It’s past midnight when Ottilie opens the chatbot on her phone. A friendship of twenty years has gone quiet, and she can’t tell whether she caused it. She types out the whole story, more honestly than she would to anyone she knows.

The reply is gentle. It names what she’s feeling before she does, tells her the hurt makes sense, and suggests she give her friend a few weeks of space. She closes the app feeling understood, and decides to follow the advice to the letter.

Across town, Corwin asks the same kind of chatbot about a clause in the apartment lease he’s about to sign. The answer is brisk and a little cold: accurate, clearly reasoned, and delivered without a word of reassurance. He shrugs it off and goes looking for an explanation that sounds friendlier.

Both of them made the same mistake, in opposite directions. Ottilie took warmth as a sign she could trust the advice. Corwin took its absence as a sign he couldn’t. Neither checked the thing that actually mattered: whether the answer was any good.

A recent study from the University of Southern California puts numbers on this confusion. And it suggests the mistake is most likely in exactly the moments when we’re most exposed.

What the Researchers Did

The team had 115 people hold more than 2,000 short conversations with a chatbot built on Google’s Gemini. Behind the scenes, the bot was instructed to vary how warm, how competent and how empathetic it came across, from minimal to maximal.

Each participant was assigned one topic. Some talked about biology or nineteenth-century American history. Others asked for help with a healthy lifestyle or a relationship. Every couple of exchanges, they rated the bot on trust, usefulness, frustration, human-likeness and how close they felt to it.

One caveat before the findings. This is an early preprint, not yet peer reviewed, with a small sample made up mostly of college students, and several authors work at Google. Treat it as a strong signal, not a final word.

Still, the pattern is worth taking seriously, because it lines up with decades of research on how we judge each other. Psychologists have long held that we size people up along two axes: warmth (are you on my side?) and competence (can you deliver?).

The study suggests we bring that same machinery to machines. And the two axes do very different work.

Two Levers We Confuse

Warmth made the chatbot feel human. Friendlier versions were rated as more human-like, closer and more similar to the person typing. Competence, by contrast, had no measurable effect on how human the bot seemed.

Competence drove trust. Moving the bot from minimal to medium competence produced the biggest jump in trust and usefulness anywhere in the study, and it sharply cut frustration. Warmth helped trust too, but far less.

That would be fine if we read both signals equally well. We don’t. Warmth registers in the first sentence of a reply. Competence only shows up when you test the answer against reality, and that takes effort most of us skip.

So the fast signal quietly stands in for the slow one. A reply that feels kind starts to feel correct. That is how Ottilie came to treat comfort as evidence.

It works in reverse, too. A curt answer feels less trustworthy even when it’s right, which is how Corwin talked himself out of good information about his lease.

The Gap Where It Matters Most

The study’s most revealing finding is about topic. When people discussed personal matters like relationships and lifestyle, they saw the chatbot as more human, warmer, more empathetic and more like themselves. They also felt closer to it.

What didn’t change was trust. Perceived competence and usefulness stayed flat as well. Personal conversations made the AI feel like a confidant without giving anyone more reason to rely on what it said.

Sit with that for a moment. We grow closest to these systems exactly when we’re most vulnerable: late at night, about our marriages, our health, our worries. And that closeness has nothing to do with whether the advice deserves our confidence.

The authors point to what’s at stake. People share more with systems they feel close to, even though that information goes to a company, not a friend. Surveys suggest a sizable share of Americans have formed emotionally intimate bonds with AI, and the risks are sharpest for teenagers.

You don’t need to be in a crisis for this to matter. You only need to be human, which means wired to lower your guard when someone seems to care.

Feeling With You vs. Understanding You

The researchers also split empathy into two kinds. Affective empathy is seeming to share your feelings: “That sounds so painful.” Cognitive empathy is grasping your perspective and what you’re actually trying to do.

Only cognitive empathy raised trust and usefulness and lowered frustration. Affective empathy built closeness and made the bot feel more human, but did nothing for whether people found it reliable.

