That was Geoffrey Hinton—whom many consider the godfather of AI—in 2016. 

Long story short, he told the world to stop training radiologists. 

"It's just completely obvious that within five years, deep learning is going to do better than radiologists."

Then they’ll all be out of jobs. 

Wile E. Coyote, meet gravity.

The press believed him. Medical students believed him—some may have even walked away from radiology careers on the strength of that one prediction.

Ten years later, the verdict is in.

Hinton was right about the technology… 

But he was dead wrong about everything else.

Meanwhile, the same prediction is being made today—about your job, your kids' jobs, everyone's jobs.

Which is why what actually happened matters.

The Cliff That Never… Cliffed 

Hinton’s logic was airtight. 

If AI can do a better job than humans, humans lose those jobs. 

And yet… 

Since Hinton’s prediction, the Mayo Clinic has grown its radiology staff 55%. Four hundred radiologists and hiring.

The American College of Radiology projects the demand will keep growing for the next three decades.

And America’s now facing the largest radiologist shortage in its history. Imaging departments at some hospitals are backlogged for months. 

Why? Because AI did what Hinton said it would and then demand exploded.

The smartest man in artificial intelligence—Nobel laureate, father of the field—looked at the data, made the right call, but got it exactly backwards.

We don't need fewer radiologists. We need more than ever.

Right About AI. Wrong About the World.

Hundreds of FDA-cleared radiology algorithms are in use today. On narrow tasks, the machines match or beat human eyes. 

Hinton called that part perfectly.

Here are three things he didn’t account for: 

Problem one: the data doesn't exist. A radiologist reads X-rays, CTs, MRIs, ultrasounds—forty different scan types a day. No training set covers them all. So AI attacks one slice at a time. Chest CTs. Mammograms. Bone fractures. An algorithm that handles one of forty daily tasks helps at the margin. 

That’s theoretically solvable. The next two are harder. 

Problem two: the money follows the doctor. Insurance pays when a licensed physician signs the report. No signature, no payment. That’s not changing. 

Problem three: somebody has to get sued. Miss a tumor, face a jury. Malpractice law wants a human name on that report. No hospital is putting an algorithm on the witness stand.

Capability is one thing. Absorption is another. Hinton had the first and ignored the second.

Why This Matters to You

AI did change radiology. 

It made imaging faster and cheaper. And when something gets cheaper, people consume more of it.

Cheaper imaging meant doctors ordered more imaging. More scans meant more reads. More reads meant more radiologists.

The technology that was supposed to destroy the job multiplied it.

You've heard the predictions. Your kids, your grandkids, you: staring down mass unemployment. 

Those making these forecasts are brilliant. Credentialed. Confident.

So was Hinton.

And they're making his exact error—dazzled by what the machine can do, blind to the friction of the real world. 

The missing data. The regulation. The liability. The institutions built brick by brick around human judgment. 

The trust that takes decades to transfer.

What actually happens looks like radiology. Throughput climbs. Costs fall. Demand explodes. 

The industries that thrive will be the ones wrapped around AI, not hiding from it. Human judgment at the edges of what the machine can't yet do. 

And there will always be edges. 

It was trading at $60.

I said: “bitcoin is a fad, or a scam, or a ponzi scheme, or worse.”

Then my friend Naval Ravikant flew to New York, stood at a whiteboard for an entire day, and showed me why I was wrong.

By the end I realized Bitcoin might be the most amazing thing I'd ever seen in finance—and I'd been in finance since 1999.

Three months after calling it a scam, I released my book Choose Yourself on a Bitcoin-only store I built myself.

I sold about 60 copies for a tenth of a Bitcoin each. Six dollars at the time. That tenth of a Bitcoin is worth about $7,000 today.

So, in a way, I sold 60 copies for $7,000 each.

I got reminded of all this last week in Paris.

I was at the Proof of Talk conference when an interviewer named Anna Tutova asked the one question I never escape:

"Didn't you call Bitcoin a scam?"

Here’s the lesson I learned back then:

When something sounds insane, that's exactly when you should sit down with the smartest person you know and let them draw on a whiteboard.

In fact, that’s exactly why I went to Paris in the first place.

