Larry Page & Sergey Brin: Google Was Always an AI Bet

Transcript

Alex: Two men own about five percent each of a four-trillion-dollar company.

Sam: Okay.

Alex: And between them, they control more than half the vote.

Sam: Wait — that's not a typo?

Alex: Not a typo. And the company in question is Google — or rather its parent, Alphabet — the one building the AI that might get to general intelligence first. Welcome back to Dan's AI Intel — the show that takes the question everyone's half-following and actually runs it all the way to ground. There's a version of the Google story that gets repeated constantly, and it treats the company as a search business that got caught flat-footed by a chatbot and had to scramble to catch up. Today we're putting that version on trial — the story of the two men who actually built it, Larry Page and Sergey Brin — because the paper trail says something close to the opposite.

Sam: Okay, I'm listening — where's the trigger for this one?

Alex: The trigger is a quote almost nobody's actually heard. Twenty-two years before anyone was saying "generative AI" out loud, Larry Page sat down for an interview and named, specifically, where he thought all of this was going. Not vaguely. In a sentence that reads like it was written this year.

Sam: So the deeper question here isn't "did they see AI coming."

Alex: Right — the deeper question is what it actually means when the two people who still, quietly, control one of the most powerful companies on Earth have believed one single big idea since before most of their own employees were born. Because that's not a strategy memo. That's closer to a religion. And today we're going everywhere that takes us — into a childhood, into a boardroom trick almost nobody has explained properly, into a leaked memo that reads like a manifesto, and into a number, twenty-thirty, that two of the most plugged-in people in the entire industry keep saying out loud, on the record.

Sam: There's a twist in here too, right? Because the tidy version of this — "founders secretly planned it all along" — feels a little too clean.

Alex: It is too clean. And by the end you'll see exactly why the honest version is messier, and a lot more interesting, than that. If you've been enjoying the show, hit follow wherever you're listening — it's free, it takes about one second, and it's the single fastest way to make sure the next one actually lands in your feed.

Sam: Okay, start me at the actual beginning — who are we talking about here?

Alex: Larry Page and Sergey Brin. Two Stanford graduate students who meet in nineteen ninety-five — Brin, already a grad student, is assigned to show Page around campus for the day, and by most accounts, they spend the entire tour arguing.

Sam: That's a great first date, honestly.

Alex: They found the friction productive enough to keep talking. Page was working on the link structure of the web — basically, which pages point at which other pages, and what that tells you about which ones actually matter. Brin was doing data mining — hunting for hidden patterns buried inside enormous piles of raw information. Different specific problems, same underlying instinct: there's real structure hiding inside a mess of data, and if you can pull it out, you understand something nobody else has figured out yet.

Sam: And the "mess of data" they eventually picked was the entire web.

Alex: The entire web. And here's the part that gets sanded off in the popular version of the story — the company that came out of that research was almost an accident. They didn't set out to start a business. They built a research project, tried to sell the technology to the big web portals of the era, and got turned down. Repeatedly.

Sam: Wait, turned down by who?

Alex: Three of the biggest names of that entire era — Yahoo, Infoseek, Lycos — the companies that basically WERE the front page of the internet at the time. All three looked at the actual search technology and passed, because their whole strategy was becoming a portal — get you to land on their homepage and stay there, clicking around news and email and horoscopes — not send you somewhere else as fast as possible with the right answer. A great search engine was, in their own business logic, actively bad for a portal. It got you off the page too quickly.

Sam: Oh, that's a genuinely good detail — the exact thing that makes the product good is the thing that made it commercially uninteresting to everyone who could've bought it.

Alex: Which is exactly why the two grad students ended up starting a company at all — mostly because nobody else would take the technology off their hands.

Sam: So the thing that eventually became one of the most valuable companies on the planet was, structurally, a plan B.

Alex: A plan B. And that detail matters, because it tells you something about the temperament underneath all of this — they didn't set out to build a search engine and accidentally discover a bigger mission along the way. The mission came first, chronologically and intellectually. Search was just the version of it that nineteen-nineties computers could actually run.

Sam: Okay, so where does the mission itself come from — because you don't just wake up at twenty-five believing you're going to build a machine that thinks.

Alex: You do, if your father spent his career studying exactly that. Page grew up with a dad, Carl Page, who was a computer science and artificial intelligence professor — genuinely one of the field's early pioneers, earning his doctorate in the mid-nineteen-sixties, at a point when computer science as an academic department barely existed anywhere on Earth. His mother taught programming. He's said his own fascination with computers started around age six, in a house that was basically wallpapered floor to ceiling in tech magazines and his father's research papers.

