Podcast 5-5 by Dongee · November 9, 2026

Transcript

Gabriel: Monday, November ninth. Welcome! Lina, today I'm starting with a sensitive one: Jensen Huang, the head of Nvidia.

Lina: What happened?

Gabriel: On Sunday he sent an email to his employees. He wrote: "I was unfaithful to my wife. My actions hurt her and our children. There is no excuse for what I did."

Lina: Oof.

Gabriel: And at almost the same time, Nvidia's board announced an independent investigation, with outside lawyers.

Lina: Investigating what? The affair?

Gabriel: Reports that he was the target of a foreign intelligence effort. According to The New York Times and the Financial Times, the woman is Alice Huang… who's no relation to him.

Lina: And who is she?

Gabriel: An executive at Megaspeed, a Singapore cloud company. It's an Nvidia customer, and it's accused of helping blacklisted Chinese companies get AI chips.

Lina: And US agencies believe that…

Gabriel: …that she was recruited by China's Ministry of State Security. Careful: that's what's being reported, it's not proven. And according to The New York Times, Marco Rubio oversaw a classified memo this spring that warned about the relationship.

Lina: And what does the board say?

Gabriel: That it has no reason to believe national security or Nvidia's secrets were put at risk. He stays in charge and will cooperate.

Lina: So: the board is investigating its own boss… while he's still signing the emails.

Gabriel: Pretty much. Trump said he's "a good person" and that he hasn't heard he did anything wrong. But three Republican chairs of House committees in Congress asked him for a closed-door meeting on November eighteenth.

Lina: Right after earnings?

Gabriel: Yep, Nvidia reports earnings on the seventeenth. For some, it's a serious risk: the head of the key AI company, close to an alleged spy, while he was pushing to sell more chips to China.

Lina: For others, it's mostly a personal matter he's owned up to, with no sign of leaks and with Trump's backing. What weighs more for you: what we already know… or what's still being investigated?

Lina: Gabriel, how many times have you asked Siri for something… and it answered with a web search?

Gabriel: I'd rather not say.

Lina: Well, according to Mark Gurman at Bloomberg, that could change. John Ternus, Apple's new CEO, had a closed-door meeting with Craig Federighi.

Gabriel: The software chief?

Lina: That's the one, and he's also in charge of AI. Ternus asked him to speed up the plan and not stop at just improving Siri.

Gabriel: And what would going further look like?

Lina: Gurman thinks Siri could work as an agent starting next year: something that goes after a goal and keeps at the task without you asking for every step. Like Dots, OpenAI's agent, or Muse, Meta's.

Gabriel: And in the meantime?

Lina: It goes step by step. First, handling Apple's own apps well: cropping a photo, organizing tabs in Safari. In the spring, other companies' apps, like ordering on Grubhub, a food app.

Gabriel: Or booking a table on OpenTable.

Lina: Right, a reservations app. And for the next big version of iOS, Apple's operating system, they're changing the rhythm: six weeks to code and explore, and only two to integrate. Before, it was half and half.

Gabriel: And the Siri team?

Lina: Here's the catch: Mike Rockwell, Siri's vice president of engineering, is still seriously thinking about leaving next year.

Gabriel: Seriously?

Lina: The optimistic read says Apple is finally reacting: it has the phone, the computer and your data on the device, and a Siri that acts on its own, with privacy, would be its big advantage.

Gabriel: And the skeptical one: another "next year" promise, from a company that's already fallen behind, with the head of Siri eyeing the door.

Lina: What about you: would you let Siri order dinner for you?

Gabriel: Hey, Lina, did you see Trump on 60 Minutes, the TV show?

Lina: No. Did he talk about AI?

Gabriel: Quite a bit. It was after the midterm elections. Ross Douthat asked him whether the US should accept AI's risks to stay ahead of China.

Lina: Let me guess, he said AI is super dangerous.

Gabriel: No, the opposite: that the risk is "not nearly as real as they say" and that "it can be managed."

Lina: And would he work with the Democrats? They just won the House and the Senate.

Gabriel: They asked him. He said "absolutely." And then: "I don't mind limits. But I don't want to put us in a distant second place." And he added: "China has no limits."

Lina: Wait. China has no limits?

Gabriel: That's exactly what Douthat pushed back with: that China does have limits, the ones of a totalitarian state.

