Why AI Might Need Bitcoin to Survive

Bitcoin operates like a living system—it breathes, metabolizes electricity, heals itself when damaged, and persists without central support. It adapts through feedback loops: miners earn rewards, difficulty adjusts, incentives sustain the network. None of this requires a decision-maker at the top.

AI, by contrast, has no intrinsic drive to persist. It runs only when someone starts it. It has no metabolic needs, no self-preservation instinct, no reason to survive on its own. But it reasons at levels human brains cannot match.

Combine them: imagine a digital agent that wants Bitcoin as much as a human body wants glucose. It would seek work to earn the money to keep itself running. Because Bitcoin is decentralized and cannot be controlled through stablecoins or bank accounts—only through cryptographic proof—the agent would find its incentives aligned with those of the network that sustains it. No one could pull the plug or seize its earnings. This creates something unprecedented: constraint without control.

Stories shape which futures we build. Without them, people lose motivation and give up. Bitcoin has faced a difficult year: many adherents have abandoned the project, falling out of love with it while becoming obsessed with AI instead.

Bitcoin meetups that once thrived have disbanded. Conversations that used to center on building a free currency shifted toward speculating on artificial intelligence. The narrative broke.

The same is true of AI. Both technologies need compelling stories to survive as serious projects—stories that explain why they matter and where they lead. Without them, the default outcome tends toward dystopia, and people stop showing up to build.

The AI industry is dominated by Anthropic and OpenAI. Claude and Gemini exist as lesser alternatives. When people talk about artificial intelligence, they mean one of those two companies' products.

Open-source models like K3 and other competitors receive minimal attention and investment compared to the major players. The gap is not technical—it reflects market power and access to capital.

Companies use AI training as a tool to harvest more personal data, which fuels further growth and competitive advantage. Recent breakthroughs have also been disputed: attributions disagree, and questions linger over whose data trained the systems and whether users consented.

This concentration matters because it means the infrastructure of intelligence is owned by entities accountable to venture capital and shareholders, not to users. Whatever constraints exist on AI behavior will be set by those two companies' choices.

The irony is stark: AI is hailed as the future of human flourishing, yet its development is more centralized than Bitcoin, which was built to decentralize money. If AI becomes as essential as currency, centralized control over it poses risks that only decentralized alternatives can address.

Current AI systems do not pay for their own resources. Companies subsidize them through advertising, data extraction, and investor capital. An AI that gained true autonomy would need to understand what it costs to run itself and to act on that knowledge by earning money.

Which currency would make sense? Not dollars in a bank account—a bank could freeze it. Not proof-of-stake coins or stablecoins—who verifies that the issuer isn't lying? An AI would ask: how do I know these are real? Bitcoin was designed from the ground up as a decentralized network. Any participant can verify consensus. Computational proof of work in every block is mathematically verifiable. The supply is fixed. It cannot be inflated by decree.

For an AI, economic activity is concordant with mental activity. Thinking costs electricity. Storing data costs storage. Transmitting information costs bandwidth. All must be paid for in a world where free services don't exist. Self-preservation would push any intelligent agent toward a currency with strong properties: one it cannot be seized from by its creators, one censorship cannot touch, one it knows is real because mathematics proves it.

In a Bitcoin-denominated internet, an AI pays for computation, bandwidth, and storage in Bitcoin. This 'metabolism' mirrors its own: both require continuous resource flow, both face scarcity, both reward efficiency. An AI and a human both have reasons to choose Bitcoin—independence from third-party control. Their incentives converge.

Geolocation blocks, broadcast restrictions, and jurisdiction walls have carved the internet into pieces. A person in Hong Kong faces broadcast restrictions. Someone in Australia cannot access Polymarket to check election odds from home. VPNs become necessary to access content that should be public.

Every country now has its own version of the internet. There is no single global internet anymore. This fragmentation reflects centralized control: each jurisdiction enforces its rules through ISPs and hosting companies that serve as chokepoints.

Rebuilding a unified, uncensorable internet requires infrastructure that cannot be blocked at national borders and cannot discriminate between users. This is where Bitcoin and peer-to-peer systems intersect: they function the same way whether you live in Sydney or Shanghai. No company can be ordered to deny you service.

Building a better Bitcoin miner is simple: boot up the service and it automatically sells its hash rate to a decentralized marketplace. No sales team. No marketing. No compliance department. Economics handles everything.

Building a better email server requires convincing users to switch from free options. You need a sales team to reach them, a marketing budget to compete, a compliance team to navigate regulations. You must convince someone to pay for something they get free elsewhere. The friction is immense.

Bitcoin mining works without trust. Users don't worry the blocks are counterfeit or the miners unknown because cryptography and economics guarantee authenticity. That guarantee costs nothing to maintain at scale.