Sam: So this isn't "smart kid discovers computers in college." This is "raised inside an actual AI research household, before AI research households were really a category."

Alex: Exactly that. And there's a detail underneath it that gives the whole story a real undertow — his father died the same year he started the research project that eventually became the company. Nineteen ninety-six.

Sam: Oh.

Alex: Brin's path is different, but it points in the exact same direction. Born in Moscow, to a mathematician father, Michael Brin, who'd actually been barred from graduate physics study in the Soviet Union for being Jewish, and a mother who went on to become a researcher at NASA. The family emigrated to the United States when he was six. He grows up in a household full of Moscow State University mathematicians, studies math and computer science, and arrives at Stanford as, by most accounts, a genuine data-mining prodigy.

Sam: So — one kid raised, essentially, by an AI professor. One kid raised by a mathematician and a space scientist. Both of them incubated inside serious, technical households before either one had picked a college major.

Alex: And both of them went to Montessori schools, which — separately — they've credited for the exact same trait: a kind of anti-authoritarian instinct to question how things are "supposed" to work, instead of just accepting the syllabus in front of you. One of them has said it almost word for word — that Montessori taught him to question the way the world works, rather than simply accept it.

Sam: Which is precisely the personality you'd want if you were about to spend a decade making an enormously expensive bet that almost nobody else around you believed in yet.

Alex: And there's one more piece of texture worth sitting with here, because it explains the register of everything that follows. Neither of these two grew up thinking of computers as a tool you use. One of them grew up watching his own father build the field from scratch, in an era when almost nobody else on the planet had the vocabulary for it. The other grew up watching his father get shut out of an entire academic discipline for who he was, then rebuild a serious career anyway, on pure ability. That's not "my parents were in tech." That's "my parents proved, personally, that the biggest problems bend if you're stubborn enough about them."

Sam: So the confidence to bet a company on a machine that "understands everything" — that's inherited, almost literally.

Alex: Which is exactly what happens next.

Sam: Okay, this is the line you teased at the very top. What did he actually say?

Alex: October two thousand. Two years after Google incorporates — so this is still a scrappy startup with no obvious, proven path to making real money — and in an interview, Page says this, close to word for word: "Artificial intelligence would be the ultimate version of Google. The ultimate search engine that would understand everything on the web. It would understand exactly what you wanted, and it would give you the right thing."

Sam: Wait, say that again — "the ultimate search engine that would understand everything." That's not describing a search engine at all. That's—

Alex: That's describing a general-purpose assistant that understands your actual intent and hands you the right answer directly, no ten blue links required. Which is, more or less, an accurate description of the AI products Google ships in twenty twenty-six. Except he said it in the year two thousand.

Sam: Okay, that is genuinely eerie. Because the standard version of this story is "search company panics, discovers AI." But what you're describing sounds like the AI was always the actual target, and search was just the piece that nineteen-nineties hardware could actually deliver.

Alex: That's the entire reframe, and he says it himself, in the very same breath — "we're nowhere near doing that now. However, we can get incrementally closer to that, and that is basically what we work on." Search was never the mission. Search was rung one of a much longer ladder, and everybody just assumed the ladder stopped at rung one, because for twenty years, that was all the outside world could actually see.

Sam: So when the AI wave hits in twenty twenty-two, and the headlines all say "wow, they finally got serious about AI" —

Alex: The more accurate version is: they got to resume something they had already named, specifically and out loud, at the very beginning.

Sam: Can you actually make that concrete, though — what does "understand everything on the web" even mean, mechanically? Because that's the kind of line that sounds profound and then evaporates the second you try to pin it down.

Alex: Fair push. Think about what a nineteen-nineties search box actually did — you typed in keywords, and it matched pages that contained those same keywords, roughly ranked by how many other pages linked to them. It had no idea what you actually meant. If you typed "jaguar," it had no way to know whether you meant the car, the animal, or the operating system. What he's describing in that quote is a system that reads your actual question — the intent behind the words, not just the words themselves — and hands you a direct, correct answer, the way a genuinely well-informed person would if you just asked them out loud. That's not a faster card catalog. That's a completely different kind of machine. And it's more or less what you get when you type a question into an AI system today.

