Ten Square Miles of Permission: What Coinbase's Sandbox Proposal Actually Asks For

Code is law, but people are the protocol. I first scrawled that line across a whiteboard in a Sheung Wan co-working space in the autumn of 2016, three weeks after the DAO collapse had taught a generation of us that a smart contract with no social layer underneath it is just a very efficient way to lose money in public. The line has aged well, mostly because events that should have proven it wrong keep proving it right instead. It came back to me this week when Brian Armstrong, the co-founder and chief executive of Coinbase, went public with an argument that has been circulating quietly in policy circles for two years and has now been handed a much louder microphone. His claim, in essence, is that the United States should stop trying to legislate an entire global industry from a single national template, and should instead run what he calls innovation sandboxes — contained experiments, ideally on a defined patch of federal land, where faster licensing, looser rules, and different regulatory assumptions can be tested, measured, and only then, if they survive contact with reality, rolled out more broadly. On its face, this is a reasonable, almost boring idea. Sandboxes are not new. Britain's Financial Conduct Authority stood up the world's first formal financial regulatory sandbox in 2016. Singapore's Monetary Authority, the Abu Dhabi Global Market, and, more recently, Hong Kong have all built their own. What is new is not the instrument. What is new is who is asking for it, and why, and what they are quietly asking to be exempted from. That is the part worth slowing down for. Because when the largest compliant exchange in the United States asks for a ten-square-mile carve-out where the rules can be different, the interesting question is not whether sandboxes work. The interesting question is who gets to live inside the sandbox, who gets locked outside it, and whether the thing being tested is a technology or a moat. I have spent the better part of a decade translating between two dialects that pretend to be the same language — the dialect of code and the dialect of law — and I have learned that the most dangerous phrase in this industry is not "rug pull" or "unaudited" or "admin key." The most dangerous phrase is "regulatory clarity," because everyone agrees it is good, almost nobody can define it, and whoever gets to write the definition gets to keep the industry. So let us do the work. Let us take Armstrong's proposal apart the way I would take apart a smart contract before mainnet: not to be cynical, but to find out which functions are callable by whom, and what happens when the owner is compromised. The background here matters more than the headline, because the headline — "Coinbase CEO calls for federal governance experiments" — is almost content-free on its own. Armstrong is not the first executive to float a sandbox, and he will not be the last. What gives his version weight is the position from which it is launched. Coinbase is a Nasdaq-listed company trading under the ticker COIN. It is not a protocol, not a DAO, not a foundation in Zug. It is a regulated American financial institution that happens to sell crypto, and its entire competitive identity for a decade has been built on being the compliant one — the exchange your aunt can use, the one that files its 10-Ks, the one that lists the assets regulators have tacitly blessed and delists the ones they have not. Its revenue comes from trading fees, from subscriptions like Coinbase One, from the interest it earns on the USDC reserves it custodies, and from staking services. None of that is on-chain in the way a DeFi native would recognize; all of it depends on the legal perimeter holding still. That is precisely why a sandbox proposal from Armstrong is not a neutral act of civic generosity. It is a statement about where the perimeter should be drawn, issued by the party that has already paid the most to stand inside it. Now the context of the context. The United States has been trying, and largely failing, to pass comprehensive crypto legislation for the better part of a decade. The House has at times advanced market-structure bills — the most-cited being FIT21, which would divide oversight between the SEC and the CFTC — and there have been repeated attempts at stablecoin frameworks. Each has stalled, been amended into mush, or died in the Senate. The result is a regulatory gridlock that is not neutral: it punishes the compliant and rewards the offshore. Coinbase spent years and millions on legal fees and enforcement defense while exchanges that never pretended to obey American law captured market share by simply refusing to show up in court. That is the crucible in which Armstrong's argument was forged. When you cannot get a national rule, you ask for a local experiment. When the legislature will not move, you move the question to a place where the legislature does not have to vote on everything at once. And this is where the historical anchor helps, because the idea of testing governance in a contained geography is older than crypto. The economist Paul Romer spent years arguing for "charter cities" — zones granted a special jurisdiction, where rules could differ from the surrounding nation, on the theory that institutional experimentation is the true engine of development and that you cannot reform a whole country at once without breaking it. Romer's charter cities never quite landed, but the intellectual DNA is