Where’s Your Turbine?

The idea arrived as an insult.

In 2025, Morris was walking a nuclear startup’s reactor unveiling with a friend on its mechanical team. Vessel, pipe, coolant loop. He asked how the heat becomes electricity. Where’s your turbine? “He was like, ‘Oh, we haven’t thought about that yet.’ I’m like — what the f**k do you mean you haven’t thought about the turbine? That’s a two-year development, a two-year manufacturing process, and probably two years of qualification.”

Your ability to deliver your product is not in your control.It is in your suppliers’ hands.

He put the same question to the other half-dozen reactor companies in the El Segundo cluster. Priority three, outsourced, a year behind, or absent. The nuclear renaissance everyone in that neighborhood is pricing has a turbine-shaped hole in the middle of it, and the incumbents, in his estimation, can no longer fill it.

“If you boil down what GE does, they assemble the turbine from bought components and write technical manuals,” he says. “They are a technical manual company.”

So the quality engineer decided to build a turbine company. Gas now, nuclear next, geothermal after that.

Out of Inventory

The heavy-duty gas turbine market is a triopoly. GE, Siemens Energy, and Mitsubishi Heavy Industries hold almost 90% of it, and they have spent decades optimizing the Brayton cycle to within a decimal of what physics allows. Morris has raised $2.5 million for Stone Power from Boost VC, Forward Deployed VC, Deepwater Asset Management, and — yes — me. Two years ago he was sleeping on an inflatable mattress on a friend’s garage floor.

The company’s entire physical output sits on a wheeled cradle on the Hawthorne floor: one secondhand jet engine, veined with new plumbing — Stone’s own fuel pump, oil pump, safety loops. In late June the team lit it, throttled it, tripped its safeties, and shut it down, on software they wrote. A company learning to run a turbine the way a surgical resident practices on a cadaver.

“Dude, we turned on an engine!” is the first thing Morris says to me, before I can ask a question. He texts me the video.

That is the whole company. The math is what makes it serious.

Compute Compounds; Electrons Don’t

Microsoft, Google, Meta, and Amazon are converging on $630 billion in capital spending for 2026, up 60% from last year’s record, building gigawatt campuses for AI training runs. Compute compounds; electrons do not. The median wait to connect a new plant to the grid runs past five years, and the historical completion rate is 13%. A hundred-megawatt data center signing a Virginia lease today sits dark toward the end of the decade unless it makes its own power, and even Texas is measured in years.

So they call the incumbents. GE Vernova closed the first quarter of 2026 with 100 gigawatts under backlog and reservation, heavy frames sold out through 2030. Siemens booked a record €146 billion and is quoting 2029 and 2030. Wood Mackenzie has turbine prices hitting $600 a kilowatt by the end of next year, nearly triple 2019. Want a new aeroderivative turbine to power a data center? Three to five years. The incumbents sell waitlists.

Jet Engines, on the Ground

Aeroderivative. The machine everyone is fighting over is a jet engine. The compact, fast-starting turbines a data center wants are derived, almost without exception, from the engines under the wings of airliners: the core that pushes a 737 down the runway, bolted to a generator instead of a fan.

The gas turbine was perfected over seventy years to solve an airplane’s problems — be light, because weight is fuel at altitude; be small, because it lives in a nacelle, the housing under the wing; run murderously hot, because heat is efficiency and efficiency is range.

None of those constraints apply to a machine sitting in a field powering servers. It can be heavy, huge, and cooler-running. But the industry only knows how to build the flying kind, so that is what it sells the data centers: a racehorse, by the thousand, to pull a plow.

Death by Supply Chain

At seventeen, in a corner of a Tennessee factory where nobody was charging rent, Morris and his father Kirk rebuilt a cotton-ball rolling machine from the 1950s — bought, he’s about 90% sure, off eBay — and started making heat-bonded polymer spheres to replace the sand in water filters. The company, FilterBalls, filed a dozen patents with Dylan as co-inventor and threw off a million dollars a year for a decade. Its first pilot customer was Duke Energy, straining faintly radioactive particulate from nuclear-plant cooling water.

His first customer in American power was a nuclear utility, and he was in high school. FilterBalls got strangled anyway — not by its product but by its suppliers, nine American fiber processors dying under it in a decade as domestic textiles collapsed.

“Your ability to deliver your product is not in your control,” he says. “It’s in your suppliers’ hands.”

At Lockheed Martin he was a quality engineer, the unglamorous trade of making the thousandth unit work like the first. He pushed hardware out the door and around anyone in the way. At a performance review his manager sat him down. “Stay in your lane,” she said, kindly. “You’re doing too much. People are getting upset.”

Morris quit.

He went to Anduril and helped scale the tube-launched Altius-600 from three units a month to fifty, shipping more than a hundred of the armed variant to Ukraine. Then he quit that too, to move to Charlotte and buy a textile mill with his father — the supply-chain holy grail for the family firm that had watched those nine suppliers die. He hired a team, poured in his savings, and came up $400,000 short of what the bank wanted. At ninety days the deal collapsed. He fired the people he’d just hired and went six months without pay.

Beaten, Not Defeated

Three months before his wedding, Morris landed at LAX with $2,000 and spent it on a Honda Accord with no air conditioning and no radio. A buddy let him sleep in the garage, on an old air mattress. Then he called his old boss and asked for his job back.

He doesn’t varnish any of it. The man selling a trillion-dollar outcome keeps an itemized ledger of his own defeats — the three Anduril rejections before the fourth interview stuck, the mill, the short tour running manufacturing at Rainmaker for a thousand drones the company turned out not to need. In an industry that runs on selective memory, the ledger is rarer than the ambition.

Your material cost is somebody else’s energy cost, and theirs is somebody else’s below that, all the way down.

Read it twice, though, because it keeps two sets of books. Morris tells his story as grit, and the grit is real. It’s also a story about rain. The first real internship came, in his phrase, “from a fluke of meeting somebody.”

He reached Lockheed early because the pandemic killed the trip he’d planned first. Anduril turned him down three times before a recruiter who liked him found the fourth job, and a staffing decision, not a plan, put him on Roadrunner, an autonomous drone powered by two jet engines — his introduction to turbomachinery. A friend’s phone call later pulled him off the garage floor and into the company where he caught the itch to found one. He saw the shortage coming, partly. He was also, over and over, a man standing where the rain was about to fall.

The God Input

Dylan Morris found the secret of the American economy inside a textile mill. Doing diligence on the factory’s books, he saw that 30% of its cost of goods was electricity. Not labor. Not fiber. Power. He tried everything to beat it — modeled rooftop solar, ran the offsets, called a nuclear startup: “Dude, when can I get a reactor? I need a megawatt.” There wasn’t one. There wasn’t anything.

Energy is like that: buried in everything else, unevenly. Take fertilizer. Natural gas is 70 to 90% of the cost of making the nitrogen kind — not the fuel that runs the plant, the actual feedstock, the molecule itself. Fertilizer isn’t made with energy; it more or less is energy, pressed into a form roots can eat. When European gas prices spiked in 2022, the plants didn’t cut back. They switched off — 70% of the continent’s capacity, dark, because the product had become worth less than the gas it was made of.

Your material cost is somebody else’s energy cost, and theirs is somebody else’s below that, all the way down.

Now follow that fertilizer up into a two-dollar loaf of bread. The wheat it grew is about seven cents of the price. The farmer’s whole cut — fertilizer, diesel, seed, land — is 3 or 4%. The rest, the other dollar-ninety, is baking and packaging and trucking and shelf space and the lit, chilled store. Because your material cost is somebody else’s energy cost, and theirs is somebody else’s below that, all the way down — and at the bottom of that stack sits a molecule of natural gas.

