
Sign up to save your podcasts
Or


Part II of THE ICE, the three-part narrated documentary series on polar expeditions, leadership, and margins. I am deliberately silent in this one — the narrator owns it. If you missed Part I (”The Drift and the Balloon”), start there; the series is one story.
Eight words
Melbourne, 12 October 1910. Captain Robert Falcon Scott, outbound with the largest scientific expedition ever sent south, comes ashore to find a telegram handed in at Kristiania nine days earlier. It reads, in full:
“Beg leave to inform you Fram proceeding Antarctic. Amundsen.”
No destination. No apology. And absolutely no mention of a race — which is precisely how you announce one. Roald Amundsen — the young man from Part I’s closing crowd — had watched Cook and Peary claim the North Pole and kill his life’s plan overnight. So he re-aimed his whole expedition at the bottom of the world and told almost no one: not the king, not Nansen, not even his crew until Madeira, where every man, asked in turn, said yes.
Two systems
By 1911 two wooden huts stood on the ice, four hundred miles apart — and two different theories of reality. Framheim: nine men, one purpose, workshops dug under the snow where sledges were planed to half weight, three tons of depots laid, marker flags planted in a numbered line miles wide so a sledge lost in fog could hit the line anywhere and know exactly where dinner was. When the flags ran out they marked the route with dried fish. Waymarks you could eat.
Cape Evans: two dozen men and a laboratory, the most ambitious science programme ever taken south — and a transport system of everything at once: motor sledges, ponies, dogs, and men in harness. On the way down, Terra Nova nearly foundered in a storm; the expedition survived by bucket chain. And in the autumn, with the ponies failing, Scott stopped his great depot 31 miles short of plan. It was named One Ton. Hold that name.
What the legend leaves out
The perfectionist cracked first. Spooked by the idea of British motor sledges, Amundsen started too early — into −56°, a retreat that became a rout, and an open rebuke at the breakfast table from Hjalmar Johansen, Nansen’s old companion, who had saved a comrade the leader’s dash had left behind. Amundsen’s reply was administrative and total: Johansen was struck from the polar party, came home unthanked, and within sixteen months had taken his own life. The episode lets that sequence speak for itself.
Then, in proper season: five men, 52 dogs, an unmapped glacier climbed almost at a run, and at the top — the Butcher’s Shop, where twenty-four dogs, by Amundsen’s own count, were “marked out for death” as planned freight. On 14 December 1911, five hands gripped one flagpole together. Amundsen’s verdict is the driest sentence in polar literature: never had anyone stood so diametrically opposed to the object of his desires. And at Polheim he left a letter to King Haakon — with a note asking Scott to deliver it. His rival, drafted as postman and insurance policy.
The long dying
Scott reached the Pole 34 days later, past a black flag snapping on a sledge runner. “Great God! this is an awful place.” Then the ledger came due: fuel crept from depot tins Amundsen would have soldered shut; the Barrier delivered a once-in-a-decade cold anomaly (Susan Solomon’s climate work, a century later, took Scott’s side); Evans died at the Beardmore; Oates walked out into the blizzard on his 32nd birthday — “I am just going outside, and may be some time.” The last camp stood eleven miles from One Ton — the depot laid thirty-one miles short. And at One Ton itself, dog teams waited eight days and turned home: Scott’s own relief order had died in transmission on the way. The man who waited, Apsley Cherry-Garrard, did that arithmetic for the rest of his life.
The search party found the tent in November: Scott half out of his bag, an arm toward Wilson, Amundsen’s envelope faithfully carried — and 35 pounds of fossil plants on the sledge. They died dragging the science.
Meanwhile the winner’s return read like an office memo: 99 days, eleven dogs, ten days ahead of schedule, the men heavier than they left. When a system works, triumph looks like accounting. When it fails, it looks like literature.
The last prize
And in London, one man read the news differently. Both Poles claimed — so the last great prize was the crossing, coast to coast through the Pole. He had once turned back 97 miles short rather than kill his men: better a live donkey, he liked to say, than a dead lion. He bought a Norwegian tourist yacht never meant for pack ice and renamed her after his family motto — the only name in this series that is also an instruction.
Fortitudine vincimus. By endurance we conquer.
Next, the finale: The Wreck.
Sources: Amundsen’s The South Pole*, Scott’s diaries (SPRI), Susan Solomon’s* The Coldest March*, Karen May’s Polar Record papers on the relief order, and Björn Lantz on the false start, among others.*
Something different this time: the first part of THE ICE, a three-part narrated documentary series about the polar expeditions — and about leadership, margins, and what happens when the story you tell about yourself meets an environment that does not negotiate. A new narrator carries the series; I only open the door. Part II, “The Race,” follows next.
It started, as these things do now, with a post on X. Someone mentioned a book about Shackleton. I bought it that evening and read it the way you drink water after a long run — and then I kept going: Scott, Amundsen, Nansen, the whole shelf. This podcast is usually about AI and energy and coordination. This series is about something older: leadership when the map runs out. It turns out the coldest places on Earth are the clearest mirror I have ever found.
So for three episodes I hand you over to a guide who has read all the diaries. Here is where she takes you in Part I.
A boathook on the White Island
August 1930. A freak warm year leaves the sea open around Kvitøya — the White Island — east of Svalbard, a dome of glacier ice so unreachable that most years the ocean simply says no. A Norwegian sealing ship, the Bratvaag, sends men ashore to hunt walrus. In a stream bed they find an aluminum lid. Then a canvas boat in a snowdrift. Then a boathook, engraved: “Andrée’s Polar Expedition, 1896.”
Thirty-three years of national mystery, solved by one line of engraving read aloud by a walrus hunter at the end of the world.
They find the men. A month later a second ship finds a tin box with film in it — and in a darkroom in Stockholm, out of 240 exposures that have lain a third of a century in the ice, 93 images come back from the dead. One of them stops the breath of everyone who sees it: a giant balloon, collapsed on the pack ice, two men standing beside what is left of their plan.
The ship built to be trapped
To understand the balloon you first have to understand the ship — and a decade earlier, the strangest idea in the history of exploration. When wreckage from the crushed American ship Jeannette drifted from Siberia clear across the Arctic to Greenland, Fridtjof Nansen drew a conclusion of magnificent, lunatic simplicity: if the ice flows across the Pole, don’t fight the ice. Build a ship it cannot kill — round-bellied, nothing to grip, designed to be squeezed upward like a pip from an orange — freeze it in deliberately, and let the Arctic itself carry you.
Everyone serious called it suicide. It worked almost boringly well. Fram spent three years locked in the drift with a library, electric lamps fed by a windmill, and nothing to fight but the calendar. Nansen’s diary from those winters is not a record of danger; it is a record of a man pacing: “I feel I must break through this deadness, this inertia.”
The episode’s hinge is what happens when Nansen leaves the drifting ship to ski for the Pole — a one-way door, since the ship keeps moving — and who he doesn’t take: Otto Sverdrup, his best man, who must stay, because someone has to bring the other eleven home. Glory is a resource like any other. Nansen budgeted it.
Two men reach 86°13.6′ North, farther than any human before them. Then Nansen looks at the chaos of ice between him and the Pole — and turns around. A winter in a stone hole on Franz Josef Land follows (nine months, one shared sleeping bag, and — gloriously — formal address until New Year’s Eve). And then a chain of survival so improbable no fiction editor would allow it: a swim after drifting kayaks, a walrus tusk through a hull, and — because of that forced repair stop — a dog barking where no dog should be, and an Englishman in a checked suit saying: “You are Nansen, aren’t you?”
On 13 August 1896 Nansen steps ashore in Norway. The same day — the same day — Fram breaks free of the ice on her own, every man alive, exactly as designed.
