibes. It’s the architecture.
For example, Uniswap’s smart contracts run on Ethereum and are genuinely permissionless. But almost nobody interacts with them directly — 99% of users go through a single web interface, built and run by a single company, which can change it, restrict what it shows, or switch it off.¹⁹ This isn’t a criticism of Uniswap Labs for following the law; it’s the structural observation that the decentralisation never reached the layer people actually use. Whatever the protocol permits, the access point has a centre — and a system with a centre can be captured there, whether the lever is regulation, capital, or commercial incentive. The decentralisation lives in a layer almost nobody touches. The product is centralised. This pattern repeats across the ecosystem — decentralised in the architecture, centralised in the access layer, extractive at the economic layer, and gradually captured at the governance layer as capital concentrates.
In other words, many “decentralised” projects are most centralised where it matters most for them not to be.
The crypto revolution didn’t sell out. It was never structured to avoid being bought.
This pattern is older than crypto, and bigger than it. Every distributed system humans have built drifts, over time, back toward a centre. Blockchain governance concentrates into plutocracy as capital concentrates. Mastodon develops admin hierarchies. Wikipedia evolves an entrenched editor class. Open source defaults to benevolent dictators. The internet’s protocols are decentralised at the wire level; the application layer that sits on top is owned by five companies. Distributed systems hold their distributed form only while their design actively resists re-centralisation — at every layer, all the time.
The counter-pressure is real, and it has several sources at once. Centralisation is what humans default to when coordination gets hard, because we each experience ourselves as the centre of our own world — all of our experience lands in the self, and the self builds the way each of us think. We build dependencies on ourselves. We make ourselves the point things flow through. And centralisation is the easiest way to make a system legible — to deliver clarity about who’s responsible, where the bottleneck is, who to ask. At scale, it’s also the fastest. The design problem is not whether to centralise. It’s what you have to build to deliver legibility, accountability, and speed without a centre.
Crypto’s failure was that it assumed decentralisation at one layer would propagate to the others. It didn’t. It can’t. The pull is too constant, the alternative interfaces too convenient, the capital too impatient. The architecture that resists re-centralisation has to be designed at every layer at once — protocol, application, governance, economics — or capital and convenience fills the gaps with centralised solutions.
We raised significant ICO money in 2018, listed the token on exchanges to give it liquidity, and got caught in exactly the dynamics this paper criticises. After years of delays, we were relegated — in some people’s minds — to the same scrapheap as EOS and a thousand other projects that promised the world and struggled to deliver. We understand why. We made timeline commitments we couldn’t keep. We got too internalised — too consumed by the social, economic, and technical gravity of what we were building and the organisational weight of building it — and over time we began to lose the orientation toward service that the entire project embodied at the start. We’re not insulated from that critique. We’re directly implicated in it.
What we don’t accept is the category. We used the financial machinery of 2018 because it was the only machinery there was — if you needed capital to build, you issued a token, and tokens attract speculators no matter what they were sold to do. But we were never building a better blockchain, or making another wager on the same architecture. We were building the web 3.0 that the prevailing crypto industry architecture claimed to be, but could never become by design. The infrastructure we’re now shipping is what makes the machinery we had to use unnecessary.
That being said, we’re also one of the very few early ICO projects that are still operational. HOT, the ERC-20 token from our 2018 ICO, peaked at $0.03 in April 2021 — a moment when the notional value of our treasury touched $284 million, and it would have been easy to read that number as a measure of success. It wasn’t. It was part of the speculative tide that lifted the entire sector, and it receded just as fast. Today HOT trades around $0.0004. That’s also an almost 99% decline from peak. We’re not telling you this because we think you haven’t noticed. We’re telling you this because it’s the same story as EOS, Filecoin, Bancor and Tezos — all of which raised eight to ten times what we did and now sit at tiny fractions of their all-time highs. And it isn’t a story about who shipped and who didn’t. Filecoin built working decentralised storage; Internet Computer put live infrastructure in the market. Both still fell more than 99% from their 2021 peaks.²⁰ The price tracked the asset class, not the delivery — which is exactly the dynamic we’re describing. It’s the same story as the majority of projects in this space, and if we’re going to ask you to pay attention to what we’ve built, we have to stand inside that economic reality without flinching.
One critical difference is that we’re still here, and we’re still building. The code is in the repositories, the protocol is stable, and the infrastructure runs. We haven’t delivered at the scale or on the timeline we hoped and promised. But we haven’t stopped. And that distinction is about to matter.
Since Holo issued its original Green Paper in 2017, the scale, complexity, and cost of the challenges we collectively face have all sharply accelerated.
What hasn’t changed is the techno-utopian view increasingly held by state, civil society, and corporate actors alike — always looking for the next silver bullet they hope will unravel the harms of the previous generation of techno-utopianism, only to further compound them.
And now we have the combined promise and threat of AI, where we’ve got one foot pressed firmly on the brake, while the other is pedal to the metal. The sovereign AI debate in 2026 has governments and enterprises racing to control their own stacks. But “sovereign AI” as currently defined means sovereignty at the level of the nation or the corporation. It replaces one centre with another. Few are seeking to answer the more difficult question of what sovereignty looks like at the level of the individual entity — the person, the community, the small business — that actually generates the data.
The security implications are worse than most of the industry acknowledges. Companies deploying AI agents across their operations are concentrating their organisational data and intelligence behind a single security perimeter. CyberArk estimates that machine identities already outnumber human employees by 82 to 1 in enterprise environments.²¹ One breach doesn’t get you one person’s email. It potentially cracks open the entire organisation.
This is not hypothetical. This is the architecture companies are building right now, as fast as they can.
Open-source models don’t fix this. Granted, they solve the access problem; anyone can run the model. What they don’t solve is the infrastructure problem. An unlocked model with total access to everything — running on a network never designed to defend against internal threats, in a world where autonomous agents now expand that internal attack surface — is a powerful AI holding the keys to your operational life, on infrastructure that may not survive the next five years of cryptographic reality.
Because regardless of what AI and security professionals may say about how “easy” it is to protect oneself, the quantum horizon makes all of it worse. “Harvest now, decrypt later” (HNDL) — collecting encrypted data today, and storing it until quantum computers can break the encryption — is already happening. The U.S. Federal Reserve published a paper in September 2025 analysing HNDL risks specifically for distributed ledger networks, concluding that data privacy of previously recorded transactions remains permanently vulnerable because an adversary can copy the ledger and wait.²² Blockchains put every transaction on one global ledger — one target. Centralised AI infrastructure puts an organisation’s operational intelligence behind encryption that has a shelf life. Corporate data doesn’t expire, and may become more valuable to an adversary five years from now than it is today.
Meanwhile, regulation is catching up with innovation. The EU AI Act’s transparency obligations took effect on 2 August 2026 — disclosure, content marking, provenance — with the high-risk regime to follow from December 2027; at the top end, non-compliance carries fines of up to €35 million or 7% of global turnover.²³ MiCA’s grandfathering expired across the EU on 1 July 2026; any crypto-asset service provider still serving EU clients without authorisation is now operating in breach.²⁴ The CLARITY Act is advancing in the United States. Thousands of AI ventures face a compliance reckoning because the infrastructure they built on was never designed to provide the accountability that a growing number of jurisdictions now demand.
And governments are making the same argument with procurement decisions. On 8 April 2026, France ordered every ministry to eliminate extra-European digital dependencies by autumn, including operating systems, collaborative tools, cloud, and AI platforms.²⁵ Germany is migrating government workstations²⁶. Austria’s military is dropping Microsoft Office²⁷. These are not fringe positions. They are governments concluding, publicly, that infrastructure dependency is a strategic vulnerability. The European Commission’s own officials have said as much.²⁸
The EU Digital Identity Wallet — rolling out across member states by the end of 2026 — is the clearest illustration. The idea is right: a portable digital credential you control. But civil society organisations including epicenter.works and the European Digital Rights network have rightly warned that the implementation could eliminate anonymity and enable pervasive tracking.²⁹ As a preview of just how disastrously wrong things can go, in April 2026, France’s centralised digital identity system was breached, with personal information for up to 19 million French citizens and residents offered for sale on the dark web.³⁰
The problem was never the idea, but the infrastructure — building it where one operator, or one breach, can see everything.
There’s a pattern here that nature already knows: concentration is what turns the beneficial toxic. Manure spread thin feeds a field; piled up in an industrial feedlot it poisons the watershed. Water that soaks in nourishes the soil; channelled into a torrent it strips it bare. Data is no different — the reflex to centralise everything doesn’t just create a single point of failure, it concentrates what was never meant to pool until the flows themselves turn destructive.
Step back from the individual headlines and the pattern is singular. The AI security exposure, the quantum shelf-life problem, the regulatory reckoning, the sovereignty scramble — these are not separate crises. They are the same crisis surfacing in different domains: too much concentrated in too few centres, on infrastructure that was never built to keep it safe. Every one of them points to the same answer — and it’s not a better centre. It’s no centre at all.
The commercial landscape has shifted as dramatically as the threat landscape. The hosting economics we proposed in 2017 have since been borne out by an entire decentralised-compute sector. LLMs barely registered as a commercial proposition when we issued the first Green Paper; Statista now puts the worldwide AI market at approximately $254 billion for 2025, with other credible estimates ranging considerably higher depending on what’s counted.³¹ Enterprise security has moved from a perimeter problem to a data-sovereignty problem — the firewall-bounded network has given way to data that moves through AI systems outside the enterprise’s control, and regulators are writing the rules to match.³²
So while our early supporters are understandably frustrated by our failure to launch on our initial proposed timeline, the delay has put us in front of a commercial opportunity significantly larger than the one we originally set out to address — and an infrastructure need the prevailing architecture can’t meet. We’re transcending and including the original hosting vision, because what this technology can do turns out to be considerably more than what we first imagined.
The shift the last section described is not only technological. Beneath the change in what we can build sits a change in what we can no longer afford to ignore: money itself. You can decentralise the infrastructure and resist the capture and still be left with an economy that measures the wrong things — because the instrument at its centre was only ever built to measure one.