Energy is 90% of the thing at the base and pennies of the thing on the shelf, hiding at every step inside a line that says “materials” or “freight.” It isn’t 6% of the economy. It’s the foundation of the economy, priced like a rounding error because we only meter it at the end.

Cheap Energy Is Cheap Everything

When energy stopped getting cheaper in 1979, America didn’t lose its industry at random. It lost it from the base of the gradient up — the smelters, the ammonia plants, the foundries, the things that were mostly congealed electricity. The bakeries stayed. We kept the part of the economy that barely uses energy and exported the part that’s made of it, and called the wreckage a transition to services. Morris’s textile mill, 30% electricity, wasn’t a weird outlier. It was a piece of the half that left.

It’s a fringe view. It might also be true — and if it’s true, you’d expect to see it on a map, in countries that bet opposite ways on the same input.

Germany decided, over two decades, that the mark of an advanced society was using less energy. It’s a rich country’s instinct, and a moral one: efficiency as virtue, restraint as progress. Then Russian gas vanished in 2022, and the industrial base cracked along its most energy-heavy seam — chemicals first, because chemicals are the most naked energy in solid form, and BASF, the largest chemical company on earth, began pulling production out of the country that invented the industry. German industrial output has now fallen four years running.

China decided the opposite, at a scale that’s hard to make an American reader feel. It is building thirty-some nuclear reactors at once, roughly half the world’s nuclear construction, at four to five years apiece — while also building the coal, the gas, the solar, the transmission, all of it, on the theory that a country’s ceiling is set by how much energy it can command.

One country treats energy as a sin to be managed. The other treats it as the substance the future is printed on. The thesis under Morris’s spreadsheet — that energy is upstream of everything, that abundant power is a growing everything — isn’t really his. It’s the operating assumption of the fastest-industrializing power in history, and the abandoned assumption of the country that used to hold the title.

The Efficiency Trap

The incumbents are trapped by their own brilliance. For thirty years GE and Siemens competed to save utilities fractions of a cent on fuel, pushing combined-cycle efficiency past 64% by driving inlet temperatures to a blistering 1,600°C. At that heat a blade can’t be forged. It has to be grown from molten nickel superalloy as a single continuous crystal, then drilled with microscopic cooling holes so it survives gas hotter than its own melting point — a masterpiece of materials science, and a hostage situation.

Maybe four or five companies on earth cast single-crystal blades well; two of them, Precision Castparts and Howmet, hold 80% of the capacity, and their books are full. A third of a turbine’s bill of materials is forged, on two-year lead times, by a handful of presses. GE can’t build turbines faster because GE requires perfect blades.

The villain of this story is not GE. It’s the forging press with the two-year waiting list.

Morris’s heresy is to walk around it. He won’t publish his spec sheet — “I’m not giving you my secret sauce” — but the doctrine is explicit. “GE and Siemens got nerd-sniped by the efficiency math,” he says. They traded near-term margins for long-term fragility, and lost the ability to make new things.

A Dumber Turbine, Faster

Chasing the last decimal made sense when America imported expensive oil. Gas is cheap now and staying cheap, and a megawatt-hour is worth what it’s never been worth before, because megawatt-hours become AI tokens and tokens become dollars at margins the grid has never seen.

80% of a machine’s cost is locked the moment you pick its architecture. So Stone picks a different one — “the architectures of old, ‘60s, ‘70s, ‘50s even,” with materials a generation behind the aeroderivative frontier. Run it cooler and the single-crystal priesthood becomes optional; the supply chain widens from a few hostage-takers to something a startup can buy from, or build. Trade points of efficiency for years of schedule. The target, borrowed from SpaceX, is the “idiot index” — a turbine costs about sixty times its raw materials, and Stone means to halve that.

A dumber turbine, by the thousand, faster than anyone.

Acceleration Through Simulation

The other half of the heresy is who designs it. For twenty years Silicon Valley has run on the assumption that you can software your way out of anything, and the hyperscalers are now colliding with the parts of the earth that assumption doesn’t cover. You can’t code a forge. You can’t prompt a fifty-ton rotor into existence. And the people who knew how to design these machines are mostly gone — retired, dead, or building rockets. By Morris’s count a few hundred master turbomachinery engineers are left in America, and a clean-sheet turbine has always eaten a few hundred of them for half a decade.

You can’t code a forge, but you can simulate a turbine. Stone’s answer is what Morris calls “building the world’s best turbine engineer” — an AI workflow that chains the industry’s existing high-fidelity simulation tools together and runs the thousand tedious button-clicks between a design decision and its aerodynamic verdict, standing in for draftsmen who no longer exist. It became possible, he says, only in the last six or eight months, thanks to AI.

The buyers have already voted for the doctrine. Crusoe ordered 29 modular GE turbines — about a gigawatt — for the Stargate site it’s building for Oracle and OpenAI, rather than wait for the grid. xAI ran its Memphis Colossus on some 35 mobile turbines because waiting meant missing a training run. Nobody asked about thermal efficiency.

The Stone Power Master Plan

1. Sell a deliberately worse gas turbine, now.

Revive a simple 1960s architecture, run it cooler, design out the single-crystal blade and the sold-out forge. Trade efficiency for schedule and sell it to a data center that would rather have a running turbine in 2027 than a perfect one in 2031.

2. Minumum viable vertical integration.

Don’t vertically integrate the whole machine. Own the forging and the specialty casting, the two-supplier bottlenecks the whole industry strangles on. Buy the sheet metal like everyone else.

3. Design around the forge until you can become the forge.

The near-term moat is dodging the bottleneck and growing the pie of gigawatts. The long-term moat is owning it — building a dozen machines, then fifty, then sixty, each one teaching a manufacturing lesson the incumbents outsourced away.

4. Profit from service contracts.

The incumbents make roughly 80% of their money on long-term service agreements, then farm out the actual servicing and keep the paperwork. Stone wants to keep it, sells the machine as a service, and feeds the data back into the design loop.

5. Go nuclear.

A turbine is 80 to 90% common across gas, nuclear steam, and geothermal. Use the gas years to build the team, the factory, and the forge, and be standing there — dozens of machines deep in process knowledge — when the nuclear renaissance finds the hole in the middle of itself. Sell every reactor company its hardest part on a silver platter. Nuclear is the prize. Gas is the tuition.

Read the two axes together and the plan is a single arrow, from the humble corner to the throne GE currently sits on. Every incumbent walked that arrow once, then pulled the ladder up behind them. Morris is betting he can walk it again before the money runs out, and that the industry has forgotten how.

Nuclear Purgatory

On September 16, 1954, the chairman of the Atomic Energy Commission, Lewis Strauss, told a room of science writers their children would enjoy electricity “too cheap to meter.”

In constant dollars a kilowatt-hour fell from $4.79 in 1902 to $0.32 by 1950, and kept falling. For seventy years electricity did what almost nothing else in economic history has done — it got cheaper every decade — and the country organized itself around the certainty that it always would. Demand doubled every ten years. At the peak the US Atomic Energy Commission projected a thousand nuclear reactors by 2000.

In October 1973 the trend broke. The three largest year-over-year electricity price jumps ever recorded came in 1974, 1980, and 1981 — the Arab embargo and the Iranian revolution, arriving at the meter. What followed was a massacre in paperwork: every one of the forty-one reactors ordered in America after 1973 was canceled.

The country built about a hundred reactors instead of a thousand and stopped. A generation of utility executives learned the lesson at bond-rating gunpoint — never build ahead of a forecast — which is why, when demand came roaring back fifty years later, nobody had.

Low Energy America

Everything peaked at once and never recovered. Per-capita energy use topped out in 1979; the average American now runs on less energy than his grandfather did.

Manufacturing employment peaked in 1979. This is not a coincidence.