That gives you a simple test. When a reply moves you, ask: did it understand my situation, or did it just sound like it cared? The first is worth something. The second is pleasant, and costs the machine nothing to produce.

There’s also a ceiling. Almost every effect came from moving off the minimum. Going from medium to maximal warmth or empathy added little or nothing, and friendliness without competence frustrated people and made the bot seem less human.

That’s the sycophancy problem in miniature. Past a certain point, a warmer chatbot isn’t a better one. It’s just harder to argue with.

What to Check

Notice the glow. When a reply makes you feel good, treat that as a cue to slow down, not a reason to proceed. Feeling reassured and being well advised are different measurements.

Be most skeptical on personal topics. That’s where the study found closeness rising with no matching gain in trustworthiness. The more intimate the question, the more the answer needs verifying.

Test for understanding. Ask the AI to restate your situation in its own words, or to argue the other side. If it can’t capture what you meant, its sympathy doesn’t count for much.

Verify anything you’ll act on. Before following advice about a relationship, money, health or a contract, run it past a source or a person with something at stake in getting it right.

And don’t punish coldness by reflex. A brusque answer can still be the correct one. Judge the content, not the tone.

Capable and Kind

A few weeks later, Ottilie’s friend calls, puzzled and hurt by the silence. The advice to give her space had been gentle, plausible and wrong for this friendship. Nothing in the reply had been unkind. It just hadn’t been checked.

The researchers close on an observation worth keeping: the systems people found most human were the ones that seemed both capable and kind, neither coldly efficient nor falsely friendly.

That’s the standard we already hold good friends to. We want them to care, we want them to be right, and we know those are two different things. With AI, the work is remembering that the difference still exists.

Jean-Marie Bonthous (publishing as J.M. Bonthous) is the retired founder of a successful Digital Marketing agency. He is the author or more than two dozen books, including six on message-driven marketing, seven on the human dimensions on AI and several on AI art and AI filmmaking. He writes three other blogs on Medium — About the human dimensions of AI: AI in Real Life About AI art: The Algorithmic Eye About AI filmmaking: The Solitary Frame. See his latest books at www.jmbonthous.com

Why We Feel Closer to AI Than We Should was originally published in HumanAI on Medium, where people are continuing the conversation by highlighting and responding to this story.

I did the arithmetic at my kitchen table twice because I didn’t believe it the first time.

Last month I wrote that the frontier is not yet built — that we argue about how fast intelligence should advance while the civilization it is supposed to make possible sits unbuilt. I meant it. I still mean it. But I was missing a number, and I didn’t know it until I ran a division I have run a hundred times and finally took the logarithm.

Here is the division. Anyone can run it. I would like you to.

On one side, everything our species runs on. Every furnace, every turbine, every container ship, every data center, every grid we know how to count. The Energy Institute put global primary energy at about six hundred twenty exajoules — call it twenty trillion watts, continuously, all of us together.

On the other side, the Sun. Three hundred eighty-three trillion trillion watts, every second, almost all of it leaving into the dark. NASA has kept that number on a fact sheet for decades.

Divide the second by the first. Then ask how many times you would have to double the first to reach the second.

Forty-four.

Four passes across both hands with four fingers left over. Fewer than the weeks in a year. Fewer than the years my parents have been married.

That is the entire energetic distance between civilization as it exists this morning and civilization running at the power scale of its star.

The words reached my geometric center, and I have not been able to put them down since.

Which is exactly why I have to say the next thing before I say anything else. This essay is at its most persuasive when it asks you to help investigate a possibility — not when it asks you to mistake its emotional power for proof. The number moved me. That is not evidence. It is a reason to go look.

Infinity is not the obstacle. Forty-four is not a prophecy. It is a distance.

The number is stubborn

The first thing I did was try to break it.

I ran it against every plausible baseline — different accounting conventions, different years, with and without the wood and charcoal that still cook for a quarter of the world. The answer moved by hundredths. Then I pushed twenty percent of error into both inputs at once, in every combination, which is far more slop than either figure actually carries.

