Resources:
Join the discussion: Got something to say? Let us know on LinkedIn and network with other AI leaders
Upcoming Episodes: Check out the upcoming Everyday AI Livestream lineup
Connect with Jordan Wilson: LinkedIn Profile
AI Week in Review: Breaking Down Gemini 3, GPT-5.1 Pro, Microsoft 365 Agent Updates & More for Business Leaders
This past week marked the most densely packed period of AI innovation and product releases since December 2024. Google, OpenAI, and Microsoft dropped major updates that carry immediate implications for business owners and decision makers. In this analysis, the focus will be on the explicit new features, integrations, and management capabilities that can alter how businesses operate, analyze, and grow. Insights are drawn directly from real product releases, critical updates, and benchmark revelations, ensuring clarity without fluff.
Gemini 3 AI Technology: Multimodal Performance and Enterprise Applications
Google officially released Gemini 3, positioning it as the most comprehensive AI model available, with benchmark dominance across LM arena, GPQA diamond, Math Arena, Apex, Arc AGI, and more. For organizations leveraging AI for text, images, video, audio, and code, Gemini 3 enables robust multimodal capability. Notably, the model excels at video understanding—even without transcripts—as well as code generation to build websites, video games, and custom workflows.
Businesses benefit from Gemini 3’s expanded context window of one million tokens, enabling richer outputs for longer-form inputs, such as legal documentation or full-market analyses. Enhanced agentic research and tool use help automate multistep workflows, increasing reliability in repetitive or complex tasks. New generative UIs can dynamically visualize data, supporting clear communication across teams.
Availability is immediate in Gemini’s web app, mobile app, AI Studio, Vertex AI, Google search’s AI mode, anti-gravity IDE, and Gemini CLI—making it accessible both for technical integrations and direct employee usage.
Nano Banana Pro: Advanced Image Generation for Professional Visuals
Google’s Nano Banana Pro, the integrated image generation model within Gemini 3 Pro, sets a new bar for AI-generated and edited visuals. The model offers sharp text, rich styles, and advanced scene controls, with capabilities such as blending up to 14 images to generate realistic composite photos.
Businesses can create executive-grade team images—even when leaders are geographically dispersed—by merging individual headshots into realistic group photos, simplifying marketing, investor presentations, and press materials. Nano Banana Pro’s accuracy means that outputs are nearly indistinguishable from reality, removing previous barriers of “AI image tells” and enabling use in high-stakes environments. Enhanced typography and style options support the production of infographics, branded content, YouTube thumbnails, and publication-ready visuals directly from research notes (via NotebookLM integration).
ChatGPT Group Chat: AI-Enhanced Collaboration for Teams
OpenAI launched group chat functionality for ChatGPT, supporting up to 20 participants in a single collaborative conversation. Unlike legacy LLM chat behaviors, group chats maintain privacy and separation from personalized memory, mitigating cross-contamination of sensitive data. For business use, this ensures project groups can brainstorm, ideate, and request AI-powered insights without risk to individual data stores.
While the underlying model is currently GPT-5.1 Auto—which has limitations for high-precision work tasks—the structure enables future selective model upgrades. Teams benefit from synchronized chat histories, emoji reactions, personalized image generation, and clear separation between group and personal workspace activities.
Microsoft 365 Agent Services: Security and Productivity Meets Real-Time Analytics
Microsoft announced two pivotal agent services for enterprise users at Ignite. Agent 365 centralizes management, access, security, and visibility for AI agents within businesses, offering a single dashboard to track usage, permissions, and agent relationships. This directly addresses historical security and access barriers to Copilot adoption, reducing risk from over-permissioned agents and giving IT leaders granular control.
Additionally, dedicated agents for Word, Excel, and PowerPoint are now integrated into the Microsoft 365 Copilot chat interface. Employees can use natural language to launch projects, generate structured files, and move seamlessly between chat and desktop editing environments. These agents leverage multi-step reasoning, real Office app generation (not mere XML assembly), and advanced grounding via Microsoft Graph and WorkIQ—improving organizational context, accuracy, and transparent citation tracking.
EU Digital Omnibus: Regulatory Easing for AI Expansion
The European Commission introduced the digital omnibus to simplify and delay the enforcement of strict AI regulations. For businesses operating in the EU or working with EU data, this means a likely extension of compliance timelines for sensitive areas such as biometric ID, health scoring, and law enforcement use—from August 2026 to December 2027. The easing of strict personal data definitions supports broader use of anonymized datasets for AI training, especially benefiting multinational firms with heavy European operations. While big tech welcomes the move, privacy advocates note significant rights rollbacks.
Anthropic Investments: Microsoft and NVIDIA Diversify the AI Landscape
Microsoft and NVIDIA’s joint multibillion-dollar investment in Anthropic strengthens competition in advanced model development and cloud AI workloads. Beyond their funding roles, Anthropic commits to substantial Azure compute purchases, creating cyclical business ties and reinforcing cloud infrastructure priorities. For decision makers, this signals a broadening of available models and compute partnerships, with practical implications for AI sourcing, vendor negotiations, and future-proofing technology roadmaps.
