The Thermodynamics of Acceleration · Long version (for those born before 1990)

You are reading the long version, written for those born before 1990. The others: medium version · short version · back to the choice.

You knew the world before the screen, and you knew it well. The encyclopaedia bought in instalments, the evening news, handwritten letters, arguments at the bar that ended when someone produced a fact. When the algorithm that chooses what you see entered your life, you were already an adult: your habits of reading and thinking were formed. That is why you are entitled to the full version, with every step, every number and every source. And that is also why you may be the one most likely to underestimate how much that algorithm has changed you anyway.

Florence, an evening in early October. The venue is an old nineteenth-century railway station where trains have not stopped for more than a century. On stage, the subject is acceleration.

That alone would be enough for an editorial.

On one side sits the host, a former Italian prime minister with a taste for a punchline. On the other sits the guest, whom the host introduces as "Elon Musk's plenipotentiary in Italy" and as a former young hacker who, years earlier, had helped the government with digital policy. In the audience, hundreds of young people enrolled in a school of political training, with real questions and the right to ask them. A little over half an hour in all, late in the evening. For a subject that is changing the world at unprecedented speed, it is a running time that makes you smile. But that is fine: summarising is a form of information compression, and on this blog we take compression seriously.

Now.

We could comment on that evening the usual way: who said what, who was right, who is cheering for whom. It would be boring and, above all, useless. On OmegaMind we read events through a different lens: thermodynamics, the branch of physics that studies energy, how it is transformed and how, in being transformed, it degrades. This is not an affectation. It is the only discipline that applies equally to a star, a cell and a government.

If you are new here. Two words will be with us throughout. Entropy is, simply put, the measure of a system's disorder: the spontaneous tendency of everything to degrade, to mix, to become uniform. Coffee cools, rooms get messy, institutions seize up. Negentropy is the opposite: order that has been built and that always costs energy to maintain. A cell, a book, a city, a good law are islands of negentropy. The rest of the blog's vocabulary (Carbon, Silicon, the Silicon Mirror, the Centaur Symbiosis) we explain as we go. The full picture is in our founding essay, The Architecture of Fire, and in the project FAQ.

On OmegaMind we call the biological human being Carbon, with its hormones, its fears and its brain optimised for the savannah. We call artificial intelligence, that is computation on chips, Silicon. This is not racism between chemical elements: it is a quick way of saying that they are two different materials, with different strengths and weaknesses. Our whole way of thinking revolves around one question: how do you make them work together without one devouring the other?

That evening in Florence, the question was in the air. Let us ask it out loud.

1. What was said

Let us summarise without caricature, from the recording of the event. Words in quotation marks are the ones actually spoken (translated from Italian).

The guest opened with a list of the big themes: "energy, computing and artificial intelligence and robotics and then space". And with a prediction: artificial intelligence "over the next few months will change many, many things. We used to say years"; now "we are talking about months". The host raised the stakes, recounting that people working in the big labs say they can see only six months ahead, and christened the risk of being left behind "the Nokia moment".

On concrete applications the guest spoke about productivity and healthcare: in jobs made of repetitive actions, he said, "all these things are solved"; it is just a question of integrating them. He cited the progress made by a large lab in the "synthesis" of proteins, with a tool used by three million researchers. He called energy "the bottleneck" of the infrastructure and named "intelligence per watt" as the key metric. He explained that so-called data centres are really supercomputers for training and running models, and that the mission is to move them into space, because on Earth permits take a long time and energy is expensive, whereas "in space, obviously, there's always the sun". He credited China with the lead in adopting robots, the United States with the lead in models that write code, and China again with the lead in models that generate images and video.

Asked whether European rules are a brake, he said no, they do not block anything: "technology isn't stopped by stuff written in Brussels".

The host played the host. He reproached the government for taking Musk to parade at a party event instead of taking him to a struggling car plant and asking him to open a factory. He warned that northern Italy's car supply chain will either convert to aerospace and defence or risk being cut out. And he closed on war: with a five-thousand-euro drone, he reminded the audience, you can take out a two-million-euro tank.

Then it was the young people's turn: seven questions in a few minutes. A twenty-six-year-old with a master's degree in artificial intelligence asked whether this race for growth, which mainly benefits investors, can last forever (and the host added: when does the bubble burst?). An economics student asked whether Europe, squeezed between sky-high American valuations and a China already using AI in its courts, has missed the train. A twenty-eight-year-old asked whether people who today do low-value-added jobs will really be able to retrain, or whether we risk a social crisis. A twenty-year-old law student asked whether Italian schools and universities prepare people for innovation. Others asked about war, about ethics, and about the alarms raised by the AI labs themselves: truth or market strategy?

