There is a sentence circulating in defense circles that deserves to be read slowly, the way you would read a line of Solidity you suspect is wrong. Volodymyr Zelensky said, in public, that the Starlink terminals keeping Ukrainian forces connected were operating under a speed limit. Elon Musk replied that any such restriction would be a decision of the U.S. military and, ultimately, the President. Donald Trump, in turn, said he had not known about it and that, in any case, the restriction had been lifted.
Three men. Three claims. No document. No timestamp. No technical report. Just the slow collision of a satellite constellation, a sovereign war effort, and a quarterly earnings call.
I read that exchange the way I read a broken contract. Not for what it says, but for what it assumes. And what it assumes is that a communication system holding a nation's survival together can be throttled from outside, by someone who does not answer to that nation, on a schedule that nation cannot see. If that is true for bandwidth, it is true for every layer of the stack I have spent the last nine years writing about. The speed limit is not a story about satellites. It is a story about who owns the switch.
The Switch Was Never Ours
Let me explain why I could not stop thinking about that exchange during a bear market that has already taken forty percent of the liquidity out of protocols I once defended by name.
Connectivity is the strangest strategic resource ever weaponized, because it cannot be stored, cannot be resold, and cannot be substituted overnight. You can stockpile oil. You can ration grain. You can build strategic reserves of rare earths, of semiconductors, of ammunition. But you cannot stockpile connectivity. It either flows in this second or it does not, and when it degrades the degradation is immediate, universal, and impossible to hedge. When you throttle a satellite link, you are not reducing an inventory. You are reducing the present tense of an army.
That is the property that makes it the highest pain-per-cost coercion tool anyone has ever built. It does not require bombs. It does not require a blockade to be declared. It does not cross the threshold of an armed attack under any treaty I know, and it does not leave craters, which means it does not leave evidence, which means it does not leave accountability. It is a downgrade you can deny.
And then I realized, sitting with my coffee at seven in the morning in a rented apartment in Los Angeles, that the crypto industry has spent a decade promising to be the exact opposite of this โ and has, in the meantime, quietly rebuilt the same switch, in the same shape, in our own image.
I graduated in cybersecurity before any of this was fashionable, and my first obsession was not price. It was provenance โ where a thing comes from, who can touch it, and what happens to the record when they do. In 2017, during the ICO chaos, I spent six months auditing the whitepapers of seventeen fundraising projects and found three smart contract vulnerabilities that were later exploited in the wild. I wrote a series called The Code Is Not the Contract, arguing that trust has to be engineered rather than promised. I believed it then. I still believe it. But I have grown suspicious of how easily we let that belief do our thinking for us.
Because here is the uncomfortable arithmetic of the last decade: we built an industry on the promise of removing the switch, and then we wired every critical function back through a small number of switches that fewer and fewer people can see.
Let me walk you through them, because the bear market is exactly the right time to do this. Bull markets hide chokepoints behind yield. Bears expose them, because in a bear, the only question that matters is not what your asset is worth. It is whether anyone can reach in and turn your asset down.
The Anatomy of a Downgrade
Start with the mechanics of what actually happened in the Starlink case, because the mechanics are where the lesson lives.
A low-orbit broadband constellation is not a military-grade star system in the traditional sense. It is a commercial network running consumer-grade hardware with enterprise-grade ambitions. A terminal has a latency of roughly twenty to forty milliseconds and a downlink somewhere between tens and hundreds of megabits per second depending on the service tier. There is no exotic technology in the throttling itself. There is a service level, an account flag, and a queue. Somewhere in a software stack, a policy is applied. That policy can be applied selectively. It can target high-bandwidth traffic like video and imagery while leaving low-bandwidth voice untouched. It can be applied to a region, or an account, or a class of terminal.
This is the part that non-technical readers miss, and the part I want to over-explain, because I would rather be accused of being slow than of being vague. A selective downgrade is not a blunt weapon. It is a precise one. If you throttle an entire link, everyone notices and everyone complains. If you throttle only the video return path โ the channel that carries drone reconnaissance back to a fire-control team โ then the infantryman's radio still works, the officer's voice call still works, and yet the specific combat pattern that depends on streaming imagery suddenly slows to a crawl.
The military term for the affected system is C4ISR โ command, control, communications, computers, intelligence, surveillance, and reconnaissance. The layperson's term is the nervous system of a modern army. And the drone-to-artillery kill chain, the thing that defined the early years of the war, runs on a bandwidth base that a commercial operator can reach into without a single hostile action.
The coercion isn't in the megabits. The coercion is in the confirmation. You do not have to degrade an ally's army much to remind it that the ceiling on its capability was written by someone else. One throttled afternoon teaches more than a thousand words of diplomacy. And when the limit is later lifted โ as the American president said it was โ the lesson remains, because the lifting proves the tool works, and a tool that works gets used again.
