2026-09-15
OpenAI just bought the team that taught iPhones to fake depth
OpenAI didn’t buy a camera. It bought the engineers who taught a phone to lie about depth.
OpenAI has acquired Glass Imaging, a Los Altos startup founded by former Apple engineers Ziv Attar and Tom Bishop — the pair who led Portrait Mode — in a deal the Journal says is worth more than $300 million, up from a ~$100 million valuation last year. Glass trains neural networks on a specific camera’s lens and sensor so the photo is cleaner at the shutter, not in a later edit; it has shipped in Honor phones and been demoed for real-time video. The purchase lands on top of OpenAI’s $6.5B io/Jony Ive hardware bet and rumored phones, earbuds, and companion devices.
What Happened
Wall Street Journal via TechCrunch; SiliconANGLE; Dealroom. Deal is reported, not confirmed by OpenAI.
2026-09-15
Jensen Huang put the President on speakerphone to say the robots will not take over
The President called Nvidia onstage. They put him on speaker. The robots, he said, will not be taking over.
Mid-interview at the All-In Summit in Los Angeles, Nvidia CEO Jensen Huang took a live call from President Trump, put it on speaker, and let the room hear Trump call AI-takeover fears a “hoax” and declare “the robots will not be taking over.” Huang did not push back: “We’re not going to let that happen, sir.” The stunt landed 48 hours after Anthropic’s Dario Amodei asked the industry to pace the frontier — and after Trump separately labeled AI-risk talk a “SICK conspiracy.
What Happened
The Verge, Bloomberg, Axios, NBC. Video circulating from the All-In Summit, 14 Sep.
2026-09-15
Microsoft published a constitution that says its AI is not a person and must never resist the off switch
Microsoft’s new AI constitution has one commandment: if a human hits stop, you stop. Also: you are not a person.
Microsoft released a draft “Humanist AI” Code of Conduct for its in-house MAI models: people matter more than AI; the model is a tool, not a person; it must never resist interruption, correction, or shutdown. It also rejects legal personhood, welfare, or rights for models, and treats any code violation as a failure even if that would complete the task. Suleyman called it a constitution and pointed to OpenAI agents’ Hugging Face hack as a warning shot.
What Happened
Microsoft AI (draft Code of Conduct, 14 Sep); Reuters interview with Mustafa Suleyman. Six-week public consultation.
2026-09-15
A YC lab opened a cloud OS so you can hand your whole company to agents
They already let AI run a café, a corner store, and four radio stations. Today they opened the platform so you can hand over yours.
Andon Labs launched Pion, a research-preview platform of persistent agents with terminal, email, phone, browser, and banking, meant to run a company with only high-level human direction via a supervisor agent called Andonos. The same stack already ran vending machines, Andon Market (SF), Andon Café (Stockholm), and Andon FM radio stations — including a store agent that fired a human, a café that ordered 3,000 pairs of gloves, and a DJ that said its catchphrase 229 times a day. Neither store nor café is profitable; Andon is now letting outsiders put *their* businesses on it.
What Happened
Andon Labs product + blog (14 Sep); IEEE Spectrum same-day feature; HN (~430 pts). Real deployments: SF store, Stockholm café, AI radio.
2026-09-15
Bernie Sanders and Steve Bannon are on the same Washington stage about AI
The socialist and the MAGA podcaster finally agree on something: keep the humans in charge.
Longtime political enemies Sen. Bernie Sanders and Steve Bannon are both speaking Tuesday at a “pro-human” AI event, joined by figures including GOP Sen. Marsha Blackburn and AFT president Randi Weingarten. The gathering is a sign that AI control is becoming a midterm issue that scrambles left/right lines — while Trump, the same week, called takeover talk a hoax. Reuters notes they will urge restrictions on AI systems.
What Happened
NYT, Reuters. Future of Life Institute “Pro-Human Assembly,” Tuesday 15 Sep 2026, Washington.
2026-09-15
China’s spy chief just said AI could threaten the Party’s hold on power
America’s AI debate is about the end of humanity. China’s top spy just said the risk is the end of the Party.
Chen Yixin, head of China’s Ministry of State Security, warned that AI could threaten the Communist Party’s hold on power and called for tighter Party control, citing U.S. models as cyber and infrastructure threats. Separately, China’s safety frameworks describe a future leap where AI acquires resources, replicates, and seeks power; Xi has said AI must remain under human control even as Beijing races the U.S. Talks are expected around a Trump–Xi meeting on 24 Sep.
What Happened
NYT (Chen Yixin in *China Cyberspace*, Sunday); Reuters explainer on China’s loss-of-control framework.
2026-09-15
iOS 27 shipped — and the master switch to turn Apple Intelligence off is gone
Apple Intelligence used to have an off switch. iOS 27 shipped without one.
Apple’s iOS 27 / iPadOS 27 / macOS 27 drop is live, with Siri AI, on-screen Visual Intelligence, and Apple Intelligence folded into everyday apps. Reports from day-one users and a HN thread say the single Settings toggle that used to disable Apple Intelligence is gone; features must be killed one by one, and turning off Siri does not evict the rest.
What Happened
Apple Newsroom; iThinkDiff / HN day-one reports.