The AI Gold Rush Is Eating Itself

---

Meta’s cutting 8,000 people. OpenAI and Musk are heading to trial. Chinese firms are allegedly vacuum-cleaning American AI models straight off the shelf. And somehow, Intel’s just had its best quarter in years.

This is what peak chaos in the AI space looks like—not the kind that kills the industry, but the kind that separates the people who actually understand what they’re building from the people who just know how to spend money really, really fast.

The Layoff Paradox Nobody’s Talking About

Here’s the maddening part: Meta’s laying off 8,000 employees while simultaneously saying it’s doubling down on AI. The company closed 6,000 open positions too. That’s not pruning—that’s reorganization at gunpoint.

Why now? Because the AI spending spree can’t go on like this forever. Meta’s been hemorrhaging money on unproven AI infrastructure. The moment you admit that the returns on those billion-dollar bets aren’t matching the hype, the stock market notices. So you cut the people who were building yesterday’s roadmap and keep the people building tomorrow’s. In theory.

The problem is that this is happening across the entire industry simultaneously. Everyone simultaneously discovered that you can’t just throw engineers at the problem anymore. You need the right engineers, and apparently you don’t need as many of them as you thought. That’s not innovation—that’s recalibration after miscalculation.

Photo by Utopix Pictures Pictures / Pexels

China Didn’t Even Knock

The White House memo about Chinese firms “wrongfully distilling” US AI models is the clearest admission yet that the entire open-source AI movement might’ve been security theater.

Think about it: American companies spent years publishing models, datasets, and techniques. We called it democratization. We said it would accelerate innovation. And now the government’s basically saying “yeah, about that—countries we’re not exactly friends with just took all of it and packaged it for themselves.”

This isn’t sophisticated espionage. There’s no dramatic heist movie here. It’s just that when you publish your homework online, people copy it. They copy it faster, cheaper, and without the safety guardrails you spent time installing.

I genuinely don’t know what the fix is here. You can’t un-open-source something. You can’t unring a bell. But what you can do is stop pretending that American AI dominance is inevitable. It’s not. It’s a temporary advantage that expires the moment anyone competent with a laptop does what Meta and OpenAI already did—train their own models.

Elon vs. Sam: The Most Expensive Grudge Match Ever

A jury trial starting Monday. Musk seeking billions. OpenAI as the defendant.

On the surface, this is a contract dispute. Beneath that, it’s a war over the definition of what OpenAI is supposed to be. The company was founded as a non-profit. Now it’s basically a profit engine with a non-profit wrapper. Musk says that’s a betrayal. OpenAI says Musk abandoned the vision first.

Here’s what actually matters: the trial outcome could reshape AI governance for the next decade. If Musk wins, you’ve just set precedent that AI companies can’t pivot their business models without extraordinary liability. If OpenAI wins, you’ve established that founders can walk away from their founding mission consequence-free as long as the lawyers are good enough.

Neither outcome is good. Both are just different flavors of bad.

My prediction? This settles before jury deliberations end. The discovery process has probably revealed things both sides would rather keep private. But the settlement doesn’t actually solve anything—it just papering over the fact that no one has any clarity on who’s supposed to be watching these companies.

Photo by nappy / Pexels

The British Data Oopsie That Tells You Everything

UK Biobank’s medical data—500,000 people—got listed for sale in China. The government says no personally identifiable information leaked, which is technically true and entirely beside the point. They have your medical history. They have your genetics. They have your pharmaceutical footprint. They don’t need your name to make that useful.

This isn’t a security failure. It’s a governance failure. Someone decided that health data could be commercialized without thinking through who “someone” might sell it to. Now China has a dataset that would take them decades to generate independently.

And Britain’s just… moving on. Noting it happened. Saying it’s handled.

This is how you lose civilizational advantages in slow motion.

Intel’s Winning Because Everyone Else Overspent

Intel just posted a 7 percent revenue increase to $13.6 billion—$1 billion better than expected. That’s happening while every other chip company is scrambling to keep up with AI demand.

Why? Because Intel invested in manufacturing capacity when everyone thought they were irrelevant. TSMC and Samsung are booked. Intel has actual fabs that can actually produce actual chips. Right now, that’s worth more than any innovation metric.

This is the one place where American strategic planning actually paid off. The CHIPS Act basically said “we’re going to pay you to make semiconductors in America.” Intel said yes. Everyone mocked them. Now they’re printing money.

That’s not genius. That’s just the market correcting for decades of outsourcing decisions. But it’s useful to remember that sometimes the boring infrastructure play beats the exciting frontier play.

What Actually Worries Me

Anthropic just released Mythos, and apparently it’s causing “emergency responses from central banks and intelligence agencies globally.” That’s either hype, or it’s a sign that we’ve built something we don’t know how to control.

I’m genuinely uncertain which one it is. And that uncertainty is worse than knowing there’s a problem.

OpenAI’s taking a “more open approach to cybersecurity” than Anthropic. That probably sounds good until you realize what it means: they’re publishing their security practices so other people can study them, which is another way of saying they’re publishing their security gaps.

We’re in the phase where every move creates a new vulnerability. More openness = more exposure. More secrecy = more vulnerability to insider threats. There’s no clean win.

What I’m Watching

The Musk-Altman verdict timing. If this settles in January 2025, the messaging matters more than the outcome. If it goes to jury deliberation, someone’s going to lose in a way that shapes how future AI companies get funded.

China’s homegrown model performance. Within 18 months, we’ll know if Chinese AI firms can match US capabilities on their own models trained on our leaked blueprints. If they can, the White House memo becomes a historical marker—the moment American AI lost its moat.

Anthropic’s Mythos access decisions. Watch who gets permission to use it and who doesn’t. That’s the real geopolitical tell. If they’re giving access based on country of origin, you’ve got a proxy war. If they’re giving access based on use case, they’re trying to thread an impossible needle.

Microsoft’s Call of Duty play. They’re pricing Game Pass lower but delaying new releases by a year. That’s a deliberate sacrifice of day-one excitement for subscriber growth. If that works—if players actually accept the delay—it signals a fundamental shift in how the industry thinks about release windows and FOMO. Watch whether this becomes an industry standard or a Microsoft-specific experiment.