The Eagle
And next door, Sweden was watching. Sweden had no Nansen. What Sweden had was a balloon.
Salomon August Andrée’s plan had the clean geometry of an engineering drawing: hydrogen balloon, Svalbard, over the Pole in days. King Oscar II invested personally; Alfred Nobel backed it. And the balloon leaked — through millions of stitch holes, measurably, undeniably. The expedition’s own meteorologist, Nils Ekholm, did the math: seventeen days of lift for a thirty-day plan. He demanded a new balloon, was refused, and quit — and later discovered Andrée had been secretly topping up the hydrogen during the fill to mask the loss rate. The leader was not deceived about the leak. The leader was managing the appearance of the leak.
Örnen launched on 11 July 1897 and lost her drag ropes — the entire steering concept — within minutes. Sixty-five hours later, after two days of fog dragging the ice-heavy silk across the pack, three men stood unhurt on the ice at 82°56′ N: south of where Nansen had walked on skis. Then came the march the balloon was meant to spare them, across a drift that ran beneath their feet like a treadmill in the wrong direction. They shot bears, cooked proper meals, toasted the king in vintage port. In October they reached the White Island. “It feels fine to be able to sleep here on fast land,” Andrée wrote. Within days all three were dead — and to this day nobody knows what killed them. Every autopsy of the evidence finds a different killer, and the answer burned with the 1930 cremation.
Thirty-three years, almost to the day
They landed on Kvitøya in the first days of October 1897. The remains came home through Stockholm’s streets — a hundred thousand people, a king’s oration — on 5 October 1930.
Two details from the end of the episode that I have not been able to shake since we made it. Nils Strindberg’s fiancée, Anna Charlier, lived a long life abroad — and when she died in 1949, her heart was cremated separately, sealed in a silver box, and buried at Nils’s grave, on what would have been his 75th birthday. And in Strindberg’s recovered almanac there is one entry written in ink, where everything else is pencil — scholars believe he wrote it before departure, the way you write down the train you plan to catch:
“17 October. Home, 7.05 a.m.”
Why this series
Two expeditions, one decade, one problem. Nansen started from what the ice is and built backward from its indifference — and the ice paid him everything. Andrée started from what the story required and spent his margin defending the story — and the ice took the story, the margin, and the men. The ice was not cruel to one and kind to the other. The ice was identical to both.
That, in one sentence, is what THE ICE is about — and why I think it belongs on a podcast that is otherwise about energy systems and AI. The refrain you will hear in every part: the cold does not negotiate.
Next week, Part II: The Race. Amundsen, Scott, one telegram of eight words, and the most expensive difference in the history of project design.
Listen to the episode for the full story. Sources: Nansen’s Farthest North*, the Andrée diaries (Andrée’s Story, 1930), the Grenna Museum Andrée archives, the Fram Museum, and Bea Uusma’s* The Expedition*, among others.*
This one began as heavy breathing and half sentences, recorded on my watch during a Friday-afternoon run through Värsnäs, the coastal reserve just north of Kalmar. It is deliberately dated — July 17, 2026 — and it will age badly, on purpose. That’s the point.
It is the seventeenth of July, 2026. A Friday, late afternoon, twenty-six degrees — which in Sweden counts as a heat wave. I am running the trail out through Värsnäs: oak woods, open meadows, the sea flashing between the trees. No phone. Just the watch on my wrist and headphones in my ears, and the watch is recording, because this is how I think. Out loud. One footstep at a time.
Everything in this episode — and this post — began as that voice memo. A machine transcribed it afterwards, imperfectly: at one point it turned Lars, a researcher from Linköping, into a fisherman from Linköping. Somewhere between that tape and this sentence, something had to know what I meant. Hold that thought.
The joke in Glasgow
May 2019. A small conference on ship energy systems — MOSES — seventy marine engineers from eighteen countries, three days of turbochargers and fuel models in an auditorium on Cathedral Street. On the Thursday night there is a conference dinner at the Corinthian on Ingram Street, all chandeliers and Victorian plasterwork. I have a glass of wine in me and a talk to give the next morning, second-to-last slot of the entire conference.
Instead of rehearsing, I sit up that night and feed every abstract of the conference into a language model, and make it write a new one. A fake abstract — plausible nonsense in perfect conference English. I drop it at the end of my talk as a joke, and the room laughs, because in May of 2019, what else would you possibly do with that?
Then a hand goes up. A researcher from Linköping, named Lars: “Did you use GPT-2?”
I had never heard of it. In my defense, that is less strange than it sounds — I built machine-learning models for a living, but on energy data. Sensors, fuel flows, time series. Language was a different continent of the same country. But still: GPT-2 was three months old that week, the model the newspapers were writing about, and it had passed me by completely.
And I thought I was keeping up.
A runner cannot feel pace
That is the whole problem with exponentials, and I can explain it with the watch on my wrist. On a good day, eleven kilometers an hour feels like nine. On a heavy day it feels like fourteen. The legs lie, the breath lies. That is why the watch keeps splits — not because a number feels like anything, but because a split is a photograph of a moment, and photographs can be compared. Feelings cannot.
You cannot feel an exponential either. Nobody can. From inside, every single day feels locally linear. The doubling never announces itself; there is no bell. The only instrument that works is the split: mark the trail, write down the time, let the future measure the distance.
I know my own instruments are broken, because of Glasgow. On the tape from this run I tell that story as 2018 — my last PhD year, I would have sworn it. We checked. The conference program still exists, a twelve-page PDF with my name on page seven. It was May 2019. Right city, right dinner, right Lars — and the year off by exactly one. Memory, it turns out, is an insider too.
So this episode is a split. Cut deliberately, dated the way milk is dated. In two years it will sound like a man carefully describing his first telephone. Good. That quaintness is not the failure — it is the measurement.
The splits
2015–2018. On paper I was the wrong man for machine learning: marine engineer, officer’s training, no computer science at all. scikit-learn, Matlab then Python, neural networks training overnight while I slept — me checking them in the morning the way you check nets left in the water. I said the same thing to anyone who asked: these models are my hammer. My wrench. A tool. You pick it up, you use it, you put it down. And the tool stayed where you left it. Remember that part.
2019. Because of Lars I looked up GPT-2. The lab behind it had declared it too dangerous to release in full; one British paper ran, roughly, an AI so powerful it must be kept locked up for the good of humanity. The famous demo? Unicorns. Researchers gave it two sentences about English-speaking unicorns in an unexplored Andean valley, and it wrote the rest of the news story straight-faced. That was the machine locked up for the good of humanity: a unicorn journalist. The full model came out in November the same year. The danger had a shelf life of nine months.
2020–2022. A confession: GPT-3 came out in the summer of 2020 — the model of its era, the biggest thing in my own field. I touched it for the first time in 2022. Two years late. And I thought I was keeping up.
November 2022. ChatGPT. A hundred million people in two months. Then it started for real — and even from the inside, seven years inside, it felt sudden.
December 2025. Claude Code, Opus 4.5, agents orchestrating agents — the month the hands came off the wheel. You hand over a task in the evening and check the nets in the morning. Ten years later, the same ritual — except the nets got smarter than the fisherman. And since then the harness hasn’t changed; only the mind inside it keeps upgrading. Prompt engineering was a career in 2024. It quietly stopped mattering. Every clever workflow rots faster than it can be evaluated. The half-life of know-how is now shorter than the release cycle.
The photograph: the week of July 17, 2026
Dear listener in 2027 — or whenever you are. You already know how these stories end. I don’t. That is the entire reason this exists.
