Chaen.ai Work with me

AI news

AI News Roundup: Claude Meetings and NVIDIA's CEO Exit

🗓 ⏱ 8 min read

中文日本語

AI News Roundup: Claude Meetings and NVIDIA's CEO Exit

An AI news roundup for Nov 9-13, 2026: Claude Meetings, Sheets, Fable 5.2, Gemini 4 Carbon and NVIDIA's CEO exit, plus what each means in Japan.

This is my weekly report, organizing by theme what I posted on X between November 9 and 13, 2026. In one week Anthropic moved into meetings and office work, Google shipped its top model and a pair of smart glasses, and NVIDIA’s CEO resigned. By the end you will have the key points of each announcement and the decisions a Japanese company should be making internally right now.

Claude moved into meetings and office work

Claude Meetings joins the call

On November 13 Anthropic announced Claude Meetings. Claude joins a video meeting and builds documents and whiteboards while people talk.

  • Claude joins as a participant and takes instructions by voice or chat
  • Everyone can see the material as it is being built and correct it on the spot
  • After the meeting it automatically assembles notes, action items and a transcript
  • Google Meet is supported first, with other meeting tools to follow soon
  • Beta on the Team and Enterprise plans

Honestly, this is impressive. It also means meeting-notes startups may be in real trouble. Claude is going after the entire business toolchain, and weeks like this remind me how strong the companies that own a model are. Products can now be built to a decent level very quickly with AI agents, so any service competing on a single feature is shaky at the foundation.

The rollout order is easy to pick. Google Meet comes first, so companies already centered on Google Meet can try it first. It is a beta on Team and Enterprise, so run it with one team before going company-wide. The criterion to watch is simple: does the “clean up the notes and turn them into a document” step after a meeting actually disappear? If it does, this stops being a tool swap and becomes a redesign of how you run meetings.

Whiteboards, spreadsheets and dashboards

On November 10, Claude Whiteboards, Sheets and Data went live. This is aimed squarely at Excel and PowerPoint.

  • Whiteboards: turn a plan you talked through into a board
  • Sheets: ask for a budget table or tracker and it builds it
  • Data: the former Dashboards, now generally available on paid plans
  • Adobe, Atlassian and Box can be operated alongside the conversation

You ask Claude for the board, the spreadsheet or the dashboard, and the team edits it from there. Work that used to be done by hand in Excel or Miro now finishes inside Claude, which is a big shift. There will be more and more situations where Claude alone is enough for office tooling. The fastest things to hand over are the ones with a fixed shape: budget tables, progress trackers, planning boards. For final versions that go outside the company, settle who can edit and where it is stored first.

Claude Fable 5.2 and Mythos 5.2: how I would split them

On November 9 Anthropic announced Claude Fable 5.2 and Mythos 5.2, positioned as beating Opus 5.5 on nearly every item.

  • On FrontierCode 2.0 it is on par with Opus 5.5 Max at roughly half the cost
  • Output is up to 30% faster than Fable 5.1
  • 1,000 agents formalized Fermat’s Last Theorem in 30 hours
  • API pricing cut to $8 input / $40 output

My split will likely be Opus 5.5 for daily work and Fable 5.2 for heavy development and research. The “roughly half the cost” line matters directly when you design processes that run continuously inside a company.

A five-minute documentary in one shot

Matt Shumer published a roughly five-minute documentary that Fable 5.2 produced in a single pass. The theme was the AI year of 2026, and research, script, footage sourcing, every frame and the music were all AI; he only gave revision instructions. In his words, “Opus looks like a toy.” Video at this level is remarkable, so I will try it myself.

Robotics was the biggest surprise

What surprised me most this time was the robotics performance. It can move a robot arm even on tasks it was not trained on.

  • 64.9% on RoboDojo, a bimanual robot simulation
  • The previous Fable 5.1 scored 27.6%, so this is more than double in one step
  • Opus 5.5 is at 39.4% and GPT-6 Astra at 29.8%
  • It also beats many AIs built specifically for robot control (company’s own measurement)

Next year physical AI looks likely to be genuinely useful on site even without post-training. I want to load it onto my Unitree G1 and try it as soon as I get home.

Google’s Gemini 4 Carbon and the Gemini glasses

On November 11 Google made Gemini 4 Carbon generally available. Google calls it its strongest model yet, built to dig deeper and handle long-running tasks and multi-step reasoning. The framing was that Argon is for complex tasks while Carbon is the model you bring in for the hardest problems. On benchmarks it was ahead of Claude and GPT, so I am going to use it and review it.

Smart glasses with Gemini

On November 13 Google announced Gemini-equipped smart glasses, co-developed with Samsung, with designs from Gentle Monster and Warby Parker and preorders now open.

  • Ask about what is in front of you with “Hey Gemini”
  • Real-time translation in more than 70 languages
  • 12MP camera and up to 3K video
  • Up to nine hours on a charge
  • IPX4 water resistance

“Look and Ask” — asking about whatever you are looking at — seems genuinely useful, and I want a pair. This becomes a serious rival to Meta Ray-Ban.

