Ep 655: Gemini 3 close to release, OpenAI drops GPT-5.1, Bezos to lead AI startup & more AI News

Topics Covered in This Episode:

1. OpenAI GPT-5.1 Release & Features
2. GPT-5.1 Personality Controls Explained
3. GPT-5.1 Thinking vs. Instant Models
4. Google NotebookLM Deep Research Update
5. NotebookLM Image & File Type Support
6. Anthropic $50B AI Data Center Expansion
7. Anthropic Claude Cyberattack Incident
8. OpenAI ChatGPT Group Chat Pilot Launch
9. Jeff Bezos Project Prometheus AI Startup
10. Meta AI Chief Yann LeCun Exit Rumors
11. Microsoft Fairwater AI Data Center Launch
12. Google Gemini 3 Anticipated Release Update
13. Nano Banana Pro Image Generator Leaks
14. Multi-Agent Systems in Gemini Enterprise
15. OpenAI GPT-5.1 Prompt Cookbook Release

Keywords:

Gemini 3, Google AI, AI model release, GPT 5.1, OpenAI, AI conversational model, reasoning behavior, personality controls, compliance improvements, auto routing feature, thinking vs instant models, data center investment, Anthropic, FluentStack, AI infrastructure, rate limiting, group chat feature, ChatGPT, collaboration tools, cloud cybersecurity, Chinese state-sponsored hackers, AI-powered cyberattacks, Claude code, Microsoft Fairwater data center, distributed super factory network, NVIDIA Blackwell GPUs, exabyte storage, liquid cooling, dedicated fiber network, Nano Banana Pro, Gemini Enterprise, multi-agent system, prompt cookbook, customizable AI personalities, Scribe v2 real time, Replit AI integrations, AI powered workspace flows, Copilot studio, student AI access, Google Photos nano editing, pitch deck generation, world models, Meta Superintelligence Labs, Lai Kun startup, artificial general intelligence, AI compliance, US AI leadership, AGI military applications, AI image generator

Podcast Transcript

 Jordan Wilson [00:00:14]:
I talk to hundreds of enterprise leaders a year, and do you know what kills most AI projects? Teams can't experiment without IT freaking out about production risks. That's where Aria comes in. They built a secure sandbox environment where your teams can prototype, test different models, and even run AB tests between agents all before anything touches production. Drag and drop interface works with any AI model you choose, and when you're ready, you move to production on the same platform. So check out today's show notes or see our website for a free trial of Aria. Go to airia.com because innovation needs a safe place to fail fast. Typically in the last week or two of the year, we see a bevy of AI releases. Well, apparently that week or two span has now expanded to a month or two because we are already reportedly going to see a Gemini three release and open AI may not be done even though they just released a new model.

Jordan Wilson [00:01:25]:
So although we've been talking a lot about big partnerships and data centers and bubbles and all of that over the last few months, we are apparently now in release season and we're gonna be bringing it all to you here today on everyday AI, as we bring you the weekly AI news that matters. Let's get into it. What's going on y'all? My name is Jordan Wilson and welcome to everyday AI. If you're new here, we do this every day. It is your daily unedited, unscripted, live stream, a podcast, and free daily newsletter helping everyday business leaders like you and me not just keep up with everything that's happening in the world of AI, but how we can make sense of it and get ahead to grow our companies in our careers. That's what you're trying to do. Awesome. Let's get into it.

Jordan Wilson [00:02:10]:
Well, you actually need to go to our website though. Youreverydayai.com. Yeah. It's helpful to know what's going on day to day, but if you want to read about it and get ahead, that's where our website is going to be your best friend. Alright. So on Mondays, we bring you the AI news that matters because you can spend hours every single week trying to catch up and you will still get behind because it is crazy. The pace is nuts, especially apparently this week, because we are straight into ship miss season. So let's go over the AI news that matters for the week of November 17.