I wanted to talk to some of the smartest people in the world about the most insane idea in AI.

Why I'm All-In on Decentralized AI

Here's what else I told Anna in Paris: centralized wins the sprint. Decentralized wins the war.

Anthropic just raised $65 billion. OpenAI will raise $100 billion.

But what happens when the models keep getting bigger? You can't raise a trillion dollars. At some point the math breaks.

Decentralized networks don't have that problem. People all over the world contribute. The network compounds while the centralized players hit walls.

This is how the internet evolved.

Thirty years ago everyone ran Windows or Mac. Meanwhile, thousands of strangers were quietly building Linux. Today Linux sits at the core of nearly everything. The decentralized operating system won.

Same with news. CBS, CNN, Fox—centralized, run by a few people telling you what matters. Then the internet turned everyone into a news source.

AI is next. And the leader in decentralized AI, by far, is Bittensor—TAO.

TAO trades around $300. I said on camera: $3,000 by year-end. Long term, $50,000 to $100,000.

A handful of the 128 subnets are already the best in the world at what they do, and the only way to invest in subnets is through TAO.

And Then I Made Everyone Mad

The interviewer asked about quantum computing cracking crypto.

Here was my answer: Quantum computing is science fiction.

It doesn't exist in any way that solves a real-world problem. I went to graduate school for computers. I work with quantum companies. Google solved a complicated math problem—good for them.

That's a demo, and a demo won't hack Bitcoin.

Maybe in a hundred years. Maybe I'm the guy calling Bitcoin a scam at $60 again.

But I doubt it. Because this time I've already been to the whiteboard.

I've made three famous calls now: the one I got wrong, the one I'm all in on, and the one nobody wants to hear.

Clip this email and grade me in a year.

Somewhere around 2017, or maybe 2021, a guy showed up at Thanksgiving talking about Bitcoin. 

A cousin. A brother-in-law. Your daughter's boyfriend with the laptop stickers. You had some fun with him. 

Fair enough—he made it easy. 

And in 2022, when the whole thing collapsed—Luna, Celsius, FTX—you got your victory lap. You passed the stuffing and asked how his magic internet beans were doing. 

He stopped bringing it up. You went back to your index fund, feeling like the last sane person in America.

Today I bring unwelcome news.

While you were enjoying the victory lap, the entire market—your market, the respectable one—quietly turned into the thing you were laughing at.

And the guy you roasted? He's the guy at the table who's trained for it.

The Casino Came to You

Look at it.

Figma popped 250% on its first day of trading, then handed most of it back within weeks. Circle popped 168%. Newsmax rose 700% and round-tripped in days.

That's the exact chart every crypto token drew in 2021—now printed on the NYSE.

Sure, our 1999 veterans have seen pops like these before. VA Linux rose 698% on day one, back when the fad was adding ".com" to your name instead of "AI."

But here's what we haven't seen before:

Zero-day options—contracts that expire the same afternoon—now make up roughly two-thirds of all S&P 500 options volume, per Cboe's own data. 

In 2016, it was 5%. 

Most of the "smart money" activity in America's flagship index now has the shelf life of a carton of milk. That’s a market that's calm almost every day, and fragile on exactly the day it matters.

And here's the part you should take personally: your index fund is in on it. 

The top 10 stocks are now 36% of the S&P 500—at the dot-com peak, the number that scared everyone was 27%—and eight of the ten are the same AI bet.

Thirty-six dollars of every hundred you contribute buys those names automatically, at any price, because they're on the list. 

The weird guy has a word for that. In 2021, he said he was “aping in”—or buying in because everyone else is, at any price, no questions asked. He did it with a dog coin. You do it every other Friday.

Point is… 

Your instincts were built for a market that ran on earnings and analyst reports. 

The new market increasingly runs on attention—which arrives in seconds, concentrates in narratives, and leaves without saying goodbye or cleaning up afterward. 

Crypto was this market with the training wheels off. The people who survived it were doing their residency.

What the Weird Guy Knows

Every scar he picked up maps onto today's tape. 

Three of them are worth stealing.

ONE: Narratives rotate, and the rotation is the trade. Crypto compressed a full hype cycle into 18 months, sometimes 18 weeks, and he sat through three of them: majors → L1s → DeFi → dog coins → exit. 

