The week in technology, sorted · Signal · Money · Noise · The Long Wave
Sunday, November 8, 2026 · Issue No. 6
Signal — What will still matter in five years
Voters say no to data centers
On Tuesday, voters favored limiting data-center development in 40 of 41 local ballot measures, Ballotpedia found. The votes spanned roughly ten states. Five places approved outright bans. In Ohio, almost 20 towns, cities and counties voted, and opponents won every race. Others passed partial bans, six-month moratoriums or rules requiring a public vote on large projects. Emporia, Kansas, banned “high impact data centers,” defined as those drawing more than 10 megawatts of power or 100,000 gallons of water a day. The lone exception was Harwood, North Dakota, where voters rejected a review requirement. Voters cited water, power bills and secrecy. As Brookings’ Mark Muro told Route Fifty, many projects arrive through deals shrouded in nondisclosure agreements. Two days later, OpenAI announced that it will pay the full cost of the electricity and infrastructure its data centers use, write a water plan for each site, largely stop asking public officials to sign nondisclosure agreements, and sign a compact with each host community.
Why it’s signal: the AI boom has assumed that the hard part was chips and money. This week suggested the hard part may be permission. A data center needs a town to say yes, and towns are discovering that they can say no.
An AI hacker, kept on a leash
On Thursday, OpenAI released GPT-6 Cyber, a model trained to find software flaws, write working attacks and suggest fixes, and rated it “Critical,” the highest level of cybersecurity risk in its safety framework. Only a small group of vetted security customers can use it, from approved network addresses, and OpenAI recommends running it in isolated environments under human supervision. In OpenAI’s tests, it found 81 previously unknown serious flaws in the Firefox browser, which Mozilla has already fixed. It also built working attacks for 948 vulnerabilities in Linux, the software that runs most of the world’s servers, more than five times what an earlier model managed, at a computing cost of about $1,279 per attack. OpenAI says it sent 304 critical findings to Linux maintainers along with suggested fixes. All of these figures come from OpenAI, The Hacker News notes, and no independent testing has been published.
Why it’s signal: last month, a bank hired a thousand people to patch flaws that AI had found. Now a working attack costs about as much as a new laptop. When finding weaknesses gets this cheap, the main safeguard left is deciding who may use the tool, and for now that decision rests with the company that built it.
Say no to AI, lose Google Search
On Tuesday, Google said that schools using its K-12 software will lose student access to Google Search unless they opt in to its AI tools. Starting early next year, Gemini and Search will be switched off by default on student accounts until administrators turn them on. Because Search now includes AI Overviews and AI Mode, which can’t be turned off, a school that wants to keep AI away from students can’t give them Search either, Fortune reports. “Search and AI are deeply integrated,” Google’s education chief, Chris Phillips, told The New York Times, and Google says the change gives “schools more choice, not less.” Former Federal Trade Commission chair Lina Khan accused Google of “using its search monopoly to coerce schools into adopting AI.” Jared Cooney Horvath, a neuroscientist and former teacher, offered schools another route: “Maybe go to a book. Maybe go to the library.” New York City and Los Angeles have already paused generative AI tools in their elementary and middle schools.
Why it’s signal: for most of computing’s history, “off” was always an option. As AI is built into the tools we can’t do without, saying no to the AI means saying no to the tool. Schools are the first to face that choice. They won’t be the last.
Follow the Money — Where investors are betting
The trillion-dollar flinch
On Thursday, the Nasdaq fell 5.5 percent and the S&P 500 fell 4.4 percent, its worst day since the tariff turmoil of April 2025, led by AI chipmakers. AMD dropped 17 percent and Nvidia 9 percent. The Associated Press pointed to doubts about whether AI spending will pay off, high valuations and growing opposition to data centers. This week, Bloomberg also reported that potential buyers of Anthropic’s shares were discussing a valuation of about $1.2 trillion to $1.3 trillion, with some as low as $1.1 trillion, down from the $1.5 trillion to $2 trillion discussed a month ago. On Friday, Bloomberg reported that OpenAI had raised $11 billion at a $1.25 trillion valuation, in a round that could reach $20 billion, well short of the $30 billion it was reported to be seeking in September.
Why it matters: a trillion dollars is still a trillion dollars. But for the first time in this boom, the people writing the checks are asking for a discount, and one of the reasons they give, local opposition to data centers, is the one voters gave on Tuesday.
Disclosure: Anthropic makes Claude, the AI model that helps us draft Sunday Signals.
Noise — Loud this week, not new
The robot maid has a pilot
MIT Technology Review’s Grace Huckins hired a humanoid robot from the San Francisco startup Tau Robotics to clean her apartment. It was “undeniably cute” and cheap, at $30 an hour when she booked. It was also driven by a person. Remote pilots watch through cameras in the robot’s head and wrists and guide its movements, and the data they collect trains future models. It stumbled on stairs, overheated and stalled during its first visit, left taps dripping and knocked over a bottle of olive oil, though it did handle the toilet. Tau has completed only a couple hundred cleanings. We file it under Noise because home-robot demonstrations keep arriving faster than home robots. To its credit, this one is honest about what it is: a remote-controlled machine gathering the data that might one day let it work alone. For now, the most capable part of the robot maid is human.
The Long Wave
Every new technology arrives carrying a quiet question: may we? For most of the internet’s history, the answer was assumed. This week, people began answering out loud. Towns in about ten states told data-center builders no, and one of the largest AI companies answered by promising to pay its own power bills and stop asking officials for secrecy. A lab judged its newest model too risky for wide release and chose for itself who may use it. Google told schools that saying no to AI now means saying no to search. And investors, asked to value two AI companies at well over a trillion dollars each, said, in effect, not at that price. The power to refuse is the oldest check on any technology. The question now is who still holds it.
The Question for the Week
Is there an AI feature you would turn off if you could, and can you?
Sourced this week from Ballotpedia, Route Fifty, ABC News, the Associated Press, The Decoder, The Hacker News, Fortune, TechCrunch, Bloomberg, Investopedia, and MIT Technology Review.
— The Editors, Intellicurious —
Share this issue