Sam: So the twenty-year gap between the quote and the product isn't "they had the idea and forgot it." It's "they had the idea and the actual computing power to build it didn't exist yet." Okay, if that's the stated thesis in two thousand — when do you actually see him act on it? A quote is cheap. What did he do?

Alex: This is where it gets genuinely fun, because between twenty eleven and twenty fourteen — years before "deep learning" or "generative AI" meant anything at all to a general audience — Google, under Page, goes on an absolute buying spree for the entire field. Twenty eleven: an internal deep-learning project called Google Brain starts, almost as a side effort by researchers Andrew Ng and Jeff Dean, and within about a year it's trained a neural network on millions of unlabeled YouTube video frames that essentially teaches itself to recognize the concept of a cat. Nobody told it what a cat was.

Sam: Wait — it just figured out, on its own, from raw video, that "cat" was a meaningful category?

Alex: On its own, no labels, no supervision. Think of it a bit like handing a kid an enormous stack of unsorted photographs and never once telling them a single name — and they still start sorting the pile into piles that turn out to correspond to real categories, purely from noticing what repeats. At the time that was a landmark result — real proof that if you throw enough computing power at a big enough network, it can find structure nobody explicitly programmed into it. Twenty thirteen: Google buys DNNresearch, a tiny startup founded by Geoffrey Hinton, a researcher who's since been called the godfather of deep learning — and folded into that deal come two of his graduate students, Alex Krizhevsky and Ilya Sutskever, the second of whom later goes on to co-found a rival AI lab that becomes one of the biggest names in the entire industry.

Sam: Wait — they hired someone who later helped start a direct competitor?

Alex: That's just the accidental irony of how talent actually flows through this industry, yes. It's genuinely one of the stranger footnotes in the whole story — the company doesn't just build the field internally, it also, unknowingly, trains and funds two of the specific people who'll go on to build one of its fiercest rivals. That's how small and incestuous the very top of this field actually is: for years, essentially everyone who mattered had, at some point, worked in the same handful of buildings. And then, January twenty fourteen, the acquisition that matters most for our whole story: a London AI lab whose stated mission, from day one, is explicitly to build artificial general intelligence.

Sam: General intelligence — meaning not "good at one narrow task," but something that's actually as broadly capable as a human, across essentially everything.

Alex: That is that lab's own founding mission statement, yes — the lab is called DeepMind, and its founder is a scientist named Demis Hassabis, a name worth remembering, because he's still running that same lab today. It's the lab that will later produce AlphaFold, the AI system that cracks the decades-old problem of predicting how proteins fold, and, eventually, Gemini itself. And Hassabis, at that exact moment in twenty fourteen, is one of the most sought-after people in the entire industry — a rival social network competes hard for the same deal, reportedly offering more money, and even a very famous entrepreneur tries to intervene from the other side. What actually tips it isn't the money at all. It's one line Page delivers over lunch: "Why don't you take advantage of what I've already created?" — meaning the compute, the data, and the patience for a project that might take a decade to pay off.

Sam: That's such a specific kind of pitch though. It's not "we'll outbid them." It's "I've already built the ocean you'd need to swim in."

Alex: And it's worth naming exactly how serious the competing bid was, because it makes the win more impressive, not less — a rival social network wasn't just casually interested, it was fighting hard for the deal, reportedly with a bigger check on the table, and a very famous, very wealthy entrepreneur separately tried to get involved from the other side to help make it happen. This wasn't a quiet, uncontested acquisition. It was a real fight, over a company with no product and no revenue, that Page won anyway.

Sam: So money alone lost that fight.

Alex: Money alone lost that fight. Hassabis, after sounding out the rival option and concluding their real enthusiasm was elsewhere — virtual reality, hardware — picks Google instead. Reported price: somewhere between five hundred million and six hundred fifty million dollars.

Sam: Which sounds like real money, until you remember what the company's worth today.

Alex: In hindsight, it's one of the great bargains in corporate history. But that's not actually the point. The point is what the deal reveals — in twenty fourteen, staring at a pre-revenue research lab with literally no product, Page saw the endgame clearly enough to personally outbid the entire industry for it.

Sam: Let's sit on the scale of that for a second, though, because "buying the field" undersells it a little. You're talking about the team behind the cat-recognizing network, plus the actual godfather of the technique, plus his two best students, plus a whole separate lab explicitly chartered to build general intelligence — all inside three years, all before the rest of the industry had woken up to any of it.

