The Great Unraveling: What Happens When Tech Giants Stop Knowing What They're Building

---

Tim Cook is leaving. OpenAI’s models are hallucinating about goblins. Meta fired over a thousand workers and won’t explain why. Google’s travel AI forgot basic packing essentials. Meanwhile, a Twitch streamer got hit by a car while literally walking across America, live-streamed to thousands.

None of these things are accidents. They’re symptoms of the same disease: a tech industry that’s moved so fast and grown so large that it’s lost basic operational control.

When the Machine Stops Knowing What It’s Making

Let’s start with OpenAI’s goblin problem. The company told its ChatGPT models to stop talking about goblins because, as they put it, the issue “crept in subtly.”

This is wild. Not because of the goblins specifically—that’s almost funny. It’s wild because OpenAI, the company that convinced the world it had built AGI, apparently can’t prevent its models from randomly discussing fantasy creatures they were never trained to obsess over. They don’t know why it happened. It “crept in.” Like a ghost in the machine. Like they’re not actually in control of their own product.

Google’s Gemini travel planner is the same story in a different key. It’ll plan your flights, book your hotels, optimize your route—but it’ll forget to put underwear on your packing list. You’re outsourcing your travel logistics to software that’s missing fundamentals. The kicker? Google probably has no idea why, either.

Photo by Berna / Pexels

These aren’t isolated bugs. They’re the inevitable result of scaling systems so complex that even their creators don’t understand what they do or why they do it. We’ve been cheerleading this as “emergence”—the moment when AI systems start doing things we didn’t program them to do. Cool, right?

Except we’re now running actual business operations on these mystery boxes.

Apple’s Leadership Roulette Wheel

Tim Cook has run Apple for 15 years. The company just posted a 17% sales jump powered by iPhones. By any measure, that’s success.

So why is he handing over to John Ternus in September?

The public story is smooth transition, careful planning, succession ready since day one. But here’s what I think: Apple’s board looked at the AI arms race, looked at the fact that they’re not OpenAI or Google, and decided they needed someone younger thinking about what comes next. Cook built the machine. They want someone to rebuild it.

Ternus spoke publicly for the first time on an earnings call after his appointment was announced. One earnings call. That’s how little preparation we’re actually seeing. This feels less like the carefully orchestrated CEO transitions that companies love to brag about and more like musical chairs with a $3 trillion company.

My read: Apple’s current model—incremental hardware improvements, services revenue, ecosystem lock-in—isn’t exciting anyone anymore, especially not the board. They need disruption or the appearance of it. Whether Ternus can deliver is a completely open question.

Meta’s Ethics Theater Just Caught Fire

Over a thousand Meta workers in Kenya lost their jobs after they said they witnessed smart glasses users having sex. Meta and its subcontractor disagree on the details—why the layoffs actually happened, what the workers saw, whether they were even supposed to report it.

This is Meta’s entire ethics problem crystallized into one incident.

The company builds surveillance hardware and calls it “reality.” They employ thousands of people in the Global South to moderate content and flag problems that American executives decided were too expensive to solve algorithmically. Then when those workers do the job they were hired to do—see something, say something—Meta distances itself through subcontractors and claims confusion about what went down.

It’s not that Meta is evil. It’s that Meta has built a structure where the company genuinely doesn’t know what its own workers experienced or why they were fired. Plausible deniability as corporate architecture.

Photo by nappy / Pexels

The Streamer Who Got Hit by a Car (And Why It Matters)

Isaiah Thomas, known as hmblzayy online, was doing a 3,000-mile walking challenge. Live-streamed. In slow motion, he got hit by a car and kept going.

This isn’t really about him. It’s about what we’ve normalized. Thousands of people watching a human risk his body for engagement metrics. The platform made it possible. The audience made it profitable. The streamer made it inevitable.

This is the culture that emerges when attention is the only currency. You can’t monetize a safe life on Twitch. You monetize the edge.

What’s Actually Broken Here

These stories don’t connect because of some grand conspiracy. They connect because of a universal condition: scale without understanding.

Apple doesn’t understand what post-Cook innovation looks like. OpenAI doesn’t understand why its models behave the way they do. Meta doesn’t understand what its own contractors are doing. Google doesn’t understand why its AI makes basic mistakes. The streaming ecosystem doesn’t understand how to price human risk.

This is what happens when you build systems so large and complex that competence becomes almost theoretical. You can measure outputs. You can’t actually control them anymore.

Back in 2011, when Steve Jobs died and Cook took over, Apple was already massive. But it was still a company where a handful of people understood every major decision. By 2024? Cook’s leaving because the job is now mostly about managing momentum and succession.

That’s not a bad thing necessarily. But it signals that the era of founder-driven, comprehensible tech leadership is genuinely over. What replaces it is bureaucracy that moves at startup speeds—which is exactly as chaotic as it sounds.

My Take

The AI stuff will get worse before it gets better. Not because the technology is advancing too fast—that’s the narrative everyone uses. But because we’re letting companies deploy systems to actual customers while openly admitting they don’t understand how they work. Gemini forgetting underwear is funny. Wait until it’s a self-driving car or a hiring algorithm making identical mistakes.

Apple’s transition is the right call but the worst possible timing. Ternus is smart, but he’s taking over during the exact moment when hardware matters less and AI matters more, and nobody at Apple has proven they can compete in AI the way Altman or Demis Hassabis have.

Meta’s labor situation will barely move the stock price. That’s not a prediction, it’s an observation about how markets price in human cost.

What I’m Watching

• Ternus’s first major product decision (by Q4 2024): Does he push Apple into generative AI aggressively, or does he double down on privacy and incremental hardware? This tells you everything about whether he was picked to evolve Apple or preserve it.

• OpenAI’s next model release: When GPT-5 or whatever comes next launches, will they admit what they don’t understand about their previous models, or will they keep the “emergence” mystique going? This determines whether AI companies are being honest about their actual control level.

• Meta’s next contractor scandal (2025): They haven’t solved the structural problem. Watch for the next incident where outsourced workers do their job and get fired for it. The pattern will repeat.

• Google’s Gemini accuracy metrics post-underwear: Do they publish actual benchmarks on basic packing list generation, or does this quietly disappear from marketing materials? How a company responds to publicly visible failure tells you if they’re actually trying to fix things.