GPT-5.1 Pro and the Next Benchmark Race
OpenAI quietly rolled out GPT-5.1 Pro to pro-tier subscribers ($200/month), featuring adaptive reasoning that allocates computational resources based on query complexity. This results in dramatically improved accuracy and response relevance for both simple and complex queries. Early reports highlight breakthroughs in scientific reasoning and mathematical problem-solving, which previously eluded AI models.
Expectations are high for industry benchmarks to validate or surpass Gemini 3’s current standings, particularly with the anticipation of API release enabling broader integration. Organizations already using pro-tier models should monitor for API access to maximize competitive advantage in data science, business processes, and research innovation.
NotebookLM: Seamless Infographic and Slide Deck Generation
The integration of Nano Banana Pro within Google’s NotebookLM elevates in-workflow infographic and slide creation. Directly from loaded notes, users can produce publication-ready visuals and multi-page slide decks, significantly cutting manual design effort. Outputs feature professional aesthetics and accurate text rendering, even across complex styles and formats. This directly enhances productivity for market analysts, sales teams, and educators needing consistent character-driven imagery and rapid turnaround documentation.
Export functions currently enable PDF output, suitable for immediate sharing, though export to formats like PowerPoint or Canva are not yet available.
Fast-Paced AI Evolution: Additional Highlights Relevant to Business
Foxconn and OpenAI announced an AI partnership, adding hardware insights to the AI innovation stack.
Gemini 3’s advanced agents are now live for Ultra subscribers, promising deeper integration with Gmail, calendar, and web workflows.
Group chat and mobile browser upgrades from Poe, Perplexity, and XAI indicate continued focus on collaborative, accessible AI experiences.
Cloud and infrastructure investments, continued partnership shifts, and regulatory rollbacks collectively set the trajectory for competitive, data-driven business growth.
Conclusion: Tactical AI Adoption for Immediate Business Impact
Across image generation, workflow automation, agent security, regulatory trends, and competitive investments, this week’s AI releases aren’t just headlines—they are tangible, accessible upgrades designed to drive measurable business outcomes. From assembling global team visuals with Nano Banana Pro, enabling secure cross-org agent management with Microsoft Agent 365, to mobilizing group AI-based collaboration and leveraging multimodal search with Gemini 3, the landscape has rapidly shifted toward integrated, enterprise-grade tools.
Leaders tracking these developments should move quickly to evaluate early adoption, pilot API integrations, and negotiate compute partnerships that align with their business strategy. The competitive value is not in the raw release, but in the precision deployment of features specifically engineered for productivity, accuracy, and scalable collaboration.
For ongoing tracking and translation of the latest in AI, targeted reviews and hands-on workflow analyses remain essential.
Topics Covered in This Episode:
- Gemini 3 Model Launch & Features
- Google Gemini 3 Benchmarks & Performance
- Nano Banana Pro Image Model Release
- Nano Banana Pro Capabilities & Use Cases
- OpenAI ChatGPT Group Chats Launched
- Microsoft Agent 365 AI Management Tools
- Microsoft Copilot Word, Excel, PowerPoint Agents
- EU Digital Omnibus Eases AI Regulation
- Microsoft & Nvidia Invest in Anthropic
- OpenAI GPT-5.1 Pro Model Rollout
- Google NotebookLM Adds Nano Banana Pro
- Slide Decks & Infographics with Gemini
- AI News: XAI Grok 4.1, Perplexity Updates
- Google Ads Integration With Nano Banana Pro
Episode Transcript
Jordan Wilson [00:00:47]:
By far, this was the most impactful week of AI releases since December. Yeah. For more than a year, we haven't seen a week like this in AI. Google, OpenAI, Microsoft, all with huge drops, and it was wild, wilder than a Waffle House at 2AM. I mean, you have now Gemini three, the most powerful model in the world by far. OpenAI snuck a model in that could end up being even more powerful, but didn't even release a blog post about it. Microsoft now has AI agents where people e people actually work, and probably the one that's getting the most buzz isn't any of those. It's nano banana pro.
Jordan Wilson [00:01:36]:
So much happening in the world of AI this week. Like I said, we haven't had a week like this truly since December 2024. So let's jump into it. Let's get straight to the AI news that matters. What's going on y'all? Welcome to Everyday AI. My name is Jordan Wilson, and, well, we do this every day. It's your daily livestream podcast and free daily newsletter helping everyday business leaders like you and me not just keep up with all the AI news because it's a ton, but how we can make sense of it, cut the BS, and leverage it to grow our companies and our careers. So it starts here with the unedited, unscripted live stream podcast.
