The Brain Is the Last Unengineered System in Society

In May 2021, Lex Fridman sat next to a machine running Ubuntu 20.04 with a device that looks like a bicycle helmet settling onto his head. Bryan Johnson, founder of Kernel and previously of Braintree, told him the plan: a few jokes, and Fridman's job was not to smile. "As a Russian you should have an edge," Johnson added. "Make the motherland proud." The first joke asked what goes through a potato's brain. The answer was "tater thoughts." Fridman broke in about a second and laughed as if the experiment had already failed.

The helmet is Kernel Flow, a functional brain imager that uses spectroscopy. Neurons fire electrically and chemically, local blood oxygenation changes, and Flow measures that change. Hardware-wise there are 52 modules, each with one laser and six sensors. The sensors operate on a scale of about 100 picoseconds: photons scatter and absorb in the brain, a few come back out, and the reconstruction is built from those returns. Altogether Johnson said there are a thousand-plus channels sampling cortical activity at once.

Fridman had never had his brain imaged. He asked whether the rig supported giant heads; it did, and it felt surprisingly comfortable. People, Johnson noted, are used to fMRI at 120 decibels in a claustrophobic tube, or sticky EEG, or surgery. Flow is meant to be worn in a podcast studio, at home, or at work, in an environment the person chooses. Beside them, Ubuntu was writing the session to disk. Fridman said the setup made him feel more important, the way a good friend feels when they are actually listening. "I feel heard," he said, and he meant the machine, not the man across from him.

Johnson said most people grow up with a sci-fi picture of brain interfaces: move a cursor, type, change the channel, skip a song. That picture is anchored on control. The more useful understanding, he argued, is that neural imaging is a measurement system. Thermometers, step counters, and glucose monitors all show the same pattern. Once a thing has numbers, an endless set of products and clinical paths grows around those numbers. The brain has not had that yet.

Fridman, a machine-learning researcher, was already excited by turning body and mind into data. A wrist wearable is crude; a clean, high-bandwidth stream from cortex is a different scale. He kept returning to the feeling of being heard: state of mind, level of focus, the parts of lived experience that words and body language cannot carry. A photograph freezes a moment you cannot step back into. A brain recording, he hoped, might be a much more personal save of that moment, and a map of the trajectory that followed.

He also wanted the stream for science, not only for a personal Fitbit of the skull. He mentioned Andrew Huberman, a neuroscientist with whom he was writing a paper on eye movement: the pairing of a neuroscientist and a data person, aimed at rigorous study of human behavior, including at the scale of many people. Johnson's larger claim sat underneath that. The species has gotten good at quantifying distant stars, calories, steps, and genomes. "We can measure and quantify pretty much everything in the known universe except for our minds."

He pushed the point into engineering. Human thought, he said, is probably the single largest raw input material into society at any moment, the conversations people have with themselves and with each other, and it has never been measured. Buy an appliance and you do not bring a tape measure to check the doorframe; manufacturing and construction already share standards. Cognition has not entered as a formal engineering discipline that can scaffold news, relationships, politics, economics, and education. Kernel's most significant contribution, in his view, would be to put cognition on that track.

Fridman pressed on the leap from brain data to cognition. We still do not understand most of what cognition is made of. Johnson called that Kernel's hardest problem, and the reason the company almost could not exist as a normal startup. Investors always asked for the killer app. He understood the heuristic: is the product solved, is there a customer, will they pay, how does it compare. When he started, there was not even a known path to build technology that could become mainstream.

He funded the first $53 million himself. To raise the first round he spoke to 228 investors. One said yes. Internally they treat go-to-market more like the Drake equation, the astronomers' stack of uncertain variables for how many civilizations might exist. If the device hits a threshold of data quality, cost, accessibility, and comfort, and can be worn in real environments, then the job is to build an algorithm of discovery so other people can find the value. They want to populate an ecosystem the way early-1990s internet did, and compress twenty years of that discovery into two or three.