Alex: Which is why, when the rest of the world finally does wake up in twenty twenty-two, Google isn't starting from zero. It's starting from a decade-deep head start that almost nobody outside the building had actually noticed yet — because for years, none of it touched the product ordinary people actually used.

Sam: Okay, here's the question I keep circling back to, though — why? What does he actually want a superintelligent machine FOR? Because "understand everything" is a nice line, but it's kind of abstract.

Alex: This is exactly where you have to look at what Page spends his personal conviction on outside the day job, because to him it's not abstract at all. Twenty twelve: he hires Ray Kurzweil, a well-known futurist — an evangelist for the idea of "the singularity," and for the idea that if medicine can just keep pace, you might live indefinitely — as a director of engineering. Twenty thirteen: he launches a biotech venture with an openly stated mission to fight aging, one the press at the time literally described as trying to "solve death."

Sam: Okay, so this isn't "AI helps me search the web slightly better." This is "AI is the master key that unlocks every other hard problem, including mortality itself."

Alex: That's almost exactly his own framing. He's talked about wanting to build "things that don't exist," about something close to a duty to deploy capital at civilizational scale. For him, artificial general intelligence isn't a product line. It's leverage. It's the lever that moves every other lever.

Sam: And the futurist he hired — that's a specific, deliberate signal too, right? Because that's someone who's spent decades publicly predicting the exact moment machine intelligence overtakes human intelligence, and betting his own life on living long enough to see it. Hiring him as a director isn't a research decision. It's a statement of which future you think is actually coming.

Alex: Which is exactly the point — this founder wasn't hiring for a product roadmap. He was assembling believers.

Sam: What about Brin — is it the same motivation?

Alex: Related, but a lot more personal, and honestly the more human of the two stories. Two thousand eight, Brin does something almost no executive would do — starts a plain personal blog, and in the very first post, discloses the results of a genetic test. He carries a specific mutation that sharply raises his lifetime risk of Parkinson's disease. His mother carries the same mutation. She has the disease.

Sam: And instead of keeping that quiet —

Alex: He turns it into a research program. Funds Parkinson's science at real scale, and later founds a separate organization aimed squarely at neurological disease and climate. This is someone who understands, in his own body, that biology is fundamentally an information problem — and who's put a fortune behind the bet that enough computation, enough data, can crack problems medicine alone hasn't cracked in decades.

Sam: So when he says the "trajectory" of AI is too exciting to sit out —

Alex: The subtext isn't chatbots. It's the possibility of pointing real machine intelligence directly at a disease that's already sitting inside his own family.

Sam: Two completely different personal stakes — mortality in the abstract for one, a specific inherited disease for the other — landing on the exact same underlying bet.

Alex: And it's worth naming the gene itself, because it's such a precise detail — the mutation is called LRRK2, and carrying it doesn't guarantee you'll develop Parkinson's, it just sharply raises the odds. He's talked about it less like a private medical fact and more like a research assignment he happened to be born into.

Sam: Which reframes the blog thing too — starting a public blog in two thousand eight and using your very first post to hand the internet your own genetic risk profile is not a normal instinct for a billionaire. That's someone deciding the information is more useful out in the open than protected. He even named the blog "too" — a little pun on being the second founder and on "in addition" — which tells you this was never meant to be a polished corporate mouthpiece. It reads like an actual person's notebook, not a press office.

Alex: Which might be the most honest explanation available for why neither of them ever really let go of this, even in the years when it looked, from the outside, like they had.

Sam: Okay, walk me through the "they retired" part, because that's the piece most people half-remember.

Alex: Twenty fifteen: Google reorganizes into a parent holding structure called Alphabet, Sundar Pichai is installed to run Google day-to-day, and Page moves up into the roomier, higher-altitude Alphabet CEO title. Then December twenty nineteen, both founders go further — Page and Brin step down from their executive roles completely and hand Pichai the top job at Alphabet as well.

Sam: So on paper, that's a total handoff.

Alex: A total handoff. And Page genuinely does become reclusive. He has a real, publicly disclosed medical condition — partial paralysis in both vocal cords, the result of a virus, which leaves his voice noticeably weak, and it's one of the reasons he essentially stops doing public appearances altogether. By various accounts he spends long stretches of this period outside the country entirely, including time in Fiji. In twenty twenty-two, he shuts down Kittyhawk, the flying-car company he'd personally bankrolled for years.

Sam: And Brin?

Alex: Goes through two divorces, recedes from public view almost entirely. By twenty twenty-one, from the outside, these look for all the world like two men who've comfortably retired into their fortunes and their hobbies.

