The AI Reckoning Has Started—And It's Messier Than We Thought

---

We’re at a hinge point. Not the kind tech magazines announce with fanfare. The kind you only recognize by looking at the wreckage six months later.

The past few weeks have dumped enough contradictions onto the table to make anyone’s head spin. You’ve got Snap laying off 16% of its workforce while explicitly betting on AI. You’ve got a startup that just sold itself for $39 million planning to rebrand as an AI company and buy computer chips instead of making shoes. You’ve got government ministers telling tech CEOs the current state of things “can’t go on.” And you’ve got a man charged with attempted murder at Sam Altman’s house with documents full of violence fantasies tucked in his pockets.

These aren’t separate stories. They’re tremors from the same earthquake.

The Worker Surveillance Problem Is Here

Let’s start with something that hasn’t gotten nearly enough attention: digital twins.

The pitch is intoxicating. Firms say they can create digital copies of you—analyzing your work patterns, your productivity, your decision-making—and use those twins to make you a “superworker.” More output, same input. Everyone wins, right?

Except here’s what nobody’s saying out loud: this is worker surveillance wearing a productivity costume.

Digital twins aren’t a new technology. What’s new is the willingness of companies to scale them. And the fact that they’re doing it without clear legal frameworks. The moment you start modeling human behavior at that level of granularity, you’re in territory that employment law hasn’t caught up to. Can a digital twin be used against you in a dispute? What happens if your twin performs better than you do—does that become evidence you’re underperforming? Can it be shared with insurance companies or used in hiring decisions?

My read: this is going to blow up in the next 18 months. Not because the technology is bad, but because someone’s going to sue and win, and then every company that’s quietly rolled these out will scramble to delete the evidence. We’ve seen this movie before—with email, with location tracking, with keystroke monitoring. The legal system always lags the tech by about three years.

Photo by Marco Milanesi / Pexels

The AI Job Question Isn’t What You Think It Is

Here’s the weird part: the data suggests AI might not be a job-killer in the way we fear.

One headline that barely registered said “That Meeting You Hate May Keep A.I. From Stealing Your Job.” The premise is simple but profound—as AI makes individual tasks easier, the human work of coordination, persuasion, and reassurance is becoming more valuable, not less. Those meetings you dread? They’re increasingly what separates humans from obsolescence.

That’s not reassuring. That’s a complete reordering of what gets rewarded in the workplace. If your job suddenly revolves around 60% coordination and 40% actual work instead of the reverse, you’re in a different job. The title stays the same. Your compensation probably doesn’t.

Snap’s 1,000-person layoff happens in that context. It’s not that AI is replacing those workers outright. It’s that Snap’s recalculating what work looks like and deciding it needs fewer people doing different things.

Contrast that with what Allbirds is doing. The shoe company just sold itself for $39 million—a fire sale for what was once a darling of sustainable fashion. Now it’s pivoting entirely: buying computer chips, rebranding as NewBird AI, and apparently betting everything on becoming some kind of AI services firm.

This is the desperation move of a company that looked at its existing business and saw a sunset. It’s not alone. You’re going to see more of this in 2024—not pivots to AI, but full-throated surrenders. Companies that realize they can’t compete in their original market so they’re cashing in the chips and buying a ticket to the AI economy. Sometimes it works. Usually it doesn’t.

The Safety Argument Is Cracking

UK Prime Minister Starmer just told tech bosses that the status quo on online safety “can’t go on.” The government’s consulting on banning under-16s from social media entirely.

This isn’t new rhetoric. But the tone has shifted. There’s no more patience for self-regulation. No more “we’re working on it.” It’s: fix it or we’ll fix it for you.

Booking.com got hacked. Reservation hijacking. The company won’t say how many customers were affected. That matters because it suggests the breach is either ongoing or large enough that disclosing it would trigger a panic.

Google’s now punishing websites that use dark patterns—specifically, back-button tricks that trap users. This is a small thing that signals a big thing: tech giants are starting to police the ecosystem because governments are threatening to.

My prediction: we’re going to see a wave of regulatory action in 2024-2025 that makes the GDPR look like a gentle suggestion. The UK’s going to move on social media bans. The EU’s going to tighten the Digital Services Act’s enforcement. Congress is going to finally pass something, probably badly. And tech companies are going to spend more on compliance than on innovation for the first time.

That’s a structural shift. It’ll slow things down. It might also save them from themselves.

Photo by nappy / Pexels

The Threat Level Is Rising

A man was charged with attempted murder at Sam Altman’s house. He had documents advocating violence against AI executives.

Let’s not dance around this: we’re seeing the emergence of anti-AI extremism as a genuine security concern. Not fringe internet nonsense—actual coordinated threats and violence plots.

This matters for two reasons. First, it’s going to make security at tech companies more expensive and more paranoid. That’s not a minor cost. Second, it’s a signal that someone, somewhere, genuinely believes AI represents an existential threat worth killing over. Those people exist now. They’re organized. They’re armed.

I don’t think AI is an existential threat in the way these extremists do. But I also think dismissing them is stupid. Terrorism—even misguided terrorism—has a way of shaping policy and investment decisions regardless of whether the threat is justified.

The Pentagon Wants a Supplier

Meanwhile, the Pentagon is asking Ford and GM to help speed up weapons production.

This is buried under everything else, but it’s the most important story of the bunch. The military-industrial complex is creaking. Production is slow. Costs are astronomical. So the Pentagon is looking outside its normal contractor ecosystem to auto manufacturers with actual supply chain expertise.

What does this have to do with AI? Everything. Because once the Pentagon gets Ford and GM building weapons components, those companies become part of the defense industrial base. They’ll start investing in faster manufacturing, more automation, AI-driven production. The weapons-AI nexus deepens. And defense spending becomes even more intertwined with the companies that are building consumer AI.

That’s not a conspiracy theory. It’s just how power consolidates.

What I’m Watching

• Digital twin lawsuits by Q3 2024. Someone’s going to get fired, use a digital twin against their former employer as evidence of discrimination, and win. Watch employment law cases with “algorithmic performance” in the description.

• UK under-16 social media ban implementation timeline. If they actually pass legislation by mid-2024, the enforcement logistics are going to be a nightmare—and it’ll set a template for other countries. Watch for the EU to announce similar plans within 3 months of UK passage.

• AI startup graveyards. How many Allbirds-style pivots fail by December? If more than 30% of companies attempting to rebrand as AI services companies are either shutting down or merging by end-of-year, we’re in hype-collapse territory.

• Snap’s Q2 and Q3 earnings. Did the layoffs actually improve margins, or did they just destroy productivity? This is the real test of whether AI-driven layoffs make business sense or if they’re just optics.

The next six months will tell us whether this chaos resolves into a new equilibrium or spins into something worse.