I have seen this pattern before, and not in geopolitics. I have seen it in DeFi governance, where an admin key that is never used is still the most important object in the protocol, because its existence governs everything the protocol can and cannot promise.
The Chokepoints We Built (And Pretended Not To)
Here is where I want to turn the lens back on my own house. Because I am not writing this to lecture a satellite company. I am writing it because the crypto industry markets itself as the antidote to exactly this problem, and the honest reader deserves to know how thin that claim has become.
Chokepoint one: the stablecoin freeze.
USDC and USDT are the settlement rails of dollar-denominated crypto. They are issued by companies that maintain blacklist functions. Over the years, those blacklist functions have been used to freeze billions of dollars tied to sanctioned entities, hacks, and seizures. This is not a rumor. It is public, on-chain, verifiable, and posted every time it happens. If your lending protocol's collateral is held in a centralized stablecoin, then a compliance decision made in a conference room in Boston can render a position in your treasury unliquidatable without a single block being reorganized.
This is the Starlink throttling in a different costume. The flow is not stopped. The flow is made conditional. And the condition is set by a party you did not vote for and cannot replace.
Chokepoint two: the read layer.
Most users believe they interact with a blockchain. They do not. They interact with a remote procedure call provider โ an RPC. Nearly every wallet, every dApp, every portfolio tracker reads the chain by asking a handful of companies what the chain says. If those providers decline to answer, or answer with a filter, the user's experience of a "permissionless" network collapses into a blank screen.
We learned this the hard way in 2022, when a major provider quietly stopped serving requests for a set of smart contract addresses following a sanctions designation. The chain kept running. The validators kept validating. But a large slice of the front end went dark, because the read path was centralized even though the write path was not. A decentralized ledger read through a centralized window is a public record kept in a private room.
Chokepoint three: the write layer.
This one is subtler and I think it is the most misunderstood. When Ethereum moved to proof of stake, block construction increasingly routed through a small number of specialized relays that bundle transactions. To win blocks, validators began complying with relay filters. Compliance became profitable. Profitability became default. Default became infrastructure.
I do not want to overstate this, and I do not want to understate it either. The network did not stop. Blocks still contained a long tail of transactions. But a meaningful share of blocks became filtered, and filtering became normal, and normal became the thing nobody argues about anymore. The switch was installed politely, by a software upgrade, with near-unanimous consent, and almost no one noticed they had consented.
Chokepoint four: the custody and bridge layer.
Bridges are the most dangerous surface in the industry, and not only because they get hacked for nine figures. A bridge is, by construction, a place where assets leave their home chain and are represented elsewhere by something that is not the asset, governed by a set of keys or a committee or a set of validators you did not audit. When that bridge pauses withdrawals, as several did during the worst days of 2022 and again in 2023, users discovered that their "ownership" of a bridged asset was really a claim on a promise, and the promise was conditional.
I have audited enough of these systems to say it plainly, and I will use the signature I use when I am talking to a room of very confident founders: code doesn't keep promises. People do, and people have switches.
Chokepoint five: the physical layer.
And now the one that ties everything back to that satellite. A blockchain node has to run on a machine. That machine has to run somewhere. And increasingly, that somewhere is one of a handful of hyperscale cloud providers. You can run a node on bare metal, and I recommend it to anyone who asks, but the economics of the bear market push operators toward the cheapest, most convenient option, and the cheapest, most convenient option is the same three companies that host everything else on Earth.
This is the quiet asymmetry. The industry that marketed itself as the escape from rent-seeking, geographically dispersed, capture-resistant infrastructure has concentrated its own persistence layer into data centers that answer to the same commercial logic that governs everything else. If the industry that built satellites colonized the orbital commons, the industry that built blockchains colonized the cloud โ and neither one owns the ground it stands on.
The Bear Market as a Stress Test
Let me be concrete about what this means for readers who are sitting on positions and wondering, honestly, whether they should be.
The bear market is not a price phenomenon. It is a liquidity phenomenon. And liquidity is where conditional supply becomes visible.
Over the past several quarters, liquidity in decentralized exchanges and lending pools has thinned unevenly. Some protocols have bled far more than the market average, and the bleeding tells you something the price does not. A pool that loses forty percent of its liquidity in a quarter is not just unpopular. It is often a pool whose depositors have quietly decided that the reward is no longer worth the tail risk of a governance or compliance action they cannot predict.
I want to draw the reader's attention to a specific class of protocol, because this is the kind of insight that I think the industry does not surface often enough. Look at the protocols that are most transparent about their admin keys, their upgrade paths, their freeze functions, and their oracle dependencies. Then look at how their deposits behaved through the drawdown. In my observation, and this is based on following these dashboards for years rather than on any proprietary dataset, the protocols that publish their switches survive drawdowns better than the protocols that hide them. Not because they are safer โ sometimes they are not โ but because depositors can price the risk. An unknown switch is not a smaller risk. It is a larger risk, because it cannot be priced.