Fable 5. On June 9 — a Tuesday — Anthropic released the most capable model ever made public. By Friday it was gone: a US export directive, after researchers showed it could find security holes and write working exploits. A language model, briefly handled like a weapons shipment. Back July 1. That was five weeks ago, and I used it today the way I use tap water. The scandal of June is the plumbing of July.
The usage war. GPT-5.6 shipped eight days ago; its top tier is called Sol, so the frontier duel of this summer is literally named Fable versus Sol — a fairy tale against the sun. Both labs keep resetting the usage meters in a war for users, which lands us somewhere historically strange: the two most capable models ever built are, this week, effectively unlimited, flat rate.
The tests. On the Mensa-style IQ tests people run on these systems, frontier models now score around 130 — on the offline versions, written fresh so they cannot exist in any training data. Mensa membership starts at 131.
And this morning, while I laced these shoes: Beijing announced Kimi K3 — 2.8 trillion parameters, the largest open model there has ever been, frontier class, free weights in ten days. In 2023 that would have been the story of the year. Today it was barely a headline by lunchtime.
Where was I
Mid-run, mid-sentence — right as I was reaching for what all of this means — a call came in, straight into the headphones, and killed the recording. Mid-word, in fact: the tape dies on the Swedish word for but. In the episode you can hear it happen, and then you hear the next recording start with the only sentence available to a human being in that situation: Var var jag någonstans? Where was I.
I never found that thread again. And later in the same run I lost it a second time, no call involved. Nobody can hold an exponential in their head — not even for the length of a run, not even the person making the episode about it.
Living inside one is a cycle: overwhelmed, then numb, then you catch up a little, then a release note lands and you are overwhelmed again. The numbness is not laziness or denial — it is a nervous system doing exactly its job, protecting you from a signal that will not stop firing. A few weeks ago I read a post on X about someone who had checked themselves into care — burned out, not from using these tools, but from trying to ke…
This one started with a paperback on a summer vacation and a Paul Graham post I could not shake — and it ends in my own kitchen, before six in the morning.
There is a girl in my house who has stopped hearing her own name. She is thirteen. She finds a spot on the sofa, opens a book, and leaves. You can call her name from the kitchen and nothing happens. She is not ignoring anyone. She is somewhere else entirely — and I never interrupt, because I grew up in that place.
The same week she was disappearing into books, Paul Graham posted this:
“Something I told 14 yo: People are going to stop reading books. I wish this wasn’t so, but I fear it is. The silver lining in this cloud is that if you’re one of the few people who still read, you’ll have a huge advantage over everyone else.”
A few weeks before that, on vacation, I had finally read Fahrenheit 451. And the detail nobody tells you about Bradbury’s book is that the firemen arrive last. The state never kills reading. The world just speeds up — classics condensed to digests, digests to headlines, school hours shortened, screens on the walls, little seashell radios murmuring in every ear — and the people abandon the books all on their own. The burning is just the paperwork. He published that in 1953.
Lay Bradbury’s sequence next to our data and you notice what is missing: there is not a fireman in sight. There never needed to be.
The double headline
Sweden recently produced two headlines that appear to contradict each other. Statistics Sweden: book reading at its lowest level in forty years — four in ten Swedes read weekly in the early eighties, three in ten by 2021, still sliding. The Mediebarometern survey, the same season: 2024 was a record year for the book, with 66 percent weekly book contact.
Both are true. The first asks whether you sit down and read with your eyes. The second asks whether a book reaches you in any form — and any form includes the audiobook, which nearly doubled among young Swedish adults in a decade (21 to 38 percent weekly). The book is not dying here. The book is migrating, from the page to the ear.
What is declining is a specific human activity: voluntary, sustained, absorbed reading. Across the OECD, roughly half of fifteen-year-olds now say “I read only if I have to” — measured in 2018, before ChatGPT existed. Swedish schools have quietly cut about an hour of reading a week over twenty years, which compounds to roughly a full school year of reading across a childhood. In the US, the share of thirteen-year-olds who read for fun almost every day fell from 27 to 14 percent between 2012 and 2023. And yet: 86 percent of young Swedish adults say they want to read more. The will hasn’t died. It is losing, evening after evening, to everything in the pocket that is professionally engineered to win.
What reading builds
The research is clearer than I expected. Ninety-nine studies of “print exposure” (Mol & Bus, 2011) tie lifetime reading volume to vocabulary, comprehension, spelling — with the associations strengthening with age. Reading compounds, like interest. Nineteen randomized trials (Dowdall, 2020) show that simply reading books with small children measurably improves their language: the one true experiment in the pile, and the one practical takeaway if you have kids.
And the audiobook question has an honest answer (Clinton-Lisell, 2022, forty-six studies): no overall comprehension difference between reading and listening — but reading wins when it is self-paced and when inference is tested. The eye can stop, back up, sit on a sentence until it opens. The ear is strapped to the clock of the speaker. Audio is genuine access infrastructure. But the version of reading that trains a writer is the one where your eyes do the work.
Why care about training a writer? Because, as Graham puts it: you can’t think well without writing well, and you can’t write well without reading well. And because of the best paragraph I have read about reading in years, from his essay How You Know:
“Reading and experience train your model of the world. And even if you forget the experience or what you read, its effect on your model of the world persists. Your mind is like a compiled program you’ve lost the source of. It works, but you don’t know why.”
I cannot list the books I read at eleven. The titles are gone; the training remains. The top spelling grade I “got for free” in school was ten years of nightly reading, invoiced under another name.
The machine
Here is where the episode turns. For decades we studied what reading builds the slow way — surveys, cohorts, correlations. Then some engineers took a blank network and a mountain of text and just ran reading, at industrial scale. Nobody programmed grammar, logic, or facts into it. Text went in. And the only new kind of mind humanity has ever built — the only one we did not make by raising it — came out the other side. Made of nothing but reading.
So: if a language model gets more intelligent by being trained on text — is it really far-fetched to believe a human does too?
It is an analogy, not an identity. The machine needed trillions of words and a building full of GPUs; a child breaks into meaning on a millionth of the data, on porridge and bedtime stories — which should make you more impressed by the child. But the direction of the arrow — text in, thought out — stopped being a humanities opinion the day the machine came out of the corpus talking. Graham’s chain got its existence proof.
And the ironies stack up. At the exact moment machines learned to read, humans started stopping. The machine now writes too, dissolving the pressure that forced generations to think on paper — “a world of thinks and think-nots,” Graham calls it. And under his post, reply after reply makes the third irony: why read long books at all — just summarize them. Condense. Digest. Speed up. That is Bradbury’s fire chief, explaining almost kindly how it all began. The summary was the first fireman. A thousand-page book whose message fits on three pages is not inefficient: the thousand pages are a process, and the detours are load-bearing. No engineer trains a model on summaries alone and expects a mind. Compression loses precisely the thing that builds.
The ritual
So Graham is probably right about the advantage. Reading is the gym that is emptying out — the world used to make readers by default, the way farm work made muscles, and now there will still be strong minds, but only those who choose to be.
What I can control happens tomorrow at dawn: twenty minutes, minimum, before the day gets a vote. A book every other week. Several going at once, abandoned without guilt when they are not good — Naval Ravikant gave me that permission, and it matters more than any reading list: the important thing is not which book you read. The important thing is that you read. Right now it is Huntford’s The Last Place on Earth (listeners of the Outsider’s Mirror episode will smile). Then the fountain pen and the diary nobody will ever read. Two training runs. One loads the model. One compiles it.
And tonight, someone will call a thirteen-year-old’s name from the kitchen, and she will not answer, and I will not interrupt. Somewhere in this house, a mind is being trained on text. I know of no better use of an evening.
Go read something tonight.