When conversations inside Kernel start hunting for someone's intuition about the app, they flag it: human-intuition alert, stop. A young associate professor told Johnson he did not care which of five data streams would matter. He had no theory of mind about regions and functions. He wanted the data for his machine-learning model. Johnson's objective is to create the most valuable brain-data collection system ever built, then apply those tools, because "science is not a business strategy" and the brain does not accept a product timeline.

The closest analog in his own life was a wrist wearable that finally told him eating near bedtime wrecked deep sleep. He could not have introspected that. Kernel ran an internal study: Johnson and 13 coworkers did four cognitive tasks across 13 sessions, looking at reaction time, impulse control, short-term memory, and resting state. His impulse control tracked his sleep independently of how well he played the games. Fridman imagined wearing Flow at a desk for deep work, or during Sam Harris's meditation app, coupling cortical data to digital life. The promise Johnson wanted was not better ads. It was more knowns, so the next product could be rebuilt around them.

Johnson's diet protocol, which he would unpack later, is already a miniature of the philosophy. He tracks more than 200 biomarkers every 90 days and lets the data write the shopping list. He calls it goal alignment inside Bryan: two hundred-plus actors get a vote, including a liver that speaks in labs. That local practice sits inside a larger claim. Evolution designed, manufactured, and distributed intelligence until it produced us. We are now doing the same with our own systems, and over some horizon the cost of that design, manufacture, and distribution heads toward zero.

When that cost curve falls, he said, existence becomes a goal-alignment problem: a negotiation of the terms of intelligent life, inside a body, between people, with bots and voice assistants, eventually among trillions of agents. Fridman asked the survival question. Does he worry that nuclear war, or a paperclip-style system that is not even AGI, throws the whole process in the trash?

Johnson found the fear biologically unsurprising. A new object enters consciousness, the amygdala labels it foreign, apprehension follows. What has not been weighted, he said, is that this is the first generation able to look out over a lifetime and see a real possibility of evolving into forms of consciousness so novel we have no words for them. We cannot point at the thing in the sky. The defining aspiration of intelligent life, in his view, should be that possibility, not only the mitigation of extinction. Fridman compared it to the fog before the internet, when no one could predict that ideas would start traveling as collective thoughts. Superintelligence, dense brain-plugs, even a life inside a game that is not Call of Duty, sit in the same fog: it might end the species, or it might be a certain kind of ego death that is still an enriching experience.

Fridman asked a narrower question: does it make Johnson sad that his brain ends, that the data source goes offline sooner than he would like. Johnson said he would have answered differently at different times. He had chronic depression for ten years and desperately wanted the lights off. Worse, he had been born into a religion that was the only reality he knew, an in-group structure that blinded him to other realities, and an afterlife that would not let the lights stay off. The hopelessness was wanting not to exist and being forced to exist anyway.

There were other seasons, when the future looked like an evolving conscious experience beyond his ability to understand, and he jumped out of bed and raced to work. The stable fact, he said, is the climate inside his head. There is no gentle good morning. The brain is ready with a list of anxieties, four hundred replays, a reel of regrets. "It's a brutal environment in my brain," and we have normalized it as what it means to be human. He does not trust his own answer about death, because he has already been handed several realities that were not real: the religious video game, the depressive hopelessness, then behavioral psychology and a long list of cognitive biases. He assumes that whatever the conscious mind delivers is wrong on landing. The work is to find where, and how much.

Advice for the dark stretch was blunt. Those states are biochemical. They are not you. The suggestions they deliver, that life is meaningless, that you should hit the eject button, are a false reality. He said he fully understands the rational logic of suicide; it is not lost on him. The key is to greet the thought with "thank you, you're not real, I know you're not real," and remember the neurochemical state can be altered. He had tried everything he could name for depression and nothing fixed it. That is why numbers on the brain give him hope. The clinical alternative is still a one-to-ten scale and a bottle.

Fridman offered the opposite chemistry. Several times a day he can look at a cup and be overcome with how amazing it is to be alive, a feeling he compared to being on drugs, disconnected from the rest of a self-critical, often late, almost self-hating life. He thinks that engine of joy is chemistry, and that depression is its inverse. The cup, he said, would be a drink of water in a desert for someone in a low. Both men then tied the two poles back to the thesis of the hour: cognition is the largest raw input, it is unquantified, we have no bearings, and the distortion makes every other social problem worse. Humans have scaffolded technology, science, and institutions. The part still unchanged is ourselves.