The AI trade is running the same sequence. When a stock goes vertical on a story with no revenue attached, he knows what inning it is while the experts declare a new paradigm. 

Your move: when the narrative reaches the no-revenue tier, you're in the dog coin phase. Sell the dog coin phase.

TWO: Size positions so you can be wrong twice. Crypto's deepest lesson: survival. Being early looks identical to being wrong and your portfolio can’t tell the difference. 

Your move: run the 60% test. If a 60% drawdown in any single position changes your life, the position is too big, whatever the thesis.

THREE: "Up only" language is a countdown clock. When your barber, your group chat, and CNBC agree something can't go down, the marginal buyer is already in. Crypto traders learned to become allergic to euphoria.

Your move: you have it easier—FactSet counts how many S&P 500 earnings calls mention "AI." When that number sets records while the no-revenue tier leads the market, tighten up.

Most have spent years learning none of this. Now they’re walking into a market where stocks gap 40% on a tweet. Unarmed. Wearing loafers.

But your loafers can be reinforced. 

This Thanksgiving, pass the stuffing. Ask the weird guy what he's buying. 

And maybe, unless it’s dogcoins… consider it. 

They picture chatbots taking desk jobs. Robots in showrooms. A singularity arriving all at once, with a press release.

Probably not.

AI changes the economy the way water changes rock. Quietly. Structurally. One industry at a time.

And right now, you can watch it happen in real time.

Start with three questions:

Why is FedEx closing 500 buildings while making more money than ever?

Why did it just pass UPS for the first time in history?

And, finally, the most important: why should you care?

The answer to all three: AI.

Pull up a chair.

Fred Smith's Big Problem

For 25 years, FedEx sent two trucks down your street because it was more efficient than just sending one.

Because Fred Smith, the founder, was stuck.

Fred built an airline. Then he bought a trucking company and thought it best not to merge them.

Different trucks, different drivers, different buildings, same houses.

The waste was known. The waste was priced. The waste was tolerated.

Why? Because nobody could fix it.

Millions of packages, thousands of routes, two incompatible networks—too big for humans with spreadsheets.

It's no longer too big.

AI is Eating it Now

FedEx is now running the biggest overhaul in its history.

They call it Network 2.0.

The plan: one van, one neighborhood. Nearly 500 stations closing—a third of the footprint. Over 200 already dark. Done by end of 2027.

And it's working.

Delivery costs in finished markets: down 10%. Savings banked: $4 billion, with $2 billion more coming. Capex: the lowest in company history.

Then, this year, FedEx passed UPS in market value.

First time ever.

For most of industrial history, growing meant building more. The exceptions were rare—they required genius.

Now AI makes the exception the rule.

At FedEx, two petabytes of data a day are feeding the machines that now draw the routes—work no room of dispatchers could do.

AI forecasts the volume, sequences the closures, reallocates planes and trucks in real time.

The proof it’s working is in the government's data.

Courier employment peaked in August 2025—then fell straight through the holiday rush, when it never falls. Transportation and warehousing shed 100,000 jobs while package volumes grew.

More output. Fewer hands.

Delete, Densify, Dominate

We've seen a similar story like this play out before.

In the 2010s, a railroader named Hunter Harrison gutted the bloated freight railroads.

Fewer trains, fuller trains, densified networks.

They called it Precision Scheduled Railroading.

Costs collapsed, margins doubled, CSX stock roughly tripled—and the railroads kept every dollar, because in an oligopoly, nobody has to give it back.

FedEx is running the same play. For parcels.

How to Play It

FedEx already ran 77% in a year. So you might think you're late. But even when CSX's run ended, the money didn't stop.

It moved.

It moved to Wabtec—the company selling the tools every railroad needed.

Wabtec tripled, and kept climbing years after the railroads stalled. The announcement gets paid first. The procurement gets paid for a decade.

FedEx got some of its announcement money.

Meanwhile players like Symbotic (SYM)—warehouse robots, newly profitable, $22.5 billion backlog—gave back 30% from its highs while the business kept improving.

That gap is worth exploring.