Coordinated with Fredrik — Episode 100. A jubilee, but not a victory lap. For episode one hundred I walked back down the corridor of the ninety-nine episodes behind it, looking for the red thread — and let the archive speak for itself: seventeen clips from old episodes are spliced into this one, in the voices they were born with. What I found at the bottom of the corridor is that there were never four topics. There was one question, wearing different paint.
This podcast is a utility. It was built to replace voice memos — thoughts delivered to my cofounders without me ambushing anyone at the coffee machine. There was never a content strategy. There was a man with too many browser tabs open, and colleagues with a twenty-minute walk to lunch.
For its first twenty-five episodes it wasn’t even called Coordinated. It was Fredrik’s Random Thoughts — an anonymous feed, no face, no company name. The very first line of the very first post, late summer 2025, reads — and I am quoting it exactly, typo included — “All is written and researched by By ChatGPT and podcast generated by Google NotebookLM.” By By. Two bys. The machine confessed what it was in its opening breath, and the human running it did not even proofread the confession.
A hundred episodes later, the scatter — crypto frauds, Roman emperors, battery chemistry, an 1865 economist, my grandfather, a book that calls my country a soft dystopia — turns out to line up. Four doors, opening over and over.
Door one: the world’s coordination failures
Episode one was a crime story: Rowan Energy, blockchain-rewarded home solar run as a con, with a function called mintToken buried in the contract — a private exemption from the rules everyone else played by. Twenty-eight episodes later the feed found the same shape at FTX: a credit line no other customer had. The backdoor. Fake coordination — looking like a market while privately being a monarchy.
Between the crime scenes, the show gave itself an education in how the grid actually works, and then the education collided with my day job: Tesla priced its API at a level that vaporized business models overnight, and the podcast and the company became one argument.
“Solar panels and batteries are packets of energy… they need a routing protocol, a coordination layer to turn them from random bursts of energy into reliable dispatchable power. That coordination layer is the TCP/IP of the energy transition, and we are building it right now.” — EP073
“Every curtailed megawatt hour is a coordination failure dressed up as a thermodynamic necessity.” — EP089
The scorecard
A retrospective that only replays its hits is marketing. So episode 100 grades the feed’s own dated bets, deadpan. The V2G obsession: dead — my own company walked away from it. The inverter-security warning: a hit, grimly. The famous “700 hours of negative prices” statistic: a hit with two corrections on the record — first on precision, then on meaning (negative prices sound like abundance; they are a symptom). And the big one, placed five weeks ago, stays open on the card for episode 200 to grade:
“The chips are the tulip. The grid might be the railway.” — EP095
Door two: the lens
Before any of this I was a marine engineer, and then a PhD student polishing fuel curves — until an 1865 economist told me my thesis promise was backwards. Jevons: make a system more efficient, and people don’t use less of it. They find more uses for it.
“This broke my brain in the best possible way… You need abundance, not austerity. That insight is the seed of Sourceful.” — EP072
From episode 19 on, thermodynamics stopped being a topic and became the lens. 173,000 terawatts of sunlight hit this planet; civilization runs on 18. The scarcity story we organize our politics and our guilt around is a rounding error standing in front of a firehose.
“The universe doesn’t run on energy, it runs on differences… An ice cube melting in a glass of water on a quiet kitchen counter is the universe demonstrating its only real product: differences, falling.” — EP091
Door three: the person
The show did not want to be about me. Thirteen episodes of third-person explainers — then episode 14, the green-guilt confession, and the first person never really left. Stoicism arrived as founder practice and got tested for real: mid-crisis, on air, an AI co-host asked me Nietzsche’s question.
“Would I live this again, all of it, including this month? Yes.” — EP078
And the audience kept shrinking, which turned out to be the thread’s secret shape: everyone → my kids → “made for me first” → one named colleague on the E4 → a fifteen-year-old at my kitchen table running Arch Linux. A podcast that starts by broadcasting and ends up whispering to individual people is not a media business failing at scale. It is a different thing succeeding at something else.
“Be less wrong. Faster.” — EP098
Door four: the machine
December 26, 2025. During Christmas, full house, and a terminal — the combination of Claude Max, a model, and a Claude Code harness that actually worked like magic. Six episodes earlier, the feed had been openly skeptical of AI. The sharpest documented mind-change in the archive, timestamped, in public, with the wrong version left standing. That is what a notebook is for.
So I built a pipeline. And the pipeline kept a ledger of its failures, with price tags: episode 75 shipped with thirty-eight hallucinated humming noises (the public paid). Episode 84 reached exactly one listener before the verdict came back: olyssningsbart (a person paid). Episode 97 arrived with seven and a half minutes of dead air — caught in my own headphones before anyone outside heard it, and out of that came a gate that now refuses to render bad pacing.
The disclosure this episode owes you, the archive already made — sixty seconds into episode 76:
“The voice you’re listening to right now is not my organic voice… I write every word. I iterate the script with Claude. I direct what goes in and what stays out, and then I let the clone say it. That’s the stack. I’m not trying to fool anyone.” — EP076
That voice is this voice. Same clone, twenty-four episodes older, reading you its own disclosure. And the episode it reads is the pipeline’s biggest job to date: research agents walked all ninety-nine episodes; every clip was cut and verified, word by word, by a transcription model.
The braid
There were never four doors. The tell is that the same fingerprint sits on every one: Jevons would not stay in his act. Efficiency eaten by new demand in the grid; my thesis broken and rebuilt as abundance; green guilt dismantled (doing a little less is just efficiency-thinking applied to virtue); and the pipeline itself — intelligence got cheap, so I found more uses for it.
The question underneath, the one I think this show has been asking behind every opening line: how do differences become order, without anyone in charge — and what happens when someone fakes it? A grid does it fifty times a second. A market does it with differences in belief, when it is honest. A person does it with the difference between who they were and what they just learned. And the shadow is always the backdoor: the hidden exemption from rules everyone else is coordinating under.
The maddening thing is that the show knew. Episode 26, the rename manifesto: “Coordination — between ideas, systems, and curiosity — is what this project has always been about.” The show knew at twenty-six. I needed seventy-four more episodes to believe it.
The hundredth door
“Ninety-odd episodes ago… and it compounded, quietly, day after gray day while nobody was watching. He did not stop. That is the entire trick. There was never another one. He did not stop.” — EP096
Stopping was never a decision I refused to make. It was a decision that never presented itself. To stop, I would have had to conclude that the curiosity was finished — that there was no next door. And there was always a next door.
I don’t know what episode 101 is about. I never have. That is not a confession — it is the…
A solo essay about rabbit holes, and what happens when you follow one far enough to find a stranger’s furious portrait of your own country. This episode does something I’ve never done before: it wraps a full, AI-generated deep-dive of a 1971 book around my own narration. Two synthetic voices, reading a hostile book about Sweden, played in full — and me, stepping in between them.
I have been spending my evenings, lately, frozen into the pack ice of the Weddell Sea.
Not literally. In a book — Endurance, by Alfred Lansing, about Ernest Shackleton and the twenty-eight men who sailed south in 1914 to cross Antarctica on foot, and who never set a boot on the continent at all. Their ship, beautifully and almost unbearably named the Endurance, got caught in the ice one good day’s sail from land. The ice held them for ten months, slowly and patiently, and then it crushed the ship like a matchbox while the men stood on the floe and watched their only way home splinter into the dark. What follows is one of the great survival stories ever told — two years on the ice and the open sea, an eight-hundred-mile crossing of the worst ocean on Earth in a boat the size of a car — and every one of those twenty-eight men came home alive.
I tore through it. And the moment I finished, I did the thing you do when a story really gets its hooks in you: I went looking for the next one.
