A few threads caught my attention this week, and they sit closer together than they first appear. China is formalising AI within the teaching profession itself. Parents are turning out in unexpected numbers for schools built on very different assumptions. A leading lab is trying to make security a shared, defensive effort before capabilities spread. And an investor is asking what education is actually optimised for. Different starting points, but each is really a question about what we value and how quickly we are willing to change.
China writes AI into the teaching profession
China plans to include AI in its teacher qualification exams and certification, as part of a wider effort to integrate AI across its education system. What strikes me is less any single measure and more the structural commitment behind it: AI is being written into both how teachers are certified and how students progress, rather than bolted on as an optional extra.
Most countries are still debating principles and running pilots. China is moving faster on making AI a formal part of the profession. Whether implementation lives up to the policy intent is, as always, the real question, and stated ambition and classroom reality can diverge sharply. But the direction of travel is hard to ignore, and it is worth watching how it plays out. You can read the summary of the plan here.
The demand for new school models
An open day at Alpha School's New York campus was planned for around 50 people and drew nearly 300 registrations, most of them young parents actively looking for alternatives. Alpha compresses core academics into a couple of hours a day and gives the rest of the time over to leadership, projects and real-world experience.
You can, and should, debate the model itself. There is plenty to question about how it works and whether it delivers what it claims. But the demand is the part I cannot dismiss. When six times the expected number of families turn up for something like this, it tells us something. It suggests parents are questioning long-held assumptions and asking what education is actually for.
That is a question worth sitting with. Those of us working in education can disagree about the answers on offer, but we should at least take seriously the fact that so many families are looking for a different one.
A defensive-first security coalition
Anthropic announced Project Glasswing, a coalition giving large software companies access to an unreleased frontier model to find and fix vulnerabilities in critical software. Before the announcement, the model autonomously found a 27-year-old vulnerability in OpenBSD, a 16-year-old flaw in FFmpeg that automated tools had hit five million times without catching, and a chained exploit in the Linux kernel that escalated to full machine control, alongside thousands of zero-days across major operating systems and browsers.
Standing up a defensive coalition before releasing these capabilities broadly is the right instinct. But the economics cut the other way: defensive work is expensive, slow to show a return, and benefits everyone including competitors, while offensive capability comes almost for free with each general improvement. That asymmetry will not resolve itself through goodwill. There is also a concentration question, since the partner list reads like a who's who of the firms that already dominate technology infrastructure. Whether defensive-first becomes a norm will depend on governments, customers and insurers making it the economically rational choice.
What AI does to education, from a fund manager's chair
A Bloomberg conversation with Songyee Yoon ranges across AI investing, gaming and workforce disruption, but there is a thread that lands squarely in education. Her argument is that schooling has spent a couple of hundred years optimised for knowledge delivery, and that knowledge is exactly what AI is now commoditising fastest.
If that is right, the implication is uncomfortable but clarifying: the thing our systems were built to do most efficiently is the thing losing its scarcity value. Her suggestion is to shift toward building adaptable, distinctly human capabilities, the things that do not commoditise as easily. It echoes the question raised by those parents queuing for a different kind of school. It is worth a listen if you are thinking about where AI actually fits in learning.