Friday, 13 November 2026
Jensen Huang Resigns From Nvidia; CFO Colette Kress Is Interim CEO
Jensen Huang Resigns From Nvidia; CFO Colette Kress Is Interim CEO
On Sunday Nvidia’s board opened an inquiry into whether Jensen Huang had been targeted by a foreign intelligence effort. On November 12 it ended with Huang gone. Nvidia announced that he has resigned as president, CEO and board member, effective immediately; the release gives no reason. CFO Colette Kress becomes interim CEO and keeps her CFO job while the board searches inside and outside the company, and former Synopsys chief Harvey C. Jones becomes Nvidia’s first-ever board chair. The trigger came a day earlier: The New York Times reported a second extramarital relationship, with a healthcare executive who reported to Huang — on top of the relationship with Alice Huang (no relation), whom U.S. intelligence officials suspect of working for China. In a memo to staff Huang wrote, “I was unfaithful to my wife,” and said he backs the investigation. Shares fell as much as 16% intraday and closed down 4.3%, about $500 billion lost over the week. President Trump suggested Apple’s Tim Cook for the job: “A Legend replaces a Legend!” Results are still due on November 17, read by a CFO who is now also the CEO. Thirty-three years of founder, and the press release runs to four paragraphs.
Google’s Gemini 4 Hydrogen: 2.5 Times Faster, at $1.50 per Million Input Tokens
Two days after making Gemini 4 Carbon generally available, Google released the model it expects people to actually use all day. Gemini 4 Hydrogen, announced November 12, is the fast tier below Carbon and Argon: one million tokens of context in and up to one million out, four thinking levels (a new “xhigh” on top), and, at its default setting, tasks finished 2.5 times faster than Gemini 3.8 Flash. Google’s own numbers: 57% on Terminal-Bench 4.1 against Flash’s 18.2%, fabricated claims down more than fivefold, a score of 51 on the Artificial Analysis Intelligence Index — one point above Argon, the slow, expensive one. Price: $1.50 per million input tokens and $7.50 per million output, flat across the whole window, against Carbon’s $2 and $10. It is live in the Gemini API, AI Studio, the Gemini app and Search. When the cheap model beats the big one on the maker’s own chart, the big one has a positioning problem.
Source: Google, November 12, 2026
Wharton Releases Primer, Open-Source Agents That Turn a Lesson Into a Simulation
Ethan Mollick’s Generative AI Labs at Wharton released Primer, a system that lets a teacher build a classroom simulation without writing code. An intake agent interviews the teacher one question at a time — the lesson, the students’ level, the usual misconceptions, how to grade. A design agent writes a plan the teacher must approve before anything gets built; then more agents cast characters, draw images and assemble the thing. Students meet a mentor, a practice partner for role-play or “teach it back”, and a feedback coach; the teacher gets reports. Prompts, templates and code are open, funded by the Bezos, Walton and Schusterman family foundations. The lab is blunt about the limits: an “experimental prototype, not a vetted classroom product”, never tested as a learning intervention in a real classroom. The paper’s last appendix is a prompt telling an AI agent to build its own Primer. Even the instructions come with homework.