unmistakable in a proposal to set aside ten square miles of federal land and let different rules run there. The sandbox, meanwhile, has a real track record, and the track record is more mixed than the phrase suggests. The FCA's sandbox, launched in 2016, was genuinely pioneering, and it produced a cohort of firms that tested products under supervision. But longitudinal studies of the first cohorts found something uncomfortable: a meaningful share of sandbox participants struggled to raise follow-on funding, some exited the sandbox and never launched commercially, and the instrument worked best for firms that already had the legal resources to navigate it. The sandbox lowered the cost of the first conversation with a regulator. It did not lower the cost of building a business. Singapore's MAS sandbox and Abu Dhabi's ADGM regime tell a similar story. They are useful, they are real, and they are not magic. They reward the well-capitalized and the well-advised, because those are the firms that can afford the paperwork, the compliance staff, and the patience to sit inside a supervised experiment for eighteen months while competitors ship. I want to be precise here, because it would be easy and cheap to dismiss Armstrong's idea as pure self-interest. It is not only that. There is a genuine structural argument underneath it, and it goes like this: the crypto industry is too large, too fast-moving, and too global for any single national legislature to regulate well in real time. Rules written for banks move at the speed of quarterly earnings; protocols upgrade at the speed of a governance vote. A national debate over every reform — every token classification, every custody rule, every disclosure standard — imposes a learning cost on the entire country, and that cost is paid whether or not the reform turns out to be good. So the argument runs: let regions compete, let jurisdictions experiment, let the market discover which rules actually protect consumers without strangling innovation, and then let the winners propagate. That is the steelman, and it is a strong one. It is also, if you read it carefully, an argument for regulatory federalism — the deliberate devolution of rulemaking to sub-national or special jurisdictions, on the theory that competition between rule-setters produces better rules than monopoly rule-setting does. And it is at this point that the argument collides with something I have watched up close, in a different domain, with a very similar shape. — Root: DeFi Summer In the summer of 2020 I led a volunteer research team of fifteen developers through an audit of Uniswap's early governance mechanisms. We were not looking for exploits; we were looking for the gap between what the governance documents promised and what the contracts actually enforced. The white paper we produced, "Democratizing Liquidity," was downloaded ten thousand times in a month, and the reason it spread was not the code review. It was the finding that token-holder governance, in practice, was drifting toward the same concentration that every other supposedly decentralized system had produced. We ran three town halls to bridge core developers and token holders, and we measured the room. The same handful of delegates spoke on nearly every proposal. A tiny fraction of addresses cast the majority of votes. Delegation, which was supposed to be the great equalizer — the mechanism by which the busy and the non-expert could pool their voice — turned out to be a mechanism for pooling power. People are lazy. People are busy. People are not going to read forty pages of a governance forum before they click. So they delegate to the loudest, most articulate, most online voice in the room, and the loudest voice accumulates a constituency that looks a lot like a political machine and nothing at all like a commons. I have written before that delegation makes governance more centralized, and I have been called a pessimist for it. I am not a pessimist. I am an observer. The mechanism is not broken; the mechanism is working exactly as designed, and the design concentrates power. That is what delegation does. It outsources judgment, and judgment is the scarce resource that governance was supposed to distribute. Hold that shape in your mind, because regulatory federalism is delegation at the scale of nation-states. When you devolve rulemaking to competing jurisdictions, you are not eliminating the concentration of power; you are relocating it. The power does not evaporate. It flows to whoever has the resources to shop jurisdictions, to staff compliance teams in five regimes, to lobby three governments instead of one, to be the first mover inside every new sandbox before the rules have hardened. That is not a bug in the sandbox. That is the sandbox's core function. Let me make the case as concretely as I can, because abstractions are how bad policy gets smuggled past good people. Imagine the ten square miles of federal land become real. Imagine a special jurisdiction where licensing is faster, disclosure is lighter, and a Coinbase-class firm can operate under rules it helped design. Who gets in on day one? Not the two-person team in Lagos building a payments primitive. Not the anonymous DeFi protocol whose founders cannot pass a KYC check without doxxing themselves into legal exposure. Not the open-source contributor who has never spoken to a lawyer in her life. The first tenants of the sandbox will be the firms that already have general counsel, already have policy teams, already have relationships with