Same with Samsara (IOT), the AI brain inside America's truck fleets: nearly $2 billion in recurring revenue, growing 30% a year.

These companies don't sell to FedEx. But they are selling to everyone watching FedEx.

But one company in particular did get its start hauling FedEx freight: Aurora Innovation (AUR). Its driverless semis run Texas highways today—no human in the seat.

I put it on my watchlist this week.

It books $2 million a quarter. The market values it at $13 billion. You're paying today for a business that doesn't exist yet.

BUT…

If that business shows up, it shows up fast.

Aurora runs 25 driverless trucks right now. By year-end, it plans to run 200. Volvo joins in early 2027. Revenue is guided toward an $80 million run-rate—40 times what it books today.

So…

FedEx is proving how AI is changing the economy.

Two trucks on one street was a 25-year accounting error—AI found it and deleted it.

Now look around. Count the industries still running two trucks.

That's your map for the next decade.

They called it Trezor. (Czech for “vault.”)

It was a way to self-custody your own crypto without knowing how to code or figure out complex encryption methods.

It was the first of what we now call "hardware wallets," or "cold wallets"—cold because they never touch the internet.

They’re the kind of thing Obama once warned about when he moaned that encryption was putting "a Swiss bank account in your pocket."

Kickstarter rejected them. They sold 13,000 anyway.

Then the crypto exchanges started dying. Mt. Gox. Bitfinex (died three times). QuadrigaCX. Much later, FTX.

Each collapse sent a fresh wave of believers running to cold storage.

In fact… 

After FTX, hardware wallets had their biggest sales day in history—twice in one week. "Not your keys, not your coins" became gospel.

Millions of devices. Billions secured. An entire industry built on one promise: the keys never leave the device.

Then, on July 30, 2026, that promise was broken. 

Someone drained $100 million from ColdCard hardware wallets in 41 minutes—without touching a single device.

The keys to the vault never left. They didn't have to. The flaw was baked in.

So let's get into it—what happened, why, and three takeaways. (The first takeaway isn't even about crypto. It's about AI.)

The Sad Truth

It didn’t take long to figure out what really happened. And it’s the stupidest thing ever. 

Here's the simple version.

A hidden bug in ColdCard wallets made the "random" keys protecting people's Bitcoin guessable. 

Hackers, likely armed with AI, guessed them.

Over $100 million gone, mostly from long-term savers who did everything right. No shady exchanges. No careless mistakes. Coins stacked patiently for years, vanishing from cold storage in minutes.

Blockchain sleuths are tracking the stolen funds—most sit untouched, for now—while the industry scrambles to audit every wallet on the market.

I’m rooting for the non-zero chance the attackers get caught and the funds are recovered. (It’s happened before.) 

But right now, the overall reaction is this: "Self-custody is dead. Keep it on Coinbase. Buy IBIT. Let Saylor hold it."

Slow down.

Should everyone run out and buy a hardware wallet? Probably not. But self-custody isn't going anywhere—and now matters more than ever.

What’s a Wallet, Anyway?

People say self-custody exists because custodians are single points of failure. 

But they have it backwards. 

Custodians exist because self-custody—doing it right, at least—is complicated and scary. 

But here’s one thing to keep in mind: Whether you're BlackRock or John Doe on Park Avenue, there's exactly one way to custody it: a wallet holding private keys. Hot, cold, hardware, paper—pick your flavor. All of it reduces to keys in a wallet. 

That's the entire technology.

BlackRock uses wallets. Fidelity uses wallets. Every Bitcoin treasury company on Earth uses wallets. Somebody holds the keys no matter what you buy. 

The only questions are who and what do you get in return? 

Buy IBIT and you own a promise. A claim on Bitcoin that BlackRock controls. Worse, retail shareholders have no redemption rights. Every other custodian sits somewhere on the same spectrum—your coins, their permission.

Remove self-custody from the equation, you remove one of the things that makes Bitcoin valuable: the exit option. And the exit option is what disciplines everyone else. 

Exchanges have to honor withdrawals because users can leave—that's what keeps the whole system honest.

Three Takeaways

There are three things I’m taking away from this ColdCard event, and the first one isn’t what you would expect.