Sam: The flying-car company is such a perfect symbol of that era, actually — the ultimate "we have infinite money and no adult supervision" side project.

Alex: It genuinely was — years of personal funding into a moonshot most people assumed would just quietly run forever on a rich man's whim. And then in twenty twenty-two, right as the AI moment is about to detonate, he shuts the whole thing down. At the time it read as one more retirement hobby ending. In hindsight, it reads more like someone quietly clearing his calendar.

Sam: But you're setting this up like that read is wrong.

Alex: It's half right, honestly. Operationally, they genuinely had stepped away — the company runs for years on the professional operator's own judgment, not theirs. But "retired" implies you gave up control of the thing. And that's precisely the part that never actually happened.

Sam: Before we get to the actual control mechanism — you keep saying there's a paper trail. What did they literally write down?

Alex: There's a nineteen ninety-eight academic paper — the founding technical document, where the two of them, still grad students at the time, lay out the ranking algorithm underneath the entire company, and almost as an aside, worry out loud about a future version of the web that gets optimized for advertisers instead of users.

Sam: Already worrying about incentive problems before they'd made a single dollar.

Alex: Right. Then the two thousand four letter to shareholders, written for the company's public listing, explicitly modeled on Warren Buffett's famous annual letters. It opens with a line that's basically become the company's unofficial motto ever since: "This is not a conventional company. We do not intend to become one." It promises to optimize for the long term instead of smoothing out quarterly numbers for Wall Street — and, this is the part that matters most for us — it explicitly defends a specific ownership structure as the mechanism that would let the founders make big, slow, unpopular bets without the stock market second-guessing them every ninety days.

Sam: So the control structure we're about to get into — that wasn't some later power grab. That was written down and openly justified, in public, at the very beginning.

Alex: In writing, in two thousand four — years before anyone could have known what that structure would eventually be used to protect, including, it turns out, a decades-long bet on general intelligence.

Sam: There's something almost poetic about modeling that letter on Buffett, too — Buffett's whole reputation is patience, decades-long holds, ignoring the market's quarterly mood swings. Borrowing that voice for a tech IPO in two thousand four was a way of saying, out loud, "judge us on a longer clock than everyone else gets judged on."

Alex: And it worked exactly as designed — nobody could force their hand on AI spending during the years it looked like a bottomless money pit with no product to show for it. That patience is the direct ancestor of the decade-long DeepMind bet we just walked through.

Sam: Okay, now the actual mechanism. How do two guys who "retired" back in twenty nineteen still run the show in twenty twenty-six?

Alex: A multi-class share structure, designed before the company ever went public. One class of stock carries ten votes per share. The ordinary, publicly traded class carries one vote per share. And the two founders hold the overwhelming majority of that high-vote class.

Sam: So it's not really about how much of the company they own.

Alex: That's the entire trick, and it produces one of the widest gaps between ownership and control anywhere in corporate America. Think of it like a ten-person dinner where everyone paid an equal share of the bill — except two of the ten showed up holding a card that lets each of them cast six votes on where the group eats next, while everyone else gets exactly one. According to Alphabet's own twenty twenty-six proxy filing with investors, Page controls a little over twenty-seven percent of all votes, Brin just over twenty-five percent — a combined fifty-two point seven percent.

Sam: But what do they actually own, economically — like, in real dollar terms?

Alex: Roughly five percent of the company, each.

Sam: Hold on. So between the two of them, they own something like a tenth of the company economically — but they can out-vote literally everyone else on the planet, combined. Every index fund, every pension fund, every ordinary shareholder, all of it, on any question that actually matters.

Alex: Every single one of them, combined. And that's the detail that reframes the entire "they retired" story. When a hired executive leaves a company, control genuinely does pass to somebody else. When these two stepped back from their day-to-day roles, control didn't move an inch. It stayed exactly where it had always been — parked, not surrendered.

Sam: So whatever the professional operator was doing with the company between twenty nineteen and twenty twenty-two, he was doing it knowing two people who could overrule literally any decision on Earth were one phone call away.

Alex: Which is exactly why, when the AI moment finally arrived, there was no governance fight required at all. No proxy battle. No board to persuade. The two of them could simply walk back in the door.

Sam: Okay, that's a lot to sit with — the idea that two guys everyone assumed had checked out never actually lost the wheel.