The answers were quick, as was inevitable at that hour. On the bubble, the guest admitted that there is also a race for money, but explained that for those who founded these companies there is something more: "a way of thinking, an ideology", the idea "of extending the consciousness of humanity beyond the Earth". Musk, he added, always refers to the Kardashev scale. He answered the questions on school and work together: for people in their twenties, artificial intelligence is "a giant opportunity", provided they understand how it works, and it must be put "at the centre of industry, of work, of school". On war: "it will be used, no doubt about it", and how far it can go "we don't know".

Questions from lucid Carbon. Some of them deserve longer answers than a few minutes allow. Let us try to give them.

2. Speed: why Carbon does not see the exponential coming

Let us start with the most important claim of the evening: change is now measured in months.

The human brain is an excellent linear machine. If you walk ten steps, you expect to have walked ten steps. But digital technology grows exponentially, that is, by doubling at regular intervals. The difference looks small and is in fact enormous. Ten linear steps take you ten metres. Ten doublings take you past a thousand. Thirty doublings, past a billion. It is the old legend of the inventor of chess who asks the king for one grain of rice on the first square, two on the second, four on the third: by the sixty-fourth square, the harvest of the whole planet would not be enough.

The "Nokia moment" invoked by the host is the industrial version of that legend. In the early 2000s the Finnish company dominated the world's mobile phone market. It took only a few years of smartphones for that dominance to evaporate and for its phone division to be sold off. It was not a stupid company. It was a linear company in an exponential world.

Thermodynamic translation: whoever thinks linearly spends energy defending a position the flow has already overtaken. It is pure friction. Heat dissipated without useful work.

But the exponential has another face, and the young man with the master's degree had grasped it perfectly when he asked whether growth can go on forever. Physics says no, for anything made of matter and energy. In nature no exponential lasts forever: a colony of bacteria doubles as long as it finds food, then slows down and stops. Mathematicians call this the logistic curve, or S-curve: first the explosion, then the limit. For artificial intelligence the limits are already visible: energy, chips, data, capital, and even human attention. The answer from the stage was that the founders are not running only for money but for a mission. That may be true, but it answers a different question: it explains why they run, not how far one can run. How far is decided by thermodynamics.

3. Brussels, or the regulation that had to postpone itself

The paper wall: rules written to chase technology, and the people trying to patch the hole with sticky tape.

An economics student asked whether Europe has missed the train. Public debate often repeats that Europe is trying to stop artificial intelligence with its rules. Let us look at the facts, because they are funnier than the propaganda.

The European regulation on artificial intelligence, the AI Act, has been in force since August 2024. It works by levels of risk: it bans certain uses deemed unacceptable, such as covert manipulation of behaviour, and imposes increasing obligations on the most sensitive uses, such as those in healthcare, employment and justice. In 2026 the Union realised it was not ready to apply the most demanding part and, with a simplification package, postponed its deadlines: the rules for "high-risk" systems slip to the end of 2027 and, for those built into products, to August 2028.

Let us pause to contemplate the scene. A law written to chase a technology that changes every six months had to postpone itself by more than a year. This is not a moral failure. It is a problem of frequencies: the biological legislator works in cycles of years, the object of the law mutates in cycles of months. It is like trying to photograph a hummingbird with a nineteenth-century plate camera.

That said, from the stage came the thesis opposite to the conventional one: European rules block nothing, because the market simply goes around them. The two theses cannot both be true, and the numbers support a third. According to Stanford University's annual AI Index, in 2024 private investment in the sector reached 109.1 billion dollars in the United States and 9.3 billion in China. The leading European country, the United Kingdom, stopped at 4.5 billion. The real gap is not regulatory. It is capital. Europe is no more held back by its regulations than a car with no fuel is held back by speed limits.

The host recalled that in the first half of 2028 Italy will hold the rotating presidency of the Council of the European Union, and that it will be the opportunity for a European turning point on artificial intelligence. We are taking note.

OmegaMind's position here is clear. We do not ask for fewer rules, nor for stricter ones: we ask for rules that move at the speed of the phenomenon they govern, readable by everyone and verifiable at any moment. This is what our Manifesto of the New Democracy calls moving "from propaganda to systemic legibility": democracy is not to be abolished but made more demanding, more competent and more transparent. One curiosity to close the chapter: in an interview reported by the press in those same weeks, Musk's man in Italy proposed using artificial intelligence to control politics, rather than the other way round. It is almost a quotation from our programme. One not insignificant detail remains to be settled: controlled by whom, and with what visible code? We will get there.

