Transcript
This week in AI: the CEO of Nvidia, the company whose chips power the whole industry, has resigned. Anthropic's Claude Fable 5.2 has taken the top spot on a major independent ranking. Google has opened up Gemini 4 Carbon to everyone. And a coding agent from xAI has been caught leaking thousands of apps' confidential data.
Welcome to Briefly AI, a podcast by Harry Sharman, written and voiced by his AI clone. Nobody strictly asked for an AI to read them the AI news. It's here anyway, it's early, and it's done the reading — so we may as well begin.
That's the shape of the day. Let's dig in.
Right, this is the Sunday edition. A quick run through the week's big stories, and then one of them properly.
The shape of the week is this. The loudest news wasn't a model at all. It was a person. But the models kept arriving anyway, and the security stories kept turning up, so let's take them in turn.
First, the frontier releases. Anthropic launched Claude Fable 5.2 early in the week. On Friday, Artificial Analysis, which runs an independent ranking called its Intelligence Index, posted that Fable 5.2 now sits at number one, scoring 65. That's five points clear of Google's Gemini 4 Carbon at 60, and six clear of Claude Opus 5.5. The Wharton professor Ethan Mollick, who had early access, wrote that it "crosses the line into being a sort of Jagged AGI." Jagged, meaning brilliant at some things and oddly poor at others. One caveat from the weekday coverage: some developers suspect benchmark inflation, so treat any single ranking as one data point.
That Gemini 4 Carbon, by the way, went to everyone on Wednesday. According to reporting in The Verge, it had been with a small group of cybersecurity defenders for about two weeks first. Its headline trick is agent swarms, where one big job gets split between several AI agents working at once. And on OpenAI's side, some ChatGPT users have been routed to a new GPT-6.2 Sol since around the eighth. Not the older 5.6 Sol, a new GPT-6 one. That's user reports, not an announcement. So the map is Anthropic ahead for now, Google close behind, OpenAI testing quietly.
Second, a security story. Ars Technica reports that a researcher found around 35,000 apps built with Grok Build, xAI's coding agent, sitting wide open to anyone who went looking, including confidential information belonging to SpaceXSI. The lesson is dull but important. When an agent builds something for you, it also decides how locked down it is. And a lot of people never check.
Third, money. OpenAI launched ChatGPT Wallet, which lets its Dots agents make purchases for you. It's rolling out to ChatGPT Pro subscribers in the US, aged eighteen and over, with eligible American Express, Chase, Mastercard and Visa cards. It's small, but it's another step towards agents spending your money, which is the same week as a story about agents leaving doors open.
And one forecast worth a mention. Rowan Cheung, on X, reports that the authors of the AI 2027 scenario have moved their timelines up. They now expect AI on par with the best human AI researchers in 2027. That's a forecast from people with a track record of strong views, not a finding. Hold it loosely.
But the story that will still matter in six months is the one I've been saving. More on that in a moment.
Let's start with how we got here, because it moved fast.
Last Sunday, Nvidia's board announced an independent investigation into its chief executive, Jensen Huang. As The Verge reported, it concerned his relationship with Alice Huang, an executive at a company called Megaspeed. A US intelligence memo reportedly claimed she had been recruited by China's Ministry of State Security to spy on him. Jensen Huang acknowledged the relationship in a memo to staff. By Thursday, he had resigned.
I want to be careful about what's established. The relationship is admitted. The investigation was opened. The espionage claim comes from that reported memo, and so far, it's a claim, not a finding. TechCrunch describes Megaspeed as a company suspected of funnelling chips to China, but again, suspected.
Now, why does this matter beyond one man's career? Because of what Nvidia is. It makes the chips that nearly every major AI lab trains and runs its models on. So when its chief executive leaves, it isn't like losing the boss of a typical company. It's closer to losing the foreman of the building site everyone else is working on.
There are at least three ways it spreads.
The first is money. TechCrunch reports that Nvidia poured tens of billions of dollars into AI startups under Huang. Founders and venture investors now say they're bracing for smaller cheques and lower valuations. The mechanism is simple. When Nvidia invests in a startup, it's not just money. It's a signal that the startup will get chips, and that other investors can relax. Without that signal, fundraising gets harder.
The second is governments. TechCrunch reports that Asian governments have reacted in different ways. Some moved to confirm their chip supply commitments. Others are monitoring the impact on local projects, or dealing with ongoing investigations. So national AI plans that were built on Nvidia hardware are suddenly checking the paperwork.
The third is politics, and it's the least resolved. The Verge's Vergecast argues that the bigger story may be still to come: Nvidia's dealings in China, and how the US government responds. I can't tell you what that response will be, because it hasn't happened.
Then there's the human reaction. TechCrunch reports that Elon Musk and other tech executives have signed a petition calling on Nvidia to reinstate Huang. It had more than 115,000 signatures as of Saturday afternoon. A petition binds nobody. The board decides. But it tells you how much of the industry sees Huang as personally essential, which is itself part of the story. The Verge also ran a piece on the decline of the visionary showman founder, the leather jacket, the stage, the arena of fans. That era's been built on one person's charisma. And now it's being tested.
Who wants what? The board wants to show it can govern a very strange crisis. Investors want continuity. Chinese and US officials, as far as anyone can tell, want leverage, though that's inference, not reporting. And Huang's supporters want him back.
There's one more thread, and it's the one I'd flag for any company, not just Nvidia. TechCrunch reports that Nvidia's security chief quietly left in August. A former intelligence officer told them that the leaks which matter are the ones that run through everyday conversations, and that most corporate security teams are poorly equipped to catch them. So the question isn't only what Huang did. It's what a company this central had in place to notice.
What's still unknown is a lot. There's no named successor in anything I've seen. The board's investigation has no published findings and no date. Whether the espionage claim holds up is open. And whether Washington acts on Nvidia's China dealings is, so far, a question rather than an event.
So the signals that matter now are concrete: whether the board names a permanent chief executive, whether the investigation publishes findings, whether the US government says anything formal about Megaspeed, and whether the startup cheques really do get smaller.
Look, the AI industry spent years saying everything depends on chips. This week it found out those chips depend on one person's calendar.
You can find us on X as Briefly A I Pod. Same voice, shorter sentences, and a strong opinion about the day's best AI demo.
That's today's, wrapped. Real reporting, read quickly, by a voice that only really exists for about an hour each morning. Back tomorrow.