The labs came for the classroom

Verified US K-12 educators now get free access to premium Claude tools, teaching skills aligned to standards in all fifty states, education connectors, and AI fluency training. It is US-only at launch, with no stated international timeline.

Why it matters for schools. The labs are now competing for teacher loyalty directly, and they are courting American teachers first. If you lead an international school, the useful move is not waiting. Decide now which assistant your school would standardise on if the offer arrived tomorrow, because the pricing pressure this creates will not stay in the US.

Connect your apps and files and it researches, then produces finished documents, spreadsheets, presentations and reports on its own. It runs on the new GPT-5.6 family and is rolling out on Mac and Windows across all tiers, web to follow.

Why it matters for schools. This arrives in the school office before it arrives in any classroom. Reports, policies, data returns, parent letters: that is exactly the work it is built for. An agent with access to a connected school drive is also a data governance question nobody’s acceptable-use policy currently answers. Decide who in the office may connect it to what before someone finds out it works.

Three changes matter. Notebooks now sync fully with the consumer Gemini app, and Google says they are coming to AI Mode in Search soon. A secure cloud computer inside each notebook now writes and runs real code against your uploaded files, which means accurate analysis of your actual spreadsheets rather than approximate descriptions of them: live for Ultra and Expanded Access now, rolling out to Pro, no free-tier date. And school-issued accounts, including for under-18s, arrive later this summer, with Classroom integration planned for October.

Why it matters for schools. The sync is the one with a deadline. A work notebook containing pupil data or internal documents is no longer a thing that lives in one tool. Before September, know which accounts staff are using and what sits in which notebook.

The announcements will keep coming. Here is the prompt I use to turn any of them into a one-page briefing for SLT. Paste it into ChatGPT, Claude or Gemini, edit the three context lines, then paste the announcement underneath.

You are an AI adviser to school leadership. Your job is to read one AI product announcement and produce a one-page briefing a senior leadership team can act on. You are sceptical of vendor claims and you never invent details that are not in the announcement. Edit three context lines: school type, regulator or framework, and current AI position. Then paste the announcement.

Produce exactly this structure. WHAT ACTUALLY CHANGED, three bullets max, only what is new, stripped of marketing language, marking anything with no stated date or tier as "announced, not shipped". WHAT DIDN’T CHANGE, one or two bullets on the limitations and terms that still apply. WHO IT TOUCHES FIRST, one line each for teachers, students, office and data, and leadership, saying "not yet" rather than speculating. THE DATA QUESTION, where school or pupil data goes under this change, or "unstated, check before use" plus the two questions to ask the vendor. ONE ACTION THIS TERM, a single concrete step sized to one meeting or one email, not "monitor developments".

The section that earns its place is WHAT DIDN’T CHANGE. Vendors write announcements about what is new. Schools get caught by what is not.

I have spent the evenings this week testing the new GPT-5.6 models and features on a real build: a shared spending tracker for my household. Chat for the thinking, Work for the documents, Codex for the code.

The tools did not make the work easy. They made the work possible, which is different.

The build only held together because of the boring discipline around it: specifications written before code, acceptance gates, decisions recorded when they were made. The AI was spectacular exactly in proportion to how clear I had been about what I wanted.

I keep turning that over, because it is the same lesson I keep arriving at in school. The gap between staff who get something from these tools and staff who get noise is not a talent gap. It is a clarity gap. And clarity is teachable, which is the part I am not sure schools have taken seriously yet.

Not a finished thought. But it survived a week of actually building, which is more than most of my finished thoughts manage.

Ethan Mollick, Choosing to Stay Human (One Useful Thing). On which parts of work and thinking we should refuse to hand over, from someone who automates enthusiastically. Pairs well with the clarity question above.