What happened: On November 12, Nvidia announced that Huang is stepping down as CEO. He co-founded the company 33 years ago and built it into a $4 trillion giant at the center of the AI boom. Huang himself has said almost nothing in public all week.

Why it happened: Huang admitted to an affair with an executive at Megaspeed, a Singapore data center firm and a major Nvidia customer. U.S. intelligence reportedly found she was a Chinese spy. The New York Times and Reuters then reported other relationships, including with Nvidia employees. China has called the spy claim a smear. Earlier in the week, the board said it had no reason to think national security or company secrets were compromised.

Who’s in charge now: CFO Colette Kress, who has run Nvidia’s finances since 2013, is interim CEO. The board will look at both internal and external candidates for the permanent job.

Market reaction: The stock fell as much as 13–14% at Thursday’s open before recovering some ground. Nvidia has lost about $490 billion in market value since last Friday. Several Wall Street analysts downgraded the stock over leadership worries.

Congress steps in: House Foreign Affairs Chair Brian Mast says Huang has agreed to testify before Congress, possibly as soon as next week.

What’s next: Nvidia still plans to report quarterly earnings on Tuesday, November 17, and Kress is expected to speak on the call. It will be the first big test of how investors feel about Nvidia without Huang.

What it is: On November 10, Google Cloud launched Gemini for Healthcare. The AI tool plugs into hospitals’ and insurers’ records, billing and insurance systems and can work across all of them.

What it can do: It combines a patient’s records from Epic and other systems into one summary and flags gaps or conflicts between them. It can also prepare prior authorizations, check insurance coverage, estimate what a patient will owe and search medical guidelines from groups like ASCO and the BMJ.

How it’s built: It launches with five preset “skills”, such as summarizing records and checking insurance benefits. It also has connectors to Epic, Medicare coverage data and ASCO/BMJ guidelines. Every finding links back to its source so clinicians can check it.

In action: In a demo, it compared a patient’s Epic chart with an outside specialist’s report in 35 seconds. It caught new lab results and an allergy the chart had missed. Gemini can read Epic records but not change them, so doctors still make every care decision.

Launch partners: Northwell Health, Highmark Health, HCA Healthcare, Optum, Blue Cross Blue Shield of Massachusetts, Hartford HealthCare and Stanford School of Medicine.

Why it matters: Admin work makes up an estimated 25–35% of U.S. health spending. Each doctor handles about 40 prior authorizations a week, roughly 13 hours of staff time. Northwell alone has about 400 people whose job is fighting insurance denials.

The catch: Google hasn’t shared pricing or availability terms. There’s no independent data yet on accuracy or real savings, so the promised benefits are still claims.

What it is: On November 9, Meta Superintelligence Labs released Muse Ember, the fourth model in its Muse family. It follows Spark (April), Glimmer (August) and Wonder (October).

What it’s for: Ember handles the repetitive jobs AI agents do all day, like tool calls, coding subtasks, routing, data extraction and long-document parsing. Meta’s pitch: “Muse Wonder plans, Muse Ember handles the volume.”

Key specs: It works with text, images, audio and video and has a 1M-token context window. It’s also the first model trained using Muse Wonder Full, Meta’s larger model.

Pricing: It costs $0.10 per million input tokens and $0.50 per million output tokens. On Artificial Analysis, a task costs about $0.14 on Ember versus $2.39 on Muse Spark 1.3, which scores only one point higher.

How it compares: Meta says Ember gives more intelligence per dollar than Claude Haiku 5.5 and GPT-6 Luna. Its benchmark scores are close to Haiku 5.5 at nearly 3x lower cost.

Availability: It’s live now in Muse Code and the Meta Model API (model ID muse-ember-1.0), plus Google Cloud and Oracle Cloud. Meta Muse itself is still limited to select markets, including the US, Canada, Singapore and Japan.

The catch: Like Meta’s other Muse models, Ember’s weights are private. It’s not open source.

Adobe AI Assistant Create and edit faster with AI across your Adobe apps

YouArt Desktop App An AI studio that handles creative work for your team

Nameshot Names your screenshots for you automatically

Yogurt Free, private meeting notes, an open-source Granola alternative

Toki Matter An AI agent that keeps your tasks moving forward