Bernard Lietaer, who helped design the mechanisms behind the Euro, spent the rest of his career arguing that we misunderstand what money is.³³ A currency, he held, is at its root an agreement to use something as a means of exchange — but its real work is to connect unmet needs with unused resources. The word carries the idea: currency shares a root with current — from the Latin currere, to run. Money is not, first of all, a store of value; it is a way of seeing and directing flow. It makes certain movements visible and rewards them, and so encodes what a system treats as worth having. Ours makes extractive financial return exquisitely visible and almost nothing else — so that is what the whole economy veers toward.
Many have named what that leaves out. E.F. Schumacher argued half a century ago that an economics blind to everything but price would consume the natural and social capital it depended on³⁴; Elinor Ostrom’s Nobel-winning work showed communities governing shared resources well with instruments the market model says shouldn’t work³⁵; Arthur Brock, one of Holochain’s co-creators, maps wealth as a spectrum of increasing dimensions of aliveness — from scarce tradable things at the base, up through the measurable health of a system, its performance, its relationships, and finally its capacity to generate new forms.³⁶ The through-line is consistent: living systems draw their health from levels our money cannot see.
An economy that can only value what is scarce will keep destroying what is abundant to manufacture more scarcity.
The instinctive response is to reach for a better single currency — a sounder money, a fairer coin. That instinct is the error, and Lietaer named why: the problem is monetary monoculture. A single national currency, optimised for large-scale trade, is extraordinarily good at what it does — and, like any monoculture, efficient but fragile, and unable to carry the kinds of value it was never designed for. Healthy ecosystems are plural; a resilient economy would be too. Not one currency, but many — national money for the trade and capital it already serves well, and alongside it the currencies it has no way to express: for local exchange, for regenerative work, for the trust between communities.
This is neither new nor untested. Mutual Credit — units issued at the moment two parties transact, one balance rising as another falls, net supply always zero, nothing mined and no scarcity manufactured — has run in the real economy for nearly a century. Switzerland’s WIR Bank has operated a mutual-credit currency among tens of thousands of small businesses since 1934, alongside the franc, not instead of it; researchers have found it tends to move counter-cyclically, steadying its members when bank credit contracts.³⁷ Ralph Borsodi and Robert Swann issued the Constant in 1973, a unit anchored to a basket of real goods that held its value while the dollar inflated³⁸; that lineage runs through the Schumacher Center to BerkShares, still circulating in Massachusetts today.³⁹ None replaced national money. Each did something national money couldn’t.
The goal was never to replace the dollar. It was to end the monoculture — to build the currencies for the value money was never designed to see.
Two harder problems kept this from scaling, and both now have answers. The first is how plural currencies hold together without collapsing back into a single reserve asset that drains value toward whoever issues it — the problem Keynes tried to solve at Bretton Woods with the bancor, a shared unit for clearing between nations⁴⁰, and the one WIR has quietly solved for its members for ninety years: multilateral mutual-credit clearing, with nothing hoardable in the middle. The second is what anchors value if not scarcity — the question Borsodi answered with a commodity basket and that today’s regenerative-finance efforts are answering with measures of real ecological value: a shared denominator that is not itself a currency. The monetary theory has never been the missing piece. The missing piece is the infrastructure to issue, verify, and clear such currencies across many communities at once — which is exactly where nearly every attempt, for fifty years, has stalled.
This is the problem Holochain was built to solve. Before it was infrastructure for anything else, it was infrastructure for currencies: a substrate for issuing plural, real-value-backed currencies and clearing between them, with no central mint, at the scale the theory always needed and never had. Everything that follows — the accounting engine, the mutual-credit currency, the marketplace — is that substrate taking shape. And it reaches maturity as serious economics reaches for the same thing again: regenerative and bioregional finance, complementary currencies, the search for money that measures real value rather than only extracting it. The ideas are old and well-founded. What is new is that the coordination layer to run them now exists.
We’re transcending and including the original hosting vision, because what this technology can do turns out to be considerably more than what we first imagined. The 2017 Green Paper described a hosting marketplace — a place where people who provided hosting and compute (either with their own hardware, or with HoloPorts purchased in our crowdfunding campaign), would earn HoloFuel. Publishers, those who needed hosting and/or compute, would spend HoloFuel. HoloFuel would be stable, asset-backed, useful — a utility currency for a specific economic context.
That vision was sound. The opportunity was real, and the architectural approach was right. It also barely scraped the surface of what was both possible — and now more necessary than ever. The 2017 paper half-saw this itself, noting that if HoloFuel were ever used as a more general-purpose currency, hosting would become a smaller part of the economy, making room for asset-backed currencies for energy, food, and transport. What was an aside then is the argument now.
The accounting infrastructure built to make distributed hosting work — mutual credit, countersigned transactions, proof-of-service validation, smart agreements — is not specific to hosting. It’s general-purpose economic coordination infrastructure. What else can you do with an accounting engine that can run on any device, denominate any unit of value, settle without a trusted intermediary, and costs a fraction of a cent per transaction?
The answer is: most of the things the current financial infrastructure does badly or doesn’t do at all.
Consider what launching a project-specific token actually requires in 2026: legal analysis in every jurisdiction. Exchange negotiations. Liquidity provision and ongoing market-making. Tokenomics management. Governance of the token itself, separate from governance of whatever you were actually trying to build. A community of speculators whose investment thesis may actively conflict with the community of people using the service. And the background noise of people measuring your project’s health through token price, regardless of whether token price reflects anything real. That cumulative overhead has effectively crushed new economy currency products and projects worldwide.
This overhead isn’t incidental. The trading pair model enforces token issuance as the price of participation. This model selected for tokenising everything, whether or not it served the project.
Almost every project that launched a token became a project that served the token.
Mutual credit currency offers a different path. Rather than each project issuing its own token and carrying the overhead that comes with it, projects can denominate their activity in a shared currency whose supply tracks real economic activity rather than speculative demand. HoloFuel — the mutual credit currency we described in the original Green Paper — is the working instance of that model. A project building on Holochain can adopt HoloFuel as its native currency — pricing its services and letting its community transact in it — without issuing a token of its own. The project participates in a token economy without having to set up and run one.
This matters most for the projects the current system has failed most thoroughly — regenerative finance initiatives trying to denominate ecological value without speculating on it; community currency projects wanting local exchange with global interoperability; mutual aid networks needing accounting without extraction; cooperative enterprises that want to share value with participants rather than speculators; and impact organisations where capital should flow toward outcomes rather than toward whoever generates the most trading volume.
Many of these projects have been built without tokens precisely because the token infrastructure was built for speculation — and since they don’t want that erosive force driving their community, ecosystem, or project, they’ve been largely undercapitalised as a result.
And then there’s AI — the use case the 2017 Green Paper design didn’t anticipate, but our architecture is already built to serve.
One breach of a centralised AI stack reaches everything behind it. That is the property this architecture removes. Holochain has no single ledger where every application’s data accumulates — each application runs its own network with its own membrane, and the contents of private entries never leave the agent’s own device. What an application publishes is visible to that network’s peers, which is what makes validation possible. But there is no layer above it where everything aggregates, and no central infrastructure to breach.
Each entity — individual, organisation, or community — can run AI agents on its own Holochain application DNA, with its own validation rules and data boundaries, its own security posture calibrated to what it’s actually protecting. You don’t build Fort Knox for a village post office. You don’t need the same security economy for internal comms AI as you do for financial operations AI. Agent-centric architecture means each function sets its own posture, because there’s no single network imposing a single model on everyone.
The practical stack mapping that we explore in the following section is straightforward. Holochain agents can run local AI models — trained on or tuned to the entity’s own data, communicating with other agents peer-to-peer through the same cryptographic mutual accountability that governs every other Holochain interaction. Our accounting software can handle the microtransaction economics when relevant: agent-to-agent service payments at fractions of a cent, that conventional payment rails can’t cost-effectively process. A clearinghouse layer between sovereign networks can settle cross-network exchanges without requiring any single network to surrender its autonomy. HoloFuel can denominate value. This is the work Holo Ventures — the commercial entity we’ve restructured to deliver it — is now pursuing.
An identity wallet on agent-centric architecture — where the agent holds its own credentials and discloses only what each interaction requires — delivers the portability and selective disclosure the EU’s digital identity wallet promises citizens, without creating the thing that makes the centralised version dangerous: a single issuer positioned to observe every use, breachable in one incident, and compellable to hand over what it holds.
Agent-centric architecture produces data provenance as a natural byproduct of how it works. Every action signed to a source chain, every interaction countersigned, data lineage cryptographically verifiable without retrofitting anything. The regulatory requirements described in Section 2 aren’t problems for this infrastructure. They’re descriptions of what it already does.
This is not a position paper on AI. We have no interest in telling anyone what models to run. What it points to is the infrastructure layer that makes any AI deployment structurally resistant to the things that keep security professionals awake at night. What the regulatory environment is now confirming is that the architecture was right before the regulations existed.
We built infrastructure for the projects the prevailing system was never going to fund — and for a future it doesn’t seem to care to protect.
Holochain, and the collective of projects it has stewarded over the years, hasn’t built a single product, but together delivers a layered architecture — each layer enabling the next, each designed without the capture dynamics that corrupted the previous generation of decentralised technology. Holochain, Holo, and HoloFuel were the original layers. But the newest layers, Unyt and the Decent Market, extend the reach far beyond the initial vision. Understanding what we built requires understanding how the layers relate.
Holochain: The Foundation
Holochain is not a blockchain. It has fundamental architectural differences.
The comparison is worth making because blockchain is the reference most readers arrive with. But it isn’t like-for-like, and the mismatch runs in our favour: blockchains, as they are actually used, do one thing — order transactions on a shared ledger. Holochain is a framework for building group applications, of which currency is one case. What follows is drawn on blockchain’s terms, which understates the range rather than overstating it.
Blockchains are decentralised: they have many nodes and no central authority. But the architecture starts from a fundamentally centralising assumption: there is one truth, and every node must agree on it. The problem isn’t that agreement is required — every distributed system needs its peers to converge on what happened. It’s the scope: consensus about a single ordering of every event that has ever happened, reached before anything can proceed. This is why blockchain is fundamentally slow — the cost of a transaction grows with network size. It takes a decentralised system running on hundreds of thousands of processors and makes it behave as a single-threaded process, everybody marching in lock-step. There is no horizontal scaling: throwing more machines — or more powerful ones — at the network buys no additional throughput, because every node processes the same sequence regardless.