Primary aluminum peaked in 1980, and the last new American smelter opened that same year. Aluminum is the cleanest confession in the record — it’s congealed electricity, and when the power stopped getting cheaper it left. Four smelters survive, under 2% of world production. A competitive one needs power around $40 a megawatt-hour; AI companies are paying $115 and up and still can’t get enough. The industry America lost for want of $40 power is watching the new one pay triple, into a shortage.

What died in the 1970s wasn’t cheap energy — American power stayed cheap by world standards. What died was the expectation that it would keep getting cheaper, and expectation is what engineers build for.

The technical culture flipped from make more to waste less, and the single-crystal turbine blade is a monument to that flinch — a masterpiece built to shave fuel in a world where fuel would be dear forever.

Morris says GE got nerd-sniped. Traumatized is closer.

The incumbents are still designing for 1979, and so is everyone around them: the utilities that won’t build ahead of demand, the bankers who remember the bond defaults, the regulators working in the shadow of Three Mile Island. Morris was born fifteen years after the wall went up. He may be the first American power engineer who never learned to see it as a wall.

The Ghost of Jack Welch

Scarcity was the first doctrine. The second one has a name, and it explains why the wall stayed up even after the energy came back.

America has lost two-thirds of its foundries since 2000. A large power transformer — the other machine every data center and every plant needs — now takes about 128 weeks to get, and roughly 80% of the big ones are imported; one American company, Cleveland-Cliffs, still makes the grain-oriented electrical steel inside them. The ultra-large rotor forgings at the heart of a heavy turbine come mostly from one firm, Japan Steel Works. The furnaces that melt aerospace superalloy are booked solid, and the next big American one doesn’t fire up until 2027. Run down the bill of materials of the American grid and every line reads the same: one supplier, two suppliers, an ocean away, sold out.

Earlier I said the villain was the forging press. Let me amend it. The press is the murder weapon. The doctrine that left a continental economy with a handful of presses has a name and a face, and the face ran the company Morris is trying to out-build.

“What I’m most scared of is not being up to the challenge.”

Jack Welch took over GE in 1981 and became the most celebrated executive in America by discovering that Wall Street would pay you for subtraction. Close a plant, the stock rose. Shed a division, the stock rose. He cut more than a hundred thousand jobs in his first years — Neutron Jack, the buildings left standing, the people gone — and grew GE Capital until a company famous for building things made a startling share of its money as a bank. Fortune named him Manager of the Century.

And none of it was stupid; all of it worked. Every CEO in America studied the playbook because the playbook made shareholders rich. Heavy factories were liabilities. The smart money went asset-light: keep the brand, the IP, the lucrative service contracts; sell the foundry, offshore the forge, fire the mechanics. Each quarter was defensible. Wall Street applauded every one. The sum is a country that can design anything and manufacture almost nothing, on the most efficient and most brittle supply chain ever built.

When Morris calls GE “a technical manual company,” he thinks he’s insulting a competitor. He’s describing the finish line of a forty-year doctrine, and the American economy. Welch died in March 2020; the next year his conglomerate announced it would split into three. GE Vernova — the turbine maker with the 100-gigawatt backlog, the one selling waitlists into the 2030s — is a fragment of the ghost’s estate.

Token Power

Then the AI boom hit, money arrived in quantities nobody had seen, and Silicon Valley discovered what had been optimized away. You can’t train a model if you can’t turn on the lights.

The trap is circular, which is what makes it systemic. Reindustrializing — now the stated ambition of both parties — needs cheap, abundant power, because energy is 10 to 40% of the cost of everything. But cheap, abundant power needs exactly the industries that left: the forges, foundries, castings, transformers, heavy steel.

Energy needs manufacturing; manufacturing needs energy; the country is locked out of its own house with the key inside. And it’s bigger than the AI story everyone’s telling. By Morris’s read, data centers were only about 30% of the net new power the grid added last year. The rest was factories, oil and gas, the grid itself — a whole economy trying to plug back in.

None of this is finally about gas. A turbine is 80 to 90% the same machine whether the steam comes from burning gas, splitting atoms, or the heat of the earth — so the cheap-gas years are tuition: they buy the team, the factory, and the forge.

The plan is to be standing there, dozens of machines deep in process knowledge, when the nuclear renaissance goes looking for the part it forgot to design. The reactor company that hadn’t thought about its turbine wasn’t an anomaly. It was the whole market, previewed. Gas is the tuition. Nuclear is the prize.

The Race

Morris is betting on a timeline, and he has his doubts.

“Frankly, what I’m most scared of is not being up to the challenge,” he says. Quality — his home discipline — “is the gating factor in scaling.” The hardware graveyard is full of companies that verticalized too fast and bled out; he names Rivian and Relativity unprompted, which is why he means to own only the true bottlenecks and buy the rest.

The critics get their strongest shot, and it has a name. Sean Hughes sells Wärtsilä’s reciprocating engines to data centers, and this spring he made the case on the record: against the simple-cycle turbines a data center would actually run, his engines are 10 to 15% more fuel efficient; they chase AI’s spiky loads, which leap “from 20 to 100 miles an hour in the blink of an eye”; and they leave the factory in about two years, against five to seven for turbines. That’s not a brochure, it’s an order book. Wärtsilä booked 412 megawatts, forty engines, for one Ohio data center this spring, part of more than 1.6 gigawatts across four American projects.

ProEnergy is delivering refurbished 48-megawatt jet-engine conversions next year, 650 megawatts of them to Crusoe alone. Boom’s Superpower holds a $1.25 billion launch order for 2027 — which Morris waves off as “a cash grab” that reshuffles the constrained supply chain instead of growing it, a claim that’s at least falsifiable, since Boom’s design lives inside the casting bottleneck and Stone’s whole brief is to live outside it.

In San Francisco, a twenty-something Thiel Fellow named John McElhone is building American Turbines to print the hot section outright — 3D-printing the wheels and combustor a data center’s swarm of small machines would need — a bet that tries to delete the casting bottleneck rather than dodge it, and stacks an unproven manufacturing process on an unproven machine to do it.

The customer doesn’t care whether the gas burns in a turbine or a piston. And every competitor named here says they’ll ship before Stone has cut first metal.

It Might Explode

Then the money, which is the shortest entry because there’s so little of it. Stone has raised about $2.5 million. Million, with an M.

Arbor Energy, down the street, raised $55 million just for a 25-megawatt pilot; Boom has raised $300 million on a decade of prior capital; Mainspring more than $800 million. GE Vernova booked $18.3 billion in orders last quarter, which means Stone’s entire capitalization flows into GE’s order book about every twenty minutes.

Morris doesn't pretend the $2.5 million gets him to a turbine. He says the plan is customer money — joint-development deals he expects to reach hundreds of millions by the middle of next year — but he admits freely that he'll need to raise more equity to get there. First-of-a-kind rotating machinery has a century-long habit of eating budgets a hundred times this size.

“Today, we’re executing violently.”

Underneath the commercial race is a question nobody on record has answered: can a turbine that gives up the perfect blade hit the reliability a hyperscaler demands — five nines, thirty years, 3,600 rpm? Throw one poorly cast blade at speed and it becomes a multi-ton shrapnel bomb. Stone has found the one door in the industry the incumbents can’t follow it through without abandoning seventy years of their own architecture. It is also attempting the thing that has historically eaten billions and left corpses.

His engineers have at least seen the inside of the cathedral: turbomachinery and rocket people out of GE’s orbit, Blue Origin, Pratt & Whitney, from a profession with a few hundred living practitioners. People with nothing to prove chose the seed-stage shop in Hawthorne over the institutions they helped build. That doesn’t prove the machine works. It’s the strongest signal a company this young gets to send.