ChatGPT’s dot now connects to multiple PCs

On November 10, ChatGPT’s dot gained support for multiple PCs. You can connect a second or third machine to the same dot and specify which one to use per task — for example, having a second Mac check an iOS build and report back with a screenshot. One thread can use one PC. Handing builds and tests to another machine while you keep developing on your laptop is useful, and it will land especially well with people who keep a Mac Mini or Studio at home.

NVIDIA’s CEO resignation changes the backdrop

On November 12 NVIDIA formally announced that CEO Jensen Huang had resigned, stepping down immediately as CEO, president and board member after leading the company for 33 years since its founding.

  • CFO Colette Kress becomes interim CEO
  • Kress has served as NVIDIA’s CFO since 2013
  • Harvey Jones becomes the new chair of the board
  • The permanent CEO will be selected from inside or outside the company
  • The November 17 earnings report proceeds as planned

The announcement came amid reports of suspected spying for China and of an affair, but the substance of those allegations is unconfirmed, and it needs to be kept separate from the confirmed facts. NVIDIA says there is no impact on its business or product roadmap, but the sudden exit of a central figure in the AI industry is a shock, and it is likely to shift the balance of power in the sector.

The month before the resignation, Huang had talked about succession in a Bloomberg documentary — that just as he himself could be fired, the next CEO has to be chosen. He said the next leader is already inside NVIDIA, that his job is to create an environment where successors grow, that he likes not being able to rest on his internal standing, and that constantly having to earn your position is a healthy state. Hearing it now, it lands very differently. Last week’s stage appearance may have been his last. His presentations and his jokes were genuinely good.

What this means in Japan: the MUJI breach and your own operations

The heaviest story on the Japanese side came on November 12: up to roughly 27.26 million customer records at MUJI were possibly accessed improperly, with around 1.09 million credit card numbers also in scope. What stunned me is that until 2021 the full credit card number was used as the customer ID, stored with neither encryption nor masking, and even after the system was changed it was not deleted and remained for about five years.

I struggle to reconstruct the thinking that leads there. Before anyone talks about cyberattacks, the obvious basics have to be in place. This connects directly to the rest of this week’s AI news. If Claude is now sitting inside your meetings, spreadsheets and dashboards, your internal data passes through more places than before. Without taking inventory of what data you hold, in what form and where it still sits, convenience arrives with extra risk attached.

For overseas AI companies bringing products to Japan, this week’s announcements are the kind where the explanation decides the reception. Which meeting tools are supported, which plans actually get access, and how meeting audio and materials are handled. Whether those three points are spelled out in Japanese up front largely determines whether a Japanese IT department can get approval internally. In Japan that gets examined before raw capability does.

How to use this week inside your own company

Two of this week’s posts were not about tools. One was about communication: I think the defining trait of people who are good at it in business is the preparation that drives the other side’s reply effort as close to zero as possible. Rather than making someone think, research or go back and forth, you organize the information yourself so that in the end they only have to choose between A, B and C. It is about reducing their time and cognitive cost, and it applies to almost every relationship, not only work.

The other was a roughly three-minute video of Elon Musk on why he keeps working. As a teenager he struggled with not seeing any meaning in life; religious texts and German philosophy gave him no answer; the turning point was the 42 in The Hitchhiker’s Guide to the Galaxy. The universe is the answer, and what is missing is the right question. It is the clearest explanation I have seen of where SpaceX and xAI are pointed.

If I were applying this week internally, the order would be: first, measure the after-meeting work. Once you know how many hours a week go into notes and turning them into documents, the value of trying Claude Meetings becomes a number rather than a feeling. Second, revisit the cost of anything you run continuously — Fable 5.2’s price cut changes the assumptions behind that design. Third, do the data inventory in the same week. These three look like separate projects, but they always arrive together once AI starts spreading inside a company. The communication point is the same idea: whether you are handing work to a person or to a model, the side that shapes it into something easy to choose from moves faster.

FAQ

Which meeting tools does Claude Meetings support?

Google Meet is supported first, with other meeting tools said to follow soon. It launched as a beta on the Team and Enterprise plans.

How should I split work between Claude Fable 5.2 and Opus 5.5?

Fable 5.2 was announced as beating Opus 5.5 on nearly every item. My own plan is Opus 5.5 for daily work and Fable 5.2 for heavy development and research.

What happens to NVIDIA's business after the CEO resignation?

CFO Colette Kress is interim CEO, the permanent successor will be selected from inside or outside the company, and the November 17 earnings report goes ahead as planned. NVIDIA says there is no impact on its business or product roadmap.

Based on a post on XOriginal post ↗

Share this article

XFacebookLINEはてブ

About the author

Masahiro Chaen
Masahiro Chaen Founder & CEO, DigiRise Inc.

CEO of DigiRise Inc. Has supported 600+ companies and 50,000+ people with AI training, consulting and development. GMO AI & Crypto advisor, Google Gemini AI advisor. 200K+ followers on X.