Global consensus itself is a central point of capture, as well as a fundamental vulnerability subject to attack. And this lock-step march is why blockchain, despite its libertarian framing, has reproduced the fee-extraction dynamics of the systems it was supposed to replace. Paying people to run infrastructure is not the problem — we intend to do exactly that. The difference is what the payment is for. Under proof-of-stake, the right to validate and the earnings that follow scale with capital already held; under proof-of-work, with hardware already bought. Both compound position rather than reward contribution, which is why they concentrate control over time⁴¹. Hosting pays for service delivered. The rich get richer under one and the useful get paid under the other.
The crypto industry uses “decentralised” the way the food industry uses “natural”: technically meaningful in some narrow sense, practically elastic, deployed primarily for marketing purposes.
Holochain is agent-centric: each agent maintains their own cryptographically secured chain of actions, validated by peers according to shared rules, published to a distributed hash table for global visibility — without requiring global consensus. There is no single ordering of events that everyone must agree on before anything can proceed. Peers on a network do converge on the same set of validated data — that convergence is what validation means — but each agent derives its own view from it, and agreement between parties is required only where a transaction actually involves them. Yet agent-centric is only half the story: Holochain starts with empowered agency but always exercises it inside social accountability — an action is valid only if it plays by the shared rules of its network, and peers validate that it acted properly. The balance that keeps social systems healthy is written directly into the architecture. Agents interact directly. They reach agreement only when agreement is necessary.
The analogy that actually works is with version control software. Subversion — one of the dominant systems before Git — assumes a central server holding the canonical truth; every developer works against it. Git is already agent-centric: each developer maintains their own repository, makes their own contributions, and chooses who to share with and what to accept. Coherence emerges from those relationships, not from a central authority imposing it. The analogy has a limit worth naming: a Git branch can be rewritten before you share it, and a source chain action cannot. Once written, it stands — which is what makes it accountable. Holochain applies the same principle to computing generally: every agent maintains their own state, and shared truth is constructed through interaction, not consensus. Most developers today use Git through GitHub — a centralised platform operating adjacent to a distributed tool — which illustrates the access-layer centralisation problem described above.
There is no centre to capture. Period.
One consequence trips up almost everyone: there is no such thing as ‘the Holochain network.’ Ethereum and Solana are each one global network with one ledger. Holochain is a pattern for spinning up as many independent networks as there are applications — each with its own rules, its own participants, and its own membrane. When something runs ‘on Holochain,’ it runs its own network built with Holochain; it does not join one big shared one.
Characteristic
Blockchain
Holochain
State model
Global consensus required for all transactions
Local state per agent; global visibility via DHT
Scalability
O(n²) — every transaction is validated by every node, often more than once per round
O(log n) — sharding means each node validates only a subset of actions, not all of them
Throughput
7–15 TPS (Bitcoin/Ethereum L1)
Limited only by aggregate peer capacity
Data sovereignty
Signed by key holder, but replicated globally; agency depends on validator inclusion
Signed and authored by the agent; private entries never leave the device; published entries are permanent and network visible; portable by licence (CAL)⁴³
Security model
Majority consensus — a dishonest majority makes corrupted state authoritative for everyone, globally and permanently
Local validation — an agent’s chain cannot be forged without their key, and any honest peer who sees invalid data can warrant its author. A colluding group can still eclipse an agent from honest peers; an agent who then builds on invalid data is warranted for it, and that isn’t currently reversible. The exposure is targeted rather than global.
Hardware requirement
Specialised (mining rigs, validators)
Any general purpose computing device, including mobile
Attack surface
Global consensus layer — one successful attack corrupts the shared record
Individual agent chains and network membership — no global layer to capture, but agents can be isolated from honest peers
There is a deeper point about security that the comparison table doesn’t capture. In a single-network consensus system, security is a collective problem — the entire network must invest the energy to protect non-monetary actions and micro-transactions for a fraction of a cent as if they are multi-million dollar transactions regardless of who those assets belong to. In a plurality-of-networks architecture, security is proportional. Each network calibrates its validation requirements, staking economics, or warrant thresholds to the value of what it is securing. Tweets don’t need (and shouldn’t pay cost in compute overhead for) the same security posture as a multi-million dollar transaction. This is how security actually works in the physical world: as we said earlier, you don’t build Fort Knox to protect a village post office. Blockchain’s monolithic design forces a single security economy on every action. Holochain’s architecture does not.
It’s also worth being precise about what Holochain’s security model actually does. Against invalid data — fraudulent records, fabricated transactions — the mechanism is detection and isolation: the network releases warrants against invalid actions, and the immune system responds. Against history rewriting — the equivalent of a blockchain fork or double-spend — there is no global consensus to attack in the first place. These are two different problems with two different solutions, and conflating them with a single ‘51% resistance’ metric obscures both. An agent who fabricates tokens or “double-spends” them by forking their chain finds themselves blocked and unable to spend or use them.
None of this is theoretical. The protocol is stable, audited, and running real applications today — from small-group collaboration in Moss, currently in preview, to Unyt’s crypto-accounting engine now in test. What we have not yet seen at scale is the immune response under sustained adversarial pressure in production. The mechanism is built and tested⁴², and the field trials are commencing.
Holo: Distributed Hosting Infrastructure
The original Holo vision as described in the 2017 Green Paper was specific: a marketplace where ordinary people could earn by hosting Holochain applications on their own hardware, and where application publishers could access distributed compute without depending on AWS or Google Cloud. The architecture was sound. The timeline was catastrophically wrong. We said we would have it running in 2019. It’s 2026.
Three full codebase refactors, sixty people down to the essential few. Building genuinely novel distributed infrastructure is harder than the roadmaps suggest — and, as we argued earlier, holding a system decentralised against the constant pull back toward a centre is harder still. We didn’t always win that fight on time — and that’s a significant part of why this took as long as it did.
What exists now: distributed hosting infrastructure, operating. Edge nodes that keep an application’s data available when none of its users are online — work that enables hosting models the original 2017 design couldn’t anticipate, where a network persists independently of who happens to be connected. The HoloPort hardware programme put dedicated hosting nodes into the hands of early believers — and those devices remain part of the network. HoloPorts run the same node software as any commodity hardware. What’s evolved is the model: distributed hosting no longer requires dedicated devices, and the network is stronger for being open to any hardware that can run the software. The hosting vision is not abandoned. It’s operational, continuing to develop, and now situated within a broader stack that gives it substantially more to do.
The original economic argument — that commodity hardware aggregated across a distributed network can substantially undercut centralised cloud pricing — has since been validated by an entire sector. Decentralised compute networks now report general compute costs up to 85% lower than centralised cloud providers⁴⁴ with GPU access running at roughly one-third the cost of equivalent AWS instances in live marketplace comparisons⁴⁵. The structural reason is unchanged from 2017: centralised cloud providers price for data centre overhead, redundancy guarantees, vendor margin, and the premium that comes with operating at monopoly scale. Thirty percent of global cloud spending — approximately $217 billion annually — is estimated to be waste: idle resources, over-provisioning, and timing inefficiencies⁴⁶ that a distributed model with no idle infrastructure simply doesn’t carry. Commodity hardware has none of that overhead — and it carries a further structural advantage: distributed hosting eliminates the cooling load that accounts for roughly half the electricity cost of a conventional data centre. In a home hosting model, that energy is simply waste heat absorbed into a living space rather than vented into the atmosphere. The argument was right. The delivery was late.
What’s evolved since 2017 is the hosting model itself, and the evolution solves a problem the original design left open. Holo will support two distinct hosting types. In micro-hosting, individual hosts set their own prices and make their nodes available for work negotiating directly with app providers. For this type of hosting, what Holo provides is the proof-of-service infrastructure for invoicing work in HoloFuel. This works well for smaller applications and developers who want flexible, cost-competitive infrastructure without platform lock-in. In macro-hosting, the customer sets their requirements and price range, and Holo acts as the single commercial counterparty for enterprise clients — aggregating capacity from across the host network and presenting it as a unified, contracted service — and the host bids for inclusion on the hosting project. A large deployment cannot practically negotiate with thousands of independent node operators at different prices; the macro tier solves that without centralising the infrastructure that delivers it. The hosting is still distributed. The commercial relationship is not.
These aren’t competing models. They serve different groups at different scales, and together they mean Holo can play credibly at both ends of the market — the developer building on cost-competitive commodity infrastructure, and the enterprise client who needs predictable pricing and a single point of accountability.
Unyt: The Accounting Engine
Unyt is the economic engine of the stack. It is the software that makes HoloFuel work — and that makes any mutual credit currency work on Holochain. Unyt is programmable economic infrastructure: an accounting engine that can denominate any economic relationship in any unit, weave value flows between parties according to rules they set, and settle with cryptographic integrity without requiring a blockchain or a trusted intermediary.
Unyt’s Smart Agreements are akin to smart contracts on a blockchain, but significantly more flexible. They are programmable economic logic that executes locally — between the parties to an agreement — rather than requiring global network consensus. And more powerful still: the agreements can be updated as a network’s needs evolve, without rebuilding anything.⁴⁸ In other words, Unyt works the way people work.
When I give my neighbour’s child $2 for lemonade, the only two people who need to know about, and validate, the transaction are me, and her. What we don’t need, and what would be an absurd waste, would be to have everyone in the entire neighbourhood see, review, and validate our transaction. Unlike traditional blockchain smart contracts — where upgradeability requires migration to a new contract address, breaking continuity — Smart Agreements are soft-coded: the rules can be updated by agreement of the parties in place, without migration. A single Unyt deployment can handle multiple currencies, and multiple accounting tokens per transaction simultaneously — flowing in both directions — alongside internal accounting units, loyalty points, reputation scores, and any other quantified value, all with the same cryptographic guarantees. One instance supports up to 65,000 units of account.
The Unyt crypto-accounting engine launched in September 2025. P2P payments, community currency tooling, and Smart Agreements shipped in Q4 2025. Bi-directional payments, blockchain bridging to EVM Layer 2 networks, and market oracles shipped in Q1 2026⁴⁷ — the infrastructure the HOT to HoloFuel migration will run on.
HoloFuel: The Utility Currency
HoloFuel is a mutual credit currency designed as a medium of exchange, not a speculative asset. Mutual credit currencies have been in use for thousands of years, but are quite different from the fiat currencies that currently surround us. In a mutual credit currency the net supply is always zero — every positive balance is exactly matched by a corresponding negative balance. No currency is created from nothing; supply expands when credit is extended and contracts when it is repaid. The currency breathes in response to real economic activity.