Jordan Wilson [00:02:13]:
But to take it to the next level, go to our website, youreverydayai.com. Sign up for the free daily newsletter. We're gonna be recapping the AI news for today. And just real quick before we go over the AI news for the week, that's what we do on Mondays. We bring you the AI news that matters. Quick little housekeeping announcement for this week. Here in The US, we have the Thanksgiving holiday. So the show and the newsletter will be off on Thursday and Friday.
Jordan Wilson [00:02:40]:
But Wednesday, gonna be a banger show. You all wanted, round two hands on AI at work on Wednesdays with Gemini three, so make sure to tune in for that. Alright. Let's get into the AI news that matters for the week of November 24, and, well, it starts with Gemini three. So Google has officially released Gemini three after months of hype. Its latest AI model drawing widespread attention for its performance in new features. So Gemini three arrives just eight months after the release of Gemini 2.5 pro, which was already considered a leading multimodal AI model. Even though it was eight months old, it was still pretty much the best AI model in the world, yet Google woke up and chose violence and released Gemini three.
Jordan Wilson [00:03:27]:
So according to Google's benchmarks, it tops nearly every major benchmarks, including the LM arena where users go and vote, as well as the more scientific benchmarks GPQA diamond, math arena, Apex, Arc AGI, and literally just about everything else. So the model is available pretty much everywhere and literally everywhere you look. It's like, oh, Gemini three is here. So whether you're using the app, so gemini.google.com or on the actual, you know, mobile app, it's there. It's in AI studio. It's in vertex AI. It's in Google searches new AI mode. Yeah.
Jordan Wilson [00:04:03]:
Day that they released it, they released it in AI mode. And then there's also new platforms as well. So it's in their new anti gravity. They're, kind of multi agentic IDE as well as their, command line interface, coder called the Gemini CLI. So, yeah, Gemini three just literally came out everywhere except notebook LM, but I'm sure that's gonna be happening soon. So a little bit more about Gemini three itself, it sets itself apart with best in class multimodal capabilities, its ability to handle text, images, video, audio, and code, and it notably excels at video understanding even without transcripts. So I'd say out of all the new capabilities, probably one of the more impressive ones in my opinion is its ability to just create great, number one, great code, but anything that code can create. You can go create video games.
Jordan Wilson [00:04:59]:
You can go create beautiful, stunning, websites. I mean, it's literally insane. What Gemini three is capable of on the coding and and on the visual end because, with the Gemini three, Pro model is also tied to it, the Nano Banana Pro. So we're gonna be getting to that later in the show. So a little bit more about what's new inside of Gemini three. So the new model features a million token context window, allowing it to process and generate richer outputs from extensive and longer inputs from the user. There's also, improved agentic research and tool used. Both of those have been enhanced enabling Gemini three Pro to autonomously plan, code, and execute multi step workflows with improved reliability and consistency.
Jordan Wilson [00:05:50]:
So Gemini three also brings new dynamic interactive dashboards in generative UIs. Yeah. So literally on the fly, Google might just create a dynamic view or a generative user interface for you to help you to help explain something a little bit better. And that also shows up in Google's new, improved, AI mode in search. So, yeah, if you toggle it on, and we went over that on Wednesday. So, just if you haven't already, make sure you go listen to episode six fifty seven. So this was just hours after, Gemini three came out. So we did do a pretty deep dive on Gemini three in three upgraded use cases.
Jordan Wilson [00:06:36]:
And one of those use cases, we showed, the more dynamic, AI mode in search. So a ton of new on the Google Gemini three live stream audience. Let me know. Have you been using it? Are you liking it? What questions do you have? Are you still using, GPT five one or Claude Sonic four five? What's your favorite model now? All right. Speaking of favorite models on the image side, there's literally no comparison anymore. That's because Google rolled out Nano Banana Pro, the image generation and editing model that lives inside of Gemini three Pro, and it promises and delivers, my gosh, sharper text and images, richer styles, and more advanced scene in lighting control. So Nano Banana Pro is the image generation and editing model tied to Gemini three Pro, and it follows the earlier Gemini 2.5 flash image, model that was announced in August. So, yeah, the the naming is much better just calling it nano banana.
Jordan Wilson [00:07:42]:
So the original nano banana was Gemini 2.5 flash image. Now with the new Gemini three pro, it is Nano Banana Pro. So stylistically, Nano Banana Pro supports a wider variety of textures, fonts, and the calligraphy, and it can produce flat vector style graphics and more expressive typography. You can also do just extremely creative, takes, right, without even really using your brain. So as an example, you can do, you know, change the camera angle on an image. Maybe you got a headshot and you're like, this isn't a flattering angle. Okay. Well, you can upload it to Nano Banana Pro, change the angle, change the focus, the scene lighting, apply advanced color grading, so many things.
Jordan Wilson [00:08:26]:
Some other new capabilities, well, it can blend up to 14 images. Jeez. Right? Another quick example. I don't know if if you're a a global team and you're never able to get, you know, some of your senior leadership together. Well, you could just take all your individual headshots and throw them in a nano banana and create a team photo, and it will look literally realistic. It is hard to tell. It is scary good. Yeah.