Killed in Action

Ask Morris where the drive comes from — the nerve to bet $2.5 million against General Electric — and he doesn’t talk about electrons. He talks about a Sunday when he was eight.

Gurnee, Illinois. April 4, 2004. The knock came before dawn: two men and a woman in dress uniform on the doorstep. Upstairs, a light sleeper heard what never happened in that house — his father’s footsteps on the stairs. He was carried down to a kitchen where, impossibly, breakfast was cooking, sat on his father’s knee, and told that his oldest brother — Private First Class Geoffrey Morris, nineteen, “Jeff” to everyone — had been killed by a rocket-propelled grenade in Iraq.

The Morris family lost it. The father fell apart, the mother fell apart, the sisters nearly dropped out of school, food went uncooked. Over the next two and a half years, seven more of the town’s young men came home the same way. The house came apart, and an eight-year-old decided the weight was his: he and his father began organizing Gurnee’s Gold Star families — the golf outings, the fundraisers, a memorial fishing tournament that became one of the biggest on the Great Lakes.

Originals is proudly sponsored by

Cleaner financings and secondaries for private companies.Learn more here.

His father tried to go further: a memorial in the center of town with the names of every local man killed back through Korea. He raised the money and gathered the names. Then he ran for office, the politics turned septic, and the permits died in the village’s hands. The statues were never cast. Not in Gurnee.

“Everything in this world is somebody’s passion project,” Morris says — every park, every memorial, every thing you see exists only because one person made it happen. He watched his father lose that fight, and drew the lesson that nobody hands you the thing you’re owed.

“Every founder’s got trauma,” Morris says. “Mine is that I had to go through that alone. And I refuse to let my team go through hardship without my leadership. I refuse.” What Stone builds, under the turbine, is what an eight-year-old built out of casseroles and golf outings when his family came apart: a team under load, led from the front. The machine is the means. The team is the point.

80% Solution. Executed Violently.

The Gurnee memorial was never built. But his father didn’t stop. Kirk Morris spent years hauling its centerpiece — a nine-ton steel column from the 91st floor of the World Trade Center’s North Tower, released by the Port Authority in his dead son’s name — around Lake County, looking for ground that would take it. In 2015, eleven years after the knock, the village of Wauconda gave the beam a home. The Heroes of Freedom Memorial stands there now, on village land next to somebody else’s police station, in somebody else’s town. The father who was denied a memorial built one anyway. It took a decade and different ground.

That’s the lesson Morris carried into every factory since: nobody grants you the permit. And if American reindustrialization happens at all, it will probably look like that beam — not the old thing rebuilt where it stood, but a workaround, late, on borrowed ground, dragged into place by someone who wouldn’t take the denial as final. At his last job he opened every standup with a line from his brother’s Marine Corps:

“Today, we’re executing violently.”

The first turbine is due on a customer’s land next year. He will own the forge.

I own equity in Stone Power. Portrait of Dylan Morris is an AI hallucination based on real photos of the man. Infographics were generated by Gemini. Chart was generated by Claude. Read about How I Work.

The robot is carving a human face, and it cannot see.

It works in a 1930s warehouse in Mt. Vernon, New York — a seven-axis arm the size of a small car, the kind that spot-welds chassis in Stuttgart, bowed here over a block of Carrara marble cut from the same mountains Michelangelo climbed to pick his own.

A diamond bit on its wrist turns at six thousand revolutions a minute. A jet of recycled water rides the cut so the diamond doesn’t burn itself away. The arm carries no cameras. It does not know there is a face. It is executing a path computed hours ago from a three-dimensional model — the same x-y-z arithmetic that mills an engine block, turned on a four-hundred-year-old art — and it would run that path with exactly this much tenderness if the file described a cylinder head instead of a brow.

We are built to find a human face in almost nothing — three marks on a wall — and the machine cutting this one, with what looks from across the room like patience, finds nothing at all: a sorted list of points in space, hit in order, all night, without fatigue. The figure doesn’t rise out of the stone so much as the stone is carried away from around it, until a shape we know is standing where a block used to be.

I wanted to be an architect in a golden age that no longer exists.

It is an astonishing way to make a statue. Micah Springut is not really in the statue business.

He is 42, wiry, restless, with the manner of a man who has spent his life taking apart systems that don’t work — and he despises almost everything about the way we build now: the thin aluminum window frames, the optimized glass, the gray concrete and the disposable panels bolted onto it. To fix it, he has bet his career on the most aggressively modern tool on earth: a giant robot arm.

He is fighting modernity with modernity. The statue is a sideshow. What he is hunting is everything around it — the city itself, reimagined and rebuilt in stone, beauty returned to the ordinary building instead of reserved for the occasional monument.

Paris Everywhere

Here is the fact the whole company turns on. Concrete is the most-used substance on earth after water, and its cement is cooked stone — limestone, torn out of the ground, then roasted past 2000 degrees until the carbon bakes out of the rock and into the sky.

Springut’s proposition is to skip the oven. Cut the limestone instead of burning it; stack the cut blocks, thread them with steel, torque it tight, and raise the structure itself — columns, beams, floors, up to fifty stories — out of solid stone, for less than a concrete frame costs. Not a stone façade hung on a concrete core.

A building made of stone: trucked to the site pre-cut, reinforced with steel and stacked like Legos.

He means to make the splendor of the nineteenth century cheap again — cheap enough to build with, by the block, by the city.

And that building does not exist yet. The robot is real and the marble is real; the city is a wager, and the people who pour concrete for a living are the first to roll their eyes at the timeline. Both things are true at once, and the gap between them is the story. To understand why a software entrepreneur is standing in a marble warehouse making this particular bet, you have to go back to the dream he gave up at seventeen.

A Teenage Dream

Springut is not a sculptor. He holds no degree in architecture or engineering. The wound is older than any of that.

His first job, two summers at an architecture firm in high school, was dragging HVAC ducts around a screen in AutoCAD. The drafting itself wouldn’t have bothered him, he says — what killed the ambition was what it was in service of: the same cheap boxes going up everywhere.

For the first time in a century, the price of carved stone is falling.

He gave it up. “What I really wanted,” he says, “was to be an architect in 1910” — when you built with stone, with ornament, with craftsmen, and the results still stand. “I wanted to be an architect in a golden age that no longer exists.” So he gave it up and did, by his own count, a million other things instead.

The story of Monumental Labs is the story of a man spending twenty years getting back to that internship — and arriving with a machine.

Labor and The Cost Disease

Almost everything humans make has gotten better and cheaper over time. Computers, clothing, food — newer is better. But the most beautiful buildings in the world were, with embarrassing consistency, built hundreds or thousands of years ago. That is backwards. What happened?

Nothing mysterious. The price of labor happened.

The economics even has a name: Baumol’s Cost Disease. In the 1960s the economists William Baumol and William Bowen noticed something strange about a string quartet: it takes four players exactly as long to perform a Beethoven quartet today as it did in 1800. The work cannot be sped up — there is no productivity to wring from it — and yet the musicians’ wages have climbed all the same, because a cellist who couldn’t earn a modern wage would leave and go do something the modern economy rewards.

That is the cost disease: the price of any work that resists automation rises relentlessly against everything that automation has touched. And it is the entire history of carved stone. A carver in 1900 cut marble at roughly the rate of a carver in 1500 — the hands never got faster — while the world around him got cheaper by the decade and his wage had to chase it. The human hour, the one input no one could automate, was simply stranded at a price that only ever rose.

Look at the two marble lions on the steps of the New York Public Library, carved in 1911 by the Piccirilli Brothers — a family of master carvers from Massa, near Carrara, working out of a Bronx studio — from clay models by the sculptor Edward Clark Potter.