HoloFuel is currently backed by computing capacity: the real-world cost of CPU cycles, bandwidth, and storage in commodity hardware markets. Not a peg to an external currency. Not an algorithmic stablecoin whose backing evaporates under stress. A grounding in real resource costs with a self-correcting feedback mechanism: if HoloFuel trades above compute cost, hosting becomes more profitable, more hosts join, supply increases, price corrects. If it trades below cost, hosts exit, supply tightens, price corrects. The commodity hardware market provides a natural floor and ceiling that no amount of speculative attack can permanently circumvent.
Most cryptocurrencies fail the most basic test of a functioning currency: stability. Bitcoin’s fixed supply was supposed to solve the trust problem inherent in fiat’s discretionary monetary policy. It solved supply predictability. It did nothing for price predictability — and a currency whose value can move 20% in an hour is unusable for any ordinary economic purpose. HoloFuel’s value is anchored to productive capacity, not scarcity. That anchoring is what makes it functional as a medium of exchange rather than a vehicle for speculation.
Credit limits are algorithmic and role-specific. Hosts earn credit capacity through demonstrated service delivery. End users require positive balances to spend. Publishers need collateral or revenue history. Holo earns through transaction fees capped at ≤1% of value. Reserve Accounts are 1:1 asset-backed. Productive activity generates credit. Gaming the system contracts it.
What’s changed since 2017 is the scope. HoloFuel was designed as a currency for distributed hosting. Given the added layer of Unyt, it’s now a more inclusive utility currency that any values-aligned project can adopt instead of launching their own speculative token. This matters because the overhead of token issuance is enormous: legal analysis in every jurisdiction, exchange negotiations, market-making, tokenomics management, governance of the token separate from governance of the project, and a speculator community whose interests may actively conflict with the project’s actual users. A project building on Holochain can use HoloFuel as its native currency without any of that.
The token infrastructure serves the project. The project doesn’t serve the token infrastructure.
Decent: The Marketplace
A marketplace is where people meet. A market is where they do business. The distinction is old and it runs in one direction: the square came before the exchange, the fair before the trading floor. People gathered, discovered who else was there and what they were doing, and trade followed from that.
So that’s the order we’re building in; Decent starts as a marketplace.
What opens with the migration. Decent launches alongside the migration, as a directory of organisations building and operating decent infrastructure — a place where people working on similar problems can find each other, be visible to those seeking them, and connect with organisations doing adjacent work. Listing is open to any organisation whose work is aligned with what this paper describes. It’s self-serve: you supply your details, and each listing is verified by a person before it goes live, so the directory can’t be turned against the people relying on it.
What a listing gets you is visibility, connection to organisations working in the same direction, and — where there’s a fit — the beginning of a conversation with Holo Ventures about doing something together. Some of the organisations listed will want to accept HoloFuel for what they offer. That’s available, and the route to it is to get in touch; but it isn’t the price of entry.
Where it’s headed. With enough organisations wanting to transact with each other, a marketplace becomes a market — and at that point the architecture has a problem to solve that’s worth describing now, because it’s the reason the Unyt accounting engine was built the way it was.
Every cryptocurrency exchange operates on trading pairs. To exchange Token A for Token B, you need a market with liquidity on both sides. This seems manageable for two tokens. The mathematics become structurally untenable at scale. To support exchange between n tokens, a trading pair system requires n(n−1)/2 markets — 45 markets for 10 tokens, 4,950 for 100, 499,500 for 1,000. Each market requires independent liquidity — capital locked on both sides, held by market makers extracting fees for providing it. Capital that could be doing useful economic work sits idle in thousands of pools. Tokens without access to major exchanges or venture-backed market-making cannot participate at meaningful depth. The “permissionless” exchange ecosystem has a permissions layer enforced by capital requirements instead of approval processes.
The trading pair model is barter economics with better infrastructure. Barter has a double-coincidence-of-wants problem — to trade, both parties must simultaneously want what the other has. Trading pairs inherit it precisely: you can only swap Token A for Token B if someone is on the other side of that specific pair, right now, with liquidity to match.
The Decent Market applies a solution that commodity exchanges have used for centuries: a clearinghouse with a common unit of account. You do not trade TFT directly for HOT. You price both via oracles and convert atomically through a common accounting layer. Ten tokens: ten oracle feeds, not forty-five bilateral markets. One thousand tokens: one thousand oracle feeds, not half a million trading pairs. The cost to add a new currency to a clearinghouse is O(1). The cost to add a new currency to a trading pair network is O(n).
Trading Pairs
Clearinghouse
Markets required for n tokens
n(n−1)/2
n
10 tokens
45 pairs
10 oracle feeds
100 tokens
4,950 pairs
100 oracle feeds
1,000 tokens
499,500 pairs
1,000 oracle feeds
New token entry cost
Liquidity across all existing pairs
One oracle feed
Liquidity behaviour
Fragments across pairs as network grows
Concentrates at hub; depth compounds
Multi-token trades
Multiple hops, compounding fees
Single atomic conversion
In that architecture, partner networks — DePIN projects, community currencies, regenerative finance instruments, impact initiatives — operate as sovereign nodes, running their own Unyt accounting instances with their own tokens and governance. Decent provides settlement services between them. Each node maintains full autonomy over its internal operations. The relationship is peer-to-peer: nodes connect to the clearinghouse when they need to exchange with other networks, and operate independently when they don’t. Unyt earns transaction fees on cross-network settlements — not on internal node operations. There’s no extraction point at the centre because the design doesn’t concentrate value at the centre.
While we’re in active design, only some of the components are already built. It’s where this goes if enough organisations find each other in the marketplace and start doing business — and if they don’t, it stays a description of an architecture rather than an account of one. We’d rather set out the direction and be judged on whether we get there than describe the destination as though we’d arrived.
The Weave and Moss: The Experience Layer
Everything described so far sits beneath the surface — the protocol, the hosting, the accounting engine, the currency, the marketplace. None of it is what a person actually opens and uses. That top layer, the one people touch, is where decentralised technology has most often come undone: the usable products were either single-purpose silos, or — as with the crypto interfaces in Section 1 — quietly re-centralised at the access layer even while the protocol beneath them stayed open. What has rarely been built is an experience layer that is itself agent-centric: general-purpose, own-your-data, groups assembling their own interoperable tools with no platform in the middle. The Weave and Moss are that layer — early, but real.
The Weave is an open standard for social context. The web got three open standards that let anyone publish and link documents — HTML, HTTP, and the URL — but it never got one that let groups own and manage their shared context, which is exactly why that function was captured by the platforms: Facebook holds your group, GitHub holds your code, Google holds your collaboration. The Weave adds the missing standard — the Weave Interaction Pattern — for creating, searching, linking, and organising units of social context, so that independent groupware “Tools,” each a small Holochain application, interoperate, embed inside one another, and can be searched across every group you belong to, with no central server anywhere. Making an existing Holochain app Weave-compatible is a task of minutes, not months.
Moss is not an application; it is a runtime. If The Weave is the standard, Moss is the browser — what Mosaic and Netscape were to the early web: the thing through which ordinary people first actually use what lies beneath. Rather than doing one thing, Moss runs a whole suite of independent Holochain applications side by side, in a single coherent space that a group owns and controls — more than twenty of them in an open, free library, and growing. Among them:
Vines for chat
Presence for peer-to-peer video
Notebooks for real-time collaborative documents
Talking Stickies for brainstorming
KanDo for kanban boards
Emergence for events
Acorn for mapping projects and outcomes
Converge for group decision-making, and
Unyt (as above) for tracking value.
Each group — and each Tool within it — is its own private peer-to-peer network; an asset created in one Tool embeds directly into another, and everything is searchable across the groups you belong to. Standing up a group’s tooling is closer to opening a web page than to procuring software.
Two features of this bear on the paper’s argument. First, it shows the move only agent-centric architecture permits: you can compose centralisation out of decentralisation — lock down permissions, predefine a group, run it privately — but you can never run that in reverse and recover decentralisation from something that was centralised to begin with. You can aggregate; you cannot disaggregate. Holochain lets a builder dial control to exactly what a use case needs while the foundation beneath stays sovereign. Second, The Weave and Moss are built by Lightningrod Labs — the studio of Holochain co-founder Eric Harris-Braun — and released as open source. The experience layer of this stack is not ours. That is the point: an ecosystem whose most important layer is built, and held, by more than one hand.
[This section details the mechanics of the accounting and transaction infrastructure described in previous sections. Non-technical readers may proceed to Section 7.]
Countersigned Transactions
In a blockchain system, transaction validity depends on network consensus. In a Unyt transaction, validity is established through a signed handshake between the specific parties. Rather than co-signing a single shared action (as described in the original HoloFuel paper), each party validates and signs their counterparty’s action when committing to their own chain — with the result published to the distributed hash table for universal verifiability.
Completing a transaction takes only the two agents involved — but they remain accountable to the group: each validates both sides, so neither can commit an invalid action without being warranted and blocked by the peers who check it.
In the simplest sequence the first agent pays some units (committing a payment to their chain), then their counterparty accepts them (committing an acceptance to their chain which references the payment to validate). The most complex handshake sequence is for two-way trades, which run through four signed steps — propose, send, accept, and receipt — to get all the unit balances updated and sent in both directions. Any of these can be completed by two parties connected only to each other, with no access to the wider network; the rest of the network validates their actions once either agent reconnects.
Attempted double-spends are detectable: the initiator’s chain history is published, and any validator can check for conflicting entries, and for forking of their chain. Sybil resistance is addressed through the relationship between agent keys and the validation rules in each application’s DNA, so you cannot join a network you haven’t demonstrated the right to join. And there is no 51% attack vector because there is no global consensus to capture. What can be attacked is an individual agent’s connection to honest peers: a colluding group can isolate an agent and feed it invalid data, and an agent that builds on it is warranted for doing so. That exposure is targeted rather than global — the shared record is never corrupted for everyone at once — but it is real, and it is not currently reversible.
Credit Limits by Role
Credit is what makes a mutual credit currency a currency. Without it, balances only describe what each agent has earned; there is no medium to circulate, no purchase made before payment received, no economy. Credit architecture is economic architecture.
