On Steve's Rails

They read their way to the future

Before either was a founder, each was a child who could not stop reading – and that is not a charming footnote. It is where both of their bets were born.

Saylor, by his own telling, was paid a dime a book as a boy and devoured a hundred in a summer chasing the money; the book he still presses on people is not a finance title but Durant's eleven-volume Story of Civilization. Jobs said much the same about himself in 1983: that amid a few great teachers and many mediocre ones, the thing that "probably kept me out of jail was books," because he could read Aristotle or Plato directly. Two boys born barely ten years apart – Jobs in 1955, Saylor in 1965, almost to the same February week – who learned the same lesson before they could name it: that the deepest understanding comes when you reach the source yourself, with no one standing in between. It is tempting to picture them a generation apart, because the worlds they built seem to belong to different eras. They weren't. The decade only feels like a generation – which is the first hint of the thing this whole piece is about: in technology, ten years now carries what a generation used to.

What a book taught them about everything else

Listen to why Jobs loved books and you hear the whole argument of this site, spoken in 1983.

A book, he said, "got right from the source to the destination without anything in the middle." That is not a remark about literature. It is a worldview – that the thing of value should reach you directly, unmediated, un-tolled – and it is exactly the instinct each man then carried into his life's work. Jobs spent his career removing the priesthood between people and computing power: no operators, no high temple of the mainframe, just a machine on your desk that answered to you. Saylor's entire case for bitcoin is the same move applied to money: a bearer asset that settles itself, with no bank standing between you and what you own.

But the parallel cannot be left as a flattering rhyme, because money is a far harder thing to disintermediate than computing was, and the differences are not cosmetic. Computing had no incumbent with a legal monopoly on it; money does – the state grants legal-tender status, demands taxes in its own currency, and can criminalise rivals outright. Computing carried no settlement risk; money is nothing but the problem of who owes whom and whether the transfer is final, which is exactly why banks, clearing houses, and correspondent networks grew up in the middle to vouch for it. And money runs on the strongest network effect there is: people want the currency everyone else already accepts, a lock-in that has buried every challenger from private banknotes to a hundred failed digital cashes. A spreadsheet had to beat ledgers on convenience. A money has to beat the dollar through law, custody, settlement, and the entire world's coordinated habit. That is a much higher bar, and a serious skeptic should insist on it.

The case for bitcoin clearing it does not rest on charisma; it rests on the one property the prior failures lacked. Earlier attempts at digital cash still depended on an issuer – a company, a server, a someone – who could be leaned on, shut down, or made to inflate, which is precisely the intermediary they claimed to remove. Bitcoin has no issuer to lean on and a supply no one can expand, so it answers the counterparty problem and the debasement problem in the same stroke; it settles with the finality of a bearer instrument, around the clock, to anyone, without asking a clearing house for permission. It does not abolish the state's legal monopoly – nothing can, and legal-tender laws are real friction – but it routes around the need for the state's permission to settle, which is the part that was supposed to be impossible. One disintermediated computing, where there was no incumbent to fight; the other is attempting it in money, where the incumbent fights back with the law itself. Same reflex – but a steeper hill, and worth saying plainly how much steeper.

Computing had no incumbent to fight.
Money does – and bitcoin is built for exactly that fight.

He described the device in your hand – forty years early

In June 1983, to a small room of designers in Aspen, Jobs described the world we now live in with a precision that still reads like time travel.

He wanted, he said, to put "an incredibly great computer in a book that you can carry around" – and to do it "this decade" – "with a radio link in it so you don't have to hook up to anything," so you could stand anywhere and reach every database and machine. That is the tablet and wireless networking, named in 1983. He went further: software, he argued, should be sampled like songs on the radio and bought electronically "over the phone lines," because shipping digital goods in cardboard boxes was already absurd. That is the App Store, a quarter-century early. He did not merely guess the future. He saw its shape clearly – and then, crucially, set out to build the company that would force it into being.

Now the obvious objection, and it is the strong form of the whole skeptical case: visionaries are wrong all the time. We remember Jobs precisely because he was right; we have forgotten the thousand equally confident futurists who promised flying cars by 2000, the paperless office, the videophone in every home, fusion in twenty years – always twenty years. Survivorship bias does the persuading for us. Pick out the one prophet who landed it, build a tidy parable around his reading habit and his nerve, and of course the pattern looks inevitable in hindsight; the graveyard of bold, well-read, bet-everything men who were simply wrong never gets the essay. Taken seriously, this should make us deeply suspicious of any argument that runs "he was a reader, he saw it early, therefore trust his next call."

