The nameplate is bigger than the door

On 22 September 2026, at the United Nations, the American president said the word “artificial” makes AI “sound fake”.¹ A week later he signed Executive Order 14434, Inaugurating the Era of Super Intelligence.² From that day, federal agencies were told to stop writing “artificial intelligence” and “AI”, and to write “Super Intelligence” (“SI” for short) instead.²

Before we begin: Anthropic, the AI company, appears in this story, and its models also helped with our research. The note at the end says how we handled that.

So before forming any opinion, I tried to understand the world around this one word. I started with a long conversation with an AI model.

What the order actually says

The order is short, and one sentence of it settles the first question. Section 3(a) says Super Intelligence means “the technologies and systems encompassed by the term ‘artificial intelligence’” as US law already defines it.² The same systems. A new name.

The real definition comes later. Section 3(b) gives the President’s science adviser 60 days to send the President “proposed legislative language” for one.² That is about 28 November. It is a proposal, not a law, and nothing says it must be made public.²

The name being retired was never chosen with ceremony either. In 1955 John McCarthy and three colleagues proposed a summer study of “artificial intelligence” at Dartmouth College.³ He explained the choice later: “I had to call it something.”⁴

An executive order directs the executive branch. It cannot change a law, it binds no company, state or country, and it appropriates no money.²⁵ It also has a twin: in 2025 another order gave the Defense Department the secondary title “Department of War”, with the same two steps, use the name now and propose a law in 60 days. A year later it is still not law.⁶ The Gulf of Mexico became the “Gulf of America” by order too.⁷ Now a technology.

The case for the new name

The rename deserves its best case. The White House says these systems are “not simply imitating or automating human intelligence”, so “artificial” is the wrong word.⁸ That is a real argument. But we found no scholar or commentator arguing that the change makes the language more accurate.⁹

What “super” would have to mean

To judge whether “super” fits, I had to ask what intelligence is. It is the hardest part of the story, because the people who study it disagree. François Chollet, who built a well-known test of machine reasoning, says it is not how much you know but how efficiently you learn something new.¹⁰ My own instinct, before I read him, was close to that: how fast something learns what it was never trained on. It was a relief to find serious people standing nearby.

People usually describe three rungs: narrow AI, which works inside one bounded problem; general AI, where the definitions split; and superintelligence, which Nick Bostrom defined in 2014 as an intellect that “greatly exceeds” human performance in virtually all domains of interest.¹¹ Researchers who graded today’s AI like a person, across ten areas of ability, gave GPT-4 27% and GPT-5 57% of a well-educated adult’s profile, and called the shape “highly jagged”.¹²

Progress is real and fast. On ARC-AGI-3, a test built to measure learning in new situations, the best systems scored 0.51% at launch in March 2026. By September, verified results were above 96% with a helper tool that keeps the model’s reasoning between steps, and about 40 to 55% without it.¹³

But the profile is uneven. In a field experiment with 758 consultants, AI made them better inside its “jagged frontier” and about 19 percentage points less likely to be right outside it.¹⁴ My reading: on breadth, we are closer to a general mind than most people assume. On learning how to learn, not yet. Real progress and a definition of “super” are two different things.

One sentence, two jobs

When I first heard about the order, I said it felt like two meanings with one goal: an announcement, and the name of a project, both asking us to trust it. I didn’t expect to find both inside one sentence of the order.

Section 1 describes today: these systems “far exceed what was envisioned”. And it promises tomorrow: the name captures “the promise, potential, and rapidly advancing capabilities”.² One sentence describes the present and sells the future. That is a reading of the text, not a claim about why it was written that way.

Why can a promise make itself true? Robert K. Merton described the self-fulfilling prophecy: a false belief, once people act on it, can come true, like a healthy bank that fails because a rumour sends everyone to withdraw their money.¹⁵ If enough people act as if Super Intelligence is coming, money and talent may move toward it. That is my inference.