This is structurally different from credit in conventional finance, where bank credit is the privilege of currency creation, gated by institutions, tied to collateral, with rents extracted as interest. Mutual credit operates on the opposite principle. The currency is created at the moment of transaction. Credit limits are tied to demonstrated productive capacity rather than wealth on hand. No interest accrues, because no one is profiting from issuance. Credit, here, is not like bearing debt. It is the operational permission to participate in exchange before having generated the balance to do so.
Role
Credit Limit
Rationale
End User
None (positive balance required to spend)
Consume services; cannot extend productive credit
Service Provider / Host
Algorithmically set; based on 3-month revenue history
Credit tied to demonstrated productive capacity; anti-gaming mechanisms prevent artificial inflation
Infrastructure Provider (Holo)
Limited to ≤1% of transaction fees earned
Covers operational costs without extractive rent-seeking
Reserve Account
1:1 asset-backed; expands with deposits, contracts with withdrawals
Bridge to external currencies requires hard collateral
Proof of Service and Microtransaction Economics
One of the harder problems in distributed infrastructure: how do you pay for services costing fractions of a cent without the payment infrastructure costing more than the service? A Stripe transaction costs 2.9% plus $0.30. That fee structure makes a $0.00025 compute payment economically impossible. This fee floor is what blocks the machine economy — IoT, DePIN, automated service ecosystems requiring billions of micropayments — from existing at scale on conventional payment rails.
Holochain’s solution is cryptographically countersigned service logs and aggregated proof-of-service invoicing — the distinction from conventional batched settlement being that neither party can fabricate or dispute the logs unilaterally.
Service Flow: Service delivery → Countersigned log (both parties) → Accumulated to billing threshold → Proof-of-Service invoice → Fraud detection → Payment queue → Settlement
Each service log is cryptographically signed by both provider and recipient. Neither party can fabricate logs — the countersignatures make unilateral fraud detectable against the other party’s chain history. Logs accumulate until reaching the billing threshold, designed to keep billing costs below approximately 1% of transaction value. The Proof-of-Service invoice includes payor ID, amount, application ID, and the sequence IDs of the service logs it covers. Pre-authorisation tokens allow frequent small transactions to proceed without per-transaction validation overhead — a running tab settling at threshold. This enables genuine microtransaction economics for machine-to-machine payments where individual transactions may be worth thousandths of a cent.
This holds at the transaction volumes real applications generate. At extreme volume the constraint moves from fee economics to storage: the more data a network accumulates, the more capable the hardware needed to participate as a full node. Sharding defers that ceiling rather than removing it.
Fee Economics in Context
Platform
Typical Fee
Minimum economical transaction
Ethereum L1
$1–50+ (gas, variable)
$100+ to justify the gas
Ethereum L2 (Arbitrum, Base)
$0.01–0.50
$5–10
Solana
\\\~$0.00025
\\\~$0.01
Traditional payment rails (Stripe)
2.9% + $0.30
$10+ to keep fees below 5%
Unyt / HoloFuel
Configurable; sub-cent capable
Fractions of a cent. True microtransaction economics are viable
Before making this argument, we have to name the irony. We raised capital in 2018 by issuing a token that was listed on exchanges. The financial systems of the time gave projects like ours no clean alternative — if you needed capital to build, you issued a token, and tokens attract speculators no matter what the documentation said the token was for. What we were proposing — a mutual-credit currency rather than a tradable token — wasn’t something mainstream investors understood, notwithstanding that raising capital from traditional sources ran against the goals of almost every early-stage Web3 project of that era. We’re making the case against speculative tokenisation as people who’ve gained and lost as a consequence of it, not as people who avoided it — and, candidly, as people whose ongoing work that very speculation has helped fund. We saw it coming, and said so: the original Green Paper told participants, in plain terms, not to acquire the token “for speculative or investment purposes with the expectation of making an investment return.” They did anyway. The market does what the market does.
That’s not hypocrisy. It’s the point: the system selects for this outcome regardless of disclosure or intent. Which is precisely why the infrastructure we’ve built matters.
Most cryptocurrency tokens exist in a speculative demand cycle: people buy them because they expect others to buy them at higher prices later. Demand increases when price increases, collapses when price falls, and is structurally disconnected from whether the underlying project delivers anything useful. The Economist’s observation that crypto is “a speculative asset class with no fundamental value or income-generating potential” accurately describes this dynamic (see note 8). It’s not a criticism of decentralised systems as a concept. It’s a description of how most decentralised systems are built.
HoloFuel is different in kind because its demand derives from use, not speculation. Hosts earn HoloFuel providing compute and spend it acquiring services. Publishers buy hosting capacity. Projects that adopt HoloFuel as their native currency create communities that transact in it for real goods and services. This is demand rooted in economic activity that exists regardless of what speculators think HoloFuel is worth. The credit mechanism itself does not produce appreciation: when demand rises, credit limits expand and supply meets it, rather than units becoming scarcer. That is deliberate — stable value is what makes economic planning possible. This describes how the currency works internally. It is not a claim that no rate will ever exist between HoloFuel and anything else. HoloFuel is self-custodial, and what people agree between themselves is their own business.
That is what utility demand looks like. It’s the demand structure of useful infrastructure — and it’s what the speculative token model, almost without exception, was not.
HoloFuel is also a poor vehicle for speculation, by design. The mutual credit structure — net supply always zero, credit limits tied to productive output, value anchored to commodity compute costs — resists the price dynamics that make speculative tokens attractive to traders. The stability is the feature. Predictable value makes economic planning possible. Economic planning makes real businesses viable. Real businesses create durable demand.
Conventional token metrics — fixed supply, market capitalisation, circulating supply ratios — are artefacts of the speculative model. They don’t apply to a mutual credit currency whose supply expands and contracts with real economic activity and whose value was never fixed to a particular asset. Evaluating HoloFuel by those metrics is like valuing a current account by the logic of a commodity futures contract.
The network economics compound. Each additional token joining the Decent Market — should we move to launch — costs O(1) to add one oracle feed. But for everyone already in it, a new token adds one more exchange destination — so each arrival makes the network more valuable to all the rest as it grows. And liquidity concentrates rather than fragments: in a trading pair model, each new token draws liquidity away from existing pools; in a clearinghouse, all liquidity is available to all participants through the common settlement layer.
Every token market gets thinner as it scales. Ours gets deeper.
We don’t sell HoloFuel. You earn it by providing value to the network, or spend it to receive value back — the way any mutual credit currency works. For those who haven’t earned it, there is one route in, and only one: migrating HOT.
HOT — the ERC-20 token sold in 2018 — was designed as a pre-sale of hosting credits. This was explicit in the original documentation: HOT was not a speculative investment in equity or revenue. It was described there as a forward purchase of HoloFuel — units to be used for network services — issued in exchange for capital to build the infrastructure.
The transition from HOT to HoloFuel is what that design always pointed to. It’s not an exchange between different assets. It’s the same unit of account arriving in the form it was always meant to take. The pattern is the same as Polygon’s migration from MATIC to POL, Fantom’s migration from FTM to Sonic, Ethereum’s transition from proof-of-work to proof-of-stake — technical upgrades delivering on original design intent, not pivots to something unrelated.
The mechanics run in two steps, and only the second is the migration in the sense people mean.
Bridging comes first. HOT is sent to a smart contract on Ethereum where it is locked, and an equivalent amount of mirrored HOT is issued to the holder’s Unyt address. What the holder has after this step is still HOT — the same pre-sale claim, on a different network. Nothing has been discharged, the supply of HOT is unchanged, and the step is reversible: mirrored HOT can be bridged back to Ethereum at any time.
Migration is the second step, and a separate decision. Mirrored HOT is exchanged for HoloFuel at one to one. This is where the unit changes, and where the pre-sale claim is delivered.
An inbound rate does not establish an outbound rate. Exchanging HOT for HoloFuel does not create an entitlement to exchange it back at the same rate, or at all.
The architecture is non-custodial throughout — no single entity holds user assets at any point in the sequence. A test migration runs first, with test currencies, to validate the infrastructure under real conditions before real value moves. Migration is voluntary, and its full terms will be published separately. Nothing in this paper opens the migration or allows anyone to act on it. Eligibility, terms, timing and where it will be available are set out in that documentation, not here.
The 2017 Green Paper said HOT would be destroyed on exchange. It is locked instead, and the reason is liquidity. HOT has established trading pairs built up across many exchanges over years. Recreating that for a new unit would cost millions and take years, and holders would carry that cost. Keeping HOT as the liquidity path serves holders better than destroying it — and that requires locking rather than burning. This is also what the largest comparable migration did: Polygon’s MATIC is not burned when it converts to POL but held in the migration contract, which retains a governance-controlled “unmigration” function, and unmigrated MATIC still exists on Ethereum with no deadline to convert.⁵² Around 99% of supply migrated within the first year regardless.⁵³ Locked HOT can only be released by the smart contract, on multi-signature proof that mirrored units have been returned to the bridge agent on the Unyt side.
HOT holders have choices, and we’re not in the business of prescribing them. Those who want to participate in what the Decent Marketplace enables — to hold and use HoloFuel as a medium of exchange for real services — have a clear path. Those who don’t can make whatever decision suits them. We’re describing what exists and how to access it. We’re not offering financial advice and we’re not making promises about price movements in either direction.
What we are saying is this: HOT was described from the outset as a pre-purchase of utility. The infrastructure is now running. The bridge is built. It can now open.
Everything described so far is infrastructure — and infrastructure has to be built, operated, and paid for by someone. The moment an organisation stands behind a network, it becomes the obvious place for capture to re-enter. A protocol with no capturable centre is worth little if the entity stewarding it can be bought, or quietly run for the benefit of its insiders. So we built the organisation the way we built the architecture: without a centre to capture.
Two entities carry the work forward. The Holochain Foundation is a non-profit steward. It holds the protocol’s intellectual property and exists to keep it open, neutral, and available to anyone — including those who will go on to compete with us. It has no shareholders and no owner to enrich; its only mandate is to keep the commons a commons. The protocol sits with the Foundation, because it belongs on no one’s balance sheet.
Holo Ventures is the ecosystem’s commercial entity. It builds and operates commercial software across the full Holochain stack — distributed hosting, peer-to-peer accounting, collaboration tooling, and cross-ecosystem settlement — alongside a professional-services arm that helps others design and ship on the same infrastructure. Every component is open source and auditable. There are no gas fees, no mandatory transaction fees, and no platform lock-in: Holo Ventures earns by building and delivering software and services that work, not by taxing the people who use them or the networks that run on it.