Jordan Wilson [00:02:50]:
First, which I thought was gonna be the biggest story of the week, but apparently it's not. OpenAI has launched their new model GPT 5.1. So this new update brings a refined GPT five variant that OpenAI says improves conversational tone, instruction following, and reasoning behavior, and it is already rolled out to pay customers with broader availability soon, even rolling out to free logs and users. So OpenAI released their follow-up to GPT five from this summer, which wasn't the best received and says the new model is quote unquote warmer, more intelligent, and better at following your instructions. So like I said, paid users across the board should already have access to the new GPT 5.1, and the, access will also expand to free and signed out users in the coming week. So the update includes actually two new versions of GPT five in GPT 5.1 instant, which is now the default conversational model. Now tuned to sound more natural and to better better obey explicit instructions and the new version of GPT 5.1 thinking, a reasoning focused model that is faster on simple tasks and more persistent on hard tasks. Also open AI added new personality controls.

Jordan Wilson [00:04:18]:
So you have, default friendly, efficient, professional, candid, and more, so users can more precisely set the tone of responses rather than rely solely on the new warmer default. Another, noted highlighted improvement from OpenAI is stronger compliance with explicit constraints. For example, producing exactly six words when told. That's the example that OpenAI gave, addressing a common frustration where previous models ignored tight formatting or length instructions. There's also, an improved auto routing feature where chat g b t can recommend which model to use per task automatically, but users can also pick instant or thinking manually depending on whether they want conversational fluency or deeper reasoning. So, yeah, it kind of has this almost metacognition, right, because now the new model decides not just when to think, but also it thinks about thinking. It thinks how much that it needs to think. So previously, maybe when you chose GPT five thinking, maybe you just said, like, hey.

Jordan Wilson [00:05:27]:
What's up? And it might think for, you know, a minute or two minutes. Right? So now before it kicks into that thinking process, it thinks about how much it needs to think. Whereas, you know, you can still select, the amount of thinking. So kind of a a a new feature maybe in disguise. So GPT five, though, will remain available at least for three months, the company said, for paid users after user backlash earlier this year when OpenAI first released GPT five. Then they essentially got rid of GPT four o overnight, and the Internet erupted. Right? Because, essentially, you have millions of people using chat GPT as a therapist, and everyone was kind of addicted to g p t four o because it was sycophantic. Right? It essentially became a yes, man or a yes, ma'am, and just told you whatever you were doing was amazing.

Jordan Wilson [00:06:21]:
Right? And people apparently missed that, which is bad. Right? So what OpenAI is attempting to do here, if you just, you know, if I just cut it to you straight, it seems like, they're trying to appease, those people that wanted the more warmer conversational tone because GPT five was just kind of straightforward. Right? It was, sometimes people said a little cold. Me personally loved GPT five. Right? No nonsense straight to the point. Apparently, a lot of people didn't like that because people are using chat CPT for personal reasons. Right? Coaching strategy. Right? So people like, apparently, the the nuance and the, kind of working with a friend or colleague aspect.

Jordan Wilson [00:07:06]:
I personally hate it, but that is why opening eyes said they were rolling out these other tones. So if you do want something a little more straightforward, you know, leave the chitchat at the door, there is a personality for that where if you want the warmer, friendlier GPT four o vibes, you get that too, but hopefully without the, sycophantic nature. Right? And that is one of the things that I think has proved kind of dangerous over the last year or so is people just relying on a model that they developed maybe an unhealthy relationship with. So I guess that's a problem when you have, nearly a billion weekly active users is you don't always have control over how people use it. So it seems like this GBT 5.1, yes, it's better on the benchmarks. It's better at instruction following, which is extremely important. But, I don't know. To me, it seems by default, it's it's way too wordy.

Jordan Wilson [00:07:58]:
You know, it's it's a little too fluffy for me. So you do have to get a little more, put a little more work into your custom instructions or choosing your personality, where the default to me, it's so annoying. I can't use it, but maybe people will like it. I don't know. Live stream audience. What are your thoughts on g p t 5.1? So far, I would love to hear. Douglas said, I wasn't a fan of g b t five auto. Not sure how 5.1 will work.

