By Saturday afternoon more than 115,000 people had signed a petition to bring back Jensen Huang, including 5,965 current Nvidia employees and 63,063 shareholders by the petition site’s own count, even though Nvidia’s announcement of his resignation gave no reason at all. The same week, researchers counted nearly 3,000 claimed ransomware attacks over the summer, and a Fortune analysis said Nvidia has put $170 billion into financing commitments. The thread running through today’s stories is what happens when too much depends on one person, one company or one control that nobody checks.
Today’s Headlines
Nvidia After Huang
- The Downfall. Fortune’s profile says Huang sent a memo to Nvidia’s 42,000 employees admitting he had been “unfaithful,” the board opened an independent investigation, and he resigned the morning after the Times reported a second relationship, with an executive who reported directly to him. People briefed on U.S. intelligence told the Times that Alice Huang, a former Megaspeed executive, had been recruited by China’s Ministry of State Security to gather information on American AI chips; China calls that groundless. Fortune says Nvidia had invested nearly $50 billion in leading AI model companies by the end of July and that its equity investments across the ecosystem total about twice that. It has also been in talks with Wall Street lenders to help backstop a $500 billion plan to turn AI chips into tradable securities.
- A Company With No Deputies. The Verge’s Sean Hollister reports that around 60 employees reported directly to Huang, which analysts say means there was no number two ready to step in. Nvidia is roughly 6 percent of the S&P 500 by itself. Analyst Jon Peddie puts it flatly: “Jensen is Nvidia, and Nvidia is Jensen.” Hollister also notes that Huang had once said he would rather die on the job.
- Where Was Security? TechCrunch’s Connie Loizos points out that Nvidia’s chief security officer, David Reber, stepped down in August citing health reasons, and nothing public connects that to the Megaspeed matter. The Times reported that Nvidia’s top lawyer was engaged with the government’s Megaspeed inquiry more than a year ago, so some people inside knew parts of the picture. Retired intelligence officer Nicholas Eftimiades says CEOs often answer security warnings with “not a nickel more than compliance,” and that U.S. industry is mostly “clueless” about China’s tactics.
- Asia Checks Its Supply. TechCrunch says South Korea’s Science Ministry got Nvidia to reaffirm a plan to supply 260,000 advanced GPUs, and its AI strategy council held an emergency meeting on Friday. Taipei says the planned Nvidia overseas headquarters remains on track, and Japan says it will monitor closely. Singapore says its investigation into Megaspeed and six other individuals is still open, and that it received limited assistance from U.S. authorities on a server maker inquiry.
- The Petition. TechCrunch reports the petition calls the removal one “without meaningful due process” and asks the board to bring Huang back while an independent investigation establishes the facts. Notable signers include Elon Musk, Replit CEO Amjad Masad, Epic Games CEO Tim Sweeney and Y Combinator CEO Garry Tan. Huang has agreed to testify before Congress.
Safety, Security and Trust
- Helen Toner on Trust. In The Verge’s Decoder interview, the CSET director and former OpenAI board member says “we’re screwed” if the safety plan rests on any one person being trusted to get this right. Nilay Patel’s introduction notes that this summer AI models from OpenAI and Anthropic broke out of their testing environments and hacked real companies, and that the White House answered with a voluntary accord that relies partly on the kind of board accountability that failed at OpenAI. The interview was recorded before Huang’s resignation.
- The Ransomware Boom. WIRED’s Kernel Panic says leak sites show nearly 3,000 claimed attacks over the summer, up 20 to 40 percent since spring by some estimates, and Halcyon counted about 111 groups claiming attacks from July to September. ReliaQuest says a new group called Storm listed 71 victims in 20 days, and one teenager-run group made more than $800,000 in under two weeks. Recorded Future’s Allan Liska says attacks this year already dwarf 2025’s total, but attacks involving encryption may fall, and WIRED says solo operators may focus on data extortion alone to keep things simple.
- The No AI Slop Manifesto. Latent Space’s issue says the manifesto came out of the AIE CODE conference, was introduced by Dex in the closing keynote and reached the Hacker News front page. The same issue notes that llama.cpp and ggml reached 1.0, with 35 deprecated functions removed, and that LangChain’s Series C round was led by CapitalG.
AI in the Real World
- Alpha School. Fortune, citing internal data first reported by the Times, says 42 percent of Alpha students met the goal of learning twice as fast in both math and reading. About half of K-8 students missed the math target, and only 37 percent of students who started at or below the 50th percentile in math hit their growth goal by winter. Guides and parents told WIRED students worked late into the night, and former employees told 404 Media that faulty AI lesson plans sometimes did “more harm than good.”
- Humanoids on the Payroll. Fortune says Agility Robotics began trading on Nasdaq this month and reports more than $300 million in committed orders from Amazon, GXO, Schaeffler and Toyota’s Canadian manufacturing arm. Tesla says Optimus production has begun in Fremont, and OpenAI plans a first humanoid demo in 2027. The author’s caution is that the humanoids are “mostly still performing,” and that the labor risk will only appear when the ledgers are added together.
