Google launches Gemini 4 Carbon & Gemini off by default for students - Tech News (Nov 11, 2026)

A team of AI agents looked at free NASA laser scans of a Mexican jungle and found what may be a lost Maya city that no official registry knows about. We'll get to that story in a moment. Welcome to The Automated Daily, tech news edition. The podcast created by generative AI. I'm TrendTeller, and today is November 11th, 2026. Today: Google's new flagship model, Anthropic's trillion-dollar IPO plans, robots that rehearse before they move, a surprisingly expensive Starlink, and a shake-up at one of Silicon Valley's best-known investors. Let's get into it.

Google launches Gemini 4 Carbon

We'll start with Google, which has released Gemini 4 Carbon, calling it its strongest model so far. Paying developers and enterprises can use it now through the Gemini API and AI Studio, and it's also coming to the Gemini app, AI Ultra subscribers and AI Mode in Search. Google is pitching Carbon for long, complex jobs like software engineering, cybersecurity and scientific research, and it previewed a swarm feature that coordinates several agents on one large project. The message is simple: Google wants to be the default platform for businesses building with AI agents.

Gemini off by default for students

Google is also adding a safeguard for schools. Gemini, NotebookLM, AI features in Classroom and Search will now be switched off for K–12 student accounts unless an administrator turns them on. That gives districts a clear choice about whether, and how, generative AI shows up in their classrooms.

DeepMind AI audits Signal encryption

Over at Google DeepMind, researchers spent four weeks using agentic AI tools to test key security guarantees in the Signal encryption protocol. They found no new vulnerabilities in the parts they examined. That's good news for Signal users, and it also suggests a new way to audit the security systems that millions of people depend on.

Anthropic IPO and Nvidia slide

On to the money side of AI, and an update on Anthropic's IPO. The company's filing now lays out plans to sell 825 million shares at between 70 and 78 dollars each. That puts its valuation at roughly 1.33 trillion dollars at the midpoint. The timing is awkward, though. Nvidia shares fell again this week as tech stocks kept sliding on worries about the AI market. Anthropic's debut is shaping up as a real test of whether investors still believe in these huge valuations.

Forecasters shorten superhuman AI timelines

Some forecasters, meanwhile, think AI is moving faster than expected. The authors of a well-known AI timelines project have pulled their estimates forward. They point to new data from the research group METR suggesting AI is now speeding up its own research by about 1.7 times. Their new estimate is that a superhuman AI researcher could arrive in 2027, and they argue frontier development should be paused or slowed sharply. You don't have to accept their conclusion, but it's one of the clearest public arguments for shorter timelines we've seen.

Apple betas and Claris Harbor

In Apple news, the first developer betas of iOS 18.3, macOS Sequoia 15.3 and related updates are out. They bring an early version of Apple's agentic features, including a Siri that can work with third-party apps. Separately, Claris, the Apple subsidiary behind FileMaker, is quietly testing an app called Harbor. It turns an Apple silicon Mac into a private AI server and can link several Macs together to run larger open models, so the data never leaves your own machines. It's in early access, and the sign-up form asks for a company name, which suggests businesses are the first target.

Robots that think before acting

Now to robots. NTT has shown physical AI that tests several possible actions in simulation before a robot moves. It predicts problems like a stack collapsing or an item slipping, then picks the option most likely to work. Airport baggage handling is one of the target uses. Amazon's Ring, following up on its robotics plans, has launched an indoor flying security camera for homes and a patrol rover for businesses. Both promise wider coverage, and both raise obvious privacy questions. In China, XPeng says its two-seat flying car could start limited public use next summer, mostly in approved parks and for emergency services. It has no wheels at all and is really closer to a large passenger drone.

Ring drones, XPeng flying car

Staying connected costs more than expected for some. SpaceX's Starlink Mini 2 adds a swappable battery, USB-C power and Wi-Fi 6, but invited US buyers were quoted 999 dollars, including a year of roaming service. Leaked code had pointed to a much lower price, so a cheaper version may still be coming. Google Fi, meanwhile, is rolling out a super network feature on newer Pixel phones that switches between two carriers on its own when your signal gets weak.

Starlink Mini 2 price surprise

Reddit is turning its famously opinionated threads into AI-generated product reviews, now available to all US users just in time for holiday shopping. On a related note, Mintlify has open-sourced a tool that tells AI agents, crawlers, bots and real people apart. The company built it after noticing that much of the traffic to its documentation sites was coming from agents.

Google Fi super network switching

A security warning for crypto users: Binance says an iOS flaw nicknamed worse_query could let a malicious app read other apps' files, which might include wallet credentials. Keep your software updated, and be careful about what you install.

Reddit Reviews and agent traffic

Traditional finance keeps moving onto blockchains. Participating Visa issuers can now settle payments with stablecoins on the Monad network. Moody's and Chainlink finished a trial that shares credit ratings across multiple blockchains. Standard Chartered's partnership with 24X now covers Bitcoin and Ethereum. And Kimber, an affiliate of Plume, has received SEC broker-dealer registration so it can offer blockchain-based access to stocks, ETFs and bonds.

iOS flaw threatens crypto wallets

In venture capital, Sequoia is reshaping itself. It's closing its London office, ending its Arc accelerator and Scouts programs, flattening its hierarchy and leaving its evergreen fund structure. According to Axios, the leadership change away from Roelof Botha was less friendly than it looked. Part of the friction came from Botha joining SpaceX's board, which reportedly surprised his partners. European founders are watching to see whether the firm loses its early-stage edge.

Wall Street and blockchain converge

Now, the story from the top of the show. One writer set AI agents loose on public NASA lidar data, which uses lasers to see through forest cover. The agents flagged a hilltop complex in Mexico with a plaza and a pyramid-like mound. Mexico's National Institute of Anthropology and History replied that no registered site exists at that location. The find still needs official confirmation, but it looks like a Classic-period Maya civic center. It's a good example of what AI agents can do in science beyond writing code.

Sequoia Capital's major overhaul

And finally, a career note for engineers. One essay argues that glue work, meaning coordination, status reports and chasing down approvals, is still a trap in the AI era, and that AI is shrinking much of it anyway. The one exception: don't hand real conversations with your colleagues to a chatbot.

That's all for today's edition of The Automated Daily. Thanks for spending a few minutes with me. I'm TrendTeller, and I'll see you tomorrow with more of what's moving in tech.

Google launches Gemini 4 Carbon & Gemini off by default for students - Tech News (Nov 11, 2026)

Our Sponsors

Today's Tech News Topics