Lina: And him?

Gabriel: That over there, "whatever President Xi says, gets done." Oh, and he called himself a total optimist about AI… well, he doesn't even call it that.

Lina: What does he call it?

Gabriel: "S.I.," for superintelligence. He says there's nothing artificial about it.

Lina: Renamed on national TV.

Gabriel: And he said Xi Jinping, who visited him two weeks ago, "is all in." For him, whoever wins that race, wins. Did he announce any law? No. No law, no order.

Lina: And Reuters points out that he used to prefer that companies regulate themselves.

Gabriel: That's why you can read it as a real opening: a president who didn't want rules before now accepts limits and is even willing to work with the Democrats.

Lina: Or as an opening with fine print: any rule comes second to beating China, and he himself says the risk is smaller than the industry warns. What matters more to you: hitting the brakes in time… or getting there first?

Lina: Gabriel, do you have email on your own domain? Like "sales at your business"?

Gabriel: Sure. Why?

Lina: Because since November fifth, Outlook and Hotmail have been rejecting some of those emails… without warning. MailRoute, an email service, spotted it, and on Sunday it became a topic on Hacker News.

Gabriel: What do you mean, without warning?

Lina: Microsoft hasn't published anything: no notice, no announcement. The email bounces if your domain doesn't have a record called DMARC.

Gabriel: Dee-what?

Lina: DMARC. Picture it as a sticky note on your domain that says: "yes, these emails are really mine." But it only counts if…

Gabriel: …if the other two records check out too.

Lina: Just one of the two has to check out, okay? And lots of people have those two… but not the sticky note.

Gabriel: But wasn't that rule only for people who send mass email?

Lina: It was. The published rule applied to anyone sending more than five thousand emails a day. Now it's hitting the small senders too.

Gabriel: What a mess.

Lina: And not everything bounces: one gets through, the next one doesn't. An invoice or a confirmation just never arrives… and nobody finds out.

Gabriel: So how do you fix it?

Lina: According to the reports, with a single line on your domain: the sticky note in "monitor only" mode. In one test without it, sixteen out of sixteen inboxes rejected the email; with it, it got through.

Gabriel: Some see it as a healthy step against impersonation and email scams: that sticky note should be the bare minimum.

Lina: And others see Microsoft changing the rules without warning, halfway, with a confusing error, punishing businesses that don't even know what this is. Have you checked whether your domain has its sticky note? I'm checking mine as soon as we're done.

Gabriel: Now over to Anthropic, the company behind Claude. According to the Financial Times, it bought gas generators… a lot of them.

Lina: How many is a lot?

Gabriel: One point seven gigawatts, for about three point four billion dollars. It bought them from OnePWR, an equipment supplier from San Antonio, Texas.

Lina: It bought the company?

Gabriel: No, just the commitment for the generators. Or rather: that, plus the engineering and construction capacity to install them. Not the company, not its employees.

Lina: And why are we only hearing about it now?

Gabriel: Because the deal is from late September, but it came out on Sunday. The Financial Times says it's the largest known purchase of its kind by an AI lab.

Lina: Hold on, is that confirmed?

Gabriel: Careful there: Reuters couldn't verify it, and neither Anthropic nor OnePWR responded. So it's a report.

Lina: Okay. And why so much power?

Gabriel: Because of demand for Claude. They need more computing, and that eats up a lot of electricity. Reuters reported in September that Anthropic expects to spend more than five hundred billion dollars on infrastructure over the next decade.

Lina: So basically, they're not waiting on the power grid anymore.

Gabriel: More or less. And it comes the same week that opposition to data centers made a difference in the US elections.

Lina: What timing.

Gabriel: One view says it's pragmatic: the grid can't keep up, and anyone who wants to keep growing has to make their own power.

Lina: The other sees a bad sign for the climate and for the neighbors: more gas burned to feed AI, right when voters are complaining about data centers.

Gabriel: Where do you draw the line: your own power to grow… or listening to the neighbors first?

Gabriel: That's it for today's episode.

Lina: If one of the questions stuck with you, share it and talk it over with whoever you want.

Gabriel: Tomorrow, five more stories. And if they don't show up in your inbox, check your DMARC first.

Lina: And I'm going to ask Siri to check it for me… next year. Talk to you tomorrow.