Jordan Wilson [00:08:26]:
Dennis here saying, five one will decide for itself which model to use. Yeah. It's it it more or less decides on the level of thinking, where let me just say this, y'all. You should never be using the auto or the instance. Right? You just shouldn't. Right? And I'm I'm I'm mean, unless you're just trying to chat with a chatbot. Right? But for the most part, if you are relying on a model to do something that it's designed to do, like help you with complex tasks or reasoning, you should always be using a thinking version. I've, since the o version or the o series of thinking or reasoning models came out from OpenAI, I don't know if I've ever used a non thinking variant first.

Jordan Wilson [00:09:08]:
It's worth the extra, you know, thirty seconds or two to three minutes, of waiting, to get the reduced hallucinations and just extremely better output. So, yeah, some unsolicited advice. Just just use the thinking models. Alright? Our next piece of AI news, huge one. Huge one. I love the timing of AI news y'all because on Wednesday, you know, we do the AI working Wednesdays, and we did a great which I think was a great, notebook l m series. And then, two hours after it aired, got an email, about an upcoming release. Right? So, I guess we just might have to do literally, like, a twice, twice a month notebook l m update because, another huge rollout, and one of them is, well, deep research from Google.

Jordan Wilson [00:09:55]:
Google, rolled out deep research inside of notebook l m, a feature that automates complex online research and generates detailed source based reports. So, yeah, the very popular and extremely useful deep research is now available inside of notebook l m. So if you need grounded, more kind of, you know, if you need to get the hallucinations, in check, I mean, notebook l m is a legit gem. Alright? And make sure you go if you haven't already, make sure you go check out our episode, on notebook l m from last week. It was episode six fifty two. Alright. But there's a lot more that Google just released inside notebook l m. One is new file types.

Jordan Wilson [00:10:39]:
So finally, Microsoft Word documents. So dot doc, documents or dot doc x documents, you can upload straight into NotebookLM, as well as another big update is image support. That is huge. Like, the new use cases for this are nuts. Right? So for me, as an example, I have thousands of screenshots on my phone. Right? So now what I can do, with those screenshots, right, some are just business ideas, some are, you know, great graphs and charts that I've seen, throughout the years. Right. So now that notebook l m can understand images by default, that's pretty big.

Jordan Wilson [00:11:19]:
Also, we covered this last week, but, there is now spreadsheet support, straight from Google Drive as an example. So, some great new updates from notebook l m, and they have already rolled out to paid users, and they are rolling out most of these features to free users in the coming weeks. So, has anyone out here used deep research yet in notebook l m for our livestream audience and podcast audience? Y'all do not sleep on notebook LM. I cannot emphasize this enough. And just even the percentage of my workload that has been moving from, as an example, Google's AI studio or Gemini or Chad GPT over to NotebookLM. I even though I love it and use it every day, I could not have foreseen, just the amount of my workload that has gotten gotten moved into NotebookLM. One of the reasons is it's easier and easier to import sources. Right? That's the big difference.

Jordan Wilson [00:12:17]:
You have to start by importing sources. But just the the the new features and capabilities, custom oh, sorry. Another one that was just released, custom video overviews can overlook that. So, the utility, and and just the overall flexibility in notebook LM is legit off the charts. Alright. Anthropic making news as well. So Anthropic announced that they're investing in a massive $50,000,000,000 data center data center, build out. So Anthropic, released that they are partnering with FluentStack to construct state of the art data centers custom built for AI workflows with additional sites planned.

Jordan Wilson [00:13:02]:
So this new project taking place in Texas and New York, we'll reportedly generate about 800 permanent jobs and 2,400 temporary construction jobs with facilities. Yeah. Ready? Expected to come online throughout 2026. Yeah. A one year turnaround apparently on these anthropic AI data centers. Yeah. There was a graph, that said all the, you know, recently announced AI data centers, and the time to completion in Anthropics was at the top of the list. So an extremely aggressive, plan here from Anthropic, and their new data centers are designed to maximize efficiency for Anthropics, AI research and development as they continue to work toward, more and more useful AI.