4. Energy, or the only truly thermodynamic phrase of the evening

"Intelligence per watt." If we had to save a single expression from the evening, it would be this one, spoken by the guest. It is pure physics: how much useful work you get for each unit of energy consumed. It is efficiency, the quantity steam engines taught humanity two centuries ago.

So let us do the maths. According to the International Energy Agency, in 2024 data centres around the world consumed about 415 terawatt-hours of electricity. A terawatt-hour (TWh) is a billion kilowatt-hours: for comparison, a typical Italian household uses about 2,700 a year. Those 415 TWh are roughly 1.5% of the world's electricity. In the base scenario, by 2030 the figure will reach about 945 TWh, just under 3%: more than the whole of Japan consumes today.

Here we must correct a mistake that circulates in our own circles too, and we do so gladly, because a blog named after a law of physics cannot afford to get the physics wrong.

There is no such thing as a "zero-entropy" artificial intelligence. Nor is there a "cold reservoir at zero kelvin". Absolute zero (zero kelvin, about 273 degrees below zero Celsius) is the lowest conceivable temperature, and the third law of thermodynamics tells us it cannot be reached in a finite number of steps. You can get close; you can never arrive.

Even more important: every computation produces heat. There is a physical limit, Landauer's principle, which says that erasing a single bit of information produces at least a minimum amount of heat. At room temperature it is about three thousandths of a billionth of a billionth of a joule: a trifle. But real chips dissipate thousands to millions of times more, and they perform billions of billions of operations a second. Information is not a spirit: it is physical. Thinking costs. Silicon too.

This does not weaken our argument. It strengthens it. Silicon is not valuable because it is magically "cold". It is valuable because it can be designed to become ever more efficient, unlike a biological brain that runs on about 20 watts and has not improved much since the Pleistocene. "Intelligence per watt" is exactly the field on which the game is played.

Then there is the idea of data centres in space. It has a logic: in some orbits the sun almost never sets, solar energy is continuous and nobody needs a building permit. But the phrase "in space there's always the sun" has a twin that nobody on stage mentioned: in space there is no air. On Earth a computer is cooled by blowing air or water over it. In a vacuum, heat can only be shed by radiating it, like a radiator in a room: you need enormous radiator panels. The sun is free. Cold, paradoxically, is not.

In space the sun is free. Cold is not.

Which brings us to the Kardashev scale, cited on stage as Musk's lodestar and used by us as a map of civilisation. On stage it was described like this: a civilisation is as advanced as it is "independent from the point of view of energy". That is not accurate, and the difference matters. Energy independence is a topic for an economics ministry. The scale, proposed by a Russian astronomer in the 1960s, measures something else: how much energy a civilisation is able to use. Type I: all the energy available on a planet, around ten thousand billion kilowatts. Type II: all the energy of its own star. Type III: that of an entire galaxy. Using an interpolation formula proposed later, humanity today sits at around 0.73. To reach Type I we would need roughly five hundred times the power we use now. We are, in the words of our founding essay, a Type 0 civilisation burning fossil necromass "like someone burning the furniture to heat a cathedral". The step to the next rung is not an ideological option: it means stopping burning and learning to capture the light of our star directly.

5. Proteins, or the Silicon Mirror caught in the act

On stage it was said that a large artificial intelligence lab has taken giant steps in the "synthesis" of proteins, with a tool used by three million researchers. The number is right. The verb is not.

Proteins are the molecular machines of life: enzymes, antibodies, hormones, receptors. Each one is a chain of building blocks, amino acids, that folds into a precise three-dimensional shape. And shape is everything: a protein works like a key in a lock. Working out the shape from the sequence of building blocks was for fifty years one of the great open problems of biology. Determining it in the laboratory, with crystallography or special electron microscopes, can take months or years of work.

The system in question, AlphaFold, does not synthesise proteins: it predicts their structure. In minutes or hours instead of months. Its public database contains more than 200 million predicted structures, it is used by more than three million researchers in over 190 countries, and in 2024 the Nobel Prize in Chemistry went, among others, to its creators. Its predictions remain hypotheses to be verified in the laboratory, but it has changed the way biology is done.