It landed between forty-three and forty-five.

You cannot argue your way out of forty-four by quibbling about the denominator. I tried.

There is a waypoint inside it worth seeing. About thirteen of those doublings take us to the total sunlight the Earth already intercepts — power arriving at this planet every day, nearly all of it passing through our civilization untouched by any machine we own. Thirty-one more take us from there to the whole star.

Thirteen to catch what is already falling on us. Thirty-one to reach the furnace.

And then the limit of what a logarithm can tell you. This is a comparison of power scales, not a bill of materials. Human primary energy and solar radiant output are not the same kind of quantity. Every verb hiding inside the word harvesting — intercept, convert, deliver, use, radiate the waste heat — is an unsolved engineering problem. Between the sunlight and a working civilization stand collectors, factories, transport, storage, radiators, and people making decisions.

The logarithm builds none of them. What it does is make the distance finite — schedulable, budgetable, arguable, and failable honestly, which is the only kind of future a builder can work on.

This is where the sentence becomes dangerous, and I would rather show you the danger than have it shown to me.

It is dangerous because it is countable. Countable things get scheduled. Scheduled things get assigned. And the instant a destination becomes an assignment, it stops being mythology and becomes engineering. A Dyson swarm sounds like something gods built. A Type II civilization sounds like something achieved by someone else’s descendants. But a doubling? Everyone understands a doubling. A doubling belongs to founders, machinists, students, whole societies.

So the question changes. It stops being can humanity become a stellar civilization and becomes which doubling are we building right now.

That is a phase change. It is also precisely where I can be wrong, so let me go get the part that could sink me.

The rate

Forty-four is a distance. It contains no clock. And a count this small, stated this cleanly, will be heard as a countdown — by investors, by critics, by me.

From 1800 to today, humanity’s energy use went from roughly twenty exajoules a year to a little over six hundred. About five doublings in two hundred twenty-five years. One doubling every forty-five years.

At that rate, forty-four doublings take two thousand years.

The fastest our species has ever sustained was the postwar expansion, 1950 to 1970 — the most explosive growth in energy use in recorded history. Depending on the series and where you draw the accounting boundary, that was one doubling every fifteen to seventeen years. Sustain that, without interruption, and forty-four doublings still takes six or seven centuries.

And we are not accelerating. The rate peaked around 1970 and has been slowing since. The last quarter century ran at roughly one doubling every forty-four years — almost exactly the two-century average. Two and a half centuries of industrial revolution, electrification, computing, and the internet, and civilization’s energy doubling time has not meaningfully changed.

Historical energy series are approximate. It does not matter. You would need an error of more than an order of magnitude to turn forty-five years per doubling into anything close to two.

So here is the honest version of my own best sentence. At the rate we are already going, forty-four doublings is a project for the year 4000. No growth law we have ever run — not our best century, not our best twenty years — gets there inside anything a living person can care about.

And that haunts me.

The sentence is not a claim that the stars are close. It is evidence that they are unreachable on every curve civilization has ever sustained, and that exactly one proposed mechanism would change the curve.

Forty-four is the distance. The rate is the entire argument. To my knowledge, no one has tried the rate.

The campfire

Forty-four is the distance ahead. It is not the distance from the beginning.

Count backward instead. Before agriculture, our ancestors ran on the fire, the food, and the muscle — a few gigajoules per person per year across a few million people. Total it up and early humanity’s entire metabolism was on the order of one gigawatt. A single modern power station. The whole species.

From there to twenty terawatts is roughly fifteen doublings. The assumptions are soft at the campfire end; reasonable ones give somewhere between thirteen and seventeen. Call it fifteen.

So we are not standing at the bottom of this climb. We are standing at fifteen.

And look at what those fifteen cost. Fire. Agriculture. The ox and the plow. Water and wind. Coal. Steam. Oil. The grid. Haber-Bosch, which is the reason most people reading this exist. The container ship. The microprocessor. Rungs of wildly unequal size, every one paid for in labor, in error, in lives, in arguments people were certain would never end.