Jordan Wilson [00:08:51]:
There's actually gonna be, it's gonna be used for a lot of bad things. I've already seen it in a very short period of time. Nano Banana has been out for, like, four days. Yeah. It's gonna be used for a lot of bad things because it is that good, if that makes sense. You can't tell anymore, that something is AI generated with Nano Banana Pro. I even think with Nano Banana, you kind of could, if you paid close enough attention. Nano Banana Pro, no.
Jordan Wilson [00:09:17]:
It looks so, so stinking real. So it is rolling out globally. I do believe that they've, rolled it out to a 100% of users in the Gemini app. Also free users get a limited quota, before falling back to the original Nano Banana. So, yeah, if you are, as an example, going into Google Gemini and you want to use Nano Banana there, that's probably the easiest spot to use it. You can also use it in Google's AI Studio. Just make sure it's not properly labeled. You do have to, choose, Gemini when you go into Gemini and you click, the, you can click the create image button or you can just click thinking with three pro.
Jordan Wilson [00:10:01]:
So if you're just chatting, with, Google Gemini and you want it to create an image, yeah, some people don't know this. You don't you don't always even have to click the image button, but make sure that you choose the model thinking with three pro. Because if you, use the now default, model that says fast, that's actually gonna be using the old nano banana. Alright? So, I did confirm that with, the Google team. So just FYI. Alright.
Midroll [00:10:36]:
This podcast is sponsored by Google. Hey, folks. I'm Amar, product and design lead at Google DeepMind. We just launched a revamped vibe coding experience in AI Studio that lets you mix and match AI capabilities to turn your ideas into reality faster than ever. Just describe your app and Gemini will automatically wire up the right models and APIs for you. And if you need a spark, hit I'm feeling lucky and we'll help you get started. I.studio/build to create your first app.
Jordan Wilson [00:11:11]:
Bring your friends because now you can all chat together with Chat GPT. So OpenAI has now globally launched group chats for Chat GPT, allowing up to 20 people to collaborate with ChatGPT in a single conversation. So ChatGPT group chats are now available globally after first being rolled out to a handful of company, countries, which do not include The US. So now it's available in The US and globally. So like I said, the group feature supports up to 20 participants and keeps group conversations separate from private chats. So personal chat g b t memories do not leak into group spaces. So, yeah, that's important. Right? One of the great things about large language models like chat g p t is their ability to remember things across chat, you know, intentionally saving, little things about you as memory.
Jordan Wilson [00:12:05]:
So think of these group chats almost as, like, in terms of yourself and your participation, it's almost like you're in an incognito or private chat. But within that chat window, ChatGPT will remember everything that goes on in there, whether there's two participants or 20 with Chad GPT. So group creators remained protected from being removed by other members unless they voluntarily leave. So, yeah, different people, you can leave at different times. The thing that's important here, well, the AI remains mostly passive, which is really good to know because, yeah, if you have 20 people, just chatting with each other, you don't obviously want ChatGPT to butt in except when it's needed. So that's the thing. Chat GPT just kind of watches the conversation and responds only when either explicitly tagged or when the context clearly calls for Chat GPT to intervene. Chat g p t can react with emojis, mention, profile pictures, and even generate personalized images that include group members on requests.
Jordan Wilson [00:13:12]:
So the underlying model is g p t five one auto, which is not good, just FYI. So I wouldn't be using this for any work tasks yet. Although I do think this is, you know, OpenAI, I know, you know, have a few listeners there. You gotta update this. Right? And and, hats off because OpenAI has been, taking a lot of feedback, and saying that they're gonna be rolling out some updates, but you you you have to, allow for a model selector. I think this was kind of rolled out and promoted as kind of like a cutesy thing. Right? It, the kind of demo that they did was, you know, some people chatting, trying to find a Italian restaurant. It's like, okay.
Jordan Wilson [00:13:50]:
Cool. Y'all, we have large language models that are smarter than people. Like, we don't wanna be chatting and asking for recommendations for where to go eat. This I think the great utility of this is working with your coworkers, but do not do it now because, the g p t five auto, sorry, GPT five one auto model is not good. So, hopefully, OpenAI will be rolling out the ability to select the model, especially when it's passive. Right? Yeah. If if if Chad GPT responded to every single chat in a group thread, you would want to use probably the auto or the instant because so you don't have to wait. Right? But, hopefully, OpenAI takes a little bit more of a business approach to this because the potential is actually really, really high with this one if they get it right.
Jordan Wilson [00:14:42]:
So, continue. We'll be covering this when and if they release new updates. Alright. Well, one of the biggest, updates on the agent front this week was from Microsoft. Actually, two separate agent releases that we're gonna be covering here. So first, they launched agent three sixty five to manage AI agents across organizations. So this was announced at Ignite, their, yearly conference from Microsoft, and it's a new service that centralizes the management, access, security, and visibility for AI agents across enterprises. So this gives organization a single registry in management plane for AI agents starting with tenants in the Frontier preview who also hold at least one Microsoft three sixty five Copilot license.