Alex: It's genuinely one of the stranger facts in corporate America once you actually sit with the numbers. A little over half the voting power of a four-trillion-dollar company, resting on an ownership stake you could round down to "one in twenty." Most of us intuitively assume those two numbers — how much you own, how much say you get — travel together. Here, they were deliberately unbolted from each other, on purpose, in writing, over two decades ago.

Sam: Take a breath on that one, because it only gets sharper from here — because now we get to watch what they actually did with that wheel, once they decided to grab it back. Okay, so what's the actual trigger for the walk-back-in?

Alex: Late November twenty twenty-two. A rival AI lab releases a chatbot that detonates across the entire culture in about a week. And Google's leadership understands almost immediately that its core business — the thing that pays for literally everything else it does — is facing its first genuinely serious threat in two decades. Pichai declares an internal "code red."

Sam: Which is a phrase that means exactly what it sounds like it means.

Alex: Exactly what it sounds like. And within weeks, Pichai does something telling — he pulls Page and Brin back in, for counsel. And Brin does not treat that as a light consulting gig.

Sam: What does he actually do instead?

Alex: Starts showing up at the office three, four days a week. Personally reviewing AI research. Weighing in on hiring decisions for individual researchers. And — this is the detail that still gets me every time — actually writing code again. Submitting work directly into the flagship AI project himself. A man worth a quarter of a trillion dollars, back at a keyboard.

Sam: In his own words — what does he actually say about why?

Alex: "I kind of came out of retirement, just because the trajectory of AI is so exciting." And he's been blunter since, telling a room of technologists something close to: if you're a computer scientist and you're currently retired, that's a mistake — you should be working on this.

Sam: That's a wild thing to say out loud to a room full of people who might genuinely be retired and perfectly happy about it.

Alex: It is. And to be fair to him, the return wasn't all triumphant. Early twenty twenty-four, the company's image-generation tool got caught producing historically absurd results — an overcorrection while trying to represent diversity that went completely sideways — and it was this same founder who fronted the apology. Flatly: "We definitely messed up on the image generation."

Sam: So he's not just lending his name to the comeback for optics. He's close enough to the actual product to own its failures in public.

Alex: Which tells you this isn't ceremonial. This is a founder genuinely back inside the machine, with real opinions about the actual code.

Sam: Okay, I want to push on this a little, because the story so far is very "these two guys saw the entire thing coming." Is that actually fair?

Alex: It's not entirely fair, and it's worth being honest about that, because three things complicate the tidy version. First — the company was genuinely caught flat-footed. If this had really been a company marching calmly toward a plan, the response to the rival chatbot wouldn't have looked like panic. February twenty twenty-three, they rush out their own competing chatbot, and its very first public demo confidently states a factual claim about a space telescope that is simply, provably wrong. That single error helps wipe something like a hundred billion dollars off the company's value in a single day.

Sam: That's not the reflex of an organization that saw this coming years in advance. That's the reflex of one that's genuinely scrambling. And a hundred billion dollars, gone in a single trading day, off the back of one wrong sentence about a space telescope — that's not a rounding error, that's the market openly saying "we don't believe you're actually ahead here."

Alex: Exactly right. Second — a real chunk of the credit for the actual comeback belongs to professional operators, not the returning founders. Pichai is the one who merges Google Brain and DeepMind into one team and bets the entire roadmap on Gemini. Hassabis is the one who actually ships it. Third — yes, the two-thousand quote is completely real, but for something like fifteen years, AI at Google lived mostly in research papers and long-shot side projects, while the actual money kept coming from advertising. The conviction was genuinely there. The urgency wasn't — not for a long time.

Sam: So what's the honest, fair synthesis, if it's not "they planned it all along"?

Alex: The fair synthesis is: Page didn't literally never leave the field — that overstates it. But the foundation he built a decade early is exactly what let Google come roaring back in eighteen months, instead of never coming back at all. Conviction laid the groundwork. Complacency nearly squandered the lead they'd built. And the founders' return is what actually re-lit the fuse. All three of those things are true at the exact same time.

Sam: It's a bit like a homeowner who spent a decade quietly reinforcing the foundation of a house, then went on a long vacation and let the roof start leaking — and still ended up better off than the neighbor who never reinforced anything at all, because when the storm actually hit, there was still something solid to rebuild on.

Alex: That's exactly the shape of it. The neglect was real. So was the foundation underneath it.

Sam: Okay, let's get into what he's actually saying now, because you mentioned a leaked memo earlier.

