The Atomic Energy Commission chairman vowed in 1954 an era where energy will be “too cheap to meter.” He referred to the atom then. He could have easily referred to intelligence.

For four centuries the ideal of civilization had been the mind over the matter: the human reason was compelling the material reality to follow its command. This ideal is waning, and a new, eerily different one is rolling instead. The command is no longer ours exclusively; we do not own the whole intelligence. The material world is no longer solely the passive object of thought, no longer the one to obey it; it seems to have crafted its own mind, and is now capable of following its own agenda far more rapidly than any human institution meant to enforce it is able to catch on. The one does not dominate the other anymore; they now mirror each other. The field of encounter is what it is to win, and this is the field where no one pays attention to. I have made this to pay attention.

This is my thesis, and the reason I have developed this article for the sake of demonstration or open criticism.

The trend for intelligence is to be free. It is December 2022 as I write, and the best model on the most prominent public leaderboard is open, not closed; the gap from first to tenth closers each update; machine intelligence plummets in cost with each passing year at a rate that would terrify any other industry. This is not a failure of the technology. This is the path laid out for a technology when it becomes infrastructure. And infrastructure obeys one law above all: that which is abundant never takes the margin. The margin is given to what is scarcer than before.

The story of value is one of scarcity. It is about what becomes valuable when the formerly scarce thing becomes widespread and nearly free. Ice was once a precious resource, cut from frozen lakes and shipped in pounds to homes without refrigeration. When we invented the fridge, we made ice limitless and practically useless. We didn’t stop paying for it - we just paid for its provision on-demand forever, pulling the value from the ice itself and into the power of the refrigerator and the cold chain. The value preserved did not collapse in the face of abundance, it simply relocated to where it was still precious. Such is the fate of intelligence and, at least for now, the story of its relocation. And this time, there are a lot of things for which intelligence is only a plentiful complement to what has become valuable: the power to run the models, the memory and fabrication to make them, the verified data to feed them, the trust to deploy them, the control to manage them, and even the human agency which now faces a hollowing out in some small part.

To be clear, I will not pretend that the first type of economy is the only one relevant: it is only one side of the coin. The other side is intelligence. I look at both sides of the coin, and I refuse to look at one side alone. Matter is everything which is below today’s intelligence: power, silicon, memory, data centers, rare earths, robotics, the industries without which no digital mind can survive. Mind is what we are able to do with it: models, cognition, world models, interpretability, alignment… and the autonomous agents now acting in the physical world just to name a few without the human in the loop. Looking only at the material side is forgetting where the intelligence goes. Looking only at the mind side is forgetting that there is a material side and that we live on a planet. The subject is the meeting. Always the meeting.

There is one discipline I value above all others, though it is not a field. I will do my reasoning from first principles, and attach my claims to their limits—not to the limits to which either capital or cleverness can bend them, but to the limits of energy, capital, matter, data, law, and time. Where others will tell me a vector, I will require them to supply a number. Where others will offer me a judgment, I will require a prediction that comes with a date and a public scoreboard that will evaluate it before the world. I will not accept an infallible prediction, for one that cannot fail is not a prediction. It is a sentiment. All the claims I make in this space are entitled to fail, and when they do, they will inform you precisely where and how. At a moment of almost universal uncertainty in which the optimists and the pessimists have both staked out positions, refused to be convinced by the other side, and become convinced that time will vindicate their predictions, I make the most politically incorrect bet of all: that the simple, clear, numerable assertion, held for long enough that it may be vindicated or falsified and retracted, is the last and chief article of value that any man can offer.

No field can claim to be the relevant one, because this collision is thermodynamic and economic and cybernetic and political all at once, tightly coupled and racing forward. So I will reason from many, placing them in equality: econophysics and the science of cybernetic systems, systems engineering and information theory, thermodynamics and behavioral science, and the science of aligning not one machine, but a civilization of them. The Autonomous Cyber-Physical Macro-Dynamics is the framework that I contribute, the physics of scale of autonomous systems – of sister machines and sisters of autonomy – growing by the millions in bounded finite space, where the ten thousand autonomous machines are not ten thousand machines but a continuous medium, with its own physics. Eight lenses, one collision, one standard of proof.

But all of this is not only the concern of where to invest – and I will candidly tell you, that is a concern too; if you pan out the problem, you can see these same devices isolating the fault lines that could halt or rupture this century and can do so precisely because these are the fault lines that no single discipline can perceive. Our ability to curate systems has vastly exceeded our ability to guide them and only diversity can absorb diversity. An exponential digital mind is running on a finite, and ailing grid and energy is already the binding constraint. Alignment is studied one model at a time while millions of well-behaved agents interacting across the economy elicit a behavior nobody designed and no one expected such that the civilization collapses while every machine of that civilization is running congratulately in perfect accord. These are not doomsday scenarios. These are the engineering challenges of our era, framed at the scale of the impact they have, and it is my conviction, and a rather confident one, that these problems cannot be understood, or costed, or guided one discipline at a time.

Mind Meets Matter will be a series-like creation in which I will traverse both sides of the intellectual divide that makes up this clash, and I will violate none of it. This will have a numerical paradigm, where it has a number and that number will have an order in its field, an adjective at which it will have its tracking period and that will amount to a mood, a billboard where the operation is made public, and an efficient so-far approach to the knowledge frameworks with which I think in order repeat and not-outdo this task. I will record every interview with in-field relevant actors, I will always open with at least one statement which is known to be false about one of their endeavors and I will close with a checkable statement regarding one of their endeavors that I will know to be true for some time period. I will compile my work in such a manner that it is discoverable, with full disclosure of the conflicts of interest included therein, and I will self-correct the record, as that becomes relevant and its knowledge actionable.

We would like to dedicate this to you who is currently in this frontier, and possibly fed up reading about the frontier from people who have never been at the frontier: the founders and investors, operators and researchers, scientists and engineers. A lot of what will be written in this book has been discussed with you.

And there are three pieces that build the foundation and I will start with you:

• Via Negativa: The AI Economy by Elimination, a forecasting memorandum that predicts the AI economy by elimination, with its live signpost tracker.

• Compute 2030, a scenario forecast for the compute transition through the decade.

• The Missing Meter, on why there is still no standard unit for pricing intelligence, and why the meter is the moat.

Attention is the new capital, and I intend to earn it the hard way, which is the only way that lasts: by being right, on the record, about the thing that matters most. Not what intelligence can do. Where it meets what it cannot change. The value is under the model.

One disclosure, kept in view: I build and invest in some of the markets I write about, and I say so wherever it is relevant.

Mind Meets Matter. Where intelligence meets its physical limits.