Starlink, in this reading, is the most expensive kind of asset to depend on: an asset whose switch is real, whose switch works, and whose switch ownership is publicly contested between a company, a military, and a president, none of whom will say clearly who owns it.
If I were advising a reader on how to think about asset safety right now, I would hand them one sentence and let them do the rest: find the switch before you find the yield. Ask who can pause the contract. Ask who can freeze the token. Ask who can reread the chain from behind a filter. Ask who can turn off the machine, the node, the relay, the RPC, the cloud bill. In a bull market these questions feel paranoid. In a bear market they are simply an audit.
The Deniability Machine
There is one more structure in the Starlink episode that I think is the most important and the least discussed, and it is the structure that the crypto industry has, unknowingly, been perfecting for years.
Call it the deniability layer.
Look at what the three statements actually accomplished. A head of state said a limit existed. An executive said the decision belongs to a military and a president. A president said he did not know and that it was over. Nobody confirmed the limit. Nobody denied it. Nobody signed it. Nobody carried the political cost of degrading an ally's fighting capacity.
This is not incoherence. This is design. It is the same architecture that a properly constructed proxy upgrade uses, or a multisig with a cooperative signer set, or a governance system where the actions are real but the authorship is diffuse. When nobody owns the switch, everybody can deny the outcome, and the victim has no forum to appeal to.
I have seen this in the Terra post-mortem I helped write at the bottom of the 2022 crash, when my own publication had lost seventy percent of its revenue and I genuinely considered leaving the industry. We spent weeks with a team of three colleagues tracing how broken promises erode trust faster than broken code. What we found, and what regulators later cited, was that the collapse was not primarily a mechanism failure. It was a provenance failure. Nobody could say clearly who had promised what, and by the time they could, the trust was already gone.
The Starlink exchange is that same failure wearing formal wear. A monetary authority, a private company, and a wartime government all describing the same service in mutually incompatible terms, and no one willing to be the author of the constraint.
And this is why the exchange matters far beyond satellites. Because the crypto industry built its entire moral case on the proposition that authorship of constraints should be transparent and verifiable โ that the rules should be legible to everyone, that no administrator should be able to change the terms of a system silently. We called it credible neutrality. We sold it as the whole point.
Then we went and built admin keys, freeze functions, upgradable proxies, filtered relays, and centralized RPCs, and we told ourselves it was fine because we were the good guys and the switches would never be used against the right people. What the satellite episode shows is what happens when that assumption meets a moment where the incentive to use the switch is real.
Soulless finance is just empty pixels. And the satellite is the proof, because the strangest thing about watching a nation's connectivity get throttled is realizing how quickly a technical abstraction becomes a human emergency. Nobody in that story was arguing about megabits. They were arguing about whether a soldier would come home. The moment we forget that every switch is ultimately attached to a person, we stop being engineers and start being landlords.
The Contrarian Case: Maybe the Switch Is Not the Problem
Here is where I want to be honest about my own bias, because the version of this essay I would naturally write โ the one that ends with "therefore we must decentralize everything" โ is the version that does not survive contact with reality, and I would rather you hear the sharper argument.
The most contrarian claim in this entire discussion is that the defect is not the existence of the switch. It is the pretense that it does not exist.
I spent years writing in the register of the ideologue, and I have come to believe that ideological crypto criticism of centralized infrastructure misses the actual lesson. A switch that everyone can see is a protocol. A switch that no one can see is a trap. The tragedy is not that SpaceX has reach into its own constellation. Of course it does; it owns it. The tragedy is the gap between the marketing claim of neutrality and the operational reality of discretion. The same is true of our industry.
We now have a generation of users who believe they hold "self-custody" assets while reading the chain through a provider that filters, transacting on rails whose relays comply, storing the record on a cloud they do not control, holding collateral in tokens a compliance department can freeze, and borrowing against bridges that can pause. Every one of those dependencies is a switch. Every one of those switches is legible to anyone who looks. And each time a switch gets used, the market is briefly surprised, which means the market has been systematically mispricing the risk.
So the contrarian conclusion, and I want this to sting a little, is this: decentralization is not a property you hold. It is a price you pay continuously, and most of us stopped paying years ago and kept the brand name.
That is an uncomfortable thing to write in a bear market, when the temptation is to crawl into maximalism and recite old certainties. But it is the honest thing. Because the reader who is sitting on assets and wants to know if they are safe deserves a real answer, and the real answer is: not because the technology is magic, but because the dependency graph around it has become thin and legible and worth auditing.
The satellite speed limit is a good thing to have happened, in a strange way, because it makes visible a category of risk that has been hiding in plain sight across every layer of the crypto stack: the risk of conditional supply. Not supply that stops, but supply that continues under terms you did not negotiate, cannot see, and cannot predict.