Alex: February twenty twenty-five, an internal memo Brin writes to the Gemini team leaks to the press. And it reads less like a corporate email and more like a manifesto. He tells staff to be in the office "at least every weekday." He argues, and I'm quoting directly here, that "sixty hours a week is the sweet spot of productivity." He warns that colleagues doing "the bare minimum" are, his words, "highly demoralizing to everyone else." And he demands the team stop what he calls "building nanny products" — his own term for AI that's, quote, "overrun with filters and punts of various kinds" — and instead "trust our users" more.

Sam: Sixty hours a week is a genuinely aggressive ask from a guy who holds a controlling stake and, technically, doesn't have to show up to the office at all. And "the bare minimum is highly demoralizing to everyone else" is such a specific line — that's not a generic pep talk, that's naming an actual internal culture problem and putting it on the record for the whole team to read.

Alex: It is. And his framing of the actual goal is completely explicit — he writes: "I think we have all the ingredients to win this race, but we are going to have to turbocharge our efforts." And since then he's said, flatly: "We fully intend that Gemini will be the very first AGI."

Sam: Not "we're hoping to be competitive." Not "we're in the conversation." "We intend to be first."

Alex: First. Named specifically, as the actual target — no hedging.

Sam: But there's a real tension buried inside that memo, right? Because "drop the safety filters, trust the users more" is basically the same instinct that produced the diversity-image mess you just described a minute ago.

Alex: That's precisely the tension, and it's worth sitting with. His worldview here is essentially accelerationist — speed and raw capability weighted heavily over caution — and it's now being applied, with genuine founder authority behind it, to the single most powerful piece of technology Google has ever built. The sixty-hour demand alone drew immediate public criticism about burnout. And on the actual timeline itself — Brin and Hassabis have separately, publicly converged on almost the same estimate: general intelligence, they've each said, is plausibly arriving around twenty-thirty.

Sam: And coming from a guy who half-controls the company and personally reviews the research every week — that's not some outside pundit's guess.

Alex: That's a statement of intent, from the inside. And it's worth being precise about what "the nanny products" complaint is actually targeting, because it's easy to hear as a throwaway line. Filters, safety refusals, the guardrails that make a model say "I can't help with that" — those exist because an unfiltered model will confidently say genuinely dangerous or wrong things with total conviction. Calling them "nanny" behavior isn't a technical critique, it's a values statement: that the cost of an occasional bad answer is worth paying for the benefit of a model that never hedges.

Sam: Which is a real trade-off, not an obviously correct call either way.

Alex: It's a real trade-off. And the person making that call, right now, isn't a safety team or an ethics board. It's a founder who controls the vote.

Sam: You know, this actually connects to something we dug into properly a couple of weeks back — our field guide episode on the five people actually steering the major AI labs, episode forty. We profiled Hassabis there too — the Nobel winner who had to rebuild his own lab just to ship products fast enough. Hearing him agree, twice now, with the guy who half-owns the company he works inside — that lands differently once you know both sides of that relationship.

Alex: It really does — worth going back to that one if you haven't heard it, because it fills in exactly who's sitting across the table from these two.

Sam: Okay, so Brin came all the way back inside. What's Page actually doing?

Alex: Characteristically, building outside the walls instead. March twenty twenty-five, it emerges that Page has been quietly running a new AI startup called Dynatomics, led by Chris Anderson, the former chief technology officer of Kittyhawk. This new venture applies large language models to physical manufacturing — using AI to generate optimized product designs that can genuinely be built in real factories, aimed at aerospace, automotive, electronics.

Sam: So — not chatbots. Actual physical objects, coming off a real production line.

Alex: Which is the whole pattern of the man, honestly, if you squint at his whole career. He's always been drawn to the hardest, most physical version of any given problem — self-driving cars, flying cars, extending human life itself — and he's always believed the actual bottleneck, underneath all of it, is intelligence. This new venture is that exact same instinct, just pointed at atoms instead of information: if AI can help design better molecules and better proteins, why not better machines, and better factories to build them in?

Sam: And it tells you something about how he actually relates to the company he built in the first place.

Alex: It does. He doesn't need to be sitting inside it to keep chasing the mission. He can fund a completely fresh attempt from the outside, while still, through the votes, owning the inside the whole time. The same guy who said, back in two thousand, that AI was the real goal is, in twenty twenty-six, personally starting a brand-new AI company — just not the one with his name over the front door.