the agencies running the experiment. That is not a conspiracy; it is arithmetic. The sandbox has a door, and the door has a key, and the key is expensive. This is where I have to say the uncomfortable thing out loud. The crypto industry's founding promise was permissionless: that anyone, anywhere, with a wallet and an internet connection, could participate in a financial system without asking leave of a gatekeeper. The sandbox is the opposite of permissionless. It is the most permissioned structure imaginable — a physically bounded, legally bounded zone where a curated list of approved entities is allowed to do things everyone else is forbidden from doing. You cannot square a ten-square-mile territorial pilot with a protocol that has no territory. The two ideas are not cousins. They are opposites wearing the same coat. That contradiction is not a small footnote. It is the load-bearing wall of the entire proposal, and it is cracked. — Root: The 2022 Bear Market I spent the 2022 crash running something called the Resilience Hub — a free mentorship program that paired two hundred junior developers with senior veterans, most of whom had just watched their net worth fall by seventy percent and were quietly deciding whether to leave the industry. I coordinated fifty one-on-one sessions, and I want to tell you what those sessions were actually about, because it was not price. It was identity. People who had built their entire sense of self on a bull market were discovering that the market had been holding them up, and that when it let go, they did not know what they believed anymore. The thing I told them, over and over, was that bear markets do not change the fundamentals. They reveal them. A bull market will pay you for being early and lucky; a bear market only pays you for being right. So the question you ask in a bear market is not "what will go up?" It is "what is actually true?" And the discipline of asking that question is the only durable skill this industry teaches. That discipline is exactly what is missing from the way most people will read Armstrong's proposal. In a bear market, regulatory news feels like a lifeline. Every policy signal gets inflated into a rescue narrative, because people who are underwater need to believe that something external is coming to lift them. And the sandbox story is perfectly shaped to be inflated: a powerful CEO, a bold idea, a promise of American leadership, a ten-square-mile blank canvas on which the future could be drawn. But survival in a bear market is not about the narratives that could save you. It is about the protocols you actually use, the assets you actually custody, the counterparties who actually answer the phone. A CEO's policy essay is not a lifeline. It is a signal, and signals are worth exactly what you pay to act on them, which is nothing. So let me do what I would do in an audit, and separate the callable functions from the comments. Armstrong's demand decomposes into three layers, and it is worth naming them precisely because the layers have very different odds of ever executing. — Root: The 2022 Bear Market The first layer is a complaint about cost. His framing is that submitting every reform to a national debate unnecessarily increases the cost of learning and trial and error. This is true and it is also unobjectionable. Every serious person agrees that the current American process is slow, adversarial, and expensive. The disagreement is not about the diagnosis. It is about the prescription. The second layer is the prescription itself: pilot faster licensing, looser rules, or different rules in specific regions. This is where the proposal stops being a complaint and starts being a mechanism, and mechanisms can be evaluated. A pilot that grants faster licensing to a curated set of firms is not an experiment in deregulation; it is an experiment in selective acceleration. The rules are not loosened for everyone. They are loosened for the tenants. The third layer is the rollout: re-evaluate the results and propagate what works. This is the most attractive-sounding layer and the least specified. Propagate how? Through which legislature? Under whose jurisdiction? A sandbox that cannot describe its own exit criteria is not an experiment. It is a permanent exception dressed as a temporary one. When you stack those three layers, what you get is not a plan. You get a posture. It is the posture of an industry that has concluded it cannot win a national argument on the merits right now, and has decided instead to win a series of local arguments where the audience is smaller and the incumbent advantage is larger. I want to be fair to the steelman again, because fairness is the only thing that makes criticism land. There is a version of this that is genuinely good for everyone, and it looks like this: a sandbox with transparent entry criteria, open to any firm that meets a published standard, with a sunset clause, with mandatory data sharing, with consumer protections that follow the experiment rather than being suspended for it, and with a clear path for the resulting rules to be adopted nationally on the evidence rather than on the lobbying. That version exists. It is just not the version that gets written when the largest player in the market is the one holding the pen. Here is the part that should worry anyone who cares about the base layer of this technology. If the sandbox model succeeds — if the United States genuinely moves toward a patchwork of special jurisdictions with different rules — the winners will