The Chaos Year: When Tech Stopped Playing It Safe

---

We’re living through the year when tech’s abstractions became concrete.

Not metaphorically. A Texas man was charged with attempted murder at Sam Altman’s home because—allegedly—he had documents advocating violence against AI executives. That’s not a thought experiment anymore. That’s a guy, actual charges, federal felonies. The debate about AI’s future just moved from conference rooms into police reports.

Meanwhile, Snap is firing 1,000 people. Allbirds, a sneaker company that sold its actual business for $39 million last month, just bought GPU chips and rebranded itself NewBird AI. Booking.com customers discovered their reservations were hijacked after a hack. And somewhere in a Slack channel, someone’s probably asking: “What if my digital twin is more productive than I am? Do they keep both of us?”

This is what happens when everyone goes all-in at once.

Photo by Markus Winkler / Pexels

The Productivity Trap Nobody Admits

Here’s the thing about digital twins that the companies pushing them won’t say outright: they work. Or at least they seem to. Firms claim staff become more productive when there’s a digital version of themselves—some AI-powered avatar handling parts of the job while the human handles other parts.

But “more productive” is doing a lot of work in that sentence.

The legal exposure is wild. If your digital twin makes an error, who’s liable? If it learns from your behavior and your behavior is protecting someone else’s intellectual property, who owns what the twin knows? And here’s the one nobody’s asking yet: if the digital twin is better at your job than you are, does the company have an obligation to keep you? These aren’t rhetorical questions. They’re the lawsuits waiting to happen in 2025.

My read: we’re about to see the first major case where a worker sues a company over a digital twin they didn’t consent to, and it’s going to be ugly. The company will argue it’s just a productivity tool. The worker will argue it’s a surveillance and replacement mechanism. Both will be right.

When Online Safety Becomes Political Theater

UK Prime Minister Starmer told tech bosses that “things can’t go on like this” with online safety. The government’s consulting on whether to ban under-16s from social media entirely.

This isn’t new policy thinking. This is a government admitting it’s lost control of the narrative and is swinging for the fences. A complete ban on social media for anyone under 16 would be the most aggressive move any Western government’s attempted.

Will it happen? Probably not in the form they’re considering. But the fact that they’re even discussing it tells you something: the regulatory ratchet only turns one direction. Every year the threshold for “acceptable online safety” gets tighter. Every year the liability shifts more toward platforms.

What’s weird is that this happens right after Booking.com got hacked, customers lost reservations, and the company wouldn’t even say how many people were affected. So we’re banning kids from TikTok for vague harms while actual financial platforms get breached and just… move on? The policy response is wildly mismatched to the actual risk.

Photo by nappy / Pexels

The Job That Can’t Be Automated (Yet)

There’s a wild article hiding in the headlines: “That Meeting You Hate May Keep A.I. From Stealing Your Job.”

It’s saying the human work of convincing, cajoling, arm-twisting, and reassuring is becoming more important as AI makes technical tasks easier. In other words: as the robots get smarter at doing the work, the humans become more valuable for doing the politics.

I think that’s exactly backwards from what people expect, and it’s the thing nobody’s gaming out properly. When AI can write the proposal, the value moves to who can sell the proposal. When AI can build the feature, the value moves to who can convince the org it matters. That’s not a job saved from automation. That’s a job description rewritten to favor the people who were already good at organizational influence.

The people who were good at technical execution but bad at meetings? Yeah, those jobs are disappearing. Snap’s laying off 16% of its workforce while “increasing reliance on artificial intelligence.” That’s a translation for “we need fewer people doing the technical work and more people managing the AI and managing the org.”

Snap’s betting that 840 of its 1,000 laid-off employees weren’t the arm-twisters and reassurers. That’s a bet that’s going to look either genius or catastrophic in about eighteen months.

When the Business Model Becomes the Headline

Allbirds going from “sneaker company” to “AI company that buys GPUs” in one month is darkly hilarious. They sold their business for $39 million—which, let me be clear, probably means they were losing money or the buyer saw no future in shoes. Then they said: we have cash now, so we’re buying chips and becoming a different company.

I’m genuinely uncertain whether this is genius or desperation. On one hand, they have capital, no operational business to run, and they’re moving into the hottest sector on Earth. On the other hand, they had a product people actually knew and wore, and they traded it for the right to compete in an insanely crowded market with zero brand recognition in AI.

This is what happens when every board room thinks the only good company is an AI company. Allbirds looked at their business and said: forget shoes, we’re doing the thing that’s working. But “the thing” is also what everyone else is doing. And most of them actually know how to build software.

Photo by Markus Spiske / Pexels

The Pentagon Moment

Here’s what’s easy to miss: the Pentagon is calling Ford and GM about weapons production. Why? “Concerned about the slow pace and high cost.”

This is not normal. The US military-industrial complex has spent eighty years figuring out how to produce weapons. The fact that they’re going to car companies means they’ve concluded their current approach is broken. They’re looking for manufacturing innovation from the one place in the US that still knows how to make things at scale.

This will become important in the next administration’s defense spending. If Ford and GM can actually help speed up weapons production, you’re going to see massive contracts. If they can’t, you’re going to see a lot of Congress people demanding answers about why the military can’t figure this out on its own.

What I’m Watching

• The first digital twin lawsuit: Betting this happens by Q3 2025. The trigger will be either a worker fired after a digital twin outperformed them, or an error the twin made that caused financial loss. Watch employment law forums and employment litigation blogs.

• Snap’s earnings call in Q1 2025: They need to show that firing 1,000 people while scaling AI actually works. If their output stays flat or improves on headcount cuts, every other tech company will do the same. If it tanks, we get the 2025 hiring story. This is the proof point everyone’s waiting on.

• The UK ban proposal’s actual vote: Starmer said it, but legislation is different from rhetoric. Watch whether they propose it in Parliament by mid-2025 and whether it survives committee. This will tell you if any Western government is actually willing to regulate social media at the utility level, or if it’s just talk.

• Allbirds’ first earnings/metrics: NewBird AI needs to announce something real—a product, a partnership, a customer—by Q2 2025 or the rebranding becomes a punchline. If they go silent, you’ll know it was desperation dressed up as strategy.