What the Suits Actually Own
Let me make one more observation that ties the geopolitical machinery to the on-chain machinery, because they are more similar than the industry admits.
Traditional defense industrial logic holds that the dangerous moment is delivery: once a weapon is delivered, control leaves the manufacturer. The manufacturer worries about export licenses, about end-use certificates, about the perimeter of the sanctions regime. After delivery, its power ends.
Commercial space services invert this. The product is not delivered; it is provided. The provider keeps the keys forever. Control is exercised not at handover but at every moment of use. This is the difference between selling a rifle and selling the ammunition subscription, except the rifle only works with your ammunition, and your ammunition only works if you keep paying, and you keep paying on terms that can be revised.
This is the future of military technology exports. It is also, and I want the crypto reader to feel the recognition here, exactly the business model of a hosted RPC, a custodial wallet, a staking-as-a-service provider, a cloud-hosted validator, a rollup sequencer run by a company, a cross-chain bridge with a committee. The service is continuous. The keys are continuous. The control is continuous.
And because the legal frameworks for military exports were designed around delivery, not provision, nobody knows how to govern a service that keeps its hand on the dial. Export controls assume a moment where the goods leave. War law assumes a moment when an attack begins. Neither anticipates a service downgrade that is neither delivery nor attack.
We have the identical gap in crypto. Our regulatory frameworks were designed around transfers of value. They struggle with the ongoing provision of critical function. Nobody has cleanly answered whether a company operating a settlement layer for a sovereign-adjacent system can downgrade that system's service, under what conditions, and with what obligation to disclose. When the industry eventually gets regulated, this is the question that will matter most, and almost nobody is structuring for it.
Where the Bleeding Actually Is
Let me bring this back to the reader's portfolio, because abstract geopolitical insight is worthless if it does not help someone make a decision in a drawdown.
In a bear market, the protocols that bleed are usually not the ones with the best technology. They are the ones with the least legible dependency graphs. Look for the following signals, and treat each as a question you can answer on-chain in an afternoon:
A protocol whose critical oracle is a single multisig with signers you cannot name is a protocol whose price depends on a private phone call. A protocol whose bridged collateral depends on a validator set you have not counted is a protocol whose collateral can be paused. A protocol whose front end depends on a single hosting provider is a protocol whose accessibility can disappear in a compliance sweep. A protocol whose governance can be upgraded by a quorum you cannot mobilize is a protocol whose rules are provisional. A protocol whose node infrastructure lives in one cloud region is a protocol whose chain has a landlord.
I am not saying avoid every protocol with a switch. That would mean avoiding every protocol that exists. I am saying that the honest move is to know which switches exist, who holds them, and what the historical use pattern looks like. A switch that has never been used is not the same as a switch that cannot be used, and the difference is exactly the difference the market fails to price.
This is the work of a cybersecurity background applied to markets, and it is where I think the industry most needs writers who are willing to be boring about it. Everybody wants to talk about the next narrative. Almost nobody wants to inventory the switches.
The Slow Crypto Takeaway
Here is what I keep returning to, and I want to leave it with you as a thought rather than a conclusion, because I do not trust anyone who claims to have finished thinking about this.
When I retreated to a cabin in Big Sur years ago to work on a project linking digital tokens to physical carbon offsets, the thing that obsessed me was provenance โ not the price of the asset but the story of its origin. Who made it. Who can touch it. What happens to the record when they do. I built non-transferable tokens tied to real-world actions because I wanted ownership to mean something that could not be flipped for a quick exit.
The satellite incident is a provenance problem at the scale of a nation. A communication service whose origin is a company, whose authority is a government, whose user is an army, and whose ledger of obligation is a set of public statements that do not agree with each other. The asset works. The provenance does not.
And the crypto industry, which was founded precisely to fix the provenance of money, has spent a decade reapportioning the trust it promised to engineer into ever-fewer hands, held ever-more-quietly. The speed limit above Ukraine is a mirror, and the reflection is not flattering.
The next narrative will not be about which chain is fastest, or which rollup has more projects deploying first, or which Bitcoin meta-layer hauls the most cargo on the most elegant chassis. The next narrative will be about legibility. Users in a long winter stop asking what an asset can do and start asking who can reach in and turn it down. The protocols that survive will be the ones that can answer that question without flinching, and the ones that cannot will keep bleeding liquidity quarter after quarter while calling it market conditions.
So here is the question I want you to sit with, and I mean it literally, not rhetorically. If a head of state said tomorrow that your position was under a speed limit, who would you call to lift it? If you cannot name the person, if you cannot point to the contract, if you cannot read the decision on a public chain, then you are not holding an asset. You are holding a promise with a switch above it, and the switch was never yours. Code doesn't keep promises. People do. And people hold the dial.