Agent Horizon

Nine countries managed to agree on an AI cooperation pact this week faster than three of the world’s biggest AI companies could agree on why their chatbots all stopped working at the same time. That is, honestly, a pretty good summary of where this industry is right now: genuine, unglamorous progress happening right alongside embarrassing, very public stumbles. A regional bloc quietly signed a real infrastructure-sharing deal. A major city made a evidence-first call on AI in classrooms instead of a knee-jerk one. And a trillion-dollar sector spent a chunk of the week apologizing to its own customers. Here’s what actually mattered.

Worth Reading

• Nine European Countries Agree Closer AI Cooperation — Czechia, Poland, Romania and six other Central and Eastern European nations signed the Prague Declaration, committing to actually link their computing infrastructure and share expertise rather than just issuing another vague joint statement about AI’s importance.
• Nation’s largest school district bans AI in the classroom through 8th grade — New York City paired a one-year moratorium on student-facing chatbots for younger kids with actual AI literacy classes for high schoolers, which is a far more useful response than either blanket bans or blanket enthusiasm.
• OpenAI unveils plan to protect critical services from AI cyberattacks — A billion dollars in subsidized cybersecurity tools for water utilities, community banks and local governments is the kind of unglamorous, genuinely useful AI deployment that never trends but actually helps.
• Uber is laying off 10% of staff, or 3,300 people — The layoffs are a real cost for real people, but the accompanying $10 billion robotaxi-partnership bet is a concrete, sizable wager on autonomous transport actually reaching scale, not another vague roadmap slide.

Spot the Hype

• Sam Altman Apologizes as GPT-6 Astra Staged Launch Denies Paid Access — OpenAI declared the dawn of the “AGI era” on a Thursday and was apologizing for a messy rollout to its own paying subscribers by Friday morning; even the future needs a working billing system.
• True AI-pocalypse as ChatGPT, Claude, and Grok all go down at once — Calling a few hours of elevated error rates an “apocalypse” is exactly the kind of headline that makes actual infrastructure incidents harder to take seriously when they matter.
• NYC Reverses Its Own AI Embrace: Generative Tools Banned for 600,000 K-8 Students — Framing a careful, exception-riddled, one-year moratorium as a dramatic “full reversal” flattens a nuanced policy decision into a headline built for maximum outrage clicks.

What’s the most overhyped AI headline you scrolled past this week — and which of these stories actually made you optimistic? Drop it in the comments. And if you missed it, catch last week’s digest here.