In today’s dollars, the pair cost about $460,000. In 2019 the city cleaned them, and the cleaning alone ran $250,000. A traditional shop asked to recreate the pair today would quote close to a million.

Cheaper than 1911

Monumental can carve, finish, and install both lions now for $250,000 to $350,000, and Springut expects that to fall toward $150,000 as the robots scale — seventy-five thousand dollars a lion, for ten feet of marble.

Read that again. For the first time in a century, the price of carved stone is falling.

The architects tell a more flattering story, and it is partly true. In 1910 the Viennese architect Adolf Loos gave a lecture called Ornament and Crime, calling decoration primitive — a thing for savages and degenerates.

The Bauhaus and the Modernists took up the creed, stripped the building down to glass and steel and the right angle, and made cheapness a moral position. But Springut thinks the ideology arrived to bless a decision the spreadsheet had already made. Ornament didn’t fall out of fashion and then vanish. It became unaffordable, and the philosophy came afterward to call the loss a virtue.

The numbers are brutal once you put them in the same dollars. Blend the skilled and unskilled hands on a civic building in 1900 and on-site labor ran about $12 an hour in today's money. The same hour of union bricklaying in New York now runs about $130.

There is the whole disease in a line: the identical human hour, for work that never once got faster, grown more than tenfold dearer in real terms while everything machines could touch got cheaper around it. Springut makes it physical — the carved ornament that runs $1,800 a square foot today was, back when labor was the thing in surplus, the cheap part of the building. An army of cheap labor built the Upper West Side he loves, and that army is gone.

The cost disease has only ever had one cure: productivity — a way to make the same thing with fewer hands.

For five centuries of carved stone there was never close to enough of it; the hand was the work, and the hand could not be hurried. That, Springut believes, is what the robot finally is — not the first machine to cut stone, but the first one powerful enough to bend the cost curve.

And concrete is built to die: the steel rebar that gives it strength eventually finds moisture, rusts, swells, and splits the structure from within. Every reinforced slab is on a clock — a lifespan measured in decades, not centuries.

A Career of Killing Middlemen

To understand why a software entrepreneur is the one doing this, follow a career that looks from a distance like a drunkard’s walk and up close is a single instinct applied to ever-larger targets: find a system where a necessary thing is absurdly expensive, locate the bottleneck, and build a machine to destroy it.

He didn’t start in startups. He studied American history at Columbia and Chinese politics at Harvard, learned fluent Mandarin, and meant to advise the president on China — a plan that dead-ended in a windowless room in Virginia, reading classified cables as an intelligence analyst. “The lack of sunlight was really depressing,” he says.

Groupon rescued him, almost as a joke. The company was about to go public and recruiting Americans to run its chaotic China unit, and Springut went to the interview figuring it would at least make a good story. The interviewer asked what number he’d need. Springut named a figure he meant in Chinese renminbi; the man heard it in dollars and slid the dollar number back across the table — 8x what Springut had in mind. “I’ll take that job,” he said.

Software is easy to copy. Twenty-ton blocks of marble, and the bureaucracy around them, are not.

He was very good at it. Within two months his sales team was the top-grossing in the country. When the foreign management was purged he was handed all of sales — two thousand people — told to fire half and sell the wreckage into the Tencent joint venture. He did, and walked away with the lesson that would define him.

“I love scaling companies. Descaling companies is not fun, but wild and exciting just the same.” He also learned the colder one: any business without a real technical moat gets cloned to death. Software is easy to copy. Marble is not.

Then legal recruiting. Headhunters took 25 to 30% of a lawyer’s salary to make an introduction; Springut built a platform to cut them out, and for a while was the most prolific placer of associates in big law in the country. The firms saved a fortune, the lawyers got a bonus, the middleman got erased. Six years in, it bored him to death — “connecting pampered lawyers to the same job down the street.” So he picked up a chisel.

Four Hours a Day

This is the hinge of the whole story, and it is small.

Around 2018, sick of the company he’d built, Springut started taking four hours out of every working day to carve stone by hand at the Art Students League on West 57th Street. It’s a strange thing for a founder to do — disappear half the workday into a studio — and he half-explains it as relaxation. But the real reason is the tell. He wanted to feel the physics of the material. He carved limestone and marble with hammer and chisel, made an eye of Michelangelo’s David, made grapes off a limestone cornucopia, and the whole time he was asking one question: how the hell do I bring this back?

He came home with an answer he didn’t want. The reason the beautiful world stopped getting built was not decadence or amnesia. It was arithmetic. The handcraft simply cost too much. The romantic had gone looking for a way to restore the chisel, and the operator had discovered the chisel had to die.

That is the entire man: a romantic who does the math, and lets the math win.

He founded Monumental Labs in 2022 on what he cheerfully admits was bluster.

“I bullshitted my way into believing it could be true.”

It turned out there were already people in Italy carving with robots, so he could pull a lot off the shelf; he asserted that machine learning could automate the rest, to people who knew enough to tell him whether it was insane. They said it wasn’t. He raised $500,000 from friends and family, got one robot through the door, and watched it work. Once people could see the statues, the money got easier.

Today the company has raised roughly $18 million from Alexis Ohanian’s Seven Seven Six, Shopify’s Tobi Lütke, Stripe’s Collison brothers, Coinbase’s Fred Ehrsam and me. I invested in June 2023, before the big names and big money arrived.

The Robot and the Rasp

Here is where most versions of this story go wrong, so it’s worth getting right. The robot does not replace the artist. It replaces the apprentice.

Springut runs two tribes under one roof. The technical side — software engineers and digital fabricators who turn a sculpted 3-D model into the machine’s toolpaths — and classically trained carvers from places like the New York Academy of Art and Grand Central Atelier, who finish by hand what the machine roughs out.

The division isn't new — it's the oldest arrangement in sculpture. The artist conceives the model and almost never cuts his own stone; from Greece and Rome through Bernini and Rodin, a team translated the master's clay into marble. By the 18th century the translation had a machine of its own: the pointing machine, a mechanical triangulator that measured a spot on the model and told the carver exactly how deep to cut. The robot, in other words, is a pointing machine with a PhD.

Those aren’t sculptors. Those are craftsmen

This is the thing that makes Springut hold his temper online. He posts a clip of an arm milling marble and the replies fill with mourning — art is dead, sculptors are finished — and he finds it both exhausting and ignorant of how sculpture has always worked.

“No,” he says. “Those aren’t sculptors. Those are craftsmen. They’re replacing the old craft industry.” It is the difference between Apple and Foxconn. He is automating the Foxconn.

The proof is already shipping. Monumental milled a six-foot nude in Carrara marble — clutching, in a private joke, an expired test credit card — for the payments company Stripe: roughed by the arm in days and finished by a few carvers in a couple of weeks, a piece that by hand would have run well into six figures and taken years. The machine did the brute labor; the carvers did the part that reads as art.

Humans in the Loop

He cannot automate all of it, at least not yet, and he’s honest about which part. The spinning bit bruises marble — leaves a murky skin where it cuts — and its rounded tip can’t reach a sharp interior fold. So a carver trained at a classical academy lifts that last bruised dusting away by hand with a rasp, and only then does the marble do the thing marble does, which is glow: the dead gray peels off and a smooth, translucent surface comes up underneath.

The robot cannot make marble glow. The hand can. Which is why a life-size marble sculpture might run 600 hours on the machine and then 900 hours of human finishing.

The hand does more than the robot. That will always be true for fine art, which is why fine art will never be the business: lovely, maybe $50 million a year someday, and unscalable.

The real bet is architectural limestone. No translucent surface required; there the ratio inverts to 300 robot hours and 30 by hand (and falling). That can be driven toward full automation, and that is where the labor cost finally collapses.

