The Year Silicon Valley Stopped Pretending to Be in Control

---

A Molotov cocktail through a gate. A federal court telling an AI company it’s a national security risk. A dancer performing through a digital avatar controlled by her thoughts. Meta admitting its new AI model lags competitors on coding. Amazon quietly killing off a decade of reading devices.

These aren’t isolated incidents. They’re the sound of an industry’s mythology cracking.

For the past fifteen years, Silicon Valley has sold us a story: technology solves everything, disruption is progress, and the smartest people in the room are building the future. That narrative is collapsing in real time.

The Violence Nobody Wants to Talk About

Let’s start with the Molotov cocktail hurled at Sam Altman’s San Francisco home. Police arrested a suspect. The device burned his exterior gate. And then—nothing. The tech press moved on. No congressional hearing. No moment of reckoning about what happens when a CEO becomes a symbol of something large enough to inspire violence.

I’m not being naive about threat assessment. One incident doesn’t make a trend. But the reaction to it does. The reaction is silence. The reaction is treating it like a weather event—unfortunate, unpredictable, not something we as a society need to actively solve.

Compare this to 2013, when executives started requiring private security. We had conversations then. We asked ourselves hard questions. Now? The narrative is: he’s in a dangerous position, he knows it, and that’s just the cost of doing visionary work.

That’s the mythology breaking down. When we stop being shocked by violence against powerful people, we’ve entered a different era.

Photo by Robert So / Pexels

The AI Companies Can’t Actually Do What They Promised

Meta just released Muse Spark from its new Superintelligence Lab. It performs better than Meta’s previous models. On coding ability? It lags rivals.

Think about that for a moment. Meta—the company that profits from engagement algorithms so sophisticated they can predict what you’ll click on before you know it—cannot match OpenAI’s ability to write code. The one thing that actually matters economically in AI right now is the one thing Meta can’t do better.

And then there’s Anthropic, which just got slapped down by a federal court over a “supply chain risk” label from the Defense Department. The court denied its motion to lift the label. Translation: the government thinks Anthropic’s AI poses a security risk, and the company can’t even successfully argue otherwise in court.

Here’s what I think is happening: the hype phase is ending. These companies spent three years telling us AGI was coming, that AI would transform everything overnight, that we were living in a historical moment. Some of that’s true. But the part that’s true—that AI works really well for narrow, specific tasks like coding or chemical discovery—that part doesn’t actually justify trillion-dollar valuations. The part that would justify those valuations is the part that isn’t real yet.

So we’re getting a slow collision between expectations and capability. Meta’s coding gap is evidence of it. Anthropic’s legal setback is evidence of it.

Regulation Is Quietly Winning

The White House just told staff not to bet on prediction markets. Sounds small. It’s not.

Prediction markets—gambling platforms where users wager on future events—have grown popular. Some people are getting rich betting on geopolitical outcomes. And the government’s response wasn’t to ban them or heavily regulate them. It was quieter: don’t use them. Don’t participate.

That’s how you kill something in Washington without having to fight Congress. You just opt out. You make it uncool. You create reputational risk for anyone who uses it.

Meanwhile, Meta pulled ads recruiting people for social media addiction lawsuits. Why? Because Meta lost a landmark trial in California on exactly this issue. The company made a strategic decision: this fight is unwinnable right now. Pull the recruitment ads. Wait for the moment to counterattack.

That’s regulation working—not through legislation that’s explicit and dramatic, but through court victories that shift the incentive structure. Companies are starting to calculate that fighting regulators costs more than just complying.

Photo by nappy / Pexels

The Disinformation Problem Is Now Institutional

London’s mayor says the city is being targeted with disinformation portraying it as “in decline.” He’s not claiming foreign interference—he’s noticing a pattern. A coordinated effort to shape perception.

This is what happens when AI can generate convincing content at scale. It’s not that any single piece of disinformation is dangerous. It’s that the volume has crossed a threshold where it changes behavior. People start believing London is unsafe. Businesses move. Tourists avoid it. The disinformation becomes true through behavioral response.

We’ve known theoretically that this could happen since 2016. Now it’s happening. And we don’t have tools to stop it at the institutional level. We have fact-checkers. We have flagging systems. What we don’t have is a way to make truth stick when falsehood is cheaper to produce.

The Analog World Is Fighting Back (Quietly)

Volkswagen is ending EV production at its Tennessee plant and pivoting back to gasoline. This is supposed to be a climate tech story. Instead, it’s becoming a story about the limits of how fast you can actually change an industry.

Amazon is ending support for older Kindles—devices released before 2013. It sounds like a technical decision. It’s actually a statement: we’re done supporting the old internet. We’re moving on. Users screamed. Amazon ignored them.

These companies are realizing something that took a decade to sink in: disruption has limits. You can’t actually force the world to change faster than it wants to. You can force it to adopt your products, but that’s different from changing its underlying behavior.

A dancer with motor neuron disease performed on stage again through a digital avatar controlled by her brainwaves. Breanna Olson said the technology re-established the expression and connection she felt she’d lost.

This is the part that actually matters. Not the hype cycle. Not the valuation games. Not the marketing. The part where technology gives someone back something they’d lost. That’s real. That’s worth paying attention to. And it’s barely mentioned in the tech press because it doesn’t move stock prices.

My Read

I think we’re entering a phase where the mythology of Silicon Valley is separating from its actual utility. The useful stuff—better medical imaging, accessibility tech, communication tools—that keeps working. But the narrative that tech solves everything? That’s done.

The violence against Altman is a symptom, not a cause. The cause is that the industry promised to solve problems it can’t actually solve. It promised to eliminate inequality. It promised to fix media. It promised to make work better. It promised to save the planet.

Instead it made some things more efficient and other things worse. That’s not a failure of tech. That’s how tech actually works. But it violates the mythology.

When mythology fails, people get angry. Sometimes they throw Molotov cocktails. Sometimes they sue. Sometimes they just stop believing.

The question now isn’t whether AI will transform the world. It’s whether the companies building AI can survive the gap between what they promised and what they’re actually delivering.

My prediction: we’ll see at least one major AI company get broken up or severely restructured by 2027, not because of regulation but because it can’t compete economically. Probably Meta. Maybe Anthropic. The hype cycle has a half-life, and we’re watching it decay in real time.

What I’m Watching

• Anthropic’s next legal battle with the Defense Department (watch for Q2 2025). If they lose again, expect investment to dry up. If they somehow win, it’ll mean the security concerns are actually overblown—which would be good news, but it’d also mean the government narrative was wrong, which the government never admits.

• Meta’s coding gap vs. OpenAI by benchmark, specific metrics by Q3 2025. If Meta closes it, the company has a real path back. If the gap widens, it’s a sign that some AI capabilities just can’t be bought or built at scale.

• Whether Amazon faces real pushback on the Kindle shutdown or just accepts user anger as the cost of doing business (decision point: March 2025). This is the test of whether companies still think they need to maintain customer goodwill or whether they’ve decided that ecosystem lock-in is enough.

• The prediction market ban becoming institutional policy across the federal government (watch for similar directives from other agencies by mid-2025). If it spreads, we’ll know the White House has decided these markets are genuinely destabilizing. If it stays isolated, it was just optics.