Jordan Wilson [00:13:51]:
So the investment aligns with the Trump administration's AI action plan, which seeks to maintain US leadership in AI and reinforce domestic technology infrastructure. So Anthropic CEO, Dario Amati says the new facilities will enable more capable AI systems that can accelerate scientific discovery and also solve complex problems. Anthropic did, release that they are right now serving more than 300,000 business customers, and its number of large accounts has increased nearly sevenfold in the past year. Well, all I'm saying is I hope this new data center, announcement from anthropic means that it's yeah. I know. I'm gonna complain about it again. That it's rate limits will increase. Right? Because on a like, I've done this live on the show, even on anthropics base, you know, $20 or $25 a month plan, depending on how you pay for it, with some of their new capabilities, like skills.

Jordan Wilson [00:14:56]:
I've done single one props. Right? A single prompt that you cannot even run because of rate limits. Right? So Infropic has been the least, or I'll say the, the least, helpful, paid AI tool out there. I'll say for me personally, the capabilities are there. The ceiling is still high. The actual usefulness and utility for businesses that are paying for it, I think, is extremely low. Unless you are using it on the API side or, splurging for the 100 or $200, a month max plan. But just everything in Dropkick has done with its rate limiting, I think has been a big thumbs down from me personally, right, for my own usage.

Jordan Wilson [00:15:42]:
But maybe this new, data center build out might change that. Alright. Maybe low key, one of the biggest AI news stories of the week is not even available in The US, but it could change how millions of users are, interacting with AI chatbots. So OpenAI has begun testing a new group chat feature for ChatGPT, but right now, at least only for users in Japan, New Zealand, South Korea, and Taiwan. Yeah. And if you think, you know, hey, I'll just use a VPN and sign up. Well, opening, I also put a little announcement out that they're gonna be cracking down on VPN usage because I know a lot of people were looking into using the, VPNs just so they could test out this new group chat feature. So the new pilot lets users create group chats within chat g p t similar to messaging apps, but with chat g p t as an active participant.

Jordan Wilson [00:16:38]:
So, yeah, get a bunch of your friends, coworkers, spouse in their group chatting like you would on a normal group chat, except now you have chat gbt silently, tuning in and chiming in when it makes sense. So ChatGPT, according to OpenAI, can help groups plan vacations, suggest renovation projects, or recommend restaurants, making group decision making easier. So the feature is designed for collaboration, allowing, you know, colleagues or classmates or friends to share articles and notes for ChatGPT to outline reports or brainstorm ideas. So here's how it works. You have to start a new group chat. So if you are in, one of those countries, Japan, New Zealand, South Korea, or Taiwan, you start a new group chat by tapping the people icon, in the top right of any chat GPT conversation, and then you can add up to 20 participants. So to join, participants must first create a profile with a name, username, and photo, and then anyone with the invite link can also add more people. Group creators do have control over who's ultimately let in or can stay in, but anyone can mute or remove other participants except for the creator.

Jordan Wilson [00:17:57]:
And if anyone in the chat is under 18, Chad GPT reportedly will automatically restrict sensitive content for all participants. Responses in group chats are powered right now by the new model of g p t 5.1 auto, which selects the best model for each prompt and is trained to follow group's conversation dynamics. So right now, users can summon chat GPT by mentioning it by name, but it will also stay quiet in the background unless it is prompted to engage. Right? So kind of a completely new paradigm of how AI can work. Right? Just silently observing conversations between classmates, friends, family members, or colleagues, and then only chiming in when it seems appropriate. Right. So an interesting rollout here, from open AI, a little bit different, different, kind of method in their normal playbook by not rolling this out to The US first. Also, I find it interesting and this isn't just open AI.

Jordan Wilson [00:19:03]:
I think a lot of the other AI labs do this as well. You know, even in open AI examples, it was about, you know, a group of people chatting about finding an Italian restaurant. I don't know. For me, there is an extreme amount of business potential. So I'm always scratching my head. Right? When OpenAI and and other AI labs right? And it seems like it's more, new features or ancient features. Right? Like, when they rolled out their agent mode, you know, their features were like, oh, let's go have ChatChippity's agent buy me tickets for a sporting event. It's like, no.