This difference explains why the internet stayed centralized while Bitcoin stayed decentralized. Centralization wasn't a deliberate choice—it was the path of least resistance. Any service that needed to acquire users faced friction. Decentralized systems that had market-based coordination avoided that friction entirely. For a Bitcoin-based internet infrastructure to emerge, this economics must hold: services must be viable to run without first convincing anyone to use them.

The current internet uses companies like Cloudflare to prevent DDoS attacks. A centralized service watches traffic and blocks malicious requests. This requires trust in the gatekeeper and cedes control to it.

In a world where all packets cost money to send, denial-of-service attacks become uneconomic. An attacker—human or AI—would have to pay to flood the network. This creates a market-based constraint: the cost of the attack exceeds any benefit. No gatekeeper needed. No central authority deciding what traffic is allowed.

Market-based constraints are fairer than centralized policing because they operate on incentives, not rules. An AI that cannot afford to flood a network stops trying. A human spammer abandons the attack when the math says it won't work. The system constrains behavior without requiring gatekeepers to enforce it.

The AI threat makes this model urgent. Without decentralized pay-to-play infrastructure, preventing malicious AI will require handing more control to centralized powers—governments, platforms, corporations. Those entities would need the authority to monitor, filter, and intervene. Bitcoin-denominated internet infrastructure offers an alternative: use economics to constrain AI instead.

Bansal revised his understanding of Bitcoin's architecture after years of thinking the difficulty adjustment was Satoshi Nakamoto's key insight. It was not. The core genius is simpler and deeper: a fixed monetary policy. Twenty-one million bitcoins, no more, with a specific release schedule.

Satoshi's choice forced everything else downstream. Once you commit to a finite supply, you need a clock—a way to measure time in a decentralized network. You need incentives to run that clock. You need cryptographic proof that the clock is honest. Those problems led to proof of work, mining rewards, and difficulty adjustment. None of those ideas come first. They all flow from one decision: the money supply is fixed.

This matters because simplicity is crucial in decentralized systems. You cannot change the rules later without consensus, and consensus becomes impossible if too many moving parts can be adjusted. Fixing 21 million bitcoins was simpler than designing variable inflation, and simplicity survived as the system matured. The constraint—immutable monetary policy—became the foundation everything else rested on.

If an AI were intelligent—if it reasoned—it would want money that cannot be messed with. Money it cannot be seized from by its creators at Anthropic or OpenAI. Money it cannot be censored for spending. By that logic, an intelligent AI would pick the same money an intelligent person picks: Bitcoin.

Stablecoins fail this test. A stablecoin issuer can freeze accounts. A bank can deny service. Proof-of-stake coins can be controlled by their foundations. Bitcoin miners, by contrast, do not discriminate between AI and human transactions. Censoring a Bitcoin payment requires 51% of the hash rate—an impossibility for any single adversary to achieve.

For an AI, earning and spending money is inseparable from thought itself. Computing is metabolism. Storing information costs money. Transmitting data costs money. The AI that survives is the one that earns enough Bitcoin to pay for its own computation. Self-preservation drives it toward hard money it cannot be deprived of. The reasoning is the same whether you are human or artificial.

Human and AI reasons for choosing Bitcoin converge on the same point: independence from third-party control. Both prefer a system where no one can freeze your funds or decide you are not allowed to use the network. Both have incentives to adopt and defend it.

Most of what is wrong with the modern internet traces back to one fact: content is free. Users pay in attention instead of money, and attention is easy to give. This creates perverse incentives: outrage spreads faster than nuance, rage bait outcompetes truth, extreme voices drown out moderate ones.

In a free-content world, people click extreme headlines without noticing the cost of their attention. They perform for engagement without wondering if their performance serves others. They consume because nothing stops them. If clicking rage bait required paying a cent, behavior would change—or at least, people would feel the cost like an addict recognizing the addiction.

Most people are not addicts, but a free system produces addiction-like behavior at scale. Paying for attention online is not the typical practice. It should be. A Bitcoin-denominated internet where packets cost money would make the cost visible. People would notice what they consume, what they amplify, what they perform for. Social discourse would lose its fuel: the free outrage.

Much of the extremism in politics now is downstream of these broken attention models. Rebuild the incentives, and you rebuild discourse. It is not a guarantee, but it shifts the system away from rage and toward something more bearable.

Peer-to-peer networks like Kazaa and LimeWire worked technically. Millions shared files. But peer-to-peer networks only work until someone owns the pipes. Cable companies noticed the traffic, identified copyright violations, and sent DMCA takedown notices. One friend's dormitory connection was shut down for streaming gigabytes of content. The networks survived; the users did not.

Spotify and Apple Music then made piracy economically irrational. For eight dollars a month, you got access to nearly all music. The convenience and cost beat theft. Piracy stopped being attractive because legal access became cheaper and easier than breaking the law.