Sam: It's almost the inverse strategy of the other founder, isn't it? One of them decides the fastest way to win is to climb back inside the machine he already built and push it from the center. The other decides the fastest way is to go build a second machine from scratch, somewhere nobody's watching, aimed at a completely different slice of the same underlying bet.

Alex: Same conviction, opposite instinct on where to point it. Which, honestly, tracks with everything we just heard about who each of them actually is.

Sam: Okay, step all the way back for a second. Zoom out. Where does all of this actually land?

Alex: Almost vertiginously well, for both of them. The company two grad students started because nobody would buy their research project is now, as Alphabet, worth somewhere around four point two trillion dollars — it crossed four trillion back in January of this year, putting it among the three most valuable companies on the entire planet. Google's AI has gotten good enough that its oldest, fiercest rival — the company that makes the phone probably sitting in your pocket right now — has turned to Gemini to power its own next-generation voice assistant.

Sam: Wait, the historic rival is now licensing the AI from the company it's been competing against for two decades?

Alex: For its next-generation voice assistant, yes. And lifted by that same AI-driven surge, the two founders currently sit second and fourth on the global billionaires ranking — worth roughly two hundred ninety-five billion and two hundred seventy-four billion dollars, respectively.

Sam: So the sentence from two thousand — "AI would be the ultimate version of Google" — is actually cashing in, in real time, on their watch.

Alex: And, crucially, under their control. Not somebody else's control. Theirs. Sit with the Apple detail for a second too, because it's easy to blow past — one of the fiercest rivals in consumer tech, a company that guards its own ecosystem as tightly as anyone in the industry, is now routing its own flagship voice assistant through Google's AI instead of something built entirely in-house.

Sam: That's a pretty loud admission, coming from a competitor that size.

Alex: It really is — and it's the clearest external proof of everything we've just walked through. Not a press release. Not a keynote slide. A rival with every incentive to build its own alternative, choosing not to.

Sam: Okay, if I'm trying to walk away from this with the actual takeaway — what is it?

Alex: Three things, and each one is a genuine hinge the future turns on. First — Google's AI position isn't an accident of clever strategy. It's the fruit of a decade-old conviction, which is exactly why it had the talent, the compute, and the appetite to run flat-out the moment it finally decided to. Second — this race is being run at founder tempo, not manager tempo. The leaked memo makes the instruction from the top completely explicit: maximum speed, minimum friction — and there's no boardroom on Earth that can actually slow that down, because the boardroom doesn't hold the votes to do it. Third, and this is the genuinely double-edged one — the exact same supervoting structure that let two visionaries make an early, deeply unpopular bet on AI now concentrates enormous, largely unaccountable power over how the most consequential technology of the century actually gets built, in the hands of two people nobody elected — one of whom is currently telling his own engineers to strip out the safety filters.

Sam: So the thing that made Google genuinely great at AI, and the thing that should probably make you a little nervous about it, are literally the exact same trait.

Alex: Founder conviction, completely unchecked. And if that twenty-thirty estimate is even roughly right, the most important open question in technology might not actually be which lab crosses the finish line first.

Sam: It might be what the two guys who never really left decide to do the moment it does.

Alex: And now, unlike almost everyone else still telling the old version of this story — you actually know exactly who those two guys are.

Sam: Which, honestly, is the whole reason this one was worth the deep-dive.

Alex: And that's it for today — thank you so much for listening. I hope you came away seeing this whole moment a little differently: it's a genuinely fast-moving picture with a brutally short knowability horizon, and understanding who's actually holding the wheel is exactly what makes it worth following this closely.

Sam: And one honest note on how this show actually gets made — it's AI-generated. It's built by one person with a custom stack of AI tools, used to research, analyse, verify and illustrate the questions that are genuinely worth understanding — mostly as a way to learn all of it himself, and then shared with anyone who wants to follow along. AI-assisted, fact-checked, worth a genuine second look.

Alex: Before we go, one genuinely useful thing you can actually do: follow the show. Whatever app you're listening in right now, there's a follow button or a plus button somewhere on this screen — it's one tap, it's completely free, and it does two things. You'll get the next episode the moment it actually lands, and for a small independent show like this one, a follow is honestly the single biggest lever there is for helping it reach other people trying to make sense of all this.

Sam: And one last thing before you go — if there's someone in your life who keeps asking you where AI is actually heading, send them this one. It's still a small, independent show, and a share genuinely does more than you'd think, both for them and for us.

Alex: We'll see you next time.