The collapsing is itself an engineering project, run by a CTO named Sebastian Marino — a man who, before he taught robots to plan their own cuts, taught pixels to behave like flesh, and won an Academy Award for scientific and technical achievement for the cloth, skin, and hair simulation behind James Cameron’s film Avatar.

His software is named Hilbert, after the mathematician. Roughing out a bust of the emperor Augustus once cost a fabricator several days at the keyboard; Hilbert does it in minutes.

And when the keepers of the craft come after him — one traditional carver called robotic routing an “abominable degradation” of the art — Springut answers not with theory but with his payroll. He employs more than a dozen classically trained sculptors and fabricators, more stone cecarvers, he says, than anyone else in New York City. The robot takes the lung-shredding roughing that historically broke poor men’s bodies; the human keeps the part that was always the art.

Cheaper Than Concrete

The lions and the busts and the Stripe nude are the wedge — real, profitable, and not the point. Springut would have been happy, he says, to run a company that just makes ornament.

But you can’t scale it, because almost nobody wants to hang heavy stone on a modern building. A concrete-framed tower is already carrying its own enormous weight; bolt a stone facade onto it and you need a steel shelf drilled into the concrete, the stone hung off the shelf and shimmed to look as if gravity were doing the work — and the code won’t even let you simply mortar it on. Ugly, and the reason no one bothers.

Most folks in the building industry will immediately roll their eyes at this claim.

Which is why the bet had to get bigger. Not cladding the building in stone. Building it out of stone — columns, beams, joists, structure, all of it solid, cut in a factory, trucked to the site, and stacked like Lego. No concrete core, no air gap, no rain screen, no caulk. Stone meeting stone.

The objection writes itself: stone is brittle. It’s magnificent in compression and miserable in tension, which is exactly why the twentieth century chose concrete, which flexes. The fix is a sixty-year-old idea borrowed from precast concrete and called post-tensioning: drill a channel through a run of stone blocks, thread a high-strength steel cable through them like beads on a necklace, and crank it tight. The compression lets the brittle stone behave, under load, like reinforced concrete.

A post-tensioned stone beam, Springut says, carries roughly three times the compressive load of concrete and most of the flex of the reinforced kind — enough to raise a building of up to fifty stories. No buttresses, no cathedral-thick walls. A near one-for-one swap for a concrete frame.

And the supply, he insists, is effectively infinite, for the reason that lands like a punchline: it’s the same limestone we already crush into cement, sitting under half the country.

The Monumental Labs Master Plan

* Ornament and restoration, now. Carnegie Hall, the Frick, the de Seversky Mansion, gargoyles for the University of Michigan, the company’s first work going into the Vatican. Real revenue, de-risked.

* High-end New York homes, next. Passion-project homes, pool houses, arches, monumental columns, a tower — projects where the client doesn’t blink at an extra $300,000 because the beauty lets him sell the penthouse for $40 million instead of $35 million.

* Whole buildings and megafactories, after. Two-hundred-thousand-square-foot plants next to rock quarries — Indiana, Texas, Utah — with redundant machines and robots moving the stone, where a building becomes a numbered kit cut in weeks, and the marginal cost of a piece of stone trends toward the price of the electricity.

* Paris everywhere, eventually. Stone at half the cost of concrete, by his projection, within a few years. “We can build new Parises left and right.”

The Establishment Strikes Back

When Springut went viral claiming stone will soon beat steel and concrete on price, Brian Potter of Construction Physics printed the industry’s reaction:

“Most folks in the building industry will immediately roll their eyes at this claim.”

Even if stone won on cost, Potter argued, rewriting codes, building supply chains, and training workers would keep it from winning within the decade.

Springut doesn’t dispute it. “We have to change the laws,” he says — no code for post-tensioned structural stone, no testing protocol, no lender who’s ever financed it. His plan is the Groupon playbook for bureaucracy: get one building department comfortable, then expand. Mass timber walked that road. It took two decades.

And on what matters most, he’s behind. The revival didn’t start in the New York; it started in London, where architect Amin Taha and engineer Steve Webb spent a decade building what Springut is still drawing. Taha’s 15 Clerkenwell Close, a six-story load-bearing stone exoskeleton, was shortlisted for the Stirling Prize in 2021. They ran the fire tests. For his own first structural building, Springut hired Taha to design it — importing the precedent he lacks. The arms are off-the-shelf KUKA units; in Italy, Robotor already sells a basic version of toolpath software his team is perfecting. His edge is the two-tribe model, the vertical integration, the salesmanship. Moat or head start, we’ll find out.

Originals is proudly sponsored by

Cleaner financings and secondaries for private companies.Learn more here.

So treat it as two companies. One is ornamental and restoration — real, deliverable today, worth hundreds of millions. The other is the structural bet, an option on how every building on earth gets made. The first is a business. The second is a wager.

The instrument that settles it is being cut right now on his floor: a thirty-foot post-tensioned stone tower, no steel frame, built to make a skeptical developer stand at its base, look up, and put a hand on the wall. Until it stands, and one structural building gets permitted in an American city, the claim is what Potter says it is — aspirational.

The Architect, Again

Late in our conversation I told Springut what his own roadmap had made obvious: he’s going to be an architect again. The childhood dream he threw out at seventeen.

“Yeah,” he said. “Directing architects.”

That’s the plan he can’t quite keep the boyhood out of his voice describing. Find the ten most brilliant young architects — the ones now paid $60,000 to move HVAC ducts around on CAD, exactly as he was — pay them whatever it takes, bring them in-house, and let them design insane buildings out of stone.

A new golden age of civic splendor that puts Parthenon to shame. Monumental Labs as the Apple of a new stone age, powered by an army of soulless machines.

He may be right. He may be a romantic who has, for the second time in his life, talked himself into something true. The arithmetic of the lions is real — it really is cheaper to carve them now than in 1911, and that genuinely is new in the world. The arithmetic of a fifty-story stone tower competing with concrete is not yet real, and the people closest to it are the first to say how far the road still runs.

He has spent his life proving that any cost can be engineered toward zero. This is the one cost he cannot model — only build, and find out. The tower will tell him whether the stone holds. It cannot tell him whether the world will let him build the rest.

Back on the floor, the robot arm lifts, resets, and finds a new angle on the marble. The water hisses. A carver waits off to the side for the machine to finish, so the real work can begin.

I own equity in Monumental Labs. Read more about How I Work.

Alex Mathiasen was seven the first time he understood that everyone he loved would die. He sat on the floor and cried, and his mother gave him the only comfort any parent has for this particular arithmetic: don’t think about it.

He ignored her.

He is thirty-two now, a Danish mathematician with a PhD in algorithmic optimization. For his entire adult life he has aimed that one skill at a single target, and it has carried him here — to a company called Vitrify Labs, and a proposition that sounds, on first hearing, insane.

Scale beats clever. Even in biology.

Mathiasen wants to build a pause button for human biology. Stripped to the studs, the idea runs like this. Take a person who has run out of time — late-stage cancer, an organ failing, the disease that took his grandmother — and freeze them in suspended animation, like Han Solo in Star Wars. Hold them there, unchanging, at 196 degrees below zero, for as long as it takes medicine to invent the cure they need. Then warm them up. The same person, resumed.

He does not think this is science fiction. He thinks it is an engineering problem the world has been too unimaginative to take seriously — and that the last obstacle between us and that pause button is a poison he can compute his way around.

Scale Is All You Need

You would be within your rights to stop reading. The dream of freezing the body and waking it later has a long and faintly disreputable history — frozen heads, desert dewars, promises that conveniently cannot be checked for a century. But Mathiasen is not a man you wave off.