One: Open-weight AI is essential. This has been a debate outside of crypto, but crypto just showed why it’s important. The “white hats” (good hackers) defending Bitcoin keep hitting safety guardrails on the frontier models—and fall back on Chinese open-source AI to fight attackers who face no such constraints. Unless the closed labs dial back the restrictions, the bad guys outgun the good guys by default and everyone’s left defenseless. That's a nonstarter.

Two: The great hardening has begun. Real people lost life savings, and that's the worst part. But every wallet company is now running deeper audits, tighter reviews, harder internal questions—with AI on code-check duty like never before. Innovation with self-custody is set to accelerate faster than any corner of this space.

Three: Self-custody isn't dead—it's about to rise. Every cycle proves the same thing: trust in middlemen decays, and the tools for holding your own keys get better. The ColdCard hack won't reverse that trend. It'll accelerate it. 

The next generation of self-custody will make today's cheap, plastic devices look primitive. And the demand isn't going anywhere, because the alternative isn't going anywhere: counterparty risk, frozen accounts, promises instead of property. 

Every exchange failure minted new self-custodians. This failure will mint better ones.

Of course, this is no solace for those who lost money. They did everything right—and got robbed anyway. That’s the part that really stings. 

But Bitcoin’s survived Mt. Gox, China bans, FTX, and a thousand obituaries. Each time, the response wasn't retreat. It was better tools, harder code, smarter custody.

This time will be no different.

Iridium. Ticker IRDM. Trading under $18 a share. 

A ~$2 billion market cap for the only truly global satellite constellation on earth—66 satellites, pole-to-pole coverage, fat Pentagon contracts, and hundreds of millions in free cash flow.

Wall Street priced it like a niche sat-phone company, but the evidence against that rating was sitting in plain sight.

Four months earlier, in August 2025, AT&T paid $23 billion for a slice of EchoStar's spectrum. Then, in September, SpaceX paid ~$17 billion for the satellite portion. In November, SpaceX came back for another $2.6 billion helping.

Suddenly, the clearing price for scarce airwaves was printed in black and white. 

Wall Street saw every deal. Analysts cited them by name: in their Iridium downgrades. 

Buyers were carving up the airwaves in the biggest land grab in decades—and Wall Street's move was to dump the company holding the rarest slice. 

They priced the threat. They never priced the asset. 

But here’s the thing: 

Iridium held 8.7 MHz of globally coordinated L-band. Jam-resistant. Weather-proof. Irreplaceable. The kind of asset that takes years of international treaty filings to assemble and cannot be built at any price.

If you ran Iridium's spectrum against the deals already on the tape last year, the airwaves alone were worth more than the entire company. The satellites, the revenue, the government relationships—all of it was virtually free.

Then the dominoes kept falling.

In April, Amazon paid $11.57 billion for Globalstar—a smaller constellation with weaker spectrum. The land grab was accelerating, but the market was still slow to connect the dots. 

Yes, the IRDM stock had run—but many waved it off as a "SpaceX proxy." A place to park money while the real asset stayed private.

On June 29, the dots connected on their own. 

Rocket Lab announced it was buying Iridium for $8 billion. $54 a share. When Reuters asked CEO Peter Beck why, he said it plainly: "And of course, the all-important spectrum."

Readers who acted on the original write-up had the chance to roughly triple their money. 

Here's why I'm telling this story today.

James’ AI-powered Buyout Indicator System flashed a "blue spike" on Iridium six days before the Rocket Lab announcement.

Now the same system is flashing again.

Same signal. Same sector. 

A tiny satellite outfit—1/200th of SpaceX's size—sitting on what James believes is the missing piece of Elon's AI ambitions.

But time’s almost up. 

Today, after the closing bell, Elon’s scheduled to issue a public update on SpaceX. It will be his first since the company shattered IPO records in June. It's on the books. It's 100% happening.

James's prediction: Elon makes an announcement that sends this company's stock soaring.

As a True Alpha subscriber, you already have access to the full details. If you haven’t already, click here for the latest reports. 

And yet, for all that firepower…

It still can't out-code Claude. And code, it turns out, is where the money’s at. 

It gets worse. 