It breaks here. Inside a company, the employer decides your title and everyone has to use it. A country’s word for a technology is different: the rest of the world has to choose to answer to it.

How names move value

I had a guess about luxury bags: the same bag costs a little to make and a lot with the right name on it. The real number was better than my guess. In 2024 Milan court documents showed that a contractor was paid as little as €53 to make a Dior bag that sold for €2,600, about 49 times more.¹⁶ The shop price also pays for design and stores. The case was about labour, and Dior was not found guilty of anything: its Italian unit was put under court administration over conditions at its suppliers. So this is no joke about cheap production. The workshops were in Italy, though Chinese-owned.

The label changes the experience, not just the price. In a 2008 brain-scan study, people rated the same wine as more pleasant when told it cost $90 rather than $10, and the brain region linked to pleasure was more active too.¹⁷ It was a small study, 20 people.

Where names fail. When buyers can test the thing, labels matter less. The price of a fixed level of AI capability has fallen between 9 and 900 times a year.¹⁸ My guess is that business buyers benchmark, so a label counts for less. One 2024 study went the other way: telling buyers a product description involved “AI” lowered their intention to buy.⁴⁰ The old name may not be the handicap the order assumes.

Open. Whether “SI” will move any price at all. I used to think it could justify higher prices for AI. I’m no longer sure.

The money behind the hurry

Why would anyone want a bigger name for AI right now? I looked at the money. The Bank for International Settlements reported that the five largest cloud companies were set to spend over a trillion US dollars on AI infrastructure across 2025 and 2026.¹⁹ Goldman Sachs estimates AI needs about $300 billion a year in revenue just to break even on that build-out.²⁰ Spending is far ahead of earnings. Whether the gap closes, and how, nobody can say.

On 12 September 2026 Dario Amodei, the chief executive of Anthropic, published an essay calling for the industry to slow the pace at which it improves its models, adding that “pacing does not mean halting”.²¹ The administration said it would not put on the brakes.²² Anthropic was in a public dispute with the Pentagon, which had labelled it a supply-chain risk. A judge has blocked that label.²³ As I said at the top, Anthropic’s models helped with our research, so here I have kept to the public record.

What I can say is about effect, not intent. A slower pace would give revenue time to catch up with spending. It does not obviously help the lab that wrote the essay either, which is why I treat this as an effect and not a motive. My reading: the money explains the hurry. It does not explain the name.

The room with China

The other half of “why now” is geopolitics. China’s president Xi Jinping was in Washington from 23 to 25 September. Analysts found no progress on chips, rare earths or Taiwan.²⁴

Then the word itself. The White House fact sheet said: “The two leaders agreed to use the term ‘super intelligence’ rather than ‘artificial intelligence’.”²⁵ The next day China’s foreign ministry said China “respects their choice of the term”. Its own text kept saying “artificial intelligence”.²⁶ So China respects the choice of the new term. It did not agree to it. Those are different sentences.

The order came four days after the visit ended. Is there a connection? I don’t know. I once thought something must have been on the table. Nothing published shows it.

What does the reporting say about motive? The order itself names no money, China or security reason. Bloomberg framed it as a response to public backlash.²⁷ That is reporting, not proof.

Has a nation done this before?

Yes, many times. History shows when naming a future works.

Apollo, 1961. On 25 May 1961 John F. Kennedy told Congress the nation “should commit itself to achieving the goal, before this decade is out, of landing a man on the moon and returning him safely to the earth”, and asked for billions to do it.²⁸ Apollo cost about $309 billion in today’s money, and the landing came in July 1969.²⁹

In a recorded meeting in 1962, Kennedy said “I’m not that interested in space”, because the Soviets had “made this a test of the system”.³⁰ And in most polls of the 1960s, less than a majority of Americans thought the moon was worth the cost.³¹ I think it worked anyway, because money, a date and a single agency did the work. Enthusiasm came after success, not before it. That gives me four tests for any name that wants to become a mission: is there money behind it, is there a date, is there one agency in charge, and is there a way to check the result? Apollo had the first three, and the moon was the fourth: it landed or it did not.