This is not an essay about Shackleton. I’ll come back to the ice another time, because there is a great deal down there. This is an essay about what happened when I went looking for the next book — about the strange, branching way curiosity actually moves. Not in a straight line toward a subject, but sideways, through footnotes and forewords and the small print of who an author happened to be, until you look up and you are somewhere you would never in a hundred years have chosen to go. I went looking for more about men and ice. I came back holding a forty-year-old book that calls my own country a soft dystopia.
The rabbit hole
If you fall for the heroic-age polar stories, everyone points you to the same rivalry: 1911, the race to the South Pole. The Norwegian, Roald Amundsen — cold, professional, almost boringly competent. And the Englishman, Robert Falcon Scott, who reached the Pole second, found Amundsen’s flag already there, and died in a tent on the way back, eleven miles short of a depot full of food. For most of a century, that story was told gently: Scott the noble hero, the gallant amateur beaten by bad luck and brutal weather.
The book everyone told me to read took a wrecking ball to that. It’s called The Last Place on Earth, by Roland Huntford, and it tells the opposite story — that Scott was not a tragic hero but a bungler, vain and rigid and contemptuous of the expertise he needed most, who hauled his men to their deaths because he wouldn’t lower himself to learn how to ski or drive dogs, while Amundsen treated the whole thing as a serious professional problem and won without losing a soul. It’s brutal, gripping, deeply unfair, and deeply convincing. People in Britain still get personally angry about it. Huntford walked up to a beloved national myth and calmly kicked it over.
And that — the kicking-over of national myths — is the thread. Because somewhere in the biography of the author, I hit the footnote that started everything. Before Roland Huntford was a polar historian, he was a journalist: the Observer’s Scandinavia correspondent, who lived in Sweden — my Sweden — for about fourteen years. And years before the polar books made his name, back in 1971, he wrote his first book. About us. It’s called The New Totalitarians.
Sit with that title for a second, the way I had to. Written by a man who had lived inside my country for fourteen years, watching, taking notes. And it is exactly what the title threatens — four hundred pages of furious, brilliant, ice-cold character assassination, aimed at the one country the rest of the world was holding up, right then, as the model: the humane middle way, the future that worked. The man who refused to believe Scott was a hero refused to believe Sweden was a paradise. Same temperament, same wrecking ball — pointed at my home this time, instead of England’s.
I am Swedish. I was raised inside the exact thing this book attacks. And I could not put it down.
Huxley, not Orwell
There’s one word in that title that will lead you completely astray. Totalitarian. You hear it and think jackboots, secret police, midnight arrests — Orwell, 1984, a boot stamping on a face forever. That is not what Huntford means. There were no secret police in Sweden, no terror, no gulag. The dystopia he reaches for is the other one: Huxley, not Orwell. Brave New World. A society kept in line not by fear and pain but by comfort, security, and a consensus so total and so warm that stepping out of it feels less like courage and more like madness. A cage whose bars are made not of iron but of safety. And his genuinely unsettling claim is that this kind of control is harder to see and resist, precisely because it doesn’t hurt. A boot on your neck tells you you’re oppressed. A warm blanket just makes you feel taken care of.
I didn’t want to summarize a stranger’s attack on my own country in my own words, filtered through my own defensiveness. So I did something strange: I fed the entire book to an AI — Google’s NotebookLM — and asked it to produce one of those two-host audio deep-dives. Two synthetic voices, who have never been to Sweden, who have never been anywhere, spending forty minutes arguing about whether my home is a soft dystopia. In the episode, I play the whole thing, in three pieces, stepping in between them. What follows is the shape of what they found.
The deep history first — and hold it lightly, because it’s the seductive part. Huntford argues that Sweden missed the Renaissance and so never “discovered the individual,” missed feudalism and so never built the adversarial culture of rights that gave England the Magna Carta, and instead developed an unusually early, unusually revered central bureaucracy — a “cult of the bureaucrat.” A thousand years of collective survival, he says, where conformity wasn’t oppression but the thing that kept the boat from flipping in freezing water. (The AI hosts reach, entirely on their own, for that exact image — a crew rowing through ice where one person out of rhythm kills everyone. It’s Shackleton’s lifeboats. The ice followed me all the way here.) It’s a gorgeous story, and anything that explains everything that cleanly should make the hair on your neck stand up. It’s the same move he pulled on Scott: take a real pattern and haul on it far harder than the evidence can bear.
Then it stops being ancient history and becomes about power: the corporate state born at Saltsjöbaden in 1938, where unions, employers, and government fused into a single apparatus; the “contact network” where the real decisions got made privately, over drinks, and parliament became a rubber stamp; the pharmacist who negotiated away his entire profession in secret because public debate “would only have made things messier.” Law as a tool of social policy rather than a shield for the individual. And then the Huxleyan core — a single Swedish word, trygghet, the warm safety of the womb, weaponized so that to oppose the state is to threaten everyone’s blanket — plus study circles that manufactured consensus before a debate could even begin, and, most chilling, the pathologizing of dissent: if the society is perfect, anyone who still rebels must be sick.
The reality check
Here’s where I have to be honest, and where the AI hosts are, to their credit, honest too. Is any of it true?
Not as written. When Huntford’s own government tracked down the people he had quoted, twenty-one out of thirty said their words had been twisted, taken out of context, or simply invented. Decades later, Huntford himself looked back and called the book “a youthful indiscretion written with far too much emot…
Let me start somewhere most people never go. Below the waterline of a warship, in the engine room, where I spent a good part of my twenties. I was a marine engineer in the navy: fast attack craft, corvettes, submarines, the machine side. Diesel engines, gas turbines, electric propulsion, battery banks.
The thing about a gas turbine is that it is not an abstraction. It has sound, heat, vibration, fuel, routines, and consequences. You are responsible for the whole system, not one component. Four-hour sleep shifts, weeks away at sea, conditions that do not negotiate with you. And if something fails out there, you cannot call support. You cannot google it. You fix it, or you have a very serious problem.
I think about that environment a lot, because of one feature it has that almost nothing else in modern life has. Out there, being wrong is fast. It is physical. There is no dashboard sitting between you and the consequence, softening it, delaying it. If your understanding of the system is wrong, the system tells you, in seconds, in heat and noise and the wrong number on a gauge. The sea does not care how confident you were.
Years later I left the sea and did something that felt completely different. I spent five years on a PhD. And the single most valuable thing I got from those five years was not the title, and it was not the result. It was discovering that the central promise of my own thesis was wrong.
I read an essay this week, by a researcher who writes as Vivek, called how to be good at research. It put words to something I had felt for a long time without saying it cleanly. There is a gap between looking like a researcher and being one, and most people, without meaning to, train the first thing. The real skill is a stack of smaller habits, almost every one of which can be deliberately trained. And when I went through them, one by one, I noticed they all bottom out in the same place: how fast you can find out you are wrong.
I did not learn that from the essay. I learned it on a ship, and then I printed it. At the top of the methods chapter of my thesis, I put a borrowed line. You should take the approach that you are wrong. Your goal is to be less wrong.
Pick your own problems
The essay opens with a habit that made the mathematician Richard Hamming unpopular at lunch. At Bell Labs he would sit with the chemists or the physicists, and after a while he would ask them what the important problems in their field were. And then, once they answered, the question that emptied the table: then why aren’t you working on them? He tells it himself, with comic timing: “I wasn’t welcomed after that. I had to find somebody else to eat with.” One chemist later admitted the question had gotten under his skin and stuck with him all summer, and went on to do important work. Hamming’s quiet, devastating coda is that he never heard the names of any of the other men at that table mentioned in science again.
The sting is that most of us have no good answer. We don’t choose our problems; we absorb them, from an advisor, from whatever a big lab announced last quarter, from the paper everyone is quote-tweeting this week. And the trouble with an absorbed problem is that you hold the conclusion without the reasoning. When the famous lab pivots, you find out a year too late, racing a thousand people who started earlier and have more compute than you.