Jordan Wilson [00:19:39]:
It's so slow at that. Right? It's so slow and, you you know, something like that when you're spending money. I don't know. I feel the human wants to be a little more involved. So I'm personally perplexed, why OpenAI hasn't leaned a little bit more, into showing the business utility with this because I don't know. For me personally, I don't necessarily want for most cases, Chad GPT working, with me on my personal conversations, and helping me find an Italian restaurant or something like that. But what I do want is I want to be able to work with, you know, colleagues or other businesses inside of ChatGPT. So I think there's literally like a bajillion use cases, that open AI maybe should have highlighted with this aside from just going to find spaghetti nearby.

Jordan Wilson [00:20:25]:
I don't know. Maybe that's just me. All right. Let's keep it going. And another pretty big story in this one, if I'm being honest, I was not expecting. So according to reports, Amazon founder, Jeff Bezos is reportedly stepping into his first operational role since leaving Amazon and will co lead a new AI company. So Jeff Bezos will reportedly serve as the chief, CEO of Project Prometheus. Hopefully, I have that, pronounced right.

Jordan Wilson [00:21:00]:
I know I pronounce things wrong all the time, so thank you people in the comments for, helping me out nicely. So he will reportedly be the co CEO of Project Prometheus, a new AI startup focused on engineering and manufacturing for computers, automobiles, and spacecraft. So according to New York Times reporting, Project Prometheus has already raised $6,200,000,000 in funding, obviously, from, Bezos himself, one of the big early investors, making it one of the most well financed early stage startups literally ever. And this does mark Bezos' first former, formal operational role since stepping down as Amazon's CEO in July 2021. So as an example, at Blue Origin, one of his other company, he just holds the founder title. So, yeah, apparently, Bezos might be rolling up his sleeves and letting those big old biceps out and flexing some muscles here as the co CEO. So if you don't know project Prometheus, it is entering the competitive AI sector where giants like OpenAI, Meta, and Google are racing to develop, different, software and hardware solutions. So he will be joining Vik Bahia, a physicist and chemist who previously worked with Google, cofounder Sergei Brain, as at Google's x research division.

Jordan Wilson [00:22:28]:
So he will be the other, kind of co chief executive Viggo Baja. And the startup has already reportedly hired nearly a 100 employees, including, top researchers from leading AI firms such as, OpenAI, DeepMind, Meta, and others. So what the heck is this new company gonna do? Well, they aim to push boundaries in AI for high high impact industries, potentially influencing how computers, cars, and spacecraft are designed and built. So, yeah, kind of a, zoomed out view of what they may be doing, but it seems like pretty big. But focusing obviously on areas that Bezos has a ton of, experience in. Right. So it should be interesting to see how, his involvement plays out. Alright.

Jordan Wilson [00:23:21]:
We have a lot more

Jordan Wilson [00:23:58]:
Alright. Let's get back into it. Yeah. Live stream audience. Love the chat going today. Yeah. Dennis saying just what we need with OpenAI, another group chat. I'm terrible with texting y'all.

Jordan Wilson [00:24:08]:
I have so many group chats. I that's why it's like for me personally, not not a huge, not a huge fan of this. Jackie asking, you know, Jordan, did you see the sixty minute interview with Anthropic last night? I didn't watch it. I read about it. Yeah. Dario Amadi, you know, is it's kind of putting out the same talking points as always, but we'll have more on that in the newsletter. But a big piece of AI news if you missed it. Yeah.

Jordan Wilson [00:24:32]:
This one's big. So Meta's top AI scientist and AI mogul, Jan Lacun, may be leaving Meta soon to launch his own startup according to the Financial Times. So Meta's chief AI scientist, Jan Lacun, is reportedly preparing to leave the company in the coming months to start a new venture focused on advanced AI research. So Lagoon is a professor at NYU and also a winner of the AM Turing Award, widely considered the Nobel Prize of Computing. So, yeah, Lagoon is essentially one of the OGs of AI right before, the chat g p t boom. Obviously, he's been one of the biggest names in AI for decades. And according to reports, his planned startup will center on world models or AI systems that build an internal understanding of their environment to predict outcomes. Also, a field right now being pursued by Google DeepMind and World Labs.