Elon's chatbot trails on all important leaderboards. It sits third in consumer share and barely registers in enterprise—where Anthropic now commands 40% of spending.

Think about that… 

Elon bought 550,000 Nvidia chips. Spent $18 billion. Built a 2-gigawatt beast in Memphis at a pace Jensen Huang called "superhuman." And it had one job: make Grok the ChatGPT killer.

It failed. 

And here's the part nobody saw coming: the failure made him richer than the success ever could have.

Because while Grok fights uphill, the machine underneath it became the hottest rental property in the history of capitalism.

The Deals Keep Coming 

Anthropic—the company beating Grok in enterprise—signed a deal to rent the entire original Colossus site. Around $1.25 billion a month. Roughly $15 billion a year. Then they expanded into Colossus 2 a few weeks later.

Google—which owns more compute than nearly anyone alive—came knocking next. $920 million a month through mid-2029. About $30 billion all-in. 

Google. Renting chips. From Elon.

Add it up: over $2 billion a month flowing into SpaceX from rivals whose AI models are beating Grok's brains out.

That's more than SpaceX's entire revenue last year. Every launch. Every Starlink subscription. Combined.

The punchline: his rivals pay him a fortune to keep beating him and Elon’s happy to see it. 

The product lost. The landlord won anyway.

If He Buys It, They Will Come 

All of this goes to show… 

Every serious AI lab is still starved for compute. 

You can have the smartest researchers on Earth—without chips at scale, you're stuck. Own the bottleneck and the market comes to you. Check in hand. In a hurry.

Which raises the question all of Silicon Valley is asking right now.

Elon just became the world's first (former) trillionaire. SpaceX absorbed xAI at a trillion-plus valuation. He already announced a $60 billion agreement to buy Cursor—when he can't build it, he buys it.

So what does he buy next?

James believes he’s spotted it. 

His AI-powered buyout indicator system—the one that flashed ahead of 87.3% of Wall Street's biggest buyouts over the past five years—just lit up on a specific tech company. One James believes Elon is about to acquire.

In fact, he believes the announcement could drop as soon as TOMORROW, after market close.

If he's right, this could hand early movers triple-digit gains by month's end. Somewhere in Manhattan, the phones already rang. 

Yours is ringing now.

Today I'll show you exactly when I plan to start the engine.

Quick refresher for anyone who skipped it…

On next Thursday’s open, some 911 million shares of SpaceX come unlocked. The entire tradeable float right now is 639 million.

As Elon likes to say…

Let that sink in.

One Thursday morning hands the market 1.4 times the existing supply of a stock that's already rolled over from $200. Two days after the first earnings report in company history.

But here’s what I didn’t mention yesterday: that's only the first step.

Another 7% drips loose every couple of weeks through late October. Then another 1.3 billion shares after the Q3 report in November.

On December 8, everything left breaks the cage. Call it 4.6 billion shares between now and Christmas.

Historically, 15% to 40% of newly freed shares get sold in the first few weeks of each unlock. Given who's holding this, lean high.

Take, for example, the venture funds. They've been locked in for maybe ten years. They’re still up big.

And, for some, selling might not be a choice.

They owe money to pensions and endowments who've spent a decade asking one question—when do we get paid? August 6 is the first exit door anybody has offered them.

They need liquidity.

Then there are the employees.

Twenty-somethings sitting on eight-figure paper. As you read this, they’re calling realtors and scrolling through different colored Lambos.

Now consider something else:

The Forced Buyer is Exhausted

Russell 1000 added SpaceX in June. Nasdaq 100 took it on July 7. The biggest forced buying is already behind us.

And the S&P 500 requires GAAP profitability, which SpaceX doesn't have—it lost $5 billion last year—so the largest automated buyer on Earth sits this one out.

Even the bulls know the math.

Morgan Stanley carries a $300 target, the highest of the major banks. Their own note has SpaceX free-cash-flow negative until 2035 and needing roughly $84 billion a year of outside capital from 2027 through 2034.

That's the bull case.

My SPCX Strategy

None of this is a knock on SpaceX.

I want to own it. It’s a monster. One company flies more than 80% of what humanity puts into orbit. Someday, it will be the most valuable company on the planet.