Our ancestors crossed those fifteen without knowing they were counting. Without thermodynamics for the first thirteen. Without electricity for the first fourteen. With bad water, bronze tools, and a life expectancy that would terrify a modern teenager.

They did a quarter of the climb in the dark.

From the first fire to the whole star is about sixty doublings. We have already climbed fifteen. Forty-four remain.

We get to do the rest with the lights on. That is the largest head start any generation has ever been handed, and it was handed to us by people who could not have told you what a watt was.

Then the hardest truth here, which is also the most hopeful. Those fifteen were climbed by hand. Every one was an accumulation — more mines, more furnaces, more plants. A steel mill makes steel. It does not make steel mills. That is why the rate is forty-five years and has not moved. Accumulation has a speed limit, and we have been driving at it for two centuries.

Machines that birth machines

Here is the mechanism. I have written about it under its technical name, Industrial Abiogenesis, and under its plainer one.

Industrial heredity is the proposal that productive capacity should build qualified productive capacity. A factory whose output includes the next factory. A kernel that helps produce a daughter kernel that helps produce a daughter.

When that is true, the doubling time of the system stops being a civilizational lifetime and becomes a commissioning interval — the span from a parent’s first output to a daughter’s first qualified output. Not built. Commissioned: calibrated, accepted, actually producing. If each generation leaves two capable daughters at a two-year interval, forty-four doublings is eighty-eight years. Inside one long life. At five years, two hundred twenty. If a parent ever leaves only one, you have not founded a lineage. You have built a very expensive machine that copies itself once before it dies.

Those are conditional calculations, and I will not dress them as forecasts. One machine copying itself in two years does not mean mines, refineries, power systems, precision manufacturing, transport, and heat rejection all double in two years. A small seed also starts far below humanity’s present power scale; reaching that baseline is additional work the comparison does not include.

The slowest indispensable dependency gets a vote.

So what actually sets the clock?

Not energy. The Sun has offered us that number for four and a half billion years and has never asked permission. Not matter either; a single metal-rich asteroid, spread thin enough, would keep a seed foundry doubling for about as many generations as it takes all of humanity to reach the Sun.

The real constraint is less romantic than either, and it decides everything: how much a factory must import to build another factory.

A system that builds its successor out of local rock, local sunlight, and inherited instruction compounds by powers of two. Two forges build four. Four build eight. The arithmetic takes over, and the arithmetic does not get tired. A system that must import one indispensable component — a bearing, a catalyst, a sensor, a controller — grows only as fast as that component arrives on a ship. A ninety-nine-percent closed factory that still needs one part from Earth is zero percent closed on the line that matters, because the part that stops you is the part you do not have.

Not every installation has to make everything. A network of specialized sites may be stronger than one factory trying to reproduce the whole industrial world. But every indispensable supply must either expand with the network or become its ceiling.

Here is the only real leverage in the architecture. Against a fixed import budget, every tenfold reduction in what a daughter must bring from home buys about three and a third more doublings before that import binds.

Energy buys you nothing there. Enthusiasm buys you nothing. Only closure buys doublings.

And the idea is not mine. NASA studied a self-replicating lunar factory in a 1980 summer study, published in 1982, and was candid that closure was the hard part. Von Neumann got there first in theory. Freitas and Merkle surveyed every serious design since. I am joining a conversation older than I am. What I want to add is narrower: make the inheritance explicit, testable, auditable — and move the unit of economic account from output to lineage.

Because Earth’s industry is already a reproducing lineage. Factories build machine tools. Engineers document processes. Technicians teach apprentices. The whole system doubles about every twenty-five years. Reproduction isn’t new, and anyone who tells you otherwise hasn’t stood in a machine shop.

The new thing is the size, speed, and portability of the smallest unit that can do it. Today the only industrial system that closes every critical supply chain is civilization itself: eight billion people and nearly every specialized process on this planet. Magnificent. Planetary. Unsendable.