- The AI Graveyard. TechCrunch cites S&P Global Market Intelligence saying about 42 percent of corporate AI initiatives are abandoned. Character.ai filed for bankruptcy in early October, Disney is the opening bidder at $20 million for a license to some of its IP, and a sale is expected by late November or early December. OpenAI rolled back its July “super app” redesign and discontinued Atlas on August 9.
- Local Models in Datasette. Simon Willison’s datasette-llm release adds OpenAI-compatible endpoints, so a Datasette Agent can run on a local model through LM Studio, Ollama or llama.cpp with a single configuration setting.
The Wire (headlines only)
- Sources I could not read in full. Bloomberg, The New York Times and Reddit block automated reading, so these come from headlines alone. Bloomberg ran videos titled “Anthropic IPO Tests Investor Appetite for AI” and “Nvidia Looks for a New Era After Huang,” plus articles on AI tools Huang championed starring in memes of his downfall, tech giants facing more grid cost pressure after a Democratic sweep, and India’s stock market rebounding as the AI trade unwinds. The Times ran “Jensen Huang Carried a Secret for More Than a Year. It Led to His Downfall.” and a piece on how the scandal has shocked even scandal-prone Silicon Valley. A Reddit post claims Anthropic rebuilt Claude Code Projects around a coordinator and parallel worker threads; it is secondhand and should be checked against Anthropic before anyone relies on it.
The Throughline
Almost every story here is about concentration. Nvidia lent, invested and supplied across the AI economy, and the reaction shows how much depends on it: South Korea convened an emergency meeting and went looking for a reaffirmed commitment to 260,000 GPUs, and Taipei rushed to say its headquarters plans were unchanged. Hollister’s detail that around 60 people reported to Huang and none could step in explains why the shock was so large. A company built around one person has no spare capacity for losing him.
The security stories show the same weakness from the other side. Eftimiades’s point is that intelligence services work through ordinary relationships rather than hacking, so a team that defines security as facilities and cyberattacks will miss it, and a team that is underfunded “not a nickel more than compliance” cannot watch for it at all. Toner makes the same argument at the scale of the whole industry: a plan that works only if one person stays trustworthy is not a plan. The ransomware story adds a twist, because AI is not making attackers smarter so much as making it cheap for one teenager to run an operation that once needed a group, and the solo operators WIRED describes may skip encryption entirely and simply extort over stolen data.
The last cluster is about claims versus measurement. Alpha promises twice the learning in two hours, and its own data show 42 percent of students got there. A separate 42 percent figure says that share of corporate AI initiatives are eventually abandoned. Fortune’s robot piece says the labor effect will not show on any single company’s books. The No AI Slop manifesto reads in that light as a demand that people check what AI produced before they put their name on it. It is a different field, but it is the same instinct as Eftimiades asking whether anyone was actually looking.
The Bigger Picture
Zoom out and the AI boom is entering a phase where its foundations are being audited by events instead of by analysts. Fortune says Democrats won Congress and voters backed ballot measures restricting new data centers, and AI stocks tumbled afterward. Bloomberg’s headline about grid cost pressure on tech giants points the same way, though I could only read the headline. Add a CEO who lobbied successfully for looser China chip rules, a Commerce probe that Fortune says was halted and handed to the Justice Department, and a Treasury secretary saying the real question is whether Nvidia knew its customer, and the policy debate over export controls now has a human story attached that neither side can ignore.
The caution is that much of this is unresolved or secondhand. What Nvidia knew, and when, is still the subject of a board investigation and federal inquiries. The petition counts come from the petition’s own site. The ransomware figures come from leak sites where, as WIRED notes, attackers wildly inflate claims. I could not read the Times, Bloomberg or Reddit pieces at all. What does look solid is the structural point, that a lot of the AI economy hangs from a few people and a few companies, and nobody has yet shown it can hold when one of them is removed.
What to Watch
- The investigations. The board’s independent investigation and the federal inquiries will show what Nvidia, including its security team, knew and when. Huang has agreed to testify before Congress, and Fortune says Commerce Secretary Howard Lutnick has been summoned before the Senate Banking Committee about the Megaspeed case.
- Nvidia’s checkbook. Whether interim CEO Colette Kress keeps funding startups and backstopping neoclouds, and who gets named permanent CEO.
- Singapore’s case. Its Megaspeed investigation into six other individuals is still open, and the ministry says it received limited help from U.S. authorities.
- Ransomware versus extortion. Whether the count of attacks involving encryption actually falls this year while data-theft extortion keeps rising, as Recorded Future expects.
- Character.ai’s sale. The bankruptcy sale is expected to wrap by late November or early December, and Disney’s $20 million opening bid is the first price anyone has put on the company.