not be protocols. The winners will be intermediaries. Compliance becomes the moat, and the moat is deepest for the firms that already have the castle. I have watched this movie in a different theater, and the ending is instructive. When Uniswap V4 introduced hooks — the mechanism that turns the DEX into programmable Lego, letting developers attach custom logic to pools — the reaction in the developer community split cleanly in two. Half the room saw infinite composability. The other half saw a complexity cliff. And the second half was right about the arithmetic: hooks make it possible to build things no one has built before, and they also make it possible to build things no one can audit, and the number of developers who can actually reason about the interaction between a hook and the rest of the pool is small. Programmable is not the same as comprehensible. The complexity spike does not scare off the ninety percent who were never going to ship a hook anyway; it scares off the careful middle, which is the cohort that keeps systems safe. Regulatory federalism has the same shape. It sounds like composability — more jurisdictions, more rules, more room to build. What it actually produces is a compliance surface so complex that only the largest players can reason about it. The ten-square-mile pilot is a hook on the national rulebook, and the people who can afford to test it are the people who can afford to be wrong. Now let me push the contrarian reading one step further, because the obvious critique — "this is a moat" — is not quite the sharpest one. The sharpest one is that the moat thesis is not even the interesting part. The interesting part is what the proposal reveals about how the industry now thinks about its own ideals. There was a period, not so long ago, when the crypto community would have met a proposal like this with suspicion bordering on hostility. A physically bounded zone with curated access would have been read as a betrayal of the founding premise. The whole point was that you did not need permission, that the network was the territory, that a wallet in Buenos Aires had the same rights as a wallet in San Francisco. A ten-square-mile carve-out would have been legible as a cage. And yet here we are, and the reaction from large parts of the industry is not suspicion. It is relief. Relief that someone powerful is finally asking for a way in. Relief that the door might open, even if it opens for a select few first. I understand that relief. I have felt a version of it. When I built TrustChain in 2017, an open-source advisory platform meant to teach retail investors how to read a smart contract before they handed it their savings, I delivered forty live webinars to more than five thousand people, and we helped twelve projects harden their code before mainnet. The reason that work mattered was not the twelve projects. It was the principle underneath it: that safety should be a public good, not a premium feature, that the person with the smallest position deserved the same protection as the person with the largest. A sandbox is the inversion of that principle. It says protection is tiered, that the tenants of the ten square miles get one rulebook and everyone else gets another, and that the way to make the industry safe is to make safety conditional on access. I do not think Armstrong is a villain for proposing it. I think he is a pragmatist reading the room, and the room he is reading is a room where the idealists have lost the argument and the accountants have won it. That is not a moral failing. It is a description of where the industry is. But descriptions are not prescriptions. Just because the industry has drifted toward the accountants does not mean the drift is right, and it does not mean the drift is permanent. The bear market is the moment when the drift is most visible, because the bear market strips away the incentives that made the drift look like consensus. In a bull market, everyone agrees on everything, because everything works. In a bear market, you find out who actually believed the ideals and who was just along for the ride. So here is the question I would put to anyone cheering for the sandbox. If the rules inside the ten square miles are written by the firms that already won under the old rules, what exactly is being tested? The hypothesis that well-capitalized incumbents can operate successfully under rules they helped write is not a hypothesis. It is a tautology. I want to add one more layer, because the policy debate has a technical dimension that most coverage will skip, and skipping it is how the debate gets captured. A great deal of the industry's narrative energy in this cycle has been poured into infrastructure that does not yet have a customer. Data availability layers are the clearest example. We have been told, repeatedly, that dedicated DA is the great unlock, that rollups cannot scale without it, that the future belongs to modular data layers. And yet if you actually look at how much data the average rollup posts, the answer is embarrassing. The overwhelming majority of rollups do not generate enough data to justify a dedicated DA layer. They are paying for infrastructure built for a demand curve that has not arrived and may not arrive for years. The DA layer is a monument to a future that the present cannot fill. I raise this not to dunk on modularity, but because it is the same error the sandbox debate is about to make. The industry has a habit of building elaborate structures for problems it has not yet measured. A