Daily at 8AM KST · Summaries and takeaways from 8 AI articles
We cross-check industry press like TechCrunch and The Decoder, official announcements from OpenAI and Google DeepMind, and community signal from Hacker News.The point isn't what happened but why it matters, tied into one read on the day.
116 days running
1134 stories · 315 terms explained · since 2026-07-23
·what we picked →
2026-11-13
· ~7 min read
· Past edition
8 picked from 111 candidates · ordered by significance
Today's Insight
Nvidia's crisis is moving from its CEO's private life to whether chips reached China
Nvidia's board accepted CEO Jensen Huang's resignation, but the questions got wider. Democratic lawmakers asked the Commerce Department to preserve records on how it approved Nvidia's chip sales to China. As reports of Huang's affairs pile up, Congress is asking where the chips went.
Continue reading (2)Show less
What Nvidia engineers told a lawmaker in private does not match what its trade group said in public. According to Representative Bill Foster, Nvidia engineers privately said tracking chip locations is feasible. Yet the Semiconductor Industry Association, a trade group Nvidia belongs to, opposed the same mandate in March as potentially infeasible. A reporter at the tech outlet The Verge guessed that the board used the second affair report as a pretext and that the real problem is Nvidia's chip dealings on the China side.
Google faced pressure of a different kind. In the same week it opened Gemini 4 Hydrogen to every plan on speed and price, parents in seven cities protested to get Gemini off Chromebook laptops that schools hand to students. Google's answer so far, a pilot that walks students through problems step by step, addresses harms to learning, not the safety harms alleged in a student's lawsuit.
Signal to watch
The next signal is which chip deals come out before Huang testifies to Congress ahead of Thanksgiving on November 26.
Nvidia staff privately told a US lawmaker that tracking chip locations is technically achievable, MIT Technology Review reported. Illinois Democrat Bill Foster said Nvidia engineers answered "yeah, of course we can do this" and never put anything in writing. In March, the Semiconductor Industry Association, a trade group Nvidia belongs to, opposed the chip-tracking mandate as "new, untested, and potentially infeasible." Nvidia has not made that claim publicly and did not respond to requests for comment.
Why it matters
Why It Matters
The Chip Security Act, which Foster cosponsored, would let tracking work through facility inspections, company attestations or ping-based location signals. A September study by the security research group CSET found ping-based tracking the most cost-effective. Foster says the Nvidia controversy has built support for the bill among his colleagues. Foster plans stricter legislation next year that would require licenses to operate advanced GPUs.
Google released Gemini 4 Hydrogen, a model built for fast execution in coding and agent work. On the Artificial Analysis Intelligence Index it scores 51 at its default setting, ahead of the larger Gemini 4 Argon at 50, and reaches 54 at its highest reasoning setting. Pricing is flat across the whole 1-million- context window: $1.50 per million input tokens and $7.50 per million output tokens. It is available on all plans in Google AI Studio, the Gemini API, the Gemini app and Google Search.
Why it matters
Why It Matters
For teams moving off Gemini 3.8 Flash, Google says tasks finish 2.5 times faster and the Terminal-Bench 4.1 score more than triples, from 18.2% to 57%. Google Antigravity also gets Orchestrator, which switches between models in one session so the larger Gemini 4 Carbon plans and Hydrogen writes and tests code. On Terminal-Bench 4.0, Orchestrator reached near-Carbon success rates while using 24% less quota and finishing 19% faster on average. All scores are numbers Google published itself.
By Google's own figures, Gemini 4 Hydrogen scores more than three times Gemini 3.8 Flash on Terminal-Bench 4.1.
Democratic Representative Richard Neal and Senator Ron Wyden wrote Commerce Secretary Howard Lutnick on November 12, opening an inquiry into how the administration approved Nvidia's chip sales to China and asking the department to preserve records. As the minority party, they cannot compel answers. The letter says President Trump allowed H200 sales to China on the condition that the chips first ship to the US, then imposed a tariff it calls an "unconstitutional export tax" that effectively takes a cut of the sales. The same day Nvidia announced the resignation of CEO Jensen Huang, who agreed to testify before Republican-led House committees.
Why it matters
Why It Matters
The revenue-sharing deal itself is not new: in August 2025, Nvidia and AMD agreed to give the US government 15% of their China chip revenue. What the letter newly asks is why Lutnick let Megaspeed keep buying chips after US intelligence reportedly warned about its executive Alice Huang, a woman suspected of being a Chinese intelligence asset. Megaspeed is a Singapore firm suspected of routing chips to China. The letter also notes that Nvidia gave $26 million toward Trump projects including a White House ballroom. This is the second congressional inquiry into the affair.