Fridman asked what Kernel-style data might do for psychedelics, including the Austin rumor mill about DMT elves. Johnson had not met them; it was a friend-of-a-friend telephone game. Fridman described the elves as critical friends, harsh like trolls but hoping you get better. The joke made a serious point. Words fail that experience. Brain data versus word data is exactly the gap.

Johnson baselined the conversation with a heart. If we cared about cardiac health, we would not ask Fridman to close his eyes and report that things feel okay. A cardiologist who skipped the EKG, the blood panel, the plaques, and the cholesterol, then handed over a statin to experiment with, would seem out of his mind. "That's what we do with our mental health today. It's kind of absurd." For psychedelics the same protocol would apply: measure a baseline, measure during, measure after, follow people longitudinally, then ask which molecule, dose, frequency, setting, and diet. That is what good sleep and HRV data already allow people to do.

He thought that sophistication could improve outcomes, safety, and efficacy, and would change how we talk. Today, anything of the mind is discussed as a modality: meditation, psychedelics, a therapy brand. We cannot point at a marker the way we point at cholesterol, arterial plaque, or HRV. Once there are numbers, solutions map to numbers, and the modality stops being the whole vocabulary.

Johnson's essay "Zeroth Principle Thinking" is an attempt to say how truly original ideas arrive. First principles, in his split, are the understanding of a system's laws, the building blocks in use. Zeroth principles are the blocks themselves. Sherlock Holmes is a first-principles thinker: once you have eliminated the impossible, whatever remains, however improbable, is true. Dirk Gently, Douglas Adams's holistic detective, does not like eliminating the impossible.

After Braintree and Venmo, Johnson had asked what single thing would most raise the odds that the species thrives beyond what we can imagine. First-principles deduction did not feel sufficient. Then he read *Zero: A Biography of a Dangerous Idea*, which he called his favorite book. He had not known that zero had to be discovered, or that it tore up philosophy and mathematics on the way to building modern society. Society took a long time to accept it. Zeroth ideas are that caliber: they are not yet part of shared consciousness, and when they arrive they reset the frame.

Einstein's general relativity, in this telling, is zeroth: on top of Newton, spacetime itself can bend. AlphaGo's moves felt to human players like playing an alien from another dimension; the moves already existed, they had simply not entered awareness. Computational intelligence, Johnson argued, has a habit of introducing zero-like insights, so the future will produce more of them, faster, as the cost of intelligence falls. First-principles thinking cannot predict that. Fridman added that society draws the circle of the impossible far too wide. Very few things are actually impossible. Joe Rogan's line, "it's entirely possible," is the discipline: shrink that set.

Fridman has been thinking about the engineering of consciousness, not as a philosopher's hard problem but as a build. He is captivated by how readily people anthropomorphize, and he suspects we already do it to each other: consciousness is created in social interaction, the way emotion is created in communication and language is created in use. Philosophers want an elf with a name tag feeding us subjective experience. His guess is that the thing is an illusion we constructed to make social life work, and that engineering the illusion in software could still make deeply satisfying experiences.

Johnson agreed and inverted a current failure. Text and emoji crush the bandwidth of human interaction; fewer social cues, more room for misunderstanding, a changed conscious experience between two people. Add a mature brain interface and the dimensionality goes the other way. Fridman can say he is happy looking at a cup; Johnson can run theory of mind and guess. If the interface handed over a fifty-dimensional vector of *this* happiness, empathy might actually reach the fact that two people do not feel joy the same way. He then flipped the subjects. Give GPT-3 a closed loop on human brains and teach it manners. Fridman wanted a Texas version that had read the founding documents and learned to say no. Johnson's punchline was flavors: free, well-behaved, meet-the-parents, Vegas, and then goal alignment so they do not manipulate each other.