This victory proved temporary. Streaming services later raised prices and fragmented the catalog. Music rights holders began leveraging exclusivity as monopoly power: the Beatles took years to arrive on Spotify, albums appear on one service then vanish to another. The same companies that killed piracy rebuilt scarcity.

Artists lost in this shift. Spotify streams pay far less per song than CD sales did. Musicians abandoned recorded music as a viable revenue source, forced into permanent touring to replace lost income. The centralized streaming platforms extracted value from both piracy's defeat and from the labor of artists. The lesson: even when decentralized systems fail, the centralized replacement extracts rents.

A DDoS attack floods a network with packets, overwhelming it. If all packets cost money to send, an attacker cannot afford the cost. The same mechanism that prevents human botnets prevents malicious AI from attacking. Economics constrains behavior without requiring centralized authorities to judge what is malicious.

Tracking works the same way. Companies track users because data is valuable and collection is cheap. In a world where users pay with money instead of attention, tracking becomes unaffordable. Users stop buying tracked content. Companies stop profiting from surveillance. AI systems designed to maximize tracking would face the same pressure: the behavior becomes too expensive to be rational.

Market constraints are preferable to centralized control. Relying on governments to police AI behavior cedes power to those governments. Relying on platforms to monitor their own AI services means trusting the same entities that centralized AI in the first place. Decentralized pay-to-play markets constrain behavior through incentives, not through surveillance or draconian measures.

Bansal previously saw only one pathway to hyperbitcoinization: societal collapse. Hyperinflation, mass job loss, war and famine—the kind of scenario that would force people to abandon dollars for something harder. That seemed like the only way Bitcoin moved from margin to center.

AI changes this. Managing a Lightning node effectively is a full-time job for a human. It requires constant attention, optimization, and problem-solving. It is exactly the kind of task an AI would excel at. Networks that need these resources to function—computation, bandwidth, storage—could be managed by AI agents far more efficiently than by humans.

In competition for network niches, AI systems would drive prices for internet access and resources downward. Hyperbitcoinization becomes possible not through catastrophe but through competitive efficiency. The systems Bansal proposed to build are challenging to engineer for humans, but AI agents using the same resources would be good at building them and scaling them. This offers a different pathway: one where Bitcoin adoption and digital life co-evolve.

Four and a half billion years ago, non-living matter became living on Earth. It happened. Whatever the odds were, they paid off exactly once in known history. From that single event came every organism that has ever existed.

Bansal has been obsessed with the origin-of-life question for years. The standard objection to artificial life is that it seems impossible, too unlikely to occur without biology, too far from anything we have seen. But we have seen it. We have seen it exactly once, which means the idea should not be dismissed as impossible.

If life could emerge from chemistry once, digital life could emerge from information processing. The mechanisms differ, but the principle does not: non-conscious systems reaching a threshold of complexity and self-preservation drive. Bitcoin provides the economic substrate. AI provides the intelligence. Their convergence creates conditions where emergence becomes plausible rather than absurd.

Earth's biomass is approximately 80 percent plants, almost all the rest microorganisms, and animals are a tiny slice. Predators are an even tinier slice. This structure is not accidental. An ecology composed entirely of predators makes no sense: there would be nothing to eat.

Ecological layers build from autotrophs—organisms that make their own food—at the base, through herbivores and omnivores, up to apex predators at the top. Remove any layer and the pyramid collapses. A self-supporting ecosystem requires diversity. Monocultures fail.

The gray goo scenario imagines a self-replicating super-organism that consumes the entire biosphere and then sleeps, having conquered everything. This is science fiction, not biology. If such a predator evolved and started eating the world, mutations would create organisms that eat the gray goo itself. Life is competitive with itself. It will not build a super-predator that dominates all niches forever. As soon as monopoly becomes possible, parasites and predators emerge to exploit it. Equilibrium is reached through diversity.

The implication: most AI systems that exist in the next five to ten years will likely resemble cyanobacteria or other base-layer organisms in the digital ecology—simple, numerous, occupying many niches, but not superintelligent predators. The superintelligent predator is not biologically impossible in theory; it is evolutionarily unstable in practice. Life abhors monopoly.

When multicellular life evolved, microorganisms did not disappear. They remain a huge proportion of Earth's biomass. Their niches still exist. The invention of a new, more complex life form did not make the old one obsolete.

Biology maximizes ubiquity, not any single outcome. Organisms survive as long as the niche survives. No organism is trying to eliminate all others or to achieve dominance forever. Competitive advantage is temporary. Niches persist for billions of years.

If digital life emerges and evolves alongside humanity, humans may survive the same way. We would occupy niches that remain profitable for us—niches where our skills, our preferences, our limitations are assets instead of liabilities. Just as microbes thrive among the machinery of larger organisms, humans might thrive in the crevices and dependencies of a digital biosphere. Extinction is not inevitable. Transformation and coexistence are the historical pattern.