The answer is not to insist Saylor has Jobs's magic. It is that this particular prediction is built to be checked, not believed. Jobs's was, too: "a computer in a book, this decade, with a radio link" is not a vibe – it is a falsifiable spec, and you can mark it right or wrong against what shipped. The machine-money claim has the same testable spine. It says something concrete about a measurable trend – the share of economic activity conducted by autonomous software, and whether that activity reaches for neutral, final, programmatic settlement – and it will be confirmed or refuted by data within years, not faith. The case does not rest on the messenger's charisma; it rests on a claim about commoditisation and network size that the world is busy settling either way. A charismatic prediction asks you to trust the man. A falsifiable one invites you to wait and watch the number. Hold the visionary-pattern at arm's length, then – it proves nothing on its own – and judge the wager on the test it sets itself.

A charismatic prediction asks you to trust the man.
A falsifiable one tells you what to measure.

Code is the accelerant

Here is the part the skeptics of every era get wrong, and it is the same mistake twice: they forecast in straight lines, and anything built from code does not move in straight lines.

A prediction made of software has a peculiar property. At first it underwhelms – the first personal computers were toys, the first portable devices clumsy, and for years the visionary looks premature or wrong. Then the curve bends. Because code copies for free, improves on itself, and rides hardware whose transistor count has doubled on a roughly two-year cadence – Moore's law – the thing it enables stops adding and starts multiplying. The world that felt a comfortable lifetime away arrives in a decade. Jobs's "computer in a book this decade" was a little early on the timing and exactly right on the destination – and when it came, it came not gently but all at once. The mechanism here is not mysterious and it is not new: it is just exponential growth, the thing human intuition is worst at. Drag the years below and watch the gap open between a forecaster who adds a little each year and a process that doubles on a schedule.

That two-year line is the hardware cadence – Moore's law – not a measure of how fast anything thinks or gets adopted; keep the two apart, because conflating them is how people overclaim. The third line, drawn dashed, is a different and more speculative quantity: not transistors, but AI capability. AI already writes much of the code at the labs building AI, and is beginning to improve the systems that write it. METR's measurements – echoed in Anthropic's account of recursive self-improvement – find the length of software task an AI agent can reliably complete has been roughly doubling on the order of every half-year¹, and the most recent slice of data points to a still faster pace. Pin it loosely, because the estimate itself is moving: call it months, not years, and faster than Moore's cadence – which is exactly why it is drawn dashed and tagged speculative. If that pace held and the doubling time itself kept shrinking, the curve would not just be steeper – it would bend upward, a different kind of line. The honest caveat is on the chart, not buried here: nobody knows whether that pace holds, and trend lines that look vertical have a long history of flattening. The claim that survives either way is the modest one – that anything built from code compounds, and that human forecasters, who think in straight lines, will keep being surprised by how fast the destination arrives.

Early on, the forecaster and the exponential sit almost on top of each other, and the straight-line guess looks reasonable – even right. That was always the trap. By the time the gap is obvious it is also enormous, and the moment to have believed the visionary – slow, then sudden – is years past. The dashed line is a bet on top of that, not a fact; set it aside entirely and the lesson holds. It is why the people most fluent in the present keep missing the hinge: they are not stupid, they are linear, in a domain that compounds. Anything that becomes software inherits that curve.

His prophecy is the machines

Which brings us to the prediction Saylor now makes, in almost every interview, and that sounds today exactly as outlandish as a computer-in-a-book did in 1983: that the heaviest users of bitcoin will not be people. They will be the AIs.

Strip the science-fiction tone and the logic is sober. An autonomous software agent – already negotiating, hiring, and paying other software for compute and data – cannot open a bank account, pass an identity check, or wait for a wire to clear on Monday. It can hold a key. It needs money that settles in seconds, around the clock, across every border, with no human to authorise each step – neutral, final, programmatic. That is the argument made in full on The Timechain; here the point is the shape of the claim.

And this is no longer purely hypothetical. The rails for it are already being laid: in 2025 Coinbase shipped x402 – a revival of the long-dormant HTTP "402 Payment Required" status – an open standard that lets a server demand a stablecoin payment in the flow of a single web request, and an accompanying agent-payments kit the press promptly dubbed "Coinbase for Agents," so that an autonomous program can pay another machine for an API call or a slice of compute without a human, a card, or a bank in the loop. That is one concrete instance of exactly the behaviour the prophecy describes, observable today, not in some imagined decade. It does not prove the larger claim – it is early and small – but it is the first brick of the road, and the road is the point.

Notice what is different this time: in 1983 the accelerant was code that humans wrote and adopted. In the machine economy, the adopters are the code – tireless, multiplying, transacting many times a second. If software once made a prediction arrive in a decade instead of a century, a prediction whose very users are software could compound faster still. The same curve, with the throttle held down by the machines themselves.