I first thought this order was the same move with the same ending. It isn’t. Score it on the four tests. Money: the order appropriates nothing itself, and the fact sheet points to more than $5 billion announced for a separate programme, the Genesis Mission, against Apollo’s roughly $309 billion in today’s money.⁸²⁹ Date: none for the thing itself, only 60 days for a proposed definition. One agency: none, since it tells every agency to change its words. A way to check: not yet, because nobody has said what would count as Super Intelligence.²

Japan, 1982. The Fifth Generation project named a ten-year mission for “epoch-making” AI computers and produced no commercial success. Yet rivals answered, with Britain’s Alvey programme and America’s Strategic Computing Initiative.³² The name worked most on rivals, who spent their own money.

“Star Wars”, 1983. Ronald Reagan proposed space-based missile defence. The next morning Senator Ted Kennedy called it “reckless Star Wars schemes”, and his nickname became the name everyone used.³³

Who answered the name

Many said no, at first. China said it “respects their choice of the term” and kept saying AI.²⁶ California’s governor ordered state agencies to keep “AI” as the official term.³⁴ Asked which name they preferred, 53% of Americans chose “artificial intelligence” and 9% chose “super intelligence” (YouGov, 28 September, 21,449 adults).³⁵

Then, five days after the order, seventeen countries said yes. On 4 October the United States and sixteen others, among them Japan, Korea, Germany, Italy and the UK, endorsed the Kyoto Vision for science, a declaration that uses “SI” and never defines it. China and India were not on the list.³⁶

The sociologist Pierre Bourdieu called the state’s power to name things “the monopoly of official naming”. But he added a condition: symbolic power “can be exercised only if it is recognized”.³⁷ So who recognises the name? Of those who answered, some are close partners of Washington and some are not, and I do not know why each one signed. California’s governor is a separate case: a refusal from inside the United States, not from a foreign government. And the public, in one poll, preferred the old name.

Maps tell the same story. In 1507 Martin Waldseemüller wrote “America” on a world map. On his 1513 map he dropped the name. Too late: everyone else kept it.³⁸ When the Gulf of Mexico became the “Gulf of America”, US agencies and map services for US users adopted it. The Associated Press, Mexico and the UK refused.³⁹

That is when the picture came together for me:

A name only works if people answer to it. And who answers depends on who needs you.

This is a hypothesis, not a law. It fits the Gulf of America, the Kyoto list and China’s polite refusal. It would be wrong if “SI” spreads among countries that don’t depend on Washington, or if the public adopts it without being asked. Apollo is the case that would break it: where enough money and institutions stand behind a name, public agreement mattered less.

What I don’t know, and what to watch

What I refuse to claim. Why they did it. What was agreed with Xi. That anyone lied, or that any safety warning was insincere. Why each of the sixteen other countries signed in Kyoto.

What to watch. Around 28 November: whether the proposed definition is published, and whether it is drawn around today’s systems or sets a harder bar. Whether “SI” spreads beyond countries that depend on Washington. And whether money follows the name. A name with a budget is a mission. A name without one is a request.

Maybe one day the job will grow into the title. Until then I’ll watch the doors: who screws the new plate on, who leaves the old one up, and what each of them needs from the people holding the screwdriver.

When a name arrives before the thing, who decides what it means: the people who say it, or the people who hear it?

We will read the definition on 28 November, if it is published. You are welcome to read it with us.

---

A note on conflict of interest. Anthropic appears in this story: its chief executive, Dario Amodei, wrote the essay about slowing down, and the company was in a dispute with the Pentagon that we mention. Anthropic’s AI models also helped with our research. We checked every claim that touches Anthropic first and kept to what is on the public record, and a model from another company checked the public facts independently.