Strip away the titles and the whole of him is a gift for the shortcut. During his PhD at Aarhus University he built a parallel algorithm for one of the slowest operations in machine learning — decomposing a dense matrix — and made it run, in the best case, 27.1x faster than the standard method. The paper went to NeurIPS, the field’s most selective conference, in 2020. Then, at the chip company Graphcore – an AI chip company bought by Softbank for $600M to compete with Cerebras and Groq – he ran into the lesson that now drives his company: in machine learning, scale wins.

Big Pharma buys drugs, not platforms.

Feed a simple model a mountain of data and it beats a clever model fed a molehill. In biology the mountain doesn’t exist, because nature is expensive to measure — so he built it, simulating molecules on Graphcore’s chips until he had a dataset of 1B examples. The previous record was 20M. It had taken two years on conventional supercomputers; he produced 50x more in days.

That is the thesis: scale beats clever, even in biology.

And here is the tell. He could have sold that dataset as the definitive tool for chemical machine learning. Instead he published its flaws — in his own hand, in the documentation, he warned other researchers that his shortcuts had introduced errors and that they must not use his data to rank their models. In a field full of people overselling, he undersold his own billion-molecule monument. Hold onto that. It is rarer than the math.

Elegant math wins academic prizes. Functional math ships products. Mathiasen builds functional math.

Why Not Just Start a Company?

During a seven-month stint at Charm Therapeutics in London, he learned the broken economics of drug discovery startups up close: Big Pharma buys drugs, not platforms. So the bio teams built narrow specialists — one deep-learning model in service of one drug against one cancer — not GPTs, not the general-purpose transformers powering Anthropic and OpenAI.

The bio people were ignoring the scaling laws.

Build a GPT that understood biology the way GPT-5 understands language, and drug discovery becomes a query against it. The only open question was the cost of the data, which — unlike writing on the internet — does not yet exist in digitized form.

Cryo scaling is beating Moore’s Law. 100x in 3 years.

But every dollar at a biotech chases the next molecule, because the next molecule is what gets the company acquired. A biology GPT would cost at least as much as starting two more drug programs — so nobody starts one. Single molecules are the business model.

“There’s a 100% chance it works,” laughed Etminan. “It’s just we can’t tell you how much it’s going to cost. It could cost a million bucks. It could cost a trillion.” Mathiasen felt stuck.

Then a venture capitalist asked him the question founders spend their lives waiting for — why don’t you just start your own company? Thus was born Vitrify Labs, a company devoted to longevity without drug discovery.

Han Solo in Carbonite

His timing is not an accident. The ground shifted under this whole field in three years, and almost none of the shifting was his.

The trouble has always been one brutal fact of chemistry. Water wants to be ice. Cool it and the molecules snap into rigid six-sided crystals — Mathiasen calls it water’s “hexagon fetish” — and the crystals shred cell membranes like microscopic shrapnel.

The escape is vitrification: replace enough of the water with antifreeze, cool it fast enough, and the liquid never crystallizes. It goes solid the way glass does, a frozen liquid with nowhere to put its hexagons. At the atomic level, nothing moves. Cellular metabolism stops. It is, he says, “time dilation, but at the level of molecular diffusion.”

Or, put more bluntly: Han Solo in carbonite.

A single cell is easy — it is how IVF clinics bank embryos. The problem was always size. Bigger things hold more heat, and to rewarm them without cracking you must heat the whole mass at once, evenly, fast, which you cannot, because heat travels from the outside in. Mathiasen calls it the burrito problem: scorched at the edges, frozen in the middle. The fix is iron. Perfuse an organ with magnetic nanoparticles before cooling; when it is time to thaw, drop it into an alternating magnetic field and the particles vibrate, heating the tissue uniformly from within.

From Science Fiction to the Journal Nature

Magnetic nanoparticle perfusion has unlocked an epic run of scientific progress. Now cryogenic scaling is beating Moore’s Law. 100x in 3 years.

In 2023, in Nature Communications, a group at the University of Minnesota vitrified a rat kidney, rewarmed it, transplanted it, and kept the animal alive.

By November 2025, in the same journal, a team had pushed vitrification to liter volumes — organs the size of a pig’s liver.

In March 2026, in PNAS, a German neurologist named, almost unbelievably, Alexander German recovered a vitrified mouse hippocampus from 196 below zero with its long-term potentiation — the cellular signature of memory — substantially intact. The machinery of remembering survived the glass. Researchers are now showing pig kidneys at conferences.

Three dominoes in three years. And every one of those papers, read past the headline, names the same surviving villain. It is not the cold that does the damage now, nor the rewarming. It is the antifreeze itself.

The Poisonous Miracle of Human Antifreeze

This is where Mathiasen comes in, and what Vitrify Labs is actually building underneath the talk of solving death: a better antifreeze. He calls it the juice. The juice in use today is a brew of industrial solvents — dimethyl sulfoxide, ethylene glycol, formamide — pumped into a body to replace roughly sixty percent of its water.

It works.

He does chemistry inside an Nvidia GPU.

It also poisons the tissue, and the longer the molecules sit there waiting to diffuse, the more they poison it. The whole game is to find a cocktail that blocks ice, slips through cell membranes, and doesn’t kill the patient — and nobody has, because the space of possible molecules is effectively infinite and a wet lab tests a handful at a time.

An infinite search space; a toxic compromise nobody has optimized. It is, structurally, the problem Mathiasen has spent his career solving. He is not a chemist mixing vials at a black epoxy bench. He does chemistry inside an Nvidia GPU.

Simulating Quantum Physics with GPUs

The tools that make this thinkable are absurdly small. There are neural networks — on the order of 27M parameters, a rounding error beside the trillion-parameter LLMs pedaled by Anthropic and OpenAI — that have effectively learned quantum mechanics, simulating reality atom by atom in steps of a single femtosecond, one millionth of one billionth of a second.

The catch is speed: such a simulation has meant months of a postdoc’s coding and weeks of supercomputer time. Mathiasen sets autonomous AI agents to write the code in an hour and runs it in a day, hunting across millions of candidate molecules for the one that permeates the membrane, blocks the ice, and leaves the cell unharmed.

When he finds it, he will farm the physical synthesis out to contract labs, test it on animal tissue, and scale. In the world of synthetic biology, they call this approach, “Lab in the Loop.”

Two of the three properties he needs — blocks ice, crosses the membrane — are physics, the thing he taught a computer to simulate at Graphcore.

The third, toxicity, is the unknown, and he is unusually frank about it. He thinks the simulation can predict it. He also says he has reasons to believe the problem is hard, and then, in the same breath, that he isn’t sure he trusts his own reasons. That is the bet, without varnish: that the poison can be predicted in a computer before it is ever mixed in glass.

The Barbecue Hiring Plan

Set against the ambition, the company is almost comically small. Two people. A registered firm in Cambridge. A stack of validated credentials, a thesis that lands squarely on the open problem of a suddenly hot field, and no announced funding, no laboratory of their own. The hiring plan, when I asked, was a barbecue: Mathiasen recently threw one for friends from his old companies and figures the excited ones are the first recruits. They debate whether they want a one-pizza team or a two-pizza team. Their titles exist as a joke. “Tom executes,” Mathiasen says, “and I tech.”

Tom Etminan is the other half of the bet, and a corrective to the idea that this is one man’s death wish.

They want this to become a medical procedure rather than “a crazy alternative funeral arrangement.”

Tom is a Oxford-trained physicist whose last company, Metavoice, pioneered real-time voice cloning with small, fast on-device models.

His engine isn’t terror but impatience — a “time anxiety,” he calls it, too many things he wants to do and too few years to do them. He won’t paraglide anymore, because the cost of missing the future, if the future is as good as he thinks, is too high to gamble on a hobby. And he is not ballast: after Mathiasen spent weeks wringing a 4x speedup out of one of their models, Etminan came along with AI coding agents and pulled another 2x out of code his partner had already called finished.