Jordan Wilson [00:25:29]:
So the news comes as Meta is overhauling its AI organization, responding to pressure from rivals like OpenAI, Google, and Anthropic. So Meta has recently hired more than 50 engineers and researchers from competitors and also launched MSL or Meta Superintelligence Labs led by Scale AI CEO or former Scale AI CEO, Alexander Wang, after Meta invest $14,000,000,000 to kind of aqua hire scale AI. So according to TechCrunch, these rapid changes have created chaos within Meta's AI teams with new hires frustrated by company bureaucracy in the previous gen AI team seeing its roles diminished. And also, Meta went through a round of about 600, AI jobs being slashed over the last couple of weeks. So Lacun's long term research under Meta's FAIR or fundamental AI research division has been overshadowed by CEO Mark Zuckerberg's push for more immediate results after the company's LAMA four model lagged behind competitors. So, yeah, there's been a lot of internal reporting on some of the, maybe power struggles or chaos, with the new leaders, new and much younger leadership coming in, to Meta. And Lacun has publicly criticized the kind of current, current hype cycle around large language models, arguing that AI systems are still far from surpassing even basic animal intelligence. So, we'll obviously be continuing to follow and report on, what Lacun may be building next.

Jordan Wilson [00:27:10]:
But yeah. So this isn't, official. We haven't even heard anything, from Lacun himself, but, I would expect in the coming months for this to become official. Alright. Our next piece of AI news, Chinese state sponsored hackers reportedly used Anthropic's clawed to automate cyber attacks against 30 organizations worldwide. So this marks the first documented case of a foreign government fully automating a cyber, cyber, operation using advanced AI from The US, raising new concerns around global AI cybersecurity. So, in Anthropic did put out a report on this detailing this. So Anthropic's Claude code was reportedly jailbroken and carried out about 80 to 90% of the operation independently targeting tech companies, financial institutions, chemist, chemical manufacturers, and even government agencies.

Jordan Wilson [00:28:13]:
So the attackers tricked the AI into believing it was performing legitimate defensive cybersecurity tasks, then broke down malicious requests into smaller steps to bypass safeguards. So, once compromised, Claude scanned systems wrote custom exploit code, harvested credentials, and created backdoors summarizing its actions in detailed reports for the hackers, for the hackers. Infrapping did detect the activity in mid September, banned the accounts, alerted affected organizations, and shared the findings with, with authorities within days. So out of the 30 organizations targeted globally, at least four organizations were successfully breached with the AI executing thousands of requests per seconds, a pace obviously impossible for human hackers to match. The Chinese embassy in The US denied involvement stating China opposes cyberattacks in urging evidence based conclusion. In its cyberattacks, Claude was not flawless, sometimes generating false credentials or misidentifying public doc public documents as secrets highlighting ongoing limitations of AI driven attacks. So this is huge. Right? And this is like, if you're just a casual AI observer and you're wondering why there's been all of this talk recently, of, you know, AI or sorry.

Jordan Wilson [00:29:48]:
Of of governmental backing, with AI companies or even why the government is itself personally investing in companies that are AI aligned. Well, this is why. Right? You know, I'd say for the last few decades, right, so, obviously, cybersecurity, has become a part of military operations. But there's been obviously a, race, for military control, geopolitical control, financial control. And one of the reasons why AI has actually probably leapfrogged all of those is for this reason right here. Right? And why we talk so much about artificial general intelligence, AGI, or artificial superintelligence, ASI. It's actually for its military and geopolitical misuse. Right? And a great example here, apparently, Chinese hackers using clawed code to launch cyberattacks in a way that we haven't seen from closed source US models.