This is the argument of the paper turned on ourselves. Capture is not caused by bad actors; it is what happens when you build a centre and run capital through it. A mission held inside an ordinary company can be sold with the company; a business that lives on transaction fees or token appreciation is pulled, over time, back toward extraction. So Holo Ventures is a mission-protected commercial entity — backed by the Foundation and bound by constitutional safeguards designed to hold its mission integrity regardless of how large it grows. The protocol stays in the commons; the commercial entity is structurally constrained from becoming the thing this paper warns against. We are not asking you to trust that we will resist the pull toward a centre. We have tried to make it structurally impossible for us to drift there — the protocol is open and forkable whatever we do.
The obvious comparison is OpenAI — also founded as a non-profit around a mission, with a capped-profit structure meant to keep commercial upside in check. In 2023, its board’s authority was overridden within days of the one time it tried to use it. By 2025, the profit cap was gone and the fundraising limits lifted⁴⁹. The non-profit still holds nominal control and a large stake, and reasonable people argue that’s enough — but the point is simpler than who’s right: governance you can renegotiate is not the same as architecture you can’t. Caps get raised and charters rewritten when the cheque is large enough. An open protocol cannot be un-opened. The final guarantee is the one the original Green Paper named: you can run Holochain without us.
That is the role we intend to play. The Foundation keeps the protocol free; Holo Ventures makes the values-aligned economy commercially real — building the software, operating the infrastructure, and earning by delivering it rather than by owning the networks that run on it. The work ahead is to put this into the hands of the people the mainstream system was never built to serve — the regenerative finance projects, the community currency organisers, the mutual aid networks, the cooperative enterprises, the impact organisations, and the developers who want to run AI locally without feeding a surveillance stack.
Those are the builders. But the unit this architecture ultimately answers to is smaller and more universal than any of them: the individual. Agent-centric infrastructure pushes sovereignty all the way down to the person — which means its real constituency is anyone who would rather not watch money and influence concentrate, year after year, at their expense. Which is to say, most of us.
The work keeps moving, and we’ve learned not to turn that into a roadmap. So, as directions rather than commitments: we’re working to ensure Holochain applications run wherever people already are — in the browser with no installation, and on mobile and desktop as easily as any other app. The decentralisation has to reach the layer people actually touch; an alternative almost no one can practically use is not yet an alternative.
And we’re building the runtime that lets agents operate on the network. The same architecture that lets a person own their data lets an AI agent run on infrastructure its owner controls — and settle in mutual credit, not on the payment rails the incumbents are racing to own before the agent economy has even arrived.
We’re not in that race. We don’t need to own this economy — we just need it to exist, and to be impossible to capture once it does.
A token is a promise. Money has always been a promise — a belief in future value, a bet on continuity, a statement that the person or institution behind it will still exist and still honour what they said. Coins were promises stamped in metal. Paper was a promise backed by gold in vaults. Fiat is a promise backed by nothing but the authority that declares it — and that authority can always issue more. Crypto was supposed to change this — to replace the need for promises with the certainty of mathematics.
Trustless, they called it. The code was the contract. The contract was everything.
But what’s wrong with this is the philosophy hiding inside the word trustless: the idea that trust itself is the problem. That the goal is a world where you don’t need to trust anyone because the system makes trust unnecessary. That obligation, relationship, mutual accountability are frictions to be engineered away rather than the actual substance of what an economy is.
A trustless world is not a liberated world. It’s an atomised one.
James Madison, at the Virginia Ratifying Convention in 1788, said it plainly: “The circulation of confidence is better than the circulation of money.”⁵⁰ The crypto industry achieved consensus — mathematical certainty that you own what you own. It never achieved confidence. And an economy runs on confidence: the ability to plan, to commit, to build something that takes longer than the next price swing. A currency whose value moves 20% in an hour provides consensus without confidence, which turns out to be worth almost nothing to anyone trying to do real work.
Every interaction becomes a transaction between strangers with no prior relationship and no expected future one. No obligation running in both directions. No accountability that doesn’t reduce to the execution of a smart contract. That’s not the architecture of a better economy. That’s the architecture of the worst parts of the economy we already have, running faster and with better uptime.
A world without promises isn’t a world without risk. It’s a world without obligation.
The risk in making promises, however, is that sometimes you break them.
Holochain should have shipped years earlier than it did. The architecture that makes capturable centres structurally impossible existed in concept before 2017. The question it answered — do we actually need global consensus for the things people want to do? — was asked before the surveillance economy completed its consolidation, before three companies captured AI, before the authoritarian stack stopped pretending to be anything other than what it is. We were building the alternative while all of that was happening. And we were too slow.
Ask yourself what the world might look like if agent-centric infrastructure had been production-ready and in the hands of developers in 2019 as intended. Whether the surveillance economy could consolidate as completely if people had genuine alternatives to platforms that extracted from them. Whether AI training data could be extracted without attribution if agents owned their own information by default. Whether self-determination would remain a philosophical aspiration or become a technical fact. We can’t answer those questions. What we can say is that the alternative didn’t exist at scale when it could have mattered most.
We’re complicit in the delay, and we’re furious about its cost — not as observers but as people inside the work, who knew what was at stake and watched the window narrow while it remained unfinished. The community that’s been with us since 2018 has lived with that same gap, often with capital at stake. Their frustration is legitimate. Eight years between the promise and the delivery is a long time.
And we’re still here, the architecture continues evolving, and most importantly, the infrastructure now exists. Not as an intention, a whitepaper, or a roadmap. As working code.
But the world that code ships into is not the world we were living in, or could have imagined, in 2017.
When this project started, the argument was about data ownership and platform capture — important, but abstract enough that reasonable people could disagree about urgency. AI has collapsed the distance between extraction and control. The enclosure of cognition is not a metaphor any more. It is an engineering objective, funded at hundreds of billions of dollars annually, pursued by companies whose explicit strategy is to make their models the substrate through which all other economic and intellectual activity passes.
Web3 promised self-determination, sovereignty, agency, and equitable distribution of value. AI promised democratised intelligence and augmented human capability. Different rhetoric, same underlying argument: technology would distribute benefit broadly. Both described something people genuinely need.
But the execution is wrong, as it’s still building on the same infrastructure that reproduced the capture dynamics it was supposed to escape. And the regulatory environment — the EU AI Act, MiCA, the CLARITY Act, the European digital identity framework — is now forcing a question the industry has been avoiding: can you actually deliver what you promised?
For most of the industry, the answer is no. The infrastructure isn’t rooted in provenance. The architecture doesn’t protect the data. The tokens don’t represent utility. The sovereignty is merely rhetorical.
For agent-centric infrastructure, the answer is yes. Sovereignty at the level of the agent — not at the level of the nation or the corporation, where “sovereign AI” currently means your government or your employer controls the stack instead of a hyperscaler — but at the level of every entity that generates data and deploys intelligence. Each maintains its own AI operations, establishes its own data boundaries, and communicates through protocols that preserve autonomy rather than requiring its surrender.
Scale that and you enable ecosystems of sovereign AI agents — collective intelligence without centralised extraction. The wisdom of crowds without the crowd being farmed. Distributed cognition without a centre that captures it.
Collective intelligence leading to wisdom, when the world so desperately requires more of it.
The infrastructure we built is what makes that possible.
As we noted at the outset, Vitalik Buterin calls this category of technology “sanctuary tech” — tools that create digital islands of stability, and enable interdependence that cannot be weaponised. Others in the movement are beginning to use a parallel term from Cory Doctorow: the walkaway stack⁵¹ — infrastructure resilient enough that people can simply leave systems that extract from them, because they have somewhere else to go. We’ve been building this since 2017, before anyone had a name for it, because we asked the right question early, and have continued to follow it to its conclusion regardless of the cost.
Holochain does not make promises legible by adding a blockchain. It makes promises legible by making agreements auditable. The countersigned transaction is not a mechanism for avoiding trust. It’s a mechanism for making trust visible — both parties on record, mutual accountability built into the cryptography, the relationship made inspectable rather than the relationship made unnecessary. Smart Agreements are not code replacing discernment. They’re parties encoding their own agreements into executable form — the promise made transparent, updatable by the people who live under it, enforced by cryptographic proof rather than institutional power.
Not trustless. Trust-full. Built on promises you can verify.
The Holochain stack — protocol, hosting infrastructure, accounting engine, marketplace — was designed from the ground up to make capture structurally impossible. No global consensus means no consensus mechanism to control. No blockchain means no gas fee extraction. Agent-centric data means you own your information — including the data that trains any AI operating on your behalf. Mutual credit means currency supply tracks real activity. A clearinghouse means new participants can join without extracting permission from incumbents.
The capture dynamics that turned the internet into surveillance infrastructure and crypto into a casino are not produced by bad actors. They’re what happens when you build systems with capturable centres and then run capital through them. Build without capturable centres and different things become possible — not guaranteed, but structurally possible in ways they weren’t before.
This is the infrastructure the world needs and the tech billionaires will never build, because it is specifically designed to make them unnecessary.
We’re not asking you to believe in us. We’ve been burnt by belief — our own and others’. What we’re asking is simpler and harder: come and build. Not on faith. On architecture. The ReFi projects, the community currency organisers, the mutual aid networks, the cooperative enterprises, the impact organisations, the AI developers who want to run models locally without feeding a centralised surveillance stack — the infrastructure exists for you. Use it.
Decent tech for the people the system was never going to serve. That’s what we built. That’s what this is.
The bridge is made of people, holding together with a trust that can be verified rather than demanded. Nothing about this requires our permission any more — the protocol is stable, the tools exist, the bridge is built.
Let’s cross it. Together.
Preface
01
Statista, “Global 500 Market Capitalisation Data,” 2025. statista.com/statistics/802938 IMF, “World Economic Outlook,” 2024. imf.org/en/Publications/WEO.
02
Cory Doctorow, “The ‘Enshittification’ of TikTok,” Wired, January 2023. wired.com/story/tiktok-platforms-cory-doctorow
03
Marc Andreessen, “The Techno-Optimist Manifesto,” Andreessen Horowitz, 16 October 2023. a16z.com/the-techno-optimist-manifesto
04
Alexander C. Karp and Nicholas W. Zamiska, The Technological Republic (Crown, 2025).