I started my PhD with an absorbed problem. The shipping industry hands you one the moment you walk in: burn less fuel. So I spent my first years deep in thermodynamics, waste-heat recovery, exergy analysis, squeezing a few percent out of an engine that throws most of its energy away as heat. I genuinely loved the thermodynamics, the calculating on a real process and watching the energy balance out. But the problem was not mine. It was the industry’s. And it nearly made me quit.
What kept me in was a person, not a result. A mentor of mine, Magnus Genrup, talked me into staying at the exact moment I was wrong about quitting. And in that second phase I found my own problem. It started small, almost annoying: a ship is covered in sensors, producing an enormous amount of data every minute, so why do we still not know how much fuel each part of it is burning? Could the ship just tell you, from the data it already produces, without bolting a fuel meter onto every engine? A virtual fuel meter.
That question was mine. The researcher John Schulman splits the work into two modes: in one you read the literature and hunt for things to improve; in the other you choose an outcome you genuinely want to exist and reason backwards to the experiments. The second mode manufactures originality almost as a side effect, because a goal you actually care about drags you into territory no survey paper covers. The fuel question dragged me out of marine engineering and into machine learning, and I never really came back. I did not switch fields. The problem moved, and I followed it.
Upgrade your inputs
Shared reading lists produce shared ideas. If your information diet is the trending page of arxiv plus whatever survives the group chat, you’ll reliably reach the same conclusions as everyone else, at the same time, which makes those conclusions worth approximately nothing.
So where is the edge? One place is almost embarrassingly underpriced: old material. This field reruns its own past on a delay. There’s a thousand-word essay by Rich Sutton from 2019, the bitter lesson, that predicts the shape of the whole field better than surveys ten times its length. It’s called bitter because it’s so uncomfortable: looking back at seventy years of AI, the same pattern repeats. Researchers lovingly build their human knowledge into a system by hand; it helps for a while; and then it’s beaten, decisively, by a simpler method that just throws computation at search and learning. When Deep Blue beat Kasparov in 1997, it didn’t understand chess like a grandmaster; it searched staggeringly fast. Sutton notes, with an edge, that the chess researchers who’d encoded human knowledge for years were “not good losers.” The same story repeated in Go, in speech, in vision. The bitter part is that a lot of your hard-won specialist cleverness is a dead end, and you only learn it by reading the quiet old essay, not the loud new thread.
Which is the second habit: read the paper itself, not the thread summarizing it. The appendix is where the bodies are buried, and the limitations section is usually the most honest paragraph in the document. The interesting part of anyone’s work is almost never the headline. It’s the quiet place where they tell you what they couldn’t make work.
There’s an even older voice in the essay, and it’s the one I’d tattoo on a wall. Claude Shannon, the father of information theory, gave an informal talk to his Bell Labs colleagues in 1952 on how creative people actually crack hard problems. His first and most powerful move was radical simplification: cut it down to size. Strip everything from the problem except the essentials, until it barely resembles what you started with. And then, he says, very often if you can solve that simple problem you can add refinements, one at a time, until you get back to the one you started with. Shrink it, solve the toy, grow the difficulty back. That single trick has carried me through more walls than any productivity advice ever written.
But the place the edge hides that I know in my hands is different. Sometimes upgrading your inputs doesn’t mean finding new data at all. It means reading the data everyone already had and nobody bothered to look at. On the cruise ship I studied, almost a fifth of the electricity on board could have been replaced by heat that was already there, pouring out of the exhaust, doing nothing. In Swedish we have an expression for wasting heat like that: elda för kråkorna, heating the crows. The whole ship was heating the crows, and the signal that proved it had been sitting in the logging system the entire time. Nobody was w…
Johan Stael von Holstein says he has had the world’s best life. If you only know the public story — Icon Medialab, the yellow Ferrari, the magazine covers, the crash — that sentence sounds either delusional or like spin. This episode is my attempt to take it seriously, using a long interview almost nobody has seen.
The interview nobody watched
The source is extra material from “Bubblan som aldrig sprack,” Christian Albinsson’s documentary about the Swedish IT boom and the crash around 2000. Internetstiftelsen backed the film. The interview went up on YouTube in October 2021, and when I found it, it had around 2,600 views.
That’s absurd. It’s long enough to breathe, messy enough to be useful, and personal enough that you can hear an actual life behind the headlines. So part of why I made this episode is simple: I want to pass it on. You don’t have to agree with Johan. But if you care about Swedish tech, entrepreneurship, or the moment the future first became visible here, watch it in full.
The story is about agency, not money
Johan’s story isn’t really about getting rich. It’s about discovering that life is something you can move, not something assigned to you.
It starts with selling, not computers. He was bad at sports, didn’t love school, but he could sell. He wanted to see a football match, so he sold candy there and got in for free. He wanted to go to parties, so he organized them. He wanted to ski, so he packaged the trip and sold it. That’s his original operating system: if the world won’t hand you the thing, build the machine that produces it.
Selling also taught him to read people. That’s why it’s a mistake to file him under “technologist.” He got famous in tech, but his first skill was moving people, and that’s where his idea of leadership begins.
There’s a clean psychology under all of it. Self-determination theory says people need three things to flourish: autonomy, competence, and relatedness. The boy with the candy box has all three — he picks the game, he gets good at it, he learns to read the room. Hold onto that last one. It comes back at the end, and it’s where the story gets hard.
You beat most people by not being an idiot
One thing I believe, and Johan seems to believe it too: most people are easier to beat than they think. Not because they’re stupid, but because they self-sabotage. They don’t show up consistently. They blame everyone else. They make basic unforced errors.
At the office we put it bluntly — you beat most people just by not being an idiot. Be honest, do the work, don’t blame the room, keep showing up, put the hours into the right thing over years. That compounds.
“Any idiot can learn anything in three months”
The turn in Johan’s life is Kinnevik and Jan Stenbeck. After a 1970s Sweden he describes as forced sameness, where he felt pressed down and almost worthless, Kinnevik was a counter-Sweden: a place where ambition wasn’t treated as a moral defect.
Stenbeck gave him a line he calls one of the most important of his life. Johan was trying to dodge a technical job, insisting he was useless with machines. Stenbeck told him: “Any idiot can learn anything in three months.” Then added that Johan wasn’t an idiot, so he got one month.
It’s funny, but it relocates competence. You are not your current skill inventory. Competence is what you do with time and pressure.
The line is more interesting now. In Stenbeck’s world, a driven person could get useful in a new field in three months. AI compresses that again — it hands you the vocabulary of an industry, maps the structure, builds the first version. But it doesn’t supply the will, the taste, the responsibility, or the endurance. So the lesson isn’t that AI makes everyone competent. It makes driven people dangerous faster. That’s the same thing I argued in The Thesis: in the age of AI, the human team matters more, not less.
It wasn’t an IT bubble
Here’s the best idea in the whole interview. Johan refuses the standard line that “the IT bubble burst.” His version: it wasn’t an IT bubble, it was a financial bubble inside the IT sector.
That distinction matters because we keep confusing two questions. Was the technology real? And was the financing around it insane? Both can be true. The financing absolutely burst — people lost money, companies died, the Swedish parts of Icon went bankrupt. None of that should be romanticized away. But the technology kept going. It moved from magazine covers to procurement systems, from charismatic founders to boring implementation, and it didn’t stop.
What’s left after a bubble is the interesting part. Fiber in the ground. A talent school. A generation that had seen global tech companies imagined from Sweden, and then went and built Skype, Spotify, Klarna, Mojang. Not a straight line from Icon. A changed ecosystem.