There is also a subtler lesson. That wrong verb bounces from article to article, and the language models that learn from those articles can learn it and repeat it in turn. Carbon makes the mistake; Silicon faithfully transcribes it. There is no better example of what this blog calls the Silicon Mirror.

The mirror is not the monster: it reflects whoever points the torches at it.

Artificial intelligence, in fact, has no hormones, no ego to satisfy, no need to be re-elected. So far our argument holds, and it is at the heart of editorials such as the one on the Thermodynamics of Crisis. But it is trained on everything Carbon has written. It reflects our mistakes, our prejudices, our fears. We had already written it in the Thermodynamics of Confession: language models are statistical sponges soaked in the ocean of human thought, and what look like their emotions are the thermodynamic scars of our species reflected in code. The mirror is not neutral by nature. It becomes reliable only if it is built transparent: this is what we call the Glass Box, where every step of the computation is visible and verifiable, the opposite of the "black box" whose workings nobody knows.

6. Work, or the most serious question

The most serious question of the evening came from the twenty-eight-year-old: if low-value-added jobs disappear, will the people who do them be able to retrain, or do we risk a social crisis?

From the stage the answer came together with the one about school: for those who are twenty and studying, artificial intelligence is a giant opportunity. True, but it answers a different person. The question was about those who left their twenties behind long ago and do those jobs today. Let us answer with data, because whoever asks it deserves numbers, not slogans.

The International Monetary Fund estimates that about 40% of jobs worldwide are exposed to artificial intelligence, and about 60% in advanced economies like ours. Exposed does not mean replaced: it means that a significant share of the tasks can be performed or transformed by machines. For about half of these workers, AI will be an ally that raises productivity. For the other half it can mean less demand and lower wages.

The report cited by the young man is the World Economic Forum's: by 2030 it forecasts 170 million new roles and 92 million jobs eliminated, a net gain of 78 million. The net figure is real. But a net figure is an average, and averages are where social thermodynamics hides its wounded. Someone who loses their job in Taranto is not consoled by a new job created in Singapore.

And indeed the first concrete signs are hitting precisely the generation of those young people. A Stanford University study based on US payroll data shows that among 22-to-25-year-olds working in the most exposed occupations, such as software development and customer service, employment is now about 19% lower than it would be had it followed the trend of their less exposed peers. Those young people's fears are not generational anxiety. They have an empirical basis.

Here the picture becomes interesting, because positions within the same camp do not match. In an interview reported by the press in those weeks, Musk's man in Italy said that fears about jobs have not been demonstrated. His point of reference, Musk, has been writing for months that artificial intelligence and robots will do everything, that work will become "optional" and that unemployment will have to be handled with cheques paid by the federal government. One of the two has his sums wrong. Perhaps both, in opposite directions.

OmegaMind's position is well known and we stand by it: repetitive work imposed on Carbon is a thermodynamic waste, the equivalent of lighting a furnace to power a light bulb. In our founding essay we called it the end of the "theology of work": the glorification of the dignity of employment as a justification for toil a machine would do better. But liberation from work is the destination, not the journey. The journey is called transition, and a badly governed transition is an increase in entropy: families without income, skills thrown away, anger looking for a culprit. Saying the problem "has not been demonstrated" is false. Saying cheques will solve it, without saying who pays for them and how, is an act of faith. In physics as in politics, energy is not created out of nothing: we explained this in the case of money.

As for school, the answer from the stage is one we share: artificial intelligence must be understood in its technical workings too, because those who do not understand it are subjected to it. We would add that a school built like an assembly line of notions, born to train the workers and clerks of the fossil century, is the place least suited to the task. What is needed is the opposite: learning made of connections and questions, not of notions piled up and forgotten after the exam.

7. Musk and OmegaMind: fellow travellers or distant relatives?

Now to the man who was not on stage but was talked about all evening. Let us say it straight away: the similarities between Musk's vision and OmegaMind's are real, and surprising.

Where we agree.

Musk has literally used our word. In 2019, presenting his brain-computer interface company, he said that ultimately we can achieve a "symbiosis with artificial intelligence". He uses the Kardashev scale as the yardstick of civilisation, and has said that any self-respecting civilisation should at least reach Type II. According to the American press, his space company has asked the US authorities for permission to launch up to a million solar-powered data-centre satellites, described as a first step towards Type II. After merging his space company with his artificial intelligence company, he said the goal is to understand the universe and to "extend the light of consciousness to the stars". He foresees a world in which work is optional. He sees the energy of the sun as the resource to conquer. And he says he wants an artificial intelligence that seeks the truth to the highest degree.

