What Actually Happened in Tech Last Week (And Why Most People Missed It) 📡

“WWDC got all the headlines. But the week’s biggest story wasn’t on a stage in Cupertino.”

I almost missed all of this myself.

Between writing the WWDC recap and going through my usual feed, I realized last week had way more going on than one keynote. A government forced a billion-dollar AI company offline in 90 minutes. A camera app company launched a $2,195 bet against the smartphone. And the infrastructure war behind AI quietly got even bigger.

None of it trended the way WWDC did. All of it matters more long-term.

Let’s go through it.

🚨 The Biggest Story Nobody’s Talking About: The White House Pulled Anthropic’s Models Offline

This one’s wild, and I want to lay out the facts carefully because there’s a lot of noise around it.

On June 9, Anthropic released Claude Mythos 5 — its newest, most advanced model — to a small group of cyberdefenders and infrastructure providers, working in collaboration with the US government.

Three days later, on June 12, the Trump administration sent Anthropic a letter under a new export control directive. The administration banned the use of two of Anthropic’s latest models, Fable 5 and Mythos 5, by foreign companies, governments, and individuals — prompting the company to revoke access altogether to comply.

The speed of it is what stood out to people tracking this closely. The White House called Anthropic on a Friday afternoon, said the models posed a national security threat, and gave the company 90 minutes to disable them.

What triggered it, according to reporting: Anthropic had given the administration a list of 111 organizations slated for advanced Mythos access. Officials reviewed and approved that list. Anthropic later disclosed the list had grown by roughly 50 additional entities that already had access by the time of disclosure — which is what reportedly broke trust and led officials to push for export controls.

This sits inside a longer-running tension. Back in February, the Pentagon had reportedly labeled Anthropic a supply-chain risk to national security — a designation usually reserved for foreign firms — after Anthropic refused to let its models be used for autonomous weapons or mass surveillance of U.S. citizens, red lines the company had written into its own usage policies.

Anthropic’s CEO Dario Amodei published a blog post days later titled “Policy on the AI Exponential,” writing that the evidence of AI’s power and its risks has become undeniable, and that the company’s own Mythos product was an emblematic example of the threat frontier models pose.

Why this matters beyond the headline drama:

This is the first time the US government has used national security export control authority to force a major American AI company to pull a flagship product from the market — not over a hack, not over a leak, but over an access-control dispute. If you’re building anything on Claude, MCP, or planning architecture around any frontier model API, this is the kind of regulatory risk that’s now real, not theoretical. The rules for advanced AI access are being written in real time, under emergency authority, and they can change in 90 minutes.

I’ll be watching how this resolves. Anthropic has said it’s working to restore access. Worth a follow-up post once there’s clarity.

👓 Snap Just Bet $2,195 Against Your Smartphone

While Apple was busy on stage in Cupertino, Snap quietly unveiled something that might matter more for the next decade of computing.

At Augmented World Expo on June 16, Snap CEO Evan Spiegel unveiled the company’s latest AR Specs, calling them “the beginning of a new era in computing.” The glasses are fully standalone — no compute puck, no phone tether required.

The specs (pun intended):

Specs run two Qualcomm Snapdragon chips — one dedicated to computer vision, one for running AR experiences. They offer a 51-degree field of view, 16 million colors via proprietary LCoS display tech, and up to 20 hours of total battery life with the charging case.

Price: $2,195, with a refundable $200 deposit to preorder. Shipping this fall in the US, UK, and France.

For context — that’s more than 15 times the price of Snap’s original $130 camera-only Spectacles from 2016.

Spiegel’s actual quote on why: “Almost 20 years since the launch of the iPhone, people are ready to think about computing differently.”

The demos shown included real-time navigation overlays, live language translation, gesture-controlled interfaces, contextual help with car repair, cooking timers, and furniture measurements — all rendered directly in your field of vision, not on a screen you have to look down at.

Why I think this matters for mobile developers:

Snap is beating Apple, Meta, and Google to the consumer AR glasses market in 2026. Apple’s Vision Pro is $3,499 and headset-based. Meta’s Ray-Ban Display is $799 but more limited. Snap is positioning Specs as the middle ground — full AR computing, standalone, consumer-priced (relatively).

If this category takes off even modestly, the apps that win will be the ones that already think beyond the rectangle. Spatial UI, gesture interaction, glanceable information design — these stop being niche skills and start being relevant roadmap items. Flutter doesn’t have a clean story here yet. Someone’s going to build one.

🏗️ The Infrastructure War Nobody’s Calling a War

This is the quiet thread running under every other story this month, and I think it’s the most important one long-term.

A clear theme has emerged: governments, investors, and technology companies are racing to build the physical infrastructure required for AI — competing not just on algorithms, but on power generation, chips, networking, and data centers.

The headline numbers from just the past two weeks: Alphabet raised $80 billion in equity specifically for AI infrastructure. SoftBank committed $52 billion to a European data-center push. Anthropic confidentially filed for an IPO at a reported valuation of approximately $965 billion, following a recent funding round.

Even Jeff Bezos is reportedly betting $12 billion on what’s being called an “artificial general engineer” project.

Meanwhile, America’s AI data center boom is running into delays — major projects slowed by power grid constraints, permitting hurdles, and supply-chain bottlenecks, with a large portion of planned 2027 data center capacity not yet under construction.

Why this is the story under all the other stories:

We talk about AI progress as if it’s purely a software and model-quality race. It’s not anymore. The constraint has shifted to electricity, chips, and physical buildings. Every API call you make, every on-device model you can’t quite fit, every “why is this rate-limited” moment — traces back to this infrastructure crunch. It’s not abstract macro news. It’s the reason your AI API bill behaves the way it does.

🧵 Smaller Stories Worth Knowing

A few more things from the week that didn’t get the spotlight but are worth filing away:

SpaceX bet $60 billion on AI coding agents — the rocket company is now a major investor in AI dev tooling, while simultaneously becoming more valuable than Amazon. Space and software are colliding in ways that would’ve sounded absurd three years ago.

Nvidia quietly reopened doors in China with its new Vera CPU — a sign the chip export tensions between the US and China are shifting again, just weeks after restrictions tightened.

Microsoft pushed AI agents into everyday computing, continuing its move toward on-device, agent-first workflows — directly relevant if you’ve been following my on-device AI series.

G7 leaders sat down with frontier model CEOs at this week’s summit, with European leaders pushing hard for AI sovereignty — discussing large-scale investment in sovereign cloud and domestically developed foundation models, driven by growing discomfort with dependence on American AI infrastructure.

🧭 My Honest Take

The Anthropic story is the one I keep coming back to.

Not because of the politics — I’m not going to pretend to have a strong take on export control policy. But because of what it reveals structurally: even a company that built explicit safety red lines into its own usage policy got overridden by national security authority in 90 minutes.

If you’re building products on top of frontier AI APIs — any of them, not just Claude — this is a reminder that the ground under “API stability” is less solid than it feels. Build with that in mind. Don’t architect yourself into a corner where losing model access for a week breaks your product.

The Snap glasses are the fun story. The infrastructure numbers are the boring-but-important story. The Anthropic situation is the one that should actually change how you think about risk in what you’re building.

That’s the week. WWDC got the keynote stage. This is what happened around it.

Tags: Tech News, AI Policy, Anthropic, Snap AR, AI Infrastructure, Mobile Development, Weekly Roundup