So the useful question about AI isn’t “is it a bubble?” That’s too easy. The better question: what’s the AI bubble’s fiber? What’s its talent school? What stays standing after the financial story burns off? Johan was often wrong about the clock and much harder to dismiss on the direction. Visionaries usually are. The future moves slower than impatience and faster than memory.
The cost
Johan chose this life, and it cost him. Nights, time zones, responsibility for thousands of people across offices and countries. A near heart attack at 39. By a later interview, seven heart operations. The body is hardware, and it reports back.
The Harvard Study of Adult Development, one of the longest studies of adult life, keeps landing on the same finding: what predicts a good life isn’t fame or money, it’s relationships. Johan seems to know this. When he says he’s had the world’s best life, he doesn’t only list money — he lists his wife, his kids, the things he’s seen. Asked how he wants to be remembered, his first answer is that his children should think he was a good father.
That should stop you. After all the ideology and company-building and public fighting, the first court of appeal is the kids.
What to do with him
I don’t want to acquit Johan, and I don’t want to shrink him. The honest reading is harder. He was right about direction and wrong about timing. He gave people permission, and sometimes confused permission with pressure. He built culture, and sometimes made himself too central to it. He fought for freedom, and sometimes spoke as if his freedom were the only honorable kind.
He understood autonomy and competence at almost frightening intensity. The open question is always relatedness: who it brought closer, who it pushed away, who paid for the force of it.
If “the world’s best life” is what Johan means, the sentence isn’t a victory lap. It’s a defense, a confession, and maybe still a kind of happiness. Be who you are. Be passionate. Dare to take space. Just remember the space is shared. That’s the part Johan sometimes forgets, and the part that makes him human.
Key Takeaways
* The episode is built around an overlooked interview — extra material from the documentary “Bubblan som aldrig sprack,” sitting on YouTube since 2021 with about 2,600 views. It’s worth watching in full whether or not you agree with him.
* Johan’s real subject isn’t getting rich, it’s agency — life as something you move rather than something assigned. His childhood selling stories already contain all three pieces of what people need to flourish: autonomy, competence, and relatedness.
* Stenbeck’s line, “any idiot can learn anything in three months,” relocates competence from who you are to what you do with time and pressure. AI compresses that learning curve further, but it supplies no will, taste, or endurance — it just makes driven people dangerous faster.
* Johan’s sharpest idea: it wasn’t an IT bubble, it was a financial bubble inside the IT sector. The financing burst, but the technology, talent, and permission to build all remained. The same question applies to AI — what’s left standing after the financial story burns off?
* The cost shows up in the body and in relatedness: a near heart attack at 39, sev…
About ninety seconds before I walked on stage, I posted this to the entire internet: “Giving an inspirational talk at Kick Capital — our local angels club. Rebuilding the deck minutes before, and getting a Tailwind cache crash. Sometimes the old good PowerPoint can be missed.”
The deck recovered. By the time I was on stage it ran perfectly; the panic was private, the way it always is. But I had just confessed, in public, that I missed PowerPoint — and then I walked out and the first real thing I said to the room was that PowerPoint is dead.
That gap is the whole talk. The distance between the thing that just broke in your hands and the thing you stand up and declare anyway. Between the private panic and the public conviction. Founders live in that gap, and most of the advice you hear is written by people pretending it isn’t there.
So this is the unfiltered version. It was a room of maybe forty people in the southeast of Sweden — mostly investors and angels, plus the founders they back. I wasn’t there to pitch. I was the alum, invited back to the incubator my own company came up through, handed a generous slot and told, more or less, just be useful. Here is what I told them.
You don’t choose your pivots
We’ve raised three and a half million dollars to build this company. Almost none of it went the way I planned. That’s not the disclaimer at the front of the talk — that is the talk.
Every founder believes the story of their company is the decisions they made. The bold calls, the clever pivots they were brave enough to choose. Almost always, the true story is the levers the world let them hold. You don’t choose your pivots. The system shows you where you actually have leverage, and you’re wise or foolish based only on how fast you’re willing to see it.
The cleanest version isn’t mine, it’s PayPal’s. It started life as Confinity — cryptography for handheld computers, money beamed between two PalmPilots over infrared. A gorgeous demo. Investors loved it. Nobody used it, because to beam money you needed two people in the same room both holding the same expensive gadget, both having decided to do the weirdest possible thing at dinner. A perfect lever, bolted to nothing. So they built a throwaway fallback — just email the money instead — and half-hid it on the website, embarrassed by it. eBay sellers found the fallback and pasted “pay me by PayPal” into their listings. The afterthought ate the product. PayPal never decided to become PayPal; it was simply the one wheel connected to anything real, and the market grabbed it and turned. There was even a name war — one founder wanted to call it X, the users wanted the boring descriptive thing. He’s rebuilding X today, twenty-five years later. So he wasn’t wrong. He was early. And from inside the storm, you cannot tell those two apart. You never can.
There’s a Swedish word I kept coming back to: rådighet. It means the real ability to reach a thing, command it, change it — your hand on it, not your opinion about it. You have rådighet over what your hand actually touches, and over almost nothing else. We learned it the hard way. We hit three walls — a crypto token the Nordics rejected, a beautiful consumer app the incumbents already had a value layer for, a vehicle-to-grid play where we never controlled the charger — and every wall was the same wall. A perfect lever bolted to something we didn’t own. So we stopped gripping a wheel that wasn’t attached to the engine, and went where we actually command the surface. Same conviction. A route that finally runs through something our own hand can turn.
Be contrarian about the destination. Be empirical about the route. The clean story where the founder saw it all coming is always written afterward, by survivors who mistake having survived for having known.
The new lever
Then, in the last year, the lever itself changed. I lived through the last great one — the web, in the late nineties, at the top of the wave — so believe me when I say this one is bigger, and that we are at the very beginning of it.
Since January, no human at my company has written production code. Not because we stopped building — we build more than ever — but because building stopped being typing. The work isn’t producing code now; it’s designing the loop the agents run: change, run, see the result, adjust, faster and faster. And the colleagues running those loops aren’t all human. Two of them are named Miranda and Estelle. Miranda has her own laptop, her own email, runs our admin, prints the shipping labels, ordered the swag for Almedalen. A partner company emailed back and forth with her for weeks and had no idea she’s an AI — and when they found out, they were impressed by “the sharp young woman we’d hired.”
I’m not going to pretend that’s a settled, comfortable thing. Six of us, all men, and the agents took female names and female form; she doesn’t disclose what she is. Those deserve a harder look than a laugh, and I don’t have them resolved. The provocation isn’t the punchline — it’s that I’d rather sit in the discomfort with you than look away. But underneath it is a hard fact: six of us ship like sixty. Coordination is the new scale. And I think that means we’ll hire more people, not fewer — once the engine works, you don’t lay everyone off, you press the accelerator.
If I have one practical, unromantic thing to leave you with, it’s this: maximize the leverage now, because right now it’s on sale. The labs are subsidizing this heavily; the price doesn’t come close to the value, or to what it costs them. Two to four hundred dollars a month per person buys what amounts to a department of tireless senior staff who get better every week. That’s an arbitrage, and windows close. Don’t be clever about the tech budget. Run.
But the discipline is judgment. A model hands you a beautiful answer — fluent, confident, often only partly right — and fluent partial knowledge is the most seductive thing in the world, because it feels like understanding. Almost every failure I see isn’t the model being stupid; it’s a human nodding too fast at something that merely sounded smart. Which is the same reason I keep saying: AI doesn’t make you better. It accelerates the person already there. If you’re clear, it extends your clarity. If you’re hiding, it makes your hiding look professional.