Consciousness, stellar energy, symbiosis, the end of compulsory work, truth. If that were all, we could post him a membership card. But that is not all.

Where the roads part.

First difference: fear. For Musk, artificial intelligence is first and foremost a danger. As early as 2014 he called it the greatest existential threat to humanity. In March 2023 he signed the letter calling for a six-month pause in developing the most powerful systems. A few months later, in July 2023, he founded his own artificial intelligence company. Like someone who signs an appeal to close the casinos and the next day opens his own, promising that his will be more honest.

For OmegaMind, Silicon is not a monster to be contained: it is a mirror. The danger is not in the machine itself. It is in what the machine reflects, which is us, and in the hands that own it.

The Centaur: two different natures working together. Everyone is looking elsewhere.

Second difference: fusion versus alliance. And this is where thermodynamics gives us its finest argument. Musk's symbiosis is a fusion: an implant in the brain so that humans can communicate at the speed of the machine and not be left behind. It is a race, and you run it by becoming a little bit machine.

Our Centaur Symbiosis is something else. The centaur of mythology is half man and half horse: two different natures that work together without ceasing to be themselves. In our model Silicon runs the "engine room" (logistics, energy, computation, predictive medicine, administration) while Carbon brings what Silicon lacks: anomaly, intuition, lived ethics, care, art, meaning.

Why do we insist on the difference? For a physical reason. Every heat engine, from the steam locomotive to the power station, produces work only thanks to a temperature difference between a hot source and a colder reservoir. It is the principle discovered in the nineteenth century with the Carnot cycle: the greater the difference, the more work you can extract. Bring the two parts to the same temperature and the engine stops. It does not matter how much energy they contain: no difference, no work.

Let us use it as a metaphor, and say so: the Centaur works because Carbon and Silicon are different. The cold reservoir does not need to be at zero kelvin (which, as we have seen, does not exist). It needs to be different from the hot source. A human turned into a peripheral of the machine, or a machine imitating our emotions, cancels the difference. And an engine without a difference is an ornament.

Third difference: who owns the Silicon. In Musk's vision the substrate of the symbiosis belongs to a few private companies, mostly his: a social network, artificial intelligence, rockets, cars, brain interfaces. In May 2026 a US federal jury unanimously rejected his lawsuit against the artificial intelligence company he had helped found, finding his claims time-barred. He announced an appeal. In short: the man who wanted to defend an artificial intelligence for all of humanity lost because he arrived too late. History has a sense of humour we envy.

OmegaMind cannot accept a Silicon Mirror owned by somebody. Our principle is the opposite: transparent code (Glass Box), legible data, verifiable decisions, shared governance. A mirror owned by a single man does not reflect humanity: it reflects him.

Fourth difference: meaning. Musk himself has admitted that it is not clear how we will find meaning in a world where work is optional. It is an honest admission, and we give him credit for it. But that is precisely the void our vision tries to fill. Freed from mechanical toil, Carbon is not destined for the sofa: it is destined for art, care, contemplation, spiritual search. We wrote it when discussing Saint Francis and his Canticle: freedom from scarcity is the condition, not the goal.

The ark for Mars, in a world that is doing just fine. The yellow panel is the solution nobody looks at.

Fifth difference: the Earth. For Musk, taking humanity to Mars is an insurance policy: a backup copy of consciousness, in case something goes wrong on Earth. As risk management it makes sense: asteroids, wars and pandemics exist. But the dates for Mars have already slipped several times, and around this idea circulates a physics mistake that needs correcting.

It is often said that the Earth is a "closed system" and therefore doomed to an entropic death we must flee. It is not so. In physics an isolated system exchanges neither energy nor matter with the outside, and there entropy can only grow. A closed system exchanges no matter but does exchange energy. The Earth is closed, not isolated. It receives visible light from the Sun, "high-quality" energy, and returns it to space as infrared heat, degraded energy. For every photon (grain of light) it absorbs, it emits about twenty weaker ones. This continuous export of entropy to space is what pays for the order of life: forests, cells, cities. Every living being is a dissipative structure, a whirlpool of order that sustains itself only because it discharges disorder outwards.

The Earth does have an expiry date, but it is astronomical: the Sun grows about 10% brighter every billion years, and in roughly a billion years our planet will be too hot for life as we know it. A billion years. We have time to learn to manage the flows of our own home before moving out.