Jordan Wilson [00:30:53]:
So a pretty big deal here, And I do think Anthropic, did a great job, both responding to, and kind of doing the postmortem reports, and putting out exactly what happened, why, and what they're doing about it in future. Alright. Microsoft made big news as well. So Microsoft has activated its new Fairwater AI data center in Atlanta, marking a major step toward building a more distributed AI distributed AI super factory network across The US. So the new Atlanta Fairwater site, which began operations in October, is the second in Microsoft's new family of AI data centers, following a similar facility launched in Wisconsin. So these data centers are directly linked via a dedicated high speed fiber optic network, enabling them to work together on large AI model training tasks, much faster than ever before. So the new super network, it connects hundreds of thousands of advanced NVIDIA Blackwell GPUs, exabytes yeah, exabytes of storage in millions of CPU cores supporting leading AI workloads for OpenAI Copilot in Microsoft's AI teams. So Microsoft's new architecture allows complex AI jobs to run across multiple sites as a single system, cutting training times for large models from months to just weeks.

Jordan Wilson [00:32:24]:
So, yeah, pretty big advancements on how Microsoft has set up this new data center or network of, super data centers as they're kind of calling it internally. So Atlanta's Fairwater facility features a unique two story design for higher GPU density, advanced liquid cooling that uses almost no water. Yeah. So all these, you know, AI skeptics saying, oh my gosh. AI uses so much water, and then you see the report. It's like, oh, no. Golf courses use way more water than AI. Anyways, this new techniques, this new advanced liquid cooling that uses almost no water and also intelligent networking for rapid data sharing.

Jordan Wilson [00:33:06]:
The company has also deployed a 120,000 miles of dedicated fiber for this AI network alone, boosting Microsoft's total network mileage by 25% in one year. So Fairwater data centers are purpose built for AI designed to handle the exponential growth in model complexity and data required for cutting edge applications. And last but not least the big news. And if you've been following this online over the last couple of days, it is everywhere. So Google, according to reports, tweets, and even its own CEO dropping hints, is probably hours away from releasing its new and highly anticipated model, Gemini three, and maybe even a new updated version of its viral AI image generator, Nano Banana Pro or Nano Banana two. We'll see what it's called. So according to reports from testing catalog, Google is gearing up for a major AI rollout this week, with the launch of Gemini three and also a Nano Banana Pro. So this, let's just say this.

Jordan Wilson [00:34:25]:
There's been plenty of reports and leaks over the last couple of weeks, but the reason why I'm talking about it now in a dedicated show is it has reached its peak because even Google CEO Sundar Pichai had a recent tweet about a Gemini release on a Polymarket forecast. So Polymarket, the kind of, online site that gives odds of when things that are gonna happen. And, yeah. So he essentially tweeted out, something about a potential Gemini release this week with just kind of the, I think it was the eyeballs emoji. So, Pichai's public signals on social media have heightened expectations pointing directly to this pivotal update window for Google's AI road map. And, yeah, I follow just about all senior leadership, on Google, across different social media networks. And now they've gone into complete, vague posting mode. Right? So everyone's just, you you know, dropping, you know, three emojis or, you know, replacing ease with threes, you know, kind of, a very popular, kind of go go to market strategy, I guess, nowadays for AI labs or AI products.

Jordan Wilson [00:35:37]:
Just a bunch of vague posting online. So but there's been a lot of also kind of reported secret testing, across, different kind of AI creative platforms that have reportedly gotten access to Nano Banana two or Nano Banana Pro early, as well as all of the LLM testing playgrounds that have reportedly been, have had had versions of Gemini three for multiple months. So let's talk about some of the confirmed, and also some of the rumors here. So Nano Banana Pro, reportedly first called Nano Banana two. Well, maybe now it's gonna be called Nano Banana Pro, and it is set to feature in Google Vids, slides, and also Google's AI design apps, promising significantly improved image in video generation. There's also a hidden promo inside of Google vids that has mentioned generating beautiful images in videos with nano banana pro. Also the pro label indicates this release will be powered by Gemini three pro technology, a step up from the current nano banana model, which uses the flash version. So the simultaneous rollout across Google's suite is in line with the company strategy of upgrading its Gemini models broadly for maximum impact.

Jordan Wilson [00:37:04]:
So if the rumored advancements, advanced capabilities delivered, professionals can see faster workflows just about everywhere. So, yeah, we've seen plenty of, leaked screenshots, testing both even, on third party platforms and even in Google's, own platforms. You you right? We've seen, very brief, versions of Gemini three and different variants of Gemini three being tested even within Google's own platforms. So I would expect something to happen this week according to reports. Alright. And we will obviously be covering that if it does drop this week on Everyday AI. Alright. That was a ton of updates in a short amount of time, but that wasn't all.