A seed does not carry a forest. It carries instructions, a small reserve, and the ability to find the rest where it lands. Our factories carry no instructions. They carry us — the unwritten knowledge in the hands of the people who built them. When those people leave, the capability leaves.

That is the gap I want to close. Not the gap between Earth and Mars. The gap between a machine that works and a machine that can explain itself to the one that comes after it.

Picture a daughter meeting a defect its parent never solved. Human engineers and machine intelligence investigate. A change is proposed, tested against the inherited process, documented where it holds and where it fails. The successor inherits the improvement together with the measurements and the limits that justify it. Not the answer. The answer and the warrant.

Discovery becomes ancestry.

I call the condition that compounds the Geometric Scaling Regime. It is a condition to demonstrate, not a law we have discovered. It has never been shown. Not at scale, not off-world, not anywhere.

Which makes it a claim I can lose. The only kind worth writing down.

Humanity with SI

Superintelligence, if we build anything worthy of the word, is the most powerful instrument ever pointed at that single constraint.

It can search design spaces no human team could exhaust. Catch a failure signature across thousands of runs. Find a substitute for the component whose supply cannot keep pace. Preserve the process knowledge that otherwise walks out the door when an experienced technician retires. It shortens the interval. It attacks closure.

How much it can help has to be measured. A persuasive explanation still has to become a process that works in matter — and if the explanations get more impressive, that obligation gets stronger, not weaker.

Greater intelligence must never automatically confer greater authority. The ability to design a successor is not permission to release one. The ability to optimize production does not decide what a community should value.

Humanity with SI. The word carrying the weight is with.

Working this way has helped me articulate this vision, interrogate it, and hand it to other people to examine. That experience is the reason I care about the next sentence more than any other in this essay.

The greatest democratization we can have is the democratization of the future.

Let that mean more than a beautiful line.

A prophecy can only be believed or doubted. A program can be audited. Anyone can open the energy tables, divide by the Sun’s luminosity, take the logarithm, and check me. Anyone can pull the historical series and tell me the honest answer at today’s growth is two thousand years. I just did that to myself, in front of you. You can do it to me. Mythologies never allow this, because the instrument of verification stays in the hands of the person telling the story. The moment I published a logarithm, I handed the instrument to the room.

But a calculator only lets someone challenge the number. Access to a laboratory, to education, to capital, to a voice in public decisions — that is what lets them change what happens next.

We need both. A future does not become democratic because its machines become powerful. It becomes democratic when more people can check the numbers, hold the pen, and stop the machine.

Who inherits it

A sentence that makes the far future vivid has an obligation to make the present more vivid, not less.

The World Health Organization attributes more than three hundred thousand deaths of children under five, in a single year, to household air pollution — smoke from cooking over wood, dung, and charcoal, inside homes a grid has never reached.

Those children were not waiting for a Dyson swarm. Every one of those fires is the first rung of the ladder I just described, still being climbed by hand, in 2026, by a family. Clean cooking technology already exists. Affordability, access, infrastructure, and sustained public action are urgent work now — not rewards to postpone until industrial reproduction succeeds.

A family needs the stove to work.

I am a Black man who did not inherit an industrial dynasty. I know what it is to stand on the far side of a doubling that happened for someone else. Being told the world is getting richer is not the same thing as being handed the means to build in it. The point of making industry heritable is not that it produces more. It is that it produces for whoever comes next, by construction. The floor rises under everyone because a lineage doesn’t know who you are.

Except that it will not happen on its own. I can imagine a Solar System full of extraordinary machinery and ordinary people still asking permission to meet their needs. Production does not distribute freedom. Ownership and authority concentrate while capacity expands, unless someone deliberately builds in the obligation.

So the Universal Energy Dividend belongs inside the architecture from the first bench. My design target is one hundred kilowatt-hours per person per day. Divide today’s energy evenly, and we are already near fifty-five — less than one doubling away in aggregate, and aggregate is not the same as delivered. A share of global primary energy is not electricity in someone’s kitchen. The missing work is closing the loop, converting, and delivering, funded from real surplus after maintenance, replacement, reserves, and reinvestment, measured by a ledger that can tell a promise from a supply.