05
“The Authoritarian Stack,” n.d. authoritarian-stack.info
06
Vitalik Buterin, post on X, 3 March 2026. x.com/VitalikButerin/status/2028913738057957433
Section 1 — The Revolution that Wasn’t
07
Loup Durand, Jaguar, English translation of Le Jaguar (1988).
08
“Why this is the coldest crypto winter yet,” The Economist, 10 February 2026. economist.com/finance-and-economics/2026/02/10/why-this-is-the-coldest-crypto-winter-yet
09
“Research: $20 Billion Raised Through ICOs Since 2017,” Cointelegraph, 10 October 2018. cointelegraph.com/news/research-20-billion-raised-through-icos-since-2017 Fabric Ventures and TokenData, “The State of the Token Market,” 2018. coindesk.com/markets/2018/04/19/63-billion-2018-ico-funding-has-passed-2017s-total
10
Satis Group, “Cryptoasset Market Coverage Initiation: Network Creation,” 2018. research.bloomberg.com/pub/res/d28giW28tf6G7T\_Wr77aU0gDgFQ
11
TokenData, “Crowdsale Tracking,” via Bitcoin.com, February 2018. news.bitcoin.com. Webisoft, “ICO Statistics,” 2025. webisoft.com/articles/ico-statistics
12
CoinGecko, “EOS Price and Historical Data,” accessed 2026. coingecko.com/en/coins/eos CoinMarketCap, “EOS,” accessed 2026. coinmarketcap.com/currencies/eos
13
U.S. Securities and Exchange Commission, “Telegram to Return $1.2 Billion to Investors and Pay $18.5 Million Penalty,” Press Release 2020-146, 26 June 2020. sec.gov/litigation/litreleases/2020/lr24820.htm
14
BlackRock, “iShares Bitcoin Trust ETF (IBIT),” accessed 2026. blackrock.com/us/financial-professionals/products/333011/ishares-bitcoin-trust
15
U.S. Congress, “GENIUS Act (S.1582), 119th Congress,” 2025. congress.gov/bill/119th-congress/senate-bill/1582/text
16
McKinsey & Company, “Asset Management 2025: The Great Convergence,” 2025. mckinsey.com/industries/financial-services/our-insights/asset-management-2025-the-great-convergence CoinGecko, “Global Cryptocurrency Market Cap,” accessed 2026. coingecko.com/en/global-charts
17
Bank of America, “Global Fund Manager Survey,” August 2025, reported by BeInCrypto. beincrypto.com/bank-of-america-survey-crypto
18
Charles Hoskinson, quoted in “Why this is the coldest crypto winter yet,” The Economist, 10 February 2026 (see note 8).
19
Uniswap Labs, “Unsupported Token List,” accessed 2026. support.uniswap.org/hc/en-us/articles/18783818770701-Unsupported-Token-List “Uniswap Labs Front-End Remains Blocked in Ukraine,” Blockworks, 2025. blockworks.co/news/uniswap-labs-blocked-ukraine
20
CoinGecko, “Filecoin Price and Historical Data,” accessed 2026. coingecko.com/en/coins/filecoin CoinGecko, “Internet Computer Price and Historical Data,” accessed 2026. coingecko.com/en/coins/internet-computer
Section 2 — The World has Changed
21
CyberArk, “2025 Identity Security Landscape,” 23 April 2025. cyberark.com/press/machine-identities-outnumber-humans-by-more-than-80-to-1 Palo Alto Networks, “2026 Cybersecurity Predictions,” 2025. paloaltonetworks.com/cybersecurity-perspectives/2026-cyber-predictions
22
Jillian Mascelli and Megan Rodden, “‘Harvest Now, Decrypt Later’: Examining Post-Quantum Cryptography and the Data Privacy Risks for Distributed Ledger Networks,” Federal Reserve FEDS Working Paper 2025-093, September 2025. federalreserve.gov/econres/feds/files/2025093pap.pdf
23
European Union, “Artificial Intelligence Act, Article 99 (Penalties),” 2024. artificialintelligenceact.eu/article/99 European Parliament and Council, Regulation (EU) 2026/1744 of 8 July 2026 (Digital Omnibus on AI), Official Journal L, 24 July 2026. eur-lex.europa.eu/eli/reg/2026/1744/oj/eng
24
European Securities and Markets Authority, “Statement on the end of transitional periods under MiCA,” April 2026. esma.europa.eu/sites/default/files/2026-04/ESMA75-113276571-1679\_Statement\_on\_the\_end\_of\_transitional\_periods\_under\_MiCA.pdf
25
“France Orders All Government Ministries to Ditch Windows for Linux in Digital Sovereignty Push,” The Next Web, April 2026. thenextweb.com/news/france-linux-windows-migration-digital-sovereignty “French State Ditches US Tech Over Data-Grab Fears,” Brussels Signal, April 2026. brusselssignal.eu/2026/04/french-state-ditches-us-tech-over-data-grab-fears
26
“German State Schleswig-Holstein Ditches Microsoft for Open Source,” It’s FOSS, 2025. itsfoss.com/news/german-state-ditch-microsoft
27
The Document Foundation, “Austria’s Military Switches from Microsoft Office to LibreOffice,” 30 September 2025. blog.documentfoundation.org/blog/2025/09/30/austrias-military-switches-from-microsoft-office-to-libreoffice
28
European Commission, “Commission Proposes Tech Sovereignty Package to Strengthen Europe’s Digital Autonomy and Resilience,” 3 June 2026. digital-strategy.ec.europa.eu/en/news/commission-proposes-tech-sovereignty-package-strengthen-europes-digital-autonomy-and-resilience
29
epicenter.works, “Five Problems the Commission Must Fix in the EU Wallet,” accessed 2026. epicenter.works/en/content/five-problems-the-commission-must-fix-in-the-eu-wallet European Digital Rights (EDRi), “The eID Wallet Still Doesn’t Deserve Your Full Trust,” accessed 2026. edri.org/our-work/the-eid-wallet-still-doesnt-deserve-your-full-trust
30
Cybernews, “ANTS Hack: 19 Million Records Exposed in French ID Agency Breach,” April 2026. cybernews.com/security/ants-hack-france-19-million-records-id-agency-breach “France Confirms Data Breach at Government Agency That Manages Citizens’ IDs,” TechCrunch, 22 April 2026. techcrunch.com/2026/04/22/france-confirms-data-breach-at-government-agency-that-manages-citizens-ids
31
Statista, “Artificial Intelligence — Worldwide Market Forecast,” 2025. statista.com/outlook/tmo/artificial-intelligence/worldwide
32
Confidencial, “2025 Data Security Review: Why the Perimeter Shifted,” December 2025. confidencial.io/post/2025-data-security-rethinking-perimeter-control
Section 3 — What Money Is For
33
Bernard Lietaer, The Future of Money (Random House, 2001).
34
E.F. Schumacher, Small Is Beautiful: A Study of Economics as if People Mattered (Blond & Briggs, 1973).
35
Elinor Ostrom, Governing the Commons: The Evolution of Institutions for Collective Action (Cambridge University Press, 1990).
36
Arthur Brock, “Metacurrency” and the “Levels of Wealth” / Deep Wealth model. artbrock.com/metacurrency
37
James Stodder, “Complementary Credit Networks and Macroeconomic Stability: Switzerland’s Wirtschaftsring,” Journal of Economic Behavior & Organization 72, no. 1 (2009): 79–95.
38
Ralph Borsodi and Robert Swann, Inflation and the Coming Keynesian Catastrophe: The Story of the Exeter Experiments with Constants (Schumacher Center for a New Economics).
39
Schumacher Center for a New Economics (founded 1980 by Robert Swann as the E.F. Schumacher Society) and BerkShares, the local currency circulating in the Berkshires, Massachusetts since 2006. centerforneweconomics.org; berkshares.org.
40
John Maynard Keynes, “Proposals for an International Clearing Union” (1942–44) en.wikipedia.org/wiki/International\_Clearing\_Union
Section 5 — The Stack
41
Art Brock, “Blockchain Blind Spots,” Medium, n.d. medium.com/@artbrock/blockchain-blind-spots-1904d490218d
42
Least Authority, “Audit of Holochain Lair Keystore,” January 2023. leastauthority.com/blog/audits/audit-of-holochain-lair-keystore Least Authority, “Audit of Holochain Deterministic Integrity,” January 2023. leastauthority.com/security-consulting/published-audits
43
Open Source Initiative, “Cryptographic Autonomy License v1.0 (CAL-1.0),” n.d. opensource.org/license/cal-1-0
44
Akash Network, “The Rise of Decentralized Compute,” n.d. akash.network/blog/the-rise-of-decentralized-compute
45
Coinposters, “Decentralized GPU Cost Arbitrage: AWS at $514 vs Akash at $84 Weekly,” n.d. coinposters.com/news/decentralized-gpu-cost-arbitrage-aws-at-514-vs-akash-at-84-weekly
46
ByteIota, “Cloud Pricing Wars 2025: AWS, Azure, GCP Cost Comparison,” 2025. byteiota.com/cloud-pricing-wars-2025-aws-azure-gcp-cost-comparison
47
Unyt, “Roadmap and Release Documentation,” accessed 2026. unyt.co/blog
48
Unyt, “Unveiling Unyt Smart Agreements,” 2025. unyt.co/blog/unveiling-unyt-smart-agreements
Section 9 — Structured Against Capture
49
OpenAI, “Evolving OpenAI’s Structure,” 28 October 2025. openai.com/index/evolving-our-structure “OpenAI’s Restructuring Deal Is Full of Holes, Critics Say,” CalMatters, October 2025. calmatters.org/economy/technology/2025/10/openai-restructuring-deal-full-of-holes-critics-say
Section 10 — Our Collective Promise
50
James Madison, “Speech on the Judicial Powers of the National Government,” Virginia Ratifying Convention, 20 June 1788. Founders Online, National Archives. founders.archives.gov/documents/Madison/01-11-02-0101
51
Cory Doctorow, Walkaway (Tor Books, 2017).