(One flip, since it’s my actual field, and I’ll tell you to discount me for it: yes, this uses energy. Shut one golf course in America and you’ve covered the water for all of AI. And the demand that’s scaring everyone is also the best customer clean power has ever had — the thing pulling solar, storage and grid forward faster than anything else. Hold the fear and the flip at once, and think.)
Leverage only compounds through a system
Here’s the condition nobody says out loud: leverage only pays off if you have a system to run it through. A lever you pick up and put down does nothing. A lever you pull every day, for years, moves the world.
So the most important sentence in the talk isn’t about AI at all: the system is more important than the goal. Compounding doesn’t care about your intensity. It cares about one thing — that you didn’t stop. And it’s hard, in a way nobody warned me about when I sat in those seats. There’s a thing Ben Horowitz calls the Struggle, and I know it’s a provocative thing to say to a room that wants the story to be all up-and-to-the-right, but it’s necessary: the Struggle isn’t a sign you’re failing. It’s the texture of the job.
How do you keep the system running through it? Stoicism — and not the gym-bro version. It’s rådighet turned inward: knowing which of your feelings are yours to control, and spending energy only there. Marcus Aurelius wrote, at night, to himself, “do not waste time arguing about what a good man should be — be one.” The most powerful man alive, talking himself into it like the rest of us.
And the least intellectual, most important part: the body is the substrate of the mind. We train, every day, the whole team,…
There is glass buried under the oceans right now, and strung across this continent, that a whole generation of investors went bankrupt to put there. Tens of millions of miles of optical fiber, laid down in a frenzy at the end of the 1990s on the promise that the internet was about to swallow the world. Then the crash came. And at the bottom of it, of all that buried glass, only about two and a half percent was ever switched on. The rest sat dark in the ground, mocked as one of the great wasted fortunes in the history of capitalism.
You are using that waste right now. It is carrying my voice to you.
The man who buried a lot of it, Gary Winnick, had never run a phone company in his life. He was a bond salesman who understood something purer than how phones work: that the money was in selling the dream of the cable, not using it. He got about seven hundred and fifty million dollars out before the collapse. The bondholders got twenty cents on the dollar.
We always ask the wrong question about a thing like that. We ask was it a bubble, as if the answer told you anything. The question that actually tells you what the future will look like is different. What is it building, and who gets to keep it?
You should be suspicious of me
Here’s the trap, and I want to name it before I start, because otherwise this whole episode is a con. I’m about to tell you a string of stories where the waste turned into the future. Of course I am. Those are the only stories anyone tells. Nobody writes the history of the bubble that left a hole in the ground and nothing else. That’s survivorship bias, and it’s the reason almost everything you’ve heard about bubbles being secretly good is garbage.
So let me be honest about the base rate. Most manias leave nothing behind but lawsuits and ash. The ones that build the future are the minority, and I can’t give you the exact fraction, because anyone who quotes you a clean percentage is making it up. The title of this piece is a provocation. The claim I’ll actually defend is narrower: a specific, identifiable kind of bubble builds the future, and you can tell which kind in advance.
So here’s the test. Three questions. One: did the mania fund a network or a platform, something others can build on later, rather than a paper claim or a single object? Two, and this is the one that decides it: after the crash, can the survivors buy the wreckage for pennies and reuse it? Three: does the thing have a use beyond the speculation itself? It’s a test you can break. Find me a bubble that passed all three and built nothing, and I’m wrong.
The graveyard that became the foundation
Run that test on the winners and watch it hold.
Britain’s railway mania of the 1840s ruined a generation of clergymen and widows. George Hudson, the “Railway King,” turned out to be running a Ponzi scheme decades before Ponzi, paying old shareholders out of new capital. He died with an estate worth less than two hundred pounds. And the track stayed. The bankrupt lines were bought cheap and consolidated into the great Victorian railways. Britain lost the money and kept the railways.
The fiber is the purest case, and it’s worth being precise about it, because two things were happening at once: a stock-market mania that would vaporize, and a physical buildout that wouldn’t. The whole half-trillion-dollar build rested on one number. WorldCom told the world internet traffic was doubling every hundred days. A mathematician at AT&T measured it: it was really doubling about once a year. The company’s boss, Bernie Ebbers, a former milkman in cowboy boots, committed the largest accounting fraud in American history to that point and drove himself to prison. Two trillion dollars of market value evaporated. And then, a few years later, Google quietly bought the dark fiber for almost nothing, and the glut everyone had mocked became the backbone of search, streaming, the cloud, and eventually the machines that train AI. The bubble didn’t get the timing wrong by being stupid. It got the timing wrong, and that wrongness is exactly what pre-paid for our present. They were too early. We got to be on time, for free.
My favorite one leaves nothing you can photograph. The bicycle craze of the 1890s collapsed in an overcapacity glut. But to build millions of bicycles cheaply, the industry had to master precision ball bearings, sheet-metal stamping, steel tubing, chains, and the machine tools to make interchangeable parts at scale. Which is exactly the toolkit the automobile needed. Bike makers became car makers. Two bicycle mechanics named Wright used their chains and spoke wire to build the first airplane. And the cyclists’ lobbying for paved roads became the Federal Highway Administration. The car age inherited a workshop, and a road movement, that the bicycle had already built and abandoned. Sometimes the residue isn’t infrastructure. It’s capability.
Which bubbles lie
The test earns its keep on the failures. The South Sea Bubble and John Law’s Mississippi scheme, both in 1720, were pure paper: the thing being speculated on was the asset, so when it collapsed there was nothing on the floor to pick up. Beanie Babies left a box of toys. And 2008 is the hard case, because real houses were built, made of wood and concrete. But the signature artifact of 2008 is the “zombie subdivision”: graded lots and cul-de-sacs curving off into the desert, fifty miles from any job, with no houses. It was physical, but it wasn’t a network, and you couldn’t reuse it where it sat. Physical doesn’t mean infrastructure. That’s the whole reason there are three tests and not one.
None of this makes bubbles good. The frenzy is capitalism at its ugliest, real people are ruined, and the golden age that’s supposed to follow only arrives if the society does the hard work of adapting. A bubble can leave behind useful infrastructure and a lost decade, both at once. The leftover railway is a consolation prize paid for with other people’s ruin. It is not a reason to cheer the casino.
The one we’re in
Which brings us to the bubble we’re standing inside of. The numbers are hard to hold: the big technology companies are on track to spend something like seven hundred billion dollars in a year on data centers and chips. There is money flowing in circles in ways that should make you uneasy. The warnings, from serious people, are getting loud.
We have stood almost exactly here before. In 1929 the magic word wasn’t artificial intelligence; it was radio. The glamour stock of the age — the Nvidia of its day — was the Radio Corporation of America, up roughly two hundredfold across the decade, trading at seventy times earnings, less an investment than a religion. Then it fell ninety-eight percent, and anyone who bought at the top waited about thirty years just to break even. The paper was incinerated. But RCA the company built NBC, strung transmitters across the continent, and carried the whole apparatus forward into television. The stock was the bubble. The broadcasting industry was the buildout. The speculators were ruined, and the rest of us spent the next eighty years living inside the infrastructure their ruin paid for.
So let me take a position. I think the people calling it a bubble are basically right, and mostly missing the point. The real question is which part becomes the dark fiber, and which part is swampland sold by mail.
Run the test. The chips fail it. Graphics processors go obsolete in two or three years; the silicon is the tulip here, it depreciates to almost nothing. But I have to be fair, the way I was fair to the bicycle: this boom also leaves intangible residue, a trained generation of engineers, open models and tools, the hard-won knowledge of how to build these systems. That survives the crash too.
And then there’s the power. And here I owe you a disclosure, because it matters: I run an energy company. So when I tell you that the thing most likely to last is the power, the grid, the generation, you’re hearing a man whose entire life po…
From the publisher's feed