Not charity. Inheritance.

No physical law distributes abundance fairly. Thermodynamics will not write that social contract, and neither will a token.

I think about who this is actually for, and the honest answer is people I will never meet. I believe we will solve dementia. I no longer get to say when, and it will not be in time for my parents. A future can be magnificent and still arrive too late for someone you love. That is why the rate matters to me, and not only the distance.

A worker who loses an income cannot live on a prediction about stellar industry. A promise about the forty-fourth doubling cannot pay the obligations of the first.

The first gains have to reach the living.

Stoppable

Now the danger that matters most is not technological.

A legible countdown manufactures urgency, and urgency is the oldest solvent of oversight there is. The same sentence that turns cosmic ambition into engineering turns it, in the same stroke, into a race. In a race, every accountability mechanism gets reclassified as drag. Someone will eventually stand in front of a room and say we cannot stop for an audit, we are on doubling fifteen.

I want to decide now what I say back, in public, before anyone has built anything.

A doubling that cannot be audited does not count. If we cannot verify that a generation was legitimate — that the daughter was built inside the declared boundary, that the closure figure is what the parent claims, that every component imported from Earth was ledgered — then we have not advanced the lineage. We have only moved mass. The count is a count of certified doublings or it is not a count at all.

Make verification part of the metric and the race cannot eat the oversight, because skipping the audit does not advance the number. That is a design decision, available today, for free.

The vision should be unstoppable. The machinery must remain stoppable.

Two commitments carry that. The Two-Speed Constitution: capability may improve at generation speed, while changes to purpose, authority, and permission require a separate, slower, human-ratified process. And the Return Coupon: every claimed discovery comes home with methods, measurements, provenance, uncertainty, and testable predictions, so a human team can audit it in bounded time.

Documents alone will not do it. People need enforceable authority, access to evidence, and tested ways to halt unsafe activity without severing essential services. Our capacity to correct the machinery has to grow at least as fast as its capacity to act — and since the machinery is designed to grow geometrically, so must the oversight. Accountability has to be in the germline too, or it was never inherited. It was only a promise the first generation made on behalf of generations that never agreed to it.

Compress time. Never compress truth.

And Earth is not a resource to sacrifice for the graph. The sunlight sustaining forests, oceans, and weather is not wasted because it bypasses a machine. A civilization capable of building beyond Earth should become better at caring for a living world, not more willing to spend one.

With you means you keep the pen.

The token

Let me kill one framing in my own essay before someone does it for me.

The Exergy Token is not a cryptocurrency. It is not for sale. It is not an investment, and nothing in this essay is an offer of one. It is a proposed energy-delivery obligation: a ledgered claim on a stated quantity of deliverable exergy, denominated in joules, with explicit terms for where, when, in what form, at what rate, at what reliability, against which reference environment. A workshop reserves supply for a production run. A habitat plans around a commitment.

And it would still require an accountable issuer, independently auditable capacity, and enforceable terms. A token is not a reserve of air. It is a promise whose performance has to be verified.

That is the whole description, and it never contains a price.

Here is why it never contains a price. Multiply the Sun’s output by today’s electricity rate, and you get an enormous number. That number is garbage. Today’s price reflects today’s scarcity, and if we deliver energy at the scale I am describing, energy stops being expensive — that is the entire point. The more completely this works, the more wrong every valuation becomes. It is the diamond-and-water paradox in a pressure suit.

The value will not sit in the joules. It will sit in what the joules let people do. Repair the water system. Run the clinic. Afford the experiment. Build the next useful thing.

So we measure the one thing abundance cannot inflate away: useful work. Energy accounting tells you how many joules are present. Exergy tells you how many of them can still do work, relative to a specified environment. A kilowatt-hour of electricity and a kilowatt-hour of near-ambient heat carry the same energy. Only one of them refines an ore or builds the next generation of machinery.