52
Polygon Developer Documentation, “POL,” accessed 2026 docs.polygon.technology/pos/concepts/tokens/pol
53
“MATIC to POL Migration Is 99% Complete: What You Need to Know,” Polygon Labs, 3 September 2025. polygon.technology/blog/matic-to-pol-migration-is-99-complete-everything-you-need-to-know
Technical Documentation and Resources
Holochain: holochain.org · Holo: holo.host · Unyt: unyt.co · Source code: github.com/holochain · github.com/unytco
The hard questions about HOT and trust
Isn’t this just a scam?
A scam takes the money and disappears. Eight years on, we’re still here, still shipping code in the open. On the selling: people who worked here were compensated in HOT. Some of them sold it. That’s what compensation is for. A developer who sold tokens owes nobody an explanation, and we’re not going to manufacture one. The suggestion underneath the question is that the people who built this should have worked for free, or held a falling asset out of loyalty. Neither is a real position. There is a serious version of the question, and it’s a different one: did anyone here trade on what they knew before you knew it, or quietly liquidate while telling holders to hold? No, on both counts. What we won’t do is publish individual wallets to prove it. People who worked here were paid in HOT, and their holdings are their own business in exactly the way a colleague’s bank balance is — compensation arriving as a token doesn’t put it in the public domain. And we have never told anyone to buy, hold, or sell HOT, because that call was never ours to make. The honest charge against us was never that we stole — it’s that we took far longer than we promised, and the delay had real costs for the people who believed early. We own that one. We won’t accept the other.
I invested in this. Why does the paper talk about utility instead of my return?
We’ll be straight, because it matters. HOT was never sold as an investment, and legally it isn’t one — no equity, no revenue share, no promised return, and the original documentation said in plain terms not to acquire it expecting one. We’re not going to hide behind that distinction, though, because it cuts the other way: you didn’t buy a stake that pays a dividend — you put money toward a utility we said we would build. We hold ourselves to that, and we take it more seriously than a dividend, not less. What you should expect from us isn’t a return; it’s the working infrastructure, and an honest account of where it stands.
HOT is down roughly 99% from its peak. Why should anyone take this paper seriously?
Because the chart is a story about an asset class, not about this project. Every major token of the 2017–18 cohort — EOS, Filecoin, Tezos, Bancor, Internet Computer — trades 80–99% below its all-time high, and that includes ones that shipped real infrastructure: Filecoin built decentralised storage, Internet Computer put live infrastructure in the market, and both fell just as far. The price tracked the speculative cycle, not delivery. We also never sold HOT as an investment — the 2018 Green Paper told participants, in plain terms, not to acquire it “for speculative or investment purposes with the expectation of making an investment return.” Judge this the way we’ve always asked: by whether the infrastructure works and gets used.
Did the team enrich itself, and who controls the token supply?
The split was disclosed from the start: 25% of tokens to the organisation and team, 75% to the community, with supply expanded during the offering by a published, demand-linked formula. That offering closed years ago; supply has been static at roughly 177.6 billion HOT since, and because HOT converts one-for-one, that figure is the ceiling on the HoloFuel that can ever enter circulation by migration. HoloFuel issued through credit, as Section 6 describes, is a separate mechanism whose net supply is always zero. Whatever the team held rode the same \~99% decline everyone else did — there was no insider exit. Supply is capped and cannot grow. The token contract’s minting function was permanently disabled when the sale closed. The contract blocks all transfers while minting is open, so the fact that HOT trades at all reflects that minting has ended. The contract is deployed at 0x6c6EE5e31d828De241282B9606C8e98Ea48526E2 and can be read by anyone who wants to verify it. Tokens still held by the organisation can enter circulation over time, which changes the circulating amount but creates no new supply.
Is HOT (or HoloFuel) a security? How does this sit with regulators?
HOT was never structured or sold as a security and confers no equity or revenue interest; HoloFuel is a utility currency for transacting, not an instrument to invest in. The architecture also produces, as a by-product of how it works, the data provenance and cryptographic auditability that frameworks like the EU AI Act and MiCA now demand — which is why we describe it as regulation-ready rather than regulation-exposed. This is a description, not legal advice or a regulatory ruling.
Delivery, and the people who backed us early
You said this would be live in 2019. It’s 2026. What happened?
We delivered previous versions, but none that lived up to our expectations or commitment. We are late, and we own it. Building genuinely novel distributed infrastructure took three full codebase refactors and proved harder than the roadmap assumed — and holding a system decentralised against the constant pull back toward a centre is harder still. What exists today is set out in Section 5. The delay carried one unearned advantage — the world this ships into, defined by AI, surveillance, and the sovereignty scramble, needs it far more now than it did in 2019.
I bought a HoloPort years ago and it just sits there doing nothing. What now?
We understand why that feels like a betrayal, and we’ve read the threads where people who put in real money in 2017, on the promise of hosting and earning by 2018, said exactly that. We’re not going to answer it by pointing you at a social-media feed, the way we sometimes did before. Here is the straight version. If you bought a HoloPort or HoloPort+, you received hardware, but for a long time there was little for it to do beyond early proof-of-concept apps — not the hosting income you were promised. This is a real failure to deliver against what we said in 2018, and no supply-chain explanation changes that. That being said, you can now use your HoloPort to earn HoloFuel via hosting our Holochain testing application. Further opportunities are still being developed, and all HoloPort owners will receive early notice moving forward.
Is HoloFuel redeemable, and do the Reserve Accounts described in the original still exist?
The Reserve Accounts described in the 2017 Green Paper — managed by Holo, holding fiat and cryptocurrency on reserve, redeeming to national currency on demand — are not the current model. No entity in the present structure takes deposits, holds a reserve of national currency, or redeems claims on demand.
What replaces them, and the terms on which anything moves in either direction, will be set out in the migration documentation rather than here. We are not going to describe a mechanism in a paper that will outlive the terms it describes.
The technology and the architecture
The original Green Paper described HoloFuel’s value in terms of compute cost. Now you say HOT converts 1:1. Which is it?
Both, because they describe different things. The 1:1 is the conversion ratio from HOT to HoloFuel — the delivery of the forward purchase, not a standing exchange rate in either direction. The compute-denomination is how HoloFuel’s value is anchored: to the real cost of processing, bandwidth, and storage, not to a fixed external exchange rate.
Holo Ventures is a company that earns fees. Isn’t that exactly the centralisation you warn against?
The protocol has no capturable centre, and the commercial entity is built so it can’t become one. The Holochain Foundation — a non-profit steward that holds the intellectual property — keeps the protocol open; Holo Ventures operates the commercial layer under licence from it, bound by constitutional safeguards, with fees capped at no more than 1% of value, a limit set in the original Green Paper. Everything is open source and auditable under the Cryptographic Autonomy Licence. And the ultimate answer is the one the 2018 paper already gave: you can run Holochain without Holo Ventures at all — fully peer-to-peer, no centre, no currency required.
Is this still the same project?
Isn’t the HOT-to-HoloFuel conversion just another token migration?
No — it’s the original design arriving. HOT was defined in the original documentation as a forward purchase of HoloFuel, not a speculative asset with a life of its own. Converting it one-for-one and non-custodially is that design carried through — not a pivot to a new token.
You started as a hosting marketplace. Now it’s AI, sovereignty, and a cross-network market. Isn’t that mission drift?
It’s the opposite — it’s the plan, stated in 2017. The original Green Paper said that if HoloFuel were ever used as a more general-purpose currency, hosting would become a smaller part of the economy, opening room for asset-backed currencies for energy, food, and transport. The accounting infrastructure was always general-purpose; what’s changed is that the world finally needs the rest of what it can do. What was an aside in 2018 is the argument now.
Unyt, Moss, and Holo Hosting are not the only applications being built on Holochain. The list below is not exhaustive — hAppenings maintains a comprehensive database of applications at various levels of production readiness. What follows are independent applications you can run or use today.
Project
Purpose
Sector
Desktop app running AI models locally, with signed, tamper-evident conversation records on Holochain
AI / privacy
Further projects are building on Holochain at earlier stages, including the Internet of Energy Network (IOEN — clean-energy microgrids), Kwaxala (Indigenous-led forest data), and the Carbon Farm Network with NY Textile Lab (regenerative fibre supply chains).
This document is published for information only. It is not a prospectus, offering document, or disclosure document, and it is not an offer to sell or a solicitation of an offer to buy any token, security, or financial instrument in any jurisdiction. No token sale, offering, or fundraising of any kind is being conducted or proposed in connection with it. The migration described in this document is not open, and nothing in this document opens it.
On HOT. HOT is an ERC-20 token issued in 2018 and traded on third-party markets over which we have no control. Nothing in this document is a recommendation to acquire, hold, or dispose of HOT. The description of the HOT to HoloFuel migration sets out how existing holdings are intended to be treated; it is not an inducement to acquire HOT, and anyone who does so acts on their own assessment and at their own risk. Digital assets carry substantial risk, including total loss of value.
On characterisation. It is our position that HOT was not structured or sold as a security, that it confers no equity, revenue, or ownership interest, and that HoloFuel is a utility currency for transacting rather than an instrument to invest in. That is our position and not a determination by any regulatory authority. No regulator has examined or approved this document or the matters described in it, and its publication does not imply that any legal or regulatory requirement has been satisfied in any jurisdiction. Readers should form their own view and take their own advice.
On statements about the future. This document describes infrastructure that exists and work that is ongoing. Statements about future development, availability, timing, functionality, or adoption are statements of current intention only — not forecasts, projections, promises, or commitments. Words such as “will”, “expect”, “intend”, “aim”, “plan”, and similar expressions identify statements of this kind. Our intentions are subject to technical, commercial, and regulatory factors, some outside our control, and outcomes may differ materially from what is described. No promise of value, performance, or continuity is made or should be inferred, and no undue reliance should be placed on any statement about the future.
On accuracy. We believe the contents to be accurate at the date of publication, but we make no representation or warranty, express or implied, as to accuracy or completeness, and we accept no liability for loss arising from reliance on them. This document cites third-party sources and industry data; while we consider those sources reliable, we have not independently verified them or the assumptions on which they rest. Content may be amended without notice.
This crypto-asset marketing communication has not been reviewed or approved by any competent authority in any Member State of the European Union. The offeror of the crypto-asset is solely responsible for the content of this crypto-asset marketing communication.
Nothing in this document constitutes financial, investment, legal, or tax advice.
Further reading
uncompromise · Decent Infrastructure in an Authoritarian World — Building the Walk Away Stack · Concept paper for Holo Ventures · 8 September 2026