Brain data is as personal as data gets, so Fridman asked about privacy. Johnson's verdict on the internet was flat: it started in the wrong place. The game for companies was to take as much information as they could without approval, induce people to give more, skip the explanation, and monetize the back end. "That was a bad starting place." The correction is that the individual always controls their data. It cannot be Hungry Hungry Hippos. When Fridman's brain was recorded that day, the first move in the Kernel app was consent and control, then building on top of clear rules: what is collected, what is done with it, who sees it, on what terms. Delete should be real. So should knowing the share list.

The products are not that fine-grained yet. Johnson sketched a future permission: Spotify may use this vector-space model, only for this window, then it must delete. That structure barely exists, and it needs to be discussed. He also floated a reversal. Large technology companies built businesses on models of people. It is not unimaginable that a Kernel user becomes a better predictor of themselves than those companies are, and that the companies come asking for brain data to complete their models. Then the person sets the terms.

Fridman brought up Elon Musk and Neuralink, the invasive opposite: surgery, an implant, a direct line to neurons. Johnson said the two of them talked a lot early on. He had started Kernel, Musk was interested in brain interfaces, they explored doing something together, including Musk joining Kernel, and it was not the right move. Musk founded Neuralink; Johnson kept building. Both men started on the invasive path. Johnson was building invasive technology at the time.

The turn came from a neurosurgeon who had operated for thirty years, among the most respected in the United States, invited in to explain the profession. At the end of three hours he said that every fifteen years or so a new technology comes along that changes everything. "It's probably already here. You just can't see it yet." Johnson's jaw dropped. He had talked with Bob Greenberg, who built Second Sight first on the optic nerve and then as a cortical array, and with people at NeuroPace, which implants for seizure detection and intervention. He had watched what an implantable actually costs: seven years becomes fifteen, $100 million becomes $300 or $400 million. He did not want invasive. It had only looked like the only possible road.

He stood up an internal effort on non-invasive options against four tests: high-quality data, low cost, accessibility, and a shot at the mainstream. Then he made a bet-the-company move and switched. Both Kernel technologies now exist; Fridman had worn one of them. It is too early, Johnson said, to declare a winner, and timescales matter. In ten or fifteen years, less invasive ways to reach neuron firings may exist, or Neuralink may have chosen exactly right, or Kernel's paths may fall short. His analogy for the present: Kernel Flow is watching a movie with a full screen, but the image is blurred and the audio muffled. Kernel Flux, their MEG system, is the full movie in 1080p. Neuralink is a circle on the screen in 4K.

He bet that Flow and Flux are the answer for the next seven years, and that they would birth an ecosystem worth more than the hardware. Off-the-shelf brain interfaces have had decades in hacker communities and have not gone mainstream; room-sized machines are on their own track. Kernel has just under a hundred people, something like 36 or 37 PhDs in specialties only a few people on earth have, full stack from the photon to the machine-learning model. Both they and Neuralink, he said, have spent over $100 million, which is what a serious attempt costs. He does not see Musk as a competitor. At the beginning of the internet, how many companies get invited in? An endless number. Fitbit was a device Google bought for $2.5 billion, with little built on top and no extra data streams. Apple sold a device and an ecosystem. If the whole stack is worth a hundred, Kernel's hardware might be five or ten. The hardware is the door.

Johnson founded Braintree in 2007, acquired Venmo in 2012, and sold to PayPal the same year. The origin was not a payments thesis. At 21 he came back from two years in Ecuador living among extreme poverty, was shocked by American opulence, and decided to spend a life helping others. The plan: make enough money by 30 never to need a job, then do work that might help people without having to be a standalone business. After failures and a pile of debt he needed a part-time job. He emailed the assistants of Utah's fifty richest. Nobody replied. Dozens of interviews went nowhere. On Monster he found credit-card sales, door to door, straight commission. If you could fog a mirror, you could start.

Owners hated him on sight, then hated him more when he said credit-card processing, because the industry had cheated them. His algorithm was a hundred-dollar bill: three minutes of your time, and if you do not say yes, the hundred is yours. Then he opened a book and explained the industry, the players, the deceit, and the fact that he was no different except for a clean statement and a human on the phone. People started saying yes. Next door he mentioned that Joe and Susie had said yes too. Social proof did the rest. He became number one of 400 salespeople nationwide, part time, still running another startup. After nine months his merchants were generating $62,504 a month. He noticed that if this were his company, that number would be passive income. He started Braintree.