Jordan Wilson [00:15:32]:
So, yeah, you have to be part of the Frontier preview program and obviously have at least one Microsoft three sixty five Copilot license. So the new agent three sixty five service provides a registry as a single source of truth for all agents within the organization, helping IT teams track which agents exist, who owns them, who's using them, and where they're even allowed to be accessed. So agent three sixty five offers access control that limits each agent's permissions to only the resources required for its task, reducing risk from over permissioned agents. There's also a new unified dashboard in advanced analytics that lets administrators visualize relationships between agents, administrate, as well as the people in data and monitor agent behavior and performance in real time. Alright. Next piece of AI news. Yeah. More agents from Microsoft.
Jordan Wilson [00:16:30]:
And these ones, I think, are probably gonna get a little more use. Although agent three sixty five, I think, is sorely needed, right, because one of the biggest things holding Copilot back is permissions, security. No one can access anything. Right? I think more than anything, it's there's a ton of promise with Microsoft and Copilot and all their agents, but everyone's like, yo. Like, I spend hours trying to figure out how to get access, to anything within my organization, and I can't. Well, maybe now with the new launch, they can because Microsoft is bringing agents to the chat. Alright? The chat interface that most people use. So Microsoft rolled out dedicated Word, Excel, and PowerPoint agents inside Microsoft three sixty five Copilot.
Jordan Wilson [00:17:17]:
And the update makes its pot makes it possible to create high quality documents, spreadsheets, and presentations from the chat interface and then move it seamlessly into desktop apps for deeper editing. So the biggest headline here is that users can now start projects in Copilot chat using natural language, chatting with these agents, and have dedicated agents generate full office files, which then they can open and continue to edit inside Word, Excel, or PowerPoint. So Microsoft says the agents use multi step reasoning models and an agentic harness to plan, draft, and validate and refine outputs across stages, producing structured results rather than just one shot text. So files are created by headless real office applications rather than just stitched together with XML, which Microsoft argues preserves styles, themes, formulas, and safety controls, such as blocking, macros, and external links. There's also enterprise grounding that's available, which is extremely important, through Microsoft Graph and WorkIQ. So agents can securely pulled organizational context, such as emails and shared docs with permission checks and transparent citations to improve relevance. So, PowerPoint, the PowerPoint agent in agent mode, which we covered a couple of weeks, a couple of months ago, actually now. Wow.
Jordan Wilson [00:18:47]:
In the apps also that Copilot make direct edits in app specific changes. So Microsoft highlights agent mode in PowerPoint as new to the Frontier program. Yeah. So finally, this is a capability I think that has, users have been wanting. So many people just use Copilot inside of Copilot chat, because, well, one of the reasons it's well, you're in Teams, you know, and people like and are most familiar with using AI products or large language models in the chat interface. Right? Like, you're going to, gemini.google.com or chatgbt.com or claw.ai. Right? So front end AI usage. So, smart move from Microsoft to bringing agents to the chat, but now we'll see if permissions are going to ever be figured out, or if you're still gonna have millions of users just looking at all these Copilot announcements and being like, oh, this is cool, but we're never gonna get access even though we pay for Microsoft three sixty five Copilot.
Jordan Wilson [00:19:53]:
Alright. Our next piece of AI news, interesting one, didn't see this one coming. The EU has moved to kind of distance itself from its, previous strict AI rules. So the European commission unveiled a new digital omnibus, which is a package of proposals to simplify the much more strict AI act. So aiming to reduce regulatory friction and speed innovation, of AI in the European Union. So the single biggest change according to reports is enforcement timelines for stricter AI controls in sensitive high risk areas are proposed to move from August 2026. So, yeah, this was supposed to be around the corner. Instead, they're gonna be potentially kicking it back, more than a year from that to December 2027.
Jordan Wilson [00:20:54]:
And this would affect biometric ID, utilities, health credit scoring, and law enforcement uses. So the commission wants clearer rules on when data stops being personal, a shift that would make it easier for companies to treat anonymized or de identified EU data as usable for AI training. Yeah. This is one of those reasons a lot of times. Right? I think you saw it a lot with Chad GbT when they had more, personalized. Right? I remember with advanced voice mode, it couldn't, be released, in Europe for many months and they still had to strip away features. The same things with a a lot of Microsoft products, Google products, etcetera. The more and more personalized AI becomes, actually the more difficult or the harder it is to actually use that in the European Union because of the AI act that was passed, last year.
Jordan Wilson [00:21:46]:
But like we just said here with the new digital omnibus, some of these things could get kicked, more than thirteen months. So, a lot of, you know, kind of, AI, implementations or, you know, new products were kind of, dead on arrival in the EU. But maybe that could change, with this, easing of the restrictions. So if adopted, the changes would explicitly enable companies such as Google, Meta, and OpenAI to use more European's personal data for training AI models, a move obviously welcomed by big tech companies and industrial firms, but condemned by privacy advocates. So privacy groups and a 120 a 127 civil organizations have called the package the, quote, the biggest rollback of digital fundamental rights in EU history, arguing it weakens protections and hands more intimate data to major tech companies. So I've been saying this for multiple years now, ever since the, EU's AI act, was just an idea, right? It's obviously been passed in parts of it have been enacted. And, you know, some of the bigger pieces were right around the corner. I said this all along.