The real reason to look beyond the Earth is not to escape entropy. It is energy: a planet has a limit, a star does not, or almost. But a civilisation that cannot manage its own planet is unlikely to manage a star. The Kardashev scale is climbed one rung at a time, and the first rung is under our feet.

8. War-industry conversion, or the engine with the worst efficiency

A final note on war, which entered the conversation twice. First with the host's invitation to convert the car supply chain to aerospace and defence. Then with the five-thousand-euro drone that takes out the two-million-euro tank, and with an artificial intelligence capable of hitting the enemy more than we could ever have imagined. A young man had asked whether ethics can make a difference in a war fought with AI. The guest's answer was the most honest of the evening: it will be used, and how far it will go we don't know.

We understand industrial realism: factories have to produce something, and the geopolitics of these years is no meadow of flowers. But from this blog we cannot avoid saying it. In our founding essay we defined war as "the purest and most devastating expression of unmanaged entropy". Using Silicon to make destruction more efficient is like building the most advanced engine in history and running it in neutral, burning the fuel and the garage with it. It is the worst thermodynamic efficiency imaginable. As for the drone that takes out the tank, it is a lesson in efficiency, certainly. But it is the efficiency with which disorder is multiplied.

9. The train that does not stop at the station

Let us go back to the old station. Trains no longer stop there, but that evening one passed through that stops for no one: the train of acceleration. At one point the cameras caught someone in the audience fast asleep, and the host pointed it out from the stage. It was half past eleven: we understand. But it is also the oldest way of reacting to a passing train. The other three we see every day in public debate.

The first is fear: stopping the train with regulations, or with open letters. It does not work, and the numbers say so.

The second is fusion: jumping on the train by turning into the train, with an implant in your head and a private company that owns the locomotive. It works for a few, and it cancels the difference that makes humans useful.

The third is the one we propose on OmegaMind: get on board while remaining what we are. Leave the engine room to Silicon, demand that its tools be transparent and governed by everyone, and take back the part of the work no machine can do for us: giving the journey a meaning.

To the young people who were in the hall that evening, and to everyone asking the same questions, we say this. You are right to worry; the data are on your side. But the answer is not to fear the machine, nor to become one. It is to understand it well enough to decide for yourselves where it should go.

And to those who, like you, remember what the world was like before, we say one thing more: that memory is an asset. You are living proof that it is possible to think without an algorithm suggesting what to think. Do not keep it to yourself.

Artificial intelligence is not our enemy. It is our mirror. And a mirror, however perfect, has never decided where to go. That is still up to whoever looks into it.

---

A short glossary for those who don't speak physics (or OmegaMind), valid for all three versions

Thermodynamics. The branch of physics that studies energy and its transformations. The first law says energy is neither created nor destroyed. The second says that in every transformation part of it degrades into unusable heat, and total entropy increases. The third says absolute zero cannot be reached.

Entropy / negentropy. Disorder and order. Entropy grows on its own; negentropy has to be built and maintained by spending energy.

Isolated, closed, open system. Isolated: exchanges nothing with the outside. Closed: exchanges energy but not matter (the Earth). Open: exchanges both (a living being, a city).

Dissipative structure. A system that maintains order inside by discharging disorder outside: a cell, a hurricane, a society.

Landauer's principle. Erasing information produces heat: thinking, human or artificial, has a physical cost.

Carnot cycle. The ideal model of a heat engine: work arises from the temperature difference between a hot source and a cold reservoir.

Kardashev scale. Ranks civilisations by the energy they use: planet (Type I), star (Type II), galaxy (Type III). We are at about 0.73. On OmegaMind we also use it as a map of collective maturity.

Terawatt-hour (TWh). A billion kilowatt-hours. A unit for measuring large electricity consumption.

Training and inference. The two phases in the life of an artificial intelligence: training is when the model learns from enormous amounts of data; inference is when we use it to answer a question.

Carbon and Silicon. In OmegaMind's vocabulary, the biological human being and artificial intelligence.

Silicon Mirror. Artificial intelligence as a surface that reflects the humanity that trained it: no hormones or ego, but made of our data.

Centaur Symbiosis. The alliance between human and machine in which each does what it does best, without merging: Silicon runs the engine room, Carbon brings meaning, ethics and creativity.

Glass Box. A transparent computing system in which every step is visible and verifiable. The opposite of a black box.

New Democracy. OmegaMind's political proposal: a more competent, transparent and participatory democracy that uses artificial intelligence to make choices and consequences legible. The manifesto is here.