PayPal, after eBay bought it around 2000, had sat still for about seven years while software moved. Engineers wanted modern tools and did not have them. Braintree's entry was software engineers would love, then customer service. The real target was PayPal's model. Early eBay users funded PayPal from checking accounts, so PayPal's cost to pay a merchant was near zero and the margin was about three percent, versus pennies for a typical processor. Two-sided marketplaces are hard to build at once; others had failed. Johnson's plan, which he had already presented at the University of Chicago, was to win the best merchants, then in year five buy a consumer payments company and join the sides. It happened on schedule. After four years the customer list included Uber, Airbnb, GitHub, and 37signals. Venmo had done the counterintuitive thing of making private money public. With Venmo, users could fund an account from checking, hold a balance, and pay. PayPal saw the best merchants on one side and influential millennials on the other, and offered a number. Johnson had never wanted the largest payments company or a Forbes list. He wanted a sum that would let him work on something that still looked valuable from 2050, 2100, and 2500, without asking permission. The check was enough. Then he surveyed the world for the highest-value thing he could uniquely do, and eventually landed on brain interfaces.

Fridman asked him not to shrug off payments, especially crypto as a political technology. Johnson said he had not meant to diminish it. Braintree was early to integrate Coinbase, in 2013. He likes the principles under cryptocurrency and thinks they apply as well to the brain, to how brains meet the world, and to goal alignment between people. Money is a scaffolding layer, not a separate essence. He is bullish because Kernel will want a similar stack. Asked whether crypto is, like his diet, a demotion of the conscious mind, he said that was accurate: an autonomous system of value exchange, a version of the autonomous self he is trying to build, against a fiat system that still runs on momentary hunger, irritation, and politics.

Johnson's current diet is the residue of thousands of experiments, reset every 90 days, with measurement running in between. He wants perfect cholesterol, glucose, DNA methylation, and sleep. For the prior two weeks his sleeping resting heart rate had been 42, where HRV peaked and mornings felt best. Eating around 8:30 a.m., after a workout, on an empty stomach, put enough distance before bed that digestion was almost gone, the heart rate fell, sleep quality rose, and willpower the next night followed. The meal is vegan: a "super veggie" pudding of 250 grams of broccoli, 150 grams of cauliflower, and other vegetables; a nut pudding; a rotating third dish, that day kale, spinach, and sweet potato; plus about twenty supplements. Fridman asked if it tastes good. Johnson said he loves it. He dreams about it.

Alfred North Whitehead's line sits over the protocol: civilization advances by extending the number of important operations we can do without thinking. Johnson wants an algorithm for perfect health that he never has to think about, then a version that scales to anyone. Grocery-aisle choosing and paper-chasing among diet books are, in this view, not the future of being human. Closed-loop devices should assess what a particular body needs. He already refuses to sample his feelings in front of a cereal box. Temporal sampling over a week, a month, a quarter smooths the oscillations so a low day does not rule the shopping list. Fridman lives a cruder version, often one meal a day, especially when traveling, with worse control at night. He said meat makes him feel better and more carbs make him feel worse, and he will not lean on nutrition science for that; the field is too confounded, so the honest unit is n of one.

Sleep, Johnson argued, is a contender for the most powerful health intervention in existence. Deep sleep the night before predicts whether "rascal Bryan" binges. Society still tells founder stories of three days in the office and a nap under the desk as if that were greatness. He thinks the next mythology will call that primitive. Fridman pushed back, seriously. He has slept under desks many times. He believes in chaos, in throwing plans away for a stretch of passion, as a signal that you have thrown everything in. He fears over-optimization, the programmer's mind becoming too cautious to live. He would rather die in battle than stay safe at home, and he gets a charge from William Wallace's speech in *Braveheart* every time.