Jordan Wilson [00:23:05]:
I said, essentially, EU has two options. They are either going to fall so far behind the rest of the world, specifically, The US, and China, or they're going to eventually realize that the EU act was, the EU AI act was probably a little too strict, and they would have to start rolling it back, kind of peel by peel. So here we see the latter, the European Union, is, seemingly, admitting, yeah, maybe we were a little too robust with the AI act. And, so we'll see, what happens in the coming months. Next piece of AI news. Jeez. This one probably hit OpenAI in the gut. So Microsoft and Nvidia, two, bigger partners of OpenAI announced a, joint investment in Anthropic.
Jordan Wilson [00:24:04]:
So Microsoft and NVIDIA announced major investments in Anthropic, a move that reduces Microsoft's sole dependence on OpenAI and also widens industry competition. So Microsoft will, invest up to 5,000,000,000 in Anthropic, and NVIDIA will invest up to 10,000,000,000. So the new investments push Anthropic's valuation to about $350,000,000,000 up from roughly a 183,000,000,000 in September. So, yeah, not quite, doubling their valuation, but nearly doubling their valuation in a two month period. Right? And, getting up to roughly about the last valuation that we saw for OpenAI. So Anthropic has also committed to buying $30,000,000,000 of Azure compute capacity from Microsoft and has contracted additional compute up to a gigawatt, the company said in a blog post. So, yeah, this is, this is only gonna fuel, more of the AI bubble talk in circular funding. So, you know, obviously, anthropic got a check.
Jordan Wilson [00:25:15]:
Right? I don't know. I know they don't get checks, but we should still go old school. Right? They should, you know, give give Dario, you know, $5,000,000,000 giant check. But what would happen is, well, Dario would have to, you know, write a big billion multibillion dollar check back to Microsoft for the, Azure compute capacity. So, yeah, this is, if if nothing else, this is just gonna set off more of the alarms, the vendor financing or circular funding talks, that have been really rattling The US economy and the stock market over the last couple of months, following, OpenAI, kind of coming with a lot of these similar deals even with NVIDIA. So NVIDIA and Anthropic will collaborate on engineering and model optimization to improve performance and efficiency and to tune in video, architectures for Anthropics workload. So this one is I think probably, most interesting to look at Microsoft's position here Because Microsoft already holds a sizable stake in OpenAI, valued at about a 135,000,000,000 or roughly 27%, as an converted diluted basis. So essentially, right now, Microsoft is the single biggest or sorry.
Jordan Wilson [00:26:36]:
The sync the single largest individual, shareholder in, equity holder in OpenAI. Yeah. No one else has more than 26% except Microsoft, which holds 27% of the new for profit corporation, that OpenAI, finalized earlier this fall. So here's, here's my take on this. Not a good week overall for OpenAI. Right. So not only does Google release Gemini three, which is extremely powerful, and a lot of the chatter online, are now saying like, hey, Gemini three seems like a daily driver model. Right? A lot of, companies, even Salesforce's CEO said, hey, I've been using, ChatGPT daily for two years.
Jordan Wilson [00:27:31]:
I just went into Gemini, and wow. I'm switching. I think a lot of, you you know, we're seeing a lot of those stories anecdotally. We'll see if it moves the numbers if, anyone can catch OpenAI's, reported 800,000,000 weekly active users. But overall, this is another news story, not very good for OpenAI because you have open in OpenAI, two of their bigger partners in Microsoft and NVIDIA signing on to, well, fun, their biggest probably one of their two biggest competitors. I'd say they're competing, in a lot of different ways with Google, but specifically on the model side with Anthropic. But it wasn't all bad news for OpenAI. All right.
Jordan Wilson [00:28:16]:
Our next AI news story, aside from OpenAI releasing the, group chats, there was another big story that interestingly OpenAI didn't even really promote this, which was funny to me. So OpenAI rolled out technically their most powerful model to date called GPT five one pro. Right now, it's only released to subscribers on the pro tier, the $200 a month, subscription, and it has some new reasoning capabilities as well. Also, it is going to replace the GPT five pro model as the default pro model, over the next ninety days, at least according to OpenAI's release notes. So, yeah, OpenAI didn't even put out a blog post on GPT five one pro, which I found interesting. So the rollout is significant though, because GPT five one pro introduces a new adaptive reasoning, which lets the model decide when to spend more computational time on a prompt, giving simple queries, faster replies, and complex tests, deeper, more accurate reasoning. So essentially, it reasons about reasoning or it thinks about how much it should think. So OpenAI highlights improved performance in writing assistance, data science, and business tasks with the company in early reports noting better consistency, instruction following, and reduced hallucination compared with g p t five.