So here's what I'm doing. $160 a day. Every trading day, no exceptions, no thinking. It keeps me in the story while I wait to start the truck.

But I'm not waiting for a specific number. I'm waiting for a specific moment—when the biggest lockups clear, the shorts start covering, and the tape stops making new lows.

That cluster is my signal. Best guess: it lands in November.

Here's the real opportunity…

The Fun Part

While SpaceX could get cheaper from here, the things SpaceX needs will not.

A company burning $84 billion a year has one way to keep its cost curve from eating it alive—own the bottleneck.

Buy the supplier. Vertically integrate the piece nobody else can make. Musk loves this move because it works.

And that short list exists. But it’s small.

One name on it has a market cap smaller than what SpaceX spends on propellant in a quarter.

At the same time, going public handed SpaceX something it never had—stock it can spend. So it buys the pieces it can't afford to keep renting.

James has been using AI to track Elon’s every move. And his AI system keeps pointing at one company.

It's $115 this morning.

You're not alone in the hole. 

Six members of Congress bought that same week, and they're underwater too. Cathie Wood has been buying more nearly every day, all the way down. 

So everybody holding this stock is asking the same thing. 

Do I sell.

Wrong question. Ask one you can actually answer today.

What did you buy?

Maybe you think you bought a rocket company. 

In reality, you bought a rocket company, an AI company, a satellite internet provider, and Twitter, all folded into one ticker over three years. 

And every one of those pieces answers to one man.

Not the board. Not the investors. Elon.

Next week, that simple fact is going to matter. And if you own SPCX—(and even if you don’t)—there are three dates you need to watch. 

First, let’s back up. 

The Rules Nobody Reads

Before a company sells a single share, somebody writes down the rules. How many shares exist. What each one is worth. Who decides what, and by how much.

That document is the corporate charter, and the founder writes it.

The law says one share, one vote. Then it says you can ignore that, and hands you the pen.

Zuckerberg took the pen. 

He wrote two kinds of stock into Meta's charter—the shares he sold to the public get one vote, the shares he kept get ten. 

Then he sold most of his stake and pocketed the money. Today he owns about an eighth of Meta and wins every vote in the building because his eighth is worth ten times that. 

Musk wrote the identical rule into SpaceX. Then he did the one thing Zuckerberg didn't.

He never sold.

He owns 42% of the company and controls 85% of the votes. No board to talk him out of anything. No shareholder who can force a question. No meeting he can lose.

He built it that way on purpose. 

What He Does With It

A man who can't be outvoted doesn't negotiate. 

He gets what he wants. When Musk needs something, he buys the company that makes the thing.

And that’s how he’s always worked. 

He paid $109 million for a German automation shop in 2017, for the machines that build his machines. He paid $218 million for a battery company in 2019, kept the one process he wanted, and runs it in Texas today. 

He bought a thirty-person satellite startup in 2021 for its engineers, and their thruster flies on Starlink now. 

He also bought Twitter for $44 billion and blew a hole in his own net worth. He sold xAI to SpaceX for a quarter of a trillion dollars—the seller and the buyer, same man, same afternoon.

That's what 85% buys you. The freedom to do all of it and not beg for permission. 

Now here’s why this is important. 

Three Dates to Watch

Tuesday, SpaceX reports earnings for the first time as a public company. Twenty-four years private, and analysts finally get to ask their questions out loud.

Thursday, $123 billion in insider stock unlocks, and everyone who got in early decides whether to run.

Both are out of your hands. The third one isn't.

Because before either date, something else is moving, and nobody's watching it.

This Sunday at 7 p.m. Eastern, James goes live to make his case for the company he believes is Musk’s next big buyout—a name a two-hundredth the size of SpaceX, holding the one thing Musk won't rent.

He's calling it Elon's Big Buy.

The event is free to attend, but here’s the problem.

Musk doesn't announce on a schedule. He's made a career of moving before anyone's ready.

So we'd rather you not depend on an email you might catch too late. On the next page, you can also opt into text alerts, which we don't open up on every event. (Only if you want them.)

There’s only one way to join the event:

Click here and sign up so your seat’s locked in. 