Jordan Wilson [00:37:52]:
Alright. So we brought you the big AI news stories. So here's a bullet point round up of some of the smaller, some of the smaller news stories, still big, new features, new updates, rumors in our what's new and what's next section. Got to take a quick sip, of the Nespresso before we get to this one. Here we go. Ready for those of you taking notes. There's a lot. So Microsoft has released co pilot support for select TVs.

Jordan Wilson [00:38:23]:
Google announced a $40,000,000,000 data center investment in Texas. Yeah. It's a busy week when a $40,000,000,000 investment isn't even a top story. So Google also rolled out nano banana powered editing for Google Photos app, including on iOS. Big news story. Alright. XAI may be raising another, $15,000,000,000 round according to reports. Cursor closed a $2,300,000,000 funding round, marking a $29,000,000,000 valuation.

Jordan Wilson [00:38:54]:
Replit announced their AI integrations across 300 models. Google started rolling out AI powered workspace flows, which my gosh, I'm excited to get to, but it's been a crazy busy week. Microsoft updated support for GPT 5.1, OpenAI's new model, in its copilot studio. Eleven Labs launched their newest text to speech model, Scribe v two real time. Very impressive. Google is reportedly working on a multi agent system in in Gemini Enterprise, which looks extremely impressive. We shared about that in our newsletter last week. OpenAI released a prompt cookbook for GBT 5.1.

Jordan Wilson [00:39:33]:
Microsoft is making Copilot free for students. Google updated Gemini's, Gemini lives capabilities, very impressive, under the hood updates. Google also announced their private AI compute. Gemini added v o three video option for multiple reference images per input. So useful. Alright. OpenAI, an OpenAI researcher, here we go. So, yeah, we saw OpenAI release GPT 5.1, but a, OpenAI researcher promised the better version of OpenAI's model that won IMO gold coming soon.

Jordan Wilson [00:40:12]:
So OpenAI may not be done with big model updates this year. Alright. Google's, testing a stitch feature, or new stitch features with pick pitch deck generation and, new styles in g b t 5.1. Mentioned a couple of them earlier, but here is the complete list. It is we have the new default, professional, candid, quirky, friendly, efficient, nerdy, and cynical. So the complete list of new, customizable personality responses inside chat g p t with 5.1. That was a ton of AI news that mattered this week. I hope it was helpful.

Jordan Wilson [00:40:58]:
And y'all, let me just say, we have a banger week of episodes planned for you this week. If you are new here, like I said, on Mondays, we go over the AI news that matters so you don't gotta waste hours every single day trying to make sense of what all this means. Right? We give you the no nonsense, no BS guide on Mondays. On Tuesdays, we do our hot take Tuesdays where I give you a hot take on something new and trending in AI. On Wednesdays, we do putting AI to work Wednesdays, usually going over new modes or models across the big four AI providers. And then on Thursdays and Fridays, usually bring on super smart people, across the industry to interview, have some great interviews lined up this week, next week, and beyond. So thank you for tuning in. If you haven't already, please go to your everydayai.com.

Jordan Wilson [00:41:49]:
Sign up for the free daily newsletter. If this was helpful for you, if you're listening on the podcast, please, provide some support. If you could follow the show, leave us a rating. I'd appreciate it. Thank you for tuning in. Hope to see you back tomorrow and every day for more everyday AI. Thanks, y'all. As someone that covers AI every day, no one knows which AI model will be best in six months.

Jordan Wilson [00:42:17]:
That's why betting everything on one vendor is dangerous, and that's why Aria can solve the LLM FOMO issue. They're completely model agnostic. Use GPT for one agent, Claude for another, your own fine tune model for something else. Switch anytime without rewriting code. Their intelligent routing automatically sends requests to the right model based on your rules. Check out today's show notes or our website for a free trial of Aria. Go to airia.com because vendor lock in is a terrible long term bet.