Jordan Wilson [00:29:53]:
So the company has not yet published it well, really anything, but especially nothing on the, technical spec side for g p t five one pro. But the public notes, emphasize better clarity, relevance, and response structure. So I'll tell you this. I can speak a lot personally because I am always using the pro model and I'm on the, the pro plan as well. I've been using GPT five one pro. I haven't been able to use it a ton just because like we said, it's been literally a wild week, in AI, but it is extremely impressive. And I'll say this, I would not be surprised when and if GVD five one pro gets benchmarked because it hasn't been benchmarked yet because, they haven't released API access. You can only access it on the pro, model.
Jordan Wilson [00:30:47]:
I would assume it's going to on most benchmarks, it's going to be up there with Gemini three, or it maybe even might leapfrog, Gemini three. The pro versions of OpenAI's models are ridiculously good. Another thing to keep in mind y'all, you really should be using these models. Well, if you really wanna understand the capabilities, you should be using them either on the API side or on at least on the playground side. Right. Because as an example, Gemini three is way better, way better inside of AI studio than it is inside of Gemini, the app. Right. Technically, Gemini three is wrapped in, the the model is wrapped inside the Gemini app.
Jordan Wilson [00:31:36]:
It is much more powerful in AI Studio. I'm using AI Studio, anytime I'm just working with text only. For me personally, I'm always in AI Studio. If I want nano banana or canvas, right, then I'm probably using the Gemini app, or, you know, Gemini three in the Gemini app. But GPT five one pro, it is so good. Right? We shared in our newsletter, there's been some, apparently some, math problems that were solved that couldn't have been solved before. Some new some some new, theories, that were validated by scientists that had never technically existed before. So this model is really, really good, but this isn't a model you're gonna be like, yo, like, I don't know.
Jordan Wilson [00:32:23]:
Here's the spreadsheet. Tell me something good. Right. Which is a lot of times my use cases, but just the agentic, capabilities and the ability for, GPT five one pro to think about thinking. And again, the, the difference between five and five pro, night and day. Right? The five one thinking models are really good. I enjoy them aside from the, obnoxiously annoying, tone and responses that are baked into five one. I know you can, you know, custom instruction them out.
Jordan Wilson [00:32:58]:
It's just really really, responses are annoying. Aside from that, the five one thinking models are really good, and the five one pro model is great. Again, for me, I don't even really, I think have, the type of use cases that I can even say, hey, here's a great example of five one pro. Aside from just showing off some agentic capabilities, the ability to work on very complex tasks, do reasoning, do research, you know, pull multiple, you know, tool calls together, etcetera. But, GPT five one pro, a model to keep an eye out on, especially when and if it's released in the API and gets benchmarked. Last but not least, you, you, you all know I saved the best for last. I'm telling you what's gotten the most talk. It's not chat GPT group chats.
Jordan Wilson [00:33:47]:
It's not GPT 5.1, which again, OpenAI didn't even put out a blog post. It's not even Gemini three. Most talked about thing in AI this week by far is Nano Banana Pro. All right. And let me first, before I get, give you all the newsy details here, let me tell you this. Stop. Like, I've seen so many comments, even, you know, people I put out a poll, in our newsletter last week asking people what they wanted to know about. You know, people said, you know, I don't care about Nano Banana Pro.
Jordan Wilson [00:34:19]:
Right? My company doesn't do visuals or I'm not a graphic designer. You guys nano banana pro is mind bogglingly good, right? Even if you aren't, maybe publicizing something, the use cases are tremendous. They have me speechless. They're that good, especially when you pair them with notebook l m. Alright. So that is our final news story, kind of the combination here of Nano Banana Pro and NotebookLM. The two of them together, my brain. My gosh.
Jordan Wilson [00:34:56]:
So Google's NotebookLM now integrates the Nano Banana Pro image model to generate infographics and slide decks directly from whatever notes that you have inside of notebook l m. So notebook l m is officially using the more powerful nano banana pro to produce infographics and slide decks that aim to render accurate, grounded information and readable text inside images, improving the notebook to presentation workflow. So you can now design complex publication ready visuals without leaving the notebook l m interface, turning your research notes or whatever you load into notebook LM into Polish, arguments with minimal manual layout work. Nano Banana Pro also keeps character consistency, and coherence across wildly different styles. The infographic feature is absolutely. Oh, gosh. I was trying not to say absolutely bananas. It's just absolutely jaw droppingly good.
Jordan Wilson [00:36:01]:
The infographics are great. Much, much better than what we saw out of nano banana normal. And one of the biggest, I think, jumps here is just the text consistency. Right, for some of my testing, I just would throw entire podcast transcripts and say create an infographic, create a whiteboard visualization. Right? Create, you know, a series of, you know, six, you know, breakaway points and visualize them. The text, I haven't gotten a single piece of text wrong yet, which with nano banana, the original, I would say, I don't know, 20 to 30% of the text. Maybe not that high, maybe, 10% of the text wasn't even usable. Right? Nano Banana Pro is amazingly accurate when it comes to using your text.