The Master Plan

The arithmetic of the company itself is small enough to check. They think $10M, five engineers and a medium-sized GPU cluster can definitely answer the first question: whether the toxicity can be predicted in a computer, and a better juice found inside it.

“$1.5M is enough to de-risk 80/20,” says Mathiasen. “$10M gets us there faster and gives us maybe 95/5.”

After that, a staircase, climbed a rung at a time.

Vitrify’s 15-Year Product Roadmap

* Incrementally better cryo-juice.Juice that’s a little less poisonous than today’s — sold to IVF clinics that bank embryos and researchers who freeze cell lines.

* Larger tissue samples. Skin grafts, corneas and heart valves.

* Whole human organs.The prize is transplantation itself: the kidneys and livers that now get a few hours in a cooler strapped into the back of an ambulance, turned into a planned procedure performed on the surgeon’s calendar.

* Whole mice.

* Whole pets.

* Whole humans.

Each rung pays for the next. Each rung also tells them whether to keep climbing — or to stop.

A Medical Procedure, Not a Crazy Alternative to a Funeral

Mathiasen and Etminan work hard to stand apart from the cryonics tradition they so easily get mistaken for. They want to build upward from the provable — an organ, then a mouse, then perhaps a pet — verifying each step before the next. They want this to become, in Etminan’s phrase, a medical procedure rather than “a crazy alternative funeral arrangement”: the thing a hospital does when it has run out of options and would rather press pause than call the time of death.

Two facts sit beside each other here, and both are true. The science is real, it is moving, and the people moving it publish in the best journals on Earth. And the problem Vitrify exists to solve — the toxicity — is unsolved, by them and by everyone, and the researchers closest to the work are the first to say how far the road still runs.

Alexander German, whose mouse brain made the headlines, was blunt that better vitrification solutions — better juice — will be needed before any of this reaches a human organ. Which is either a warning or a job description, depending on where you sit. Others noted that even his success rate on a whole brain was low, and that cracking and heat transfer grow worse, not better, as things scale.

Originals is proudly sponsored by

Cleaner financings and secondaries for private companies. Learn more here.

That leaves Mathiasen at the edge of a thing that might be visionary and might be naive, with no instrument that can tell him which — an uncomfortable place for a man whose entire profession is measurement.

The commercial logic on the way there is sound enough. Better juice means more viable eggs banked per IVF cycle, sparing women repeated rounds of hormones for worse odds than they deserve. It means organs that survive longer than the few hours they now get in a cooler strapped into an ambulance, turning transplantation from a frantic race into a planned procedure. Large, real markets — enough to fund the climb.

Life Saver… or Lifestyle Choice?

But Mathiasen is not doing this to optimize IVF.

What he is selling, if it ever exists, is consent: the option, for the patient who has run out of time, to wait in glass until medicine catches up.

People will just freeze themselves because they’re annoyed.

It would also, he and Etminan freely admit, get strange. With a real pause button, people wouldn't reach for it only when they were dying. They'd reach for it when they were bored.

"Imagine a world where you have perfectly reversible freezing," Etminan says. "People will just freeze themselves because they're annoyed. Say a war's been going on for, like, twenty years. People will say, 'I'm so sick of this,' and freeze themselves with a note that says: 'Wake me up when the war's finished.'"

What Mathiasen cannot do is make it come faster. His parents are sixty. The juice that might save them does not exist, and the man whose whole gift is making slow things fast has no shortcut for this one — no parallel algorithm, no chip to throw at it. He is betting his career that fifteen years will be soon enough. It is the one calculation he cannot check.

Meet the humans building the future of civilization.

All images generated by AI. Portrait rendered from reference photographs of Alex Mathiasen. Not a photograph. Read more about How I Work.

The New Yorker for two guys in a garage. Fortune for founders you’ve never heard of.

Traditional tech media doesn’t lead. It follows. It waits for the funding round, the valuation, the IPO — for someone richer to certify that a thing matters — and only then sends a reporter to tell the story, usually one who can’t tell a transformer from a transmission.

Originals is the opposite. Fortune for founders you've never heard of. The New Yorker for two guys in a garage. A magazine about the founders, scientists, and assorted weirdos working on the things that will matter enormously in a decade — and that almost no one is paying attention to today.

Technology is the Substrate of Human Civilization

Bronze. Antibiotics. Automobiles. Computers.

Technology drives human progress.

Technology is the thing behind almost every gain in how long, how well, and how freely a human being can live.

A civilization that vilifies its visionaries is a dying civilization.

And yet the inventors of technology, the people pushing hardest on that frontier, on ideas most people think are crazy, have become the designated villains of the age.

Old Media Traded Truth Seeking for Activism

It wasn’t always like this.

Once upon a time in America, the press celebrated inventors. Steve Jobs was a national treasure. Sara Blakely was someone to emulate. The journalist’s job was to be curious — to go find out what these weird people were actually up to, and come back and tell you.

Somewhere in the last decade, that job changed.

Some of it was rational, of course. In the decade after the 2008 crisis, technology came to dominate the U.S. economy, driving the vast majority of GDP growth. The press, meanwhile, was collapsing: 2018 was the worst year for media layoffs since 2009 — and the year Apple crossed a trillion dollars.

One industry was inheriting the earth; the other was laying off the people paid to cover it. Seeing their industry implode, tech journalists traded curiosity for a verdict, and called it speaking “truth to power.”

Along the way, the distinction between Big Tech and the two guys in a garage flattened.

Now the story is almost pre-written: the founder is a proto-billionaire, presumed guilty of crimes not yet committed, slotted into a morality play with two roles: villain and victim.

The press stopped reporting on technologists and started prosecuting them.

This isn’t a complaint that the coverage is mean. It’s that it stopped being true.

I do journalism like a venture capitalist.

A civilization that vilifies its visionaries is a dying civilization.

America is sick — and our politics are fucked.

Faith in government has collapsed from 73% in 1958 to 17% in 2025. Faith in media fell from 72% in 1972 to 28% today – a new all-time low.

The majority of Americans have confidence in only three major U.S. institutions:

* Small business: 70%

* Military: 62%

* Science: 61%.

Americans haven’t turned on the people who build things. They’ve turned on the people who narrate them.

Originals Takes Technologists Seriously

Originals treats the crazy visionaries of our age — the humans pushing forward the bleeding edge of civilizational progress — with the seriousness the establishment reserves for billion dollar companies and the craft it reserves for high art.

Not press releases. Not hit pieces. Not heroes, not villains — subjects, worth the same rigorous, curious attention The New York Times give a senator or a novelist. Warm to the person. Cold and curious to the claim.

I take the people I write about seriously enough to tell you the truth about them, which is a harder and higher form of respect than flattery.

Who I Am

My name is Matt Mireles. I am the founder of multiple VC-backed companies, an inventor with 9 issued US Patents, and small-time investor in 30+ startups.

Once upon a time, I was an award-winning journalist and photographer for the New York Times and Newsweek. I paid my way through Columbia University as a 911 paramedic in the South Bronx.

I have done CPR on dead babies and dined with billionaires.

So I approach these stories the way a first-check investor does, not the way the press does: I don’t wait for someone else to certify that a company matters before I’ll cover it. I use my own judgment. And between the access that comes from being a peer and a working grasp of the actual science and engineering, I can go where most reporters can’t — into the technical details where the real story lives.

The first installment of Originals launches on Tuesday…

Death is an Engineering Problem

Alex Mathiasen wants to build a pause button for human biology – so that you can live forever. All he needs is enough GPUs to simulate quantum physics.

Meet the humans building the future of civilization.