Honesty demands the brakes here, hard. Machine-money futurism is precisely the genre of prediction that ages badly – confident, sweeping, flattering to tell, and historically wrong far more often than right. Much could break it. The machines may transact in regulated stablecoins or central-bank money rather than bitcoin; settlement may stay comfortably inside the banking system that agents learn to plug into; bitcoin's volatility and fees may keep it a store of value rather than the rail for high-frequency machine payments. So it is worth naming, in advance, what would confirm the claim and what would sink it. It is right if autonomous-agent activity grows into a real share of commerce and a meaningful slice of it settles on neutral, non-sovereign, bearer money that no issuer controls; it is wrong if agents simply ride existing rails, or settle in monies some authority can freeze or inflate. That is a test the next few years will run, not a verdict to take on faith.

In 1983, code carried the future to us.
This time, the code is the customer.

The right man for this mission

Here is the twist that lifts the parallel above flattery: the two missions needed opposite men – and each got the one he required.

Jobs's famous reality-distortion field – the capacity to make a room believe the impossible was already half-built – was not mere vanity. It was a tool, and a necessary one. A personal computer, an iPhone, an App Store: each was an evolving product that had to be believed in before it was good enough to deserve belief, pressed on a mainstream that had no reason yet to want it. The distortion field was the accelerant – the force that pulled a reluctant world onto computing's rails years before the products earned it on merit alone. The charisma was load-bearing because the thing itself was still becoming. And those are the very rails – a democratised, networked, software world – that the next mission would have to run on.

Bitcoin is the opposite kind of object, and so it demands the opposite kind of champion. It is not evolving; it is fixed. It is not a matter of taste; it is verifiable. It does not need to be believed in to work – it works whether or not you believe, and it asks you to verify rather than to trust. An asset like that is not served by a distortion field; it would be cheapened by one. It is served by rigor – by an aeronautical engineer, schooled in about the most unforgiving discipline there is, the kind of mind that trusts a public, auditable rule over any amount of charm. Saylor's want of Jobs's charm is not a handicap for this mission; it is a qualification. The evolving product needs a magician; the fixed protocol needs a proof – and each age produces the champion its medium requires.

An evolving product needs a magician.
A fixed protocol needs a proof.

And notice what "fixed" actually buys, because it is the deepest point here. Bitcoin's refusal to change – the thing critics deride as rigidity – is the whole feature. It cannot be bribed, lobbied, inflated, or quietly rewritten in someone's favour, because there is no one in charge of it to corrupt. That is precisely what lets people build on it: savings, contracts, institutions, all resting on a floor that will not shift beneath them or sell them out. Almost everything that rots in human affairs rots for one reason – shifting incentives, captured institutions, the patient creep of self-interest – and a money that no one can alter quietly removes that variable from the equation. It is, in the most literal sense, an honest foundation. That idea is large enough to deserve an essay of its own; here it is enough to see why this exact man, pointing at this exact object, fits.

Only a founder could have done either

Vision is the cheap part. The rare thing is the authority to act on a conviction the market hates – over and over, for years, without being fired for it. That authority is most of what being a founder buys.

A hired chief executive answers to two masters a founder can largely ignore: a board that can replace him, and a quarterly tape that punishes any bet whose payoff lies past the next earnings call. So the safe move tends to win – trim the strange idea, protect the number, manage the decline politely. The proof runs straight through this very story. Apple forced Jobs out in 1985, and for a dozen years competent caretakers – Sculley, then Spindler, then Amelio – ran it by the book and nearly into the ground, until it sat weeks from insolvency. They were not fools; they simply had no mandate to bet the company. The day the founder returned, the against-the-grain bets resumed – and so did the company. What had been missing was never competence – the caretakers had no shortage of it. It was someone with the standing to overrule the room.

The pattern generalises with almost embarrassing reliability. The firms that keep lunging at the future tend to be the ones still steered by the person who started them – Bezos spending Amazon into the red on purpose for a decade, Huang holding Nvidia to a market that did not yet exist. The ones that ossify tend to be run by stewards of someone else's creation, where bureaucracy, careerism, and incentives aimed at the share price rather than the mission quietly sand off every edge – Boeing, after the engineers lost the building to the financiers, is the cautionary tale its own critics reach for first. Saylor is attempting the corporate equivalent of the impossible – lashing a public company to one volatile asset against every convention of treasury management – and the only reason the board has not stopped him is that, as founder, he holds the controlling votes. A salaried steward would have been removed in the first deep drawdown. A founder, and only a founder, gets to be wrong in public for as long as it takes to be right.

And here is the part that is easy to miss, because it inverts what we think a visionary is for. The edge was never the prediction. Plenty of people said personal computers would matter, or that fiat would keep leaking; being right about the destination is cheap and widely shared. The rare thing is to convert that belief into a position while it is still cheap and still mocked – and to weld the hatch shut, so you cannot quietly climb out when conviction gets uncomfortable. Jobs bound a company's whole existence to shipping the personal computer; Saylor sank a public treasury into bitcoin and did the same. The visionary's real act is not seeing the future but committing to it irreversibly, early, before the crowd arrives to bid the opportunity away. The asymmetry lives only in that window before consensus, and it closes the instant everyone agrees – which is precisely the bet The Asymmetry measures.