Jordan Wilson [00:36:52]:
Right? So now that makes these new features inside notebook LM, specifically infographic production and slide deck creation. My gosh. Right? We already saw this out of Gemini, and we covered that, about two weeks ago. So Gemini, released a new slide feature. So I'm gonna be retesting that, you know, and seeing if that has improved at all. And if it has the nano banana features in it, but you can now create these slide decks inside notebook LM using nano banana. And they are extremely, extremely well done. Right.
Jordan Wilson [00:37:31]:
And it's, it's not all, you know, cutesy, artsy. No, I'm talking like very professional looking something like, oh, I paid a consultant seat to put all this information together. That's what they look like. Alright. So, enough of me gushing, on Nano Banana, inside of No Pig Ellem. Just know it's there. It's good. So some more details.
Jordan Wilson [00:37:56]:
Infographic outputs are suitable for really anything, but using them for featured images in your blog, YouTube thumbnails, quick visual summaries that reduce manual design work for creators and communicators. Also, slide generation creates multi page slide decks that right now you can only export as PDF. Unfortunately, you can't, export them as, you know, for Canva or PowerPoint slides yet, but it does, have the ability to export as a PDF, and you can also have presenter slides in there as well. So both features are intended to streamline workflows by replacing part of the repetitive slide in graphic design tasks normally handled by humans well with AI in Nano Banana Pro. Alright. That was a ton. Like I said, a wild week in AI, but that's not all. Here's our quick bullet points of what's new and what's next in y'all.
Jordan Wilson [00:38:53]:
In previous weeks, a lot of these stories might have been the biggest story of the week, but not this week. Alright. Ready? So, OpenAI and Foxconn announced an AI partnership. So the, Apple hardware supplier, that one will be interesting to watch. The new Gemini three agent launch, yeah, with advanced reasoning in Gmail calendar web access. I can't wait to try that. That's only available right now for Ultra subscribers, unfortunately. And a report from the information show that Sam Altman told staff that Google's AI surge is a headwind, not a knockout, but reportedly, it was enough that OpenAI, and Sam Altman had to address.
Jordan Wilson [00:39:35]:
Yeah. Google is really good. Nano Banana Pro. Yeah. More Nano Banana Pro updates here. It's available everywhere. It is now live in flow for AI Ultra users. It's also been added to Nano Banana Pro, has been added to Google searches AI mode, and it's also here's a big one, integrated into Google Ads.
Jordan Wilson [00:39:57]:
My gosh. It's everywhere. Alright. Speaking of other releases, well, XAI, released Grok 4.1. Perplexity launched, some new updates, to their Comet browser mainly by making it available for Android, so the mobile version. Poe, which we don't talk about a lot on this show, but thought it's worth noting that they introduced group chat as well, with real time history syncing. Google has reportedly been testing ads, in the AI search, so in the, AI mode. Perplexity is reportedly building a meetings hub, which combines all your call notes, schedules, and summaries.
Jordan Wilson [00:40:40]:
Google rolled out their Gemini agent, like I said, for ultra subscribers, had that on there twice, maybe to remind myself that I should review it. AI legend, Yann LeCun from Meta is reportedly leaving Meta for a new startup. One of the godfathers of AI, CloudFlare acquired Replicates merging the biggest AI model catalog with their global network. Warner struck a deal with the AI music company, UDO. Speaking of AI music, Suno raised $250,000,000 at a $2,400,000,000 valuation. Former board member Larry Summers quit the OpenAI board amid scrutiny over the Epstein, Jeffrey Epstein email revelations. Perplexity partnered with the federal government to provide enterprise, pro access for free. And then Google released anti gravity, their multi agent IDE.
Jordan Wilson [00:41:40]:
My gosh. A ton happening in the world of AI news. Alright. I hope this is helpful. Like I said, we do this almost every single Monday, go over all of the most important AI news, try to give you no BS, no marketing, no fluff, no spin. Just tell you here's what's new, here's what it means, and here's where you should be focused on. Alright. I hope this was helpful.
Jordan Wilson [00:42:04]:
If so, tell someone about it. Alright. Don't let this be your little secret. The only thing this everyday AI thing lives on, right, making AI understandable for nontechnical business leaders like you and me, is if you tell someone about it. So please, if you're listening, on LinkedIn, still listening. Thanks. Make sure to repost this. If you're on the podcast, please make sure to subscribe to the show.
Jordan Wilson [00:42:26]:
Leave us a rating. That would be great. Again, Wednesday, we're going hands on AI at work on Wednesdays with Gemini three. And then because of the holiday Thanksgiving holiday here in The US, there will be no show or no newsletter on Thursday or Friday. So thank you for tuning in. Hope to see you back tomorrow and every day for more everyday AI. Thanks, y'all.