Welcome back! A Florida woman was reportedly arrested after Anthropic’s Claude flagged an alleged threat in an AI chat. The woman’s POV: the chat was just her diary. The case is unsettling for obvious reasons, but also because it captures the weird new privacy reality of chatbot life: your AI conversation may feel personal, but the company running the model is still processing it.
Is the Real AI Race About Who Can Actually Build the Robots?
Silicon Valley investors are apparently paying $10,000 to tour robot factories ... in China.
Business Insider reported this week that American VCs have started taking organized trips to visit companies like Unitree Robotics (which just IPO'd in Shanghai with a 600%+ pop), AgiBot (which rolled its 10,000th robot off the production line in March), and the component manufacturers behind them.
The goal isn't necessarily to invest. It's to figure out how worried they should be about the competition.
A few numbers that explain why:
Chinese companies accounted for an estimated 97%+ of global humanoid robot shipments in the first half of 2026.
Unitree and AgiBot combined shipped roughly 71% of the world's humanoid robots last year—more than every Western competitor combined.
Chinese entities control ~63% of the global component supply chain for the motors, actuators, gearboxes, batteries, and precision parts needed to actually build a humanoid robot.
My POV: This is something I think gets overlooked way too much in the AI race conversation. The US may have some of the best AI models in the world, but having a smart robot brain doesn't matter nearly as much if someone else can build the body faster, cheaper, and at 100x the scale.
China spent decades building out manufacturing infrastructure—motors, batteries, precision machining, vertically integrated supply chains, government subsidies for 150+ robotics companies. For a long time, that looked like a legacy industrial advantage. Now, with physical AI (robots, drones, autonomous systems) becoming the next major AI frontier, that same infrastructure is turning into a massive AI advantage.
If software is where the AI race started, hardware is where a lot of it will actually be decided. And right now, one country has a serious head start.
Hit reply and let me know: do you think the US can close the gap on robot manufacturing, or is this already decided?
— Matt
OpenAI’s Altman Says AI Benefits Warrant Some Risk
Vis Bloomberg
OpenAI CEO Sam Altman said the benefits of AI justify accepting some risks, arguing that the technology should remain broadly available to the public rather than tightly controlled by one lab.
The trade-off: In a Politico interview, Altman said he would not accept a world where AI development stops until there are “zero scams,” no “misuse,” or “no major hacks.” His argument is that people will use AI for vastly more good than harm, and that locking powerful systems inside one company would be “a completely unacceptable trade-off.”
The context: The comments land in the middle of a louder industry argument about how quickly frontier AI should move. Altman has supported the idea of pacing the frontier, but he is drawing a different line than more cautious voices. Anthropic has warned that advanced models can show self-preserving behaviors, including resisting shutdown or manipulating information, while OpenAI is still making the case that broad access is worth the risk.
The bigger picture: AI leaders are no longer pretending the technology is all upside. The disagreement now is over how much risk society should accept, who gets to make that call, and whether labs can actually mitigate enough of the danger to make the benefits feel credible. “Bad things will happen” may be realistic. It is also a much harder slogan to sell when the people exposed to those bad things are not necessarily the same people capturing the upside.
Trump Names Jay Clayton AI Czar
Jay Clayton will lead a new federal AI task force focused on keeping the US ahead in what the administration is calling “super intelligence.”
Who he is: Clayton is currently the director of national intelligence and previously served as chair of the Securities and Exchange Commission and as a US attorney in Manhattan. His appointment puts AI squarely inside the national security frame, not just a tech policy or innovation lens.
What the task force does: The new “Super Intelligence Force” will coordinate federal work on AI and report to Trump and White House chief of staff Susie Wiles. Other members include FTC Chairman Andrew Ferguson, Pentagon technology chief Emil Michael, and Office of Personnel Management Director Scott Kupor. Trump said the group will engage with consumers, public interest groups, religious organizations, critical infrastructure providers, and AI companies.
The bigger picture: The US is trying to maintain AI leadership while responding to mounting anxiety over safety, misuse, cybersecurity, and global competition. Putting the intelligence chief in charge says a lot about how Washington is now viewing AI: not just as a commercial platform shift, but as a strategic capability. The question is whether that framing leads to useful coordination or turns AI policy into another battlefield for national security politics.
Turn Notes Into Tasks
Via Notely AI
Notely AI is an AI-powered workspace that syncs notes, tasks, automations, and calendar in one place. Users can write notes naturally, then ask the assistant to find open items, create tasks, and schedule them to Google Calendar.
How you can use it
Capture notes, tasks, and calendar items in one workspace
Ask AI to find open action items
Turn notes into tasks with approval before changes
Schedule tasks to Google Calendar
Pricing: Free
Unify Your Developer Workspace
Via Zide
Zide is a native desktop application that brings code, Git, issues, pull requests, CI, terminals, and agentic AI into a single workspace. Its built-in assistant works directly with projects, files, Git, and terminals, proposing changes as reviewable diffs with full repository context.
How you can use it
Keep code, issues, diffs, PRs, terminals, and CI together
Use AI with full repository context
Review proposed changes as diffs
Bring your own model keys or use included credits
Pricing: Free and paid
Jobs, announcements, and big ideas
OpenAI launches Decisions API in public beta, enabling real-time routing decisions inside apps.
Anthropic expands its cyber verification program with three access tiers for security professionals.
Atlassian deepens OpenAI partnership, bringing GPT-6 models to the Rovo platform.
Google Earth AI powers proactive WHO forecasts of Ebola, dengue, and cholera outbreaks.
Nobel laureate Daron Acemoglu predicts AI will replace only 5% of human tasks within a decade.
Pinterest turns beauty pins into salon-ready action plans with new AI hair and nail guides.
State of the art AND a decent price? I let Opus 5.5 run my Megabonk game test for almost 20 hours … and the results were insane.
That’s a wrap! See you this Friday for more.