Ten and a half hours. 1.3 million tokens. Working physics. Crisp graphics. Every art asset generated out of thin air.

Total cost: sixty dollars.

I’m bringing it up again today for two reasons. 

One, because I want you to understand how remarkable this is. 

Second, because it has everything to do with why I jumped on Mark Jeffrey’s podcast Hash Rate this week. 

First things first. 

Why This is Crazier Than Anyone Realizes

A “token” is about three quarters of a word. 

So the AI thought its way through a million words to hand me a video game. That’s about the length of the complete Harry Potter series. 

And it did it in about ten hours. 

I couldn't read a million words in ten hours. Nobody could. Let alone program a game from scratch. 

And how it did it is even more remarkable. 

I lifted the method from Matt Schumer. You tell the model to screenshot the real Call of Duty, build its own version, screenshot that, compare the two, split the work across agents working in parallel, and keep going until they match. 

I'm a coder by background. I remember trying to make games twenty years ago. It would’ve been impossible to make anything like this by myself, let alone in 10 hours. Even two years ago this project was a team of programmers, six months, a million dollars. I would have signed the check.

Today it costs sixty bucks. 

Everyone is arguing about inflation. Here's a number nobody has in a spreadsheet: one job, a million dollars down to sixty, in twenty-four months.

I got into all of this with Mark. And it went somewhere uncomfortable for many software stocks.

So What's Left? 

If I can clone a billion-dollar franchise in a day for the price of a decent dinner, then the code protects nothing. 

If the software doesn't protect a company anymore, something has to. 

Three things survive. 

Data. An AI is only as smart as what it's been shown. And there's information it will never see. A bank's fraud records. A hospital's scans. A police department's body cam footage. Nobody is emailing that to an AI lab. Whoever holds the most useful data holds something all the cheap intelligence in the world can't copy. 

Distribution. Being where the customer already is. I can build a better product this afternoon and never reach a single person using the incumbent. Ask anyone who built a better search engine.

Trust. The one everybody underrates. I could write my own tax software. I use H&R Block anyway. Mine would probably be better. But when the audit letter arrives, I want somebody else's name on the box. Nobody ever got fired for buying IBM.

Here's what all three look like in one company.

There's an AI project called Score that teaches machines to tell a forest fire from a guy with a flashlight on a camera feed. 

Boring. Enormously valuable. 

It works the way I just described—train on the specialized footage a general-purpose model will never be shown. And it reaches customers through PricewaterhouseCoopers, so it arrives wearing a suit clients already trust.

Score isn't a startup in Palo Alto.

It's a subnet on a crypto network called Bittensor.

The Trade

Bitcoin talked strangers all over the world into plugging in computers and burning electricity, and paid them in Bitcoin for the trouble. 

Though it’s down from all-time highs, few people can argue it hasn’t been a massive success. 

Bittensor pointed the same trick at AI. Run the machines through subnets (which are small AI companies on Bittensor), do useful AI work, get paid in a token called TAO. Best work earns the most. 

Meanwhile, Kimi K3 released its open weights on Monday. These stories are related. 

Because the moat everyone assumed the frontier labs would enjoy for years evaporated in about three seconds. State-of-the-art intelligence is rapidly becoming a commodity.

Commodities have no brand loyalty and no network effects. The cheapest producer wins. Full stop.

So who's cheapest? Bittensor. 

There's another subnet on Bittensor running frontier models on consumer graphics cards instead of million-dollar racks, on the cheapest power on Earth. Cheapest anywhere. Market cap of about twenty-five million dollars.

Meanwhile OpenRouter carries a valuation around ten billion. A Bittensor subnet called Chutes supplies a great deal of what routes through OpenRouter and books roughly a million in revenue.

Bittensor, the network all of these subnets run on, has been dead money around $200 for over a year, waiting on a breakout nobody has delivered yet.

Cheap frontier intelligence might be it.

And here's what makes it interesting at its current price. 

Bittensor is 128 AI companies wearing one ticker. Vision. Translation. Bank fraud detection. Several selling raw AI horsepower cheaper than anyone on Earth. 

Together they're valued at roughly two billion dollars, at a moment when a single AI company can carry a trillion. 

That’s the opportunity.