OpenAI Engineers Design a Device With Astra
Two OpenAI team members used GPT-6 Astra to design a test device in under a week. A contract manufacturer built a small batch of prototypes in under a month. Getting the design ready for manufacturing required several rounds of review to address production limits, tolerances and assembly details.
Texas governor win tanks AI stock
AI infrastructure stocks tumbled in overnight trading after Democrat Gina Hinojosa won the Texas gubernatorial race, defeating Republican incumbent Greg Abbott. Hinojosa campaigned on implementing a moratorium on new AI data center development in Texas, which hosts 27% of planned U.S. data center capacity and serves as a critical hub for AI infrastructure buildout.
- Nasdaq-100 futures dropped 0.8% overnight, with CoreWeave down 7%, SpaceX off 6%, and Oracle falling 4% on Robinhood as investors priced in the risk of data center restrictions in the nation’s second-largest buildout market.
- Texas represents the largest concentration of planned AI infrastructure in the U.S., accounting for 27% of all planned data center gigawatts according to Wells Fargo estimates. Democrats also won or retained governorships in Georgia (7%), Ohio (5%), and Pennsylvania (5%).
- Wells Fargo’s chief equity strategist warned the Texas result “could potentially be a left-tail event to the AI trade,” as Hinojosa explicitly campaigned on blocking new data center permits until environmental and grid impact studies are completed.
For enterprise AI buyers, the political shift in Texas signals infrastructure deployment risks that could cascade into pricing and availability constraints. Companies planning private deployments or evaluating cloud providers should verify datacenter location strategies and assess backup capacity outside Texas and other states with new Democratic leadership hostile to datacenter expansion.
OpenAI Internal Model Leads on BenchCAD 2.0
OpenAI reports that an unnamed internal model scored 0.875 on BenchCAD 2.0, compared with 0.594 for GPT-6 Astra at its highest reasoning setting. The evaluation covers tasks such as reconstructing 3D parts and assemblies and identifying connections on printed circuit boards.
Stanford Launches Digital Brain Project
Stanford launched a project to build foundation models that predict human brain activity. Researchers plan to combine high-resolution scans from a 7-tesla MRI with millisecond measurements from a wearable MEG system, capturing where and when brain activity unfolds.
Experiment Supports a Route to Topological Qubits
University of Cologne researchers observed Andreev states behaving as predicted at a junction between a topological insulator and a superconductor. The result supports a proposed route toward more robust qubits. The team still needs to measure whether the energy gap closes and reopens as predicted.
Google will cut off search access for schools that reject AI
Google announced Tuesday that K-12 schools using Google Classroom can now individually toggle AI tools on or off. The catch: search is bundled with AI, so schools that opt out of Gemini lose access to Google search entirely.
The move reverses Google’s August decision to enable Gemini for many K-12 students by default, which triggered backlash from educators and parents. Under the new policy, all tools start off by default, and administrators choose what to enable. But the company’s integration of search with AI means opting out of AI means opting out of search.
Former FTC Chair Lina Khan called the policy coercive. “Google is using its search monopoly to coerce schools into adopting AI,” she wrote on social media. “This highlights the danger of allowing an unregulated monopolist to control an essential service.”
Google dominates education tech, with 150 million students and teachers using Classroom globally. The bundling decision forces school administrators to choose between AI exposure and losing core functionality, right as New York City and Los Angeles have issued temporary AI bans in elementary and middle schools.
July Ruling Denied Interim Relief to Meta Workers
In July, a federal judge denied a temporary restraining order sought by Meta workers who alleged the company used AI to select employees for layoffs. Meta denied using AI for the decisions. The judge asked Meta to explain its selection of four workers whose immigration status could be affected.
Apple lets any app plug into Siri AI
Apple just released a developer preview of Siri AI in its latest software betas, giving apps of any kind a way to build their own features into the assistant — though developers can only test it for now, and customers are still months away.