Johnson did not deny the feeling. He translated it. Those words trigger a neurotransmitter state, and an engineer of consciousness should be able to have that state without sacrificing long-term health. The trade-off is not mandatory. He added a biological constraint: biology now looks able to replace organs, including by xenotransplantation, except the brain. You can lose a hand or live with an artificial heart. You cannot operate without the brain, and we do not know what modern stress, sleep, and diet are doing to it. Fridman still wanted a seam for not giving a damn, because strictness itself raises stress when he breaks it; a night of brisket and too much diet soda in Los Angeles had wrecked his sleep with bathroom trips and then avalanched into a messy day. The word he offered for deep work was sensitivity: hard thoughts arrive like a telescope that has to be aligned just so, a sneak peek that vanishes if you move. Phone, social media, fifteen tasks, and a bad night drive that sensitivity to zero. You remain functional. You do nothing interesting. Johnson said sensitivity was the right word. Sleep feeds willpower, willpower feeds the night, the night feeds tomorrow's sensitivity.

The climb sat at the darkest point in Johnson's life. He was ending a thirteen-year marriage, leaving the religion he had been born into, freshly on the other side of selling Braintree, and at the end of his rope with depression. He was invited to Tanzania with a group raising money for clean-water wells. He donated $25,000, the first money he had ever given beyond tithing, and found the wells and the people energizing. Then they climbed. On day three he got a stomach flu on top of altitude sickness. By day five, walking into camp at 15,000 feet, he was falling, wrecked, and thinking about three young children. Whether he lived was no longer only his decision. He felt strapped in a straitjacket.

At midnight the guide asked where he was. Johnson said he thought he could try. They left at midnight and he summited at 5 a.m. The mountain became the problem itself, every struggle given a slope. The pain got ferocious; he put on Eminem, the only voice that spoke to his soul, anger and vitality enough to get a little air. A hundred yards from the top he went dizzy and thought he would fall backward off a cliff. The guide, Ike, said he knew where Johnson was, and that he had it in him: look up, take a step, take a breath. Johnson did, made it, sat down, and cried like a baby. Then the pulse oximeter read something in the 50s, a danger zone. No helicopter, no evacuation. He had to hike back to 15,000 feet, then ride a metal one-wheel stretcher while six people took him down, his head slamming the metal every couple of seconds for hours. He called it the demarcation. He would reconstruct Bryan: his picture of reality, his existential settings, what he would go after. It had felt like a death match with reality. Summit meant he would come out on top. Giving in meant he was done. He did not design that. Mountains, he said, are magical to him.

Asked for advice to young people, he offered a method rather than a playbook. Listen to advice, see it as a mirror of that person, and remember your future sits in zeroth-principle land. What you hear today may be principles that worked before and are depreciating fast. He is a strong proponent of asking for advice and not taking it. The value is not doing the thing. The value is inspecting the assumption-and-knowledge stack underneath. Advice is the tip of the iceberg; depending on the stack it may be right or completely wrong. Fridman added the other side: the advice-giver needs empathy, especially with an eighteen-year-old, and as a parent, not to play the authority who cannot see a different brain. Johnson's most important memories are one-to-one moments like Ike, a stranger in Tanzania who seemed to see his soul in a death match and gave him three words of instruction. He would rather become that for someone else than a retail advertiser of life tricks.

On the meaning of life, he first treated the question as a mirror of the asker's technological and political moment; a hundred years ago the answer would have been different, and it is hard to climb out of context. Inside this moment he offered three layers. Evolution scaffolded intelligence until it produced us; we now scaffold higher-dimensional systems, and the cost of that is going to zero. When cost hits zero, meaning becomes goal alignment, a negotiation of conscious and unconscious existence. If we think of ourselves as explorers, technology is getting to a place where we can at least aspire to find out what is really going on, even if poking through takes a hundred, five hundred, or a thousand years. He added James Carse's *Finite and Infinite Games*: the only game worth playing is to keep playing. Against diet soda and brisket, that is practical. Possibility is changing faster than old risk algorithms. It is rational to err on the side of still being at the table. Fridman closed with Diane Ackerman: our brain is a crowded chemistry lab, bustling with nonstop neural conversations.