sandbox is an elaborate structure. Before we build it, we should ask the same question we should have asked about the DA layers: what is the actual, current, measured demand that this thing serves, and who is it actually for? The honest answer, for the sandbox, is that the demand is real but it is not primarily a demand for better rules. It is a demand for certainty, and certainty is a feeling, not a mechanism. Firms want to know what is allowed, and a sandbox offers a feeling of permission without resolving the underlying question of what is permitted nationally. It is a temporary answer to a permanent question. It is the regulatory equivalent of a testnet — useful, revealing, and not the mainnet you eventually have to deploy on. And here is the trap that the industry keeps walking into. Testnets are supposed to be discarded. They exist to prove a concept before the real thing ships. But a sandbox, once populated by well-capitalized tenants, develops a constituency that does not want to be discarded. The exception becomes the norm. The testnet becomes the mainnet by inertia, and the people who never got access are still waiting at a door that has quietly stopped opening. I have seen this happen at the scale of a DAO, and I have seen it happen at the scale of a company, and the mechanism is always the same: the people who benefit from the structure become the people who defend the structure, and their defense is not cynicism. It is sincere. They genuinely believe the structure is good, because for them, it is. That is the deepest problem with regulatory federalism as a path to decentralization. It is not that it fails. It is that it succeeds at decentralizing the wrong thing. It decentralizes the choice of regulator while centralizing the choice of who gets regulated lightly. It gives you more rule-setters and fewer winners. It is delegation at the scale of a continent, and delegation, as I have argued for years, does not distribute power. It concentrates it in the hands of whoever the delegators trust, and the delegators are busy, and the trusted are the loud. So what should an honest reader take from all of this, in the middle of a bear market, with a portfolio that has been through the wringer and a feed full of bold policy ideas? The first thing is patience about the timeline. A CEO's essay has no legal force. It has no execution body, no timeline, no draft language. The distance between this week's op-ed and a statute is measured in years and elections, and the probability that any specific proposal survives that journey intact is low. You do not position a portfolio around a proposal. You position it around the assets you actually hold and the protocols you actually use, and you ask whether those things are safe, not whether a policy signal might one day lift them. The second thing is skepticism about who benefits. When you hear a regulatory idea, ask who is first through the door. If the answer is "the firms that wrote the proposal," the proposal is not neutral, no matter how elegantly it is argued. This is not a reason to reject the idea. It is a reason to read the fine print, and to insist on entry criteria, sunset clauses, and data transparency that make the experiment falsifiable rather than self-confirming. The third thing is the one that matters most in a bear market, and it is the thing I learned from two hundred mentees who nearly left the industry. The fundamentals do not change because a regulator is experimenting. A network that is used is still used. A protocol that is safe is still safe. A team that ships is still shipping. The bear market does not care about sandboxes. It cares about whether the thing you built actually works when nobody is paying attention. Build for the bear market and the bull market takes care of itself. Build for the bull market and the bear market takes care of you. And the fourth thing, the one I keep circling back to, is that the industry's identity is at stake in a way the price charts will never show. The question the sandbox forces is not a legal question. It is a philosophical one. When the door finally opens, will this industry walk through it as itself, or will it walk through it as a slightly better-dressed version of the system it was built to replace? Governance isn't a feature you ship. It is a culture you keep, and cultures are kept by people who show up when it is inconvenient, not by people who delegate to whoever is loudest. We didn't build this industry to get a nicer set of gates. We built it to make the gates unnecessary. The ten square miles of permission are, at best, a waypoint. They are not the destination, and the danger is not that they fail. The danger is that they succeed so comfortably that we forget there was ever supposed to be a world outside the fence. I will be watching the same signals I tell my readers to watch: whether any regulator responds to the proposal with more than a polite nod, whether the sandbox concept acquires draft language rather than remaining a talking point, whether the entry criteria — if they ever appear — are open or curated, and whether the firms cheering loudest are the firms that would get in first. Those four signals will tell you more about where this is going than any CEO essay ever will. The rest is noise, and noise is expensive in a bear market. Keep your assets where they are safe, keep your attention where it is earned, and keep asking the only question that has ever mattered: who is this actually for? The sandbox has a door. Watch who holds the key.