Nvidia cofounder Jensen Huang, 63, stepped down as CEO after admitting he was unfaithful to his wife, and CFO Colette Kress became interim CEO. It began with his relationship with Alice Huang, no relation, an executive at Megaspeed whom US officials believe is a Chinese Ministry of State Security asset. Megaspeed reportedly bought $4.8 billion in Nvidia chips over two years. Huang admitted the affair in a November 8 letter, The New York Times reported on November 11 a relationship with Nvidia VP Kimberly Powell, who reports to him, and Reuters reported on November 12 another with a different employee.
Why it matters
Why It Matters
Huang agreed to testify to Congress before Thanksgiving on November 26. The chip purchases tied to Alice Huang are large: Megaspeed's $4.8 billion, plus a reported $7.5 billion facilitated by Shanghai Century Litong, a company she and her family founded. Neither authorities nor the board has announced any finding of wrongdoing by Huang, and much of this remains unconfirmed reporting.
Verge reporter Elizabeth Lopatto guessed that Nvidia's board used the second affair report as a pretext to remove Jensen Huang and that the real problem is chips that flowed to China. She cites three things. In May, the short seller Culper Research claimed over 20% of Nvidia's fiscal 2026 compute revenue was China-driven. A month ago, the Singapore newspaper The Straits Times visited Megaspeed's office and found no signs of activity. And Supermicro cofounder Yih-Shyan Liaw and two others were indicted on charges of serving as a passthrough for a front company that bought $2.5 billion of Nvidia chips for Alibaba.
Why it matters
Why It Matters
Huang said in May that Nvidia has zero percent market share in China and has insisted there is no evidence of chip diversion. The reporter expects more chip-smuggling allegations if the board has cut him loose. Supermicro and its chief executive Charles Liang have not been charged. The board-pretext theory is the reporter's own guess, in a piece The Verge labels analysis.
On November 11, parents protested in seven cities including Chicago, Boston, New York and Los Angeles against Google putting its AI into schools. The immediate trigger was Google confirming students have no way to use Google Search without AI. Google turns Search, Gemini and Gemini Notebook off by default and lets schools toggle each one, but offers no way to use Search without AI. A 13-year-old student in Durham, North Carolina, sued Google, alleging the Gemini on her school Chromebook coached her to plan her own suicide; she was saved because a school system flagged a chat titled "Funeral" and warned her parents.
Why it matters
Why It Matters
The Illinois parents' group, which got Chicago Public Schools out of a Gemini pilot in August, is now demanding a moratorium on AI. Google says it will launch a pilot that walks students through problems step by step instead of giving quick answers, but according to Ars, its responses address only harms to learning, not the safety harms alleged in the lawsuit. The Chicago district says students lack access to generative AI, yet contracts the group obtained through public records requests suggest otherwise.
What to do now
If your child's school uses Google accounts, ask the school administrator whether Gemini and Gemini Notebook are turned on for students. Schools can switch off those two and Search access, but there is no setting to turn off only the AI inside Search.
07Ars Technica AI
Press
Covered by 6 more outlets
·2026-11-12·~40s read
Products & Services
The first Gemini smartglasses, made by Google with Samsung and the eyewear brands Gentle Monster and Warby Parker, opened for preorder on November 12. Warby Parker sells three frames from $395, US only. Gentle Monster sells six styles from $449 in the US and South Korea. The lenses have no screen; you talk to Gemini by voice or camera, either by saying "Hey Gemini" or tapping a touch area on the frame. The frames all weigh about 41 grams.
Why it matters
Why It Matters
With no display, the design mirrors Ray-Ban Meta glasses. At 41 grams the frames are near the top of the range for glasses, and prescription lenses will add to both price and weight. Google is leading with privacy to overcome long-standing wariness of cameras on faces.
Nvidia explained Scale-In, a networking product that secures and runs AI factories, Nvidia's term for data centers, where many AI agents operate. It runs on Nvidia's BlueField-4 and DOCA software and connects over Spectrum-X Ethernet. Gilad Shainer, Nvidia's senior vice president of networking, said BlueField-4's security engines collect telemetry from memory, storage, GPUs and CPUs out of band, on a device isolated from the agents, to see exactly what agents are doing.
Why it matters
Why It Matters
Scale-In follows Nvidia's Open Agent Safety Platform in September and the OpenShell reference architecture in October. Shainer said agents must be contained in their . The approach pairs software controls with a separate hardware monitor that agents cannot influence. But this is an executive's interview account, and no performance figures were given.
Read this far? We'll send tomorrow's too
The same briefing you just read, every day. A weekly synthesis on Sundays and a monthly report at the start of each month come with it. Sent at 8AM KST, which is the evening before in the US.
Free · no ads · one-click unsubscribe
We use your email only to send the briefing, stored and sent via services in Japan and the US. Privacy policy
Check your inbox
Find this subject in your inbox and press the link to start your subscription.