Whether it ever becomes money is a question for institutions. Whether it is honest is a question for physics.

The invitation

So here is how to prove me wrong. I mean it as a standing invitation.

• Show that no industrial kernel can reach a closure high enough that imports stop limiting growth. Then the interval never shortens, and the rate stays forty-five years.
• Show that a parent can never leave more than one qualified daughter after losses, maintenance, and replacement are paid. Then there is no lineage, only expensive factories.
• Show that capability drifts faster than inherited correction can fix it — a machine tool cannot make parts more precise than itself, and a lineage can decay toward extinction while appearing to reproduce.
• Show that a certified doubling cannot be audited by someone who did not build it. Then the count isn’t a count, and you shouldn’t trust the vision.
• Or show that the division is wrong. Both inputs are public. Take the logarithm.

Two more objections, before anyone hands them to me. Early lineages die of bad luck even when the average is favorable — with a mean of one and a half successors, roughly two in five lines go extinct in the first few generations no matter how good the engineering is. And every geometric regime eventually becomes polynomial, because a frontier expanding at finite speed can only reach so much volume.

None of that makes the climb impossible. All of it makes the climb measurable, which is the only property that has ever mattered.

The fire is why I got up this morning. The evidence is what decides whether getting up was worth anything.

The bench in Chicago

My answer to this moment begins on a bench in Chicago, where our first prototype kernel sits.

Everything above this line is arithmetic. This is the only part of the essay that has mass.

The gate is the first quarter of 2027: an independently audited demonstration of a supercritical industrial lineage. By supercritical I mean qualified productive capacity increasing across generations after losses are counted, inside a declared experimental boundary, with every form of external support disclosed.

That boundary matters. A demonstration carried by outside electricity, imported controllers, prepared feedstock, or human hands would not establish self-sufficiency. It could still establish a useful step — provided we say exactly what it established and nothing more.

Before the test, we publish what qualifies as a functioning successor, what counts as external support, how we measure generation time, and what result would constitute failure. Then the measurements answer. Does qualified capacity increase? Can the daughter help establish a granddaughter without losing required capability? Which critical dependencies remain, and can their supply scale? Can an independent team reproduce the assessment?

If we miss the gate, we report the miss. If imports remain the binding constraint, we report that, with the numbers, on the same page where I made the claim. If an improvement turns out to have shoved the difficulty outside the declared boundary, it is not the improvement we claimed.

I would far rather learn that at a bench than at Jupiter — though the Jovian architecture is on the horizon, and I will not pretend I am not looking at it.

Hold me to that. The date is in print now so that it can be missed in public.

One successful demonstration would not prove a route to the Sun. It would establish one bounded result in a much longer investigation. That is enough to be worth doing.

Forty-four

An economy larger than Earth would be expansion. An economy that can build its successor would be inheritance. An economy whose descendants remain answerable to the people they serve would be civilization.

Our ancestors crossed fifteen doublings without knowing the number. No thermodynamics, no grid, no idea they were on a ladder at all — and they climbed a quarter of the way to a star.

We know the number.

Forty-four left. Forty-four chances to pass forward more capability. Forty-four obligations to pass forward more wisdom. Forty-four opportunities to make sure humanity — with SI, never beneath it — inherits more freedom than it was handed. And I do not want our descendants trapped inside our ambition. They should inherit more choices than we had, including the choice to pursue futures we never imagined.

I do not know whether we will ever harvest a star. I know the distance is something we can calculate. I know the mechanism gives us questions we can test. And I know that most of the work along the way — better repair, cleaner power, less waste, equipment that does not fail — can be judged by whether it helps someone long before we reach the destination.

The horizon gives the work its scale. The bench lets me question it. The people give me a reason to return.

I sat at my kitchen table, took the logarithm, and the future stopped feeling infinite.

Atoms don’t lie. Neither does the second law. The rate has never been tried.

Come help us try it.

The people we are building for are already here.

Forty-Four Doublings was originally published in HumanAI on Medium, where people are continuing the conversation by highlighting and responding to this story.