- Siri AI is Apple’s rebuilt assistant, able to see a user’s screen and read messages, and has been a waitlisted beta since iOS 27 shipped in September.
- Developers can now add their own actions instead of just Apple’s initial preset list of options, and teach Siri multi-step jobs through new “skills.”
- For now, app versions stay with a developer’s internal testers, App Store submissions start in spring, and Apple says rules for developers will come later.
- The betas also bring China its first Apple Intelligence tools (minus Siri AI) and test new Apple Watch tools that listen in the background to transcribe nearby talk.
Apple letting any app build into Siri is a real crack in its walled garden, which is a big deal. Still, it leaves the work on developers, and between Siri AI’s confusing rollout and the progress other frontier labs have made in areas like computer use, the best AI features still feel a long way off for the world’s most popular phone.
Google won’t give schools Search without AI
Google announced that K-12 student accounts will start with Search, Gemini, and Gemini Notebook all off early next year, potentially forcing districts that want AI out of classrooms to also give up Google Search for their students.
- Education VP Chris Phillips said to The New York Times that “Search and A.I. are deeply integrated,” and students at AI-free schools will see a notice that it’s off.
- NYC and LA, the two largest U.S. districts, restricted AI for ~1M students this semester, and Fairfax County, VA, built a plugin to hide Google’s AI summaries.
- Google’s announcement also comes one day after the Nov. 2 deadline senators gave the company to answer questions about pushing AI into schools.
- Ex-FTC Chair Lina Khan said Google is using “its search monopoly to coerce schools” and will “punish those who don’t fall in line.”
- Google later defended the move on X as “more choice, not less,” but its post only describes Search with AI Mode and AI Overviews, never without them.
Google is essentially saying out loud what the trend has shown — Google Search is AI search now — and schools aren’t getting the old version back. The stranger situation will be AI-free districts with kids on Chromebooks, turning in work through Classroom, and writing in Docs, yet unable to use Search.
Turn one product photo into three studio shots
In this guide, you’ll build a product-shoot app in Runway using an image of a vintage Polaroid camera. You’ll create coral, sunlit, and blue-lit scenes, then save the app so you can repeat the process.
Step-by-step:
- Save the guide’s Polaroid photo and its credit. Open Runway App Studio and check that your plan and credit balance support a build
- Paste the build prompt to create an app with three shoot presets. Check the product upload, scene menu, and download controls before running it
- Upload the original photo and list the camera details to preserve. Make one coral studio shot and inspect the lens, rainbow stripe, buttons, and lettering
- Run the same photo through the sunlit and blue-lit presets. Keep the camera’s shape and viewpoint consistent, and save each approved result
- Publish the app for yourself, then open all three downloads. Keep the source credit and a record of any defects with the set
Pro tip: Use the original product photo for every scene. Starting from a flawed AI image can carry its mistakes into the next shot.
CrowdStrike’s ‘kill switch’ to stop rogue AI agents
Cybersecurity firm CrowdStrike unveiled the AI Kill Switch, adding an emergency stop to its Falcon Guardian security platform for businesses worried about their AI agents going rogue.
- The switch targets an agent’s link to its AI model, not the model itself, while Falcon Guardian separately monitors tools like Claude Code and Codex.
- A human admin has to pull the trigger, and every use gets logged, while shutting agents down directly on work devices and in the cloud is “coming soon.”
- CEO George Kurtz explained it like a race car’s brake, saying it sits outside the AI so an agent “can’t argue with it, rewrite it, or wait it out.”
- CrowdStrike is also buying monitoring firm Fiddler AI and security tester Dreadnode for a reported $495M, and added Claude Cowork protection in beta.
Politicians have spent the summer fighting over an AI kill switch, with a pair of bills and a California executive order aimed at rogue AI agents. CrowdStrike is bringing the same concept to enterprises… where the likelier scenario isn’t a frontier model escaping a lab, but an agent holding the company’s logins going off script.