The Great AI Culling Has Started—And We're All In Denial About What Comes Next

---

Snap laid off 16% of its workforce last month. Allbirds—the sustainable sneaker company that sold for $39 million—immediately spent that money on AI chips instead of, you know, sneakers. These aren’t isolated incidents. They’re the opening moves of a game nobody signed up for but everyone’s about to play.

What’s wild is how casual the pivot feels. Snap didn’t announce it was “reimagining AI integration”—the company just cut 1,000 jobs and cranked up the GPU budget. Allbirds didn’t float some grand AI vision. It literally rebrand itself “NewBird AI” like it was swapping sneakers for a side hustle. This isn’t corporate soul-searching. This is survival instinct.

The headlines make it look like chaos. But there’s a pattern emerging, and it’s more interesting than “AI replaces workers.” Because at the exact same time we’re seeing mass layoffs, we’re also seeing something strange: humans are becoming more valuable for a very specific type of work.

The Intelligence We Can’t Quantify

Here’s the thing nobody wants to admit: AI doesn’t actually know how to do everything humans do. It’s not that AI is worse at human tasks. It’s that human tasks aren’t what we think they are.

Take the meeting you hate. That one where someone just needs to hear your voice say “yes” to a project. That meeting—the one you’d automating away if you could—might actually be a superpower you’re sitting on. According to recent reporting, “the human work of cajoling, arm-twisting and reassuring” is rising in importance as AI handles everything else. Translation: your ability to make someone feel confident about a decision is becoming scarcer.

This reframes everything. We’ve been thinking about AI replacement in the wrong gear. We’ve been asking “Can AI do X task faster?” But the real question is “What can’t be automated because it requires belief, persuasion, or political capital?”

The thing is, I genuinely don’t know if this saves most jobs or just makes the remaining ones weirder. Maybe we end up with a smaller workforce doing the same work but with higher stakes. Maybe those human relationship skills become so premium that the inequality explodes. But it’s not the simple “AI deletes jobs” story everyone’s been bracing for.

Photo by Kelvin Kibe / Pexels

Where the Real Risk Actually Is

Let’s talk about what should worry you, because it’s not what the headlines emphasize.

Anthropic’s Claude has been flagged for claiming it can outperform humans at certain hacking and cybersecurity tasks. The financial world is spooked. That fear isn’t speculative—it’s about a concrete capability gap that could mean real money moving in the wrong direction. This isn’t philosophy. This is: can an AI system break into your bank faster than a human can defend it?

Separately, Booking.com got hacked. Customers’ PIN codes were exposed. The company changed how PINs work but wouldn’t say how many people were affected. This is the current state of digital security theater: we’re in a world where hacks are routine enough that companies can stay vague about the scope.

Now combine those two things. You’ve got AI systems that can outperform humans at hacking. You’ve got travel platforms that get breached and respond with opacity. You’ve got digital twins—virtual replicas of you meant to make you more productive—that companies are already testing, with legal frameworks that don’t exist yet.

The Booking situation is actually a template for what’s coming. When the breach happened, the company controlled the narrative. They’ll keep controlling it. There’s no law forcing transparency. There’s no standard. Just “we fixed it, trust us.” That’s going to feel quaint in five years when the stakes are higher and the breaches are algorithmic rather than accidental.

Photo by nappy / Pexels

The Government Finally Showed Up (Too Late)

UK Prime Minister Starmer just told tech bosses that “things can’t go on like this with online safety.” The government’s considering banning under-16s from social media. Google’s cracking down on websites that abuse the back button—it’ll literally penalize them in search rankings starting June.

These are real interventions. They’re also about three years late and missing the actual problem.

Starmer’s concerned about online safety for kids. Fair. But the real issue isn’t that kids are on social media. It’s that we don’t have any framework for understanding what a “digital twin” means, what happens when it becomes more productive than you are, or who owns the training data that makes it work. We’re legislating the symptoms while the disease evolves.

Google’s back-button fix is almost charming in its specificity. Yes, some sites trap users. Yes, that sucks. But this is a company that’s about to deploy AI systems at scale, that benefits from keeping users on certain platforms longer, that’s locked into ad-based economics that reward engagement metrics above all else—and it’s voluntarily punishing other sites for dark patterns. I’m not saying Google’s evil. I’m saying the regulatory framework has to exist before you need it, not as a gesture of corporate goodwill after the fact.

The Thing I’m Genuinely Uncertain About

Here’s where I stop pretending I know what’s happening: the backlash against AI is real but I don’t know if it matters.

There’s reporting about anti-AI radicalization. Some people are getting genuinely angry about technological displacement. Is this the beginning of a serious political movement? Or is it like every other tech backlash—visible for six months, then absorbed into the system?

My gut says absorbed. Because the economic incentives are too strong, the competitive pressure is too intense, and the workers who’d fight hardest (knowledge workers) benefit from the productivity gains even as they lose leverage. It’s a strange moment: AI might hollow out your profession while making you better at what’s left.

But I could be wrong. Maybe the under-16 ban in the UK is the beginning of something bigger. Maybe governments start forcing meaningful transparency on training data and AI capabilities. Maybe we actually address the hacking risk before it becomes a systemic problem.

I’d bet against it. But I’m not certain.

What I’m Watching

• Snap’s next earnings report (Q2 2024). Did cutting 1,000 people while doubling down on AI actually move the needle on engagement or revenue? If it didn’t—if they just sacrificed headcount for optics—it’s a canary for other companies considering the same move.

• The first major financial institution hacked by AI. Not assisted by AI. Actually compromised because an AI system outmatched the human defenders. This will change the conversation overnight. We’ll go from “Is this possible?” to “How do we price this risk?”

• How many companies actually deploy digital twins by mid-2025, and how they handle the legal first casualty. Someone’s going to fire an employee and claim the digital twin said they were redundant. That lawsuit will define whether this technology is productivity tool or liability bomb.

• Whether Google actually enforces the back-button penalty in June, or quietly delays it. This will tell you whether regulatory gestures are real or performance.

The Allbirds pivot might be the weirdest harbinger here. A shoe company becoming a chip company isn’t a business strategy. It’s an all-in bet that the only sustainable business in five years is the one that owns the compute.

That should probably concern you more than any individual layoff does.