In this solo episode of the AI Cookbook podcast, Jeff reviews the major AI announcements from Google I/O, highlighting Gemini 3.5 Flash and Omni as the fastest, most capable multimodal models, and previewing Gemini Spark, a personal AI agent that automates tasks across Google’s ecosystem. He stresses that AI is now embedded in YouTube, Gmail, Docs, and search, declaring a shift to AI‑first search that forces businesses to adapt their SEO and geo‑targeting strategies. Jeff then discusses Anthropic’s recent $900 billion valuation, profitability, and its competitive rivalry with OpenAI and Google, warning of a possible market consolidation. He covers Claude’s new Opus 4.8 release (Claude Mythos), noting its coding strengths and recent security improvements. The episode turns to China’s rapid model development—Kimi, Deepsea, Quinn, GLM 5.1—emphasizing their cost‑effectiveness and open‑source availability, which could dominate lower‑budget markets. Jeff predicts imminent U.S. government action: executive orders and pre‑release safety requirements, with agencies like the NSA already evaluating models. He concludes that AI is a platform war, with massive financial stakes, urging listeners to experiment with the new capabilities to stay competitive and grow their businesses.
Chapters
Welcome to the AI Cookbook podcast, This Week in AI. Uh, this is Jeff. Nathan is on vacation with his family in the UK, so it's just me today. Uh, so let's get in it right away, and hopefully, uh, this is entertaining and informative for you. So the big news, and it's really from last week, but Google I/O, um, was huge and it took over everything. Uh, Gemini 3.5 Flash and Omni are now the most fastest, most capable models yet. Multimodal reasoning built in. Been playing with it this week, it's pretty interesting. Uh, we were told Gemini Spark's coming. Uh, it's a personal AI agent which automates tasks across the entire Google ecosystem. This is in response to where, um, Anthropic CoWork had been going and things like that. But perhaps the largest information here is the Ask YouTube, Gmail Live, and Docs Live. AI is now woven into everything, including search, and we're now fully AI search first, which means that everything— and anybody who hasn't done anything about geo or AI search or AI AEO or whatever you want to call it these days, you're going to fall behind right away. So, and, and honestly search as we know it is no longer there. So what we've been doing for SEO is now massively changed. You need to get on board, whether you're local or you're, or you're national. You need to get on board with the new geo capabilities and get that done for all of your clients as well. Anthropic continues to grow. I talk about this a lot. They just completing a funding round at $900 billion valuation. They are profitable, um, and they are moving forward very, very fast, which is, which is, you know, what we've been saying. There is a war between them and OpenAI and Google and things like that. I don't think that, that all three are going to live. Um, I think that, you know, if you look at all the frontier models that we have right now, there's already You know, Llama and Facebook are falling behind. XAI is already aligning with Anthropic. So I think what you're going to see as we go forward is, you know, there could be a bubble, and this is not financial advice, blah, blah, blah, blah, blah, but, you know, there could be a bubble here, and usually there's— that won't stop the progress of what's going on, but there may be a sacrificial lamb here. Well, you can decide who that's going to be. You know, getting back to Claude Mythos, uh, they— Claude announced and released Opus 4.8 this week, which has a lot of the Mythos stuff in there. This was a really powerful model. I've been using it already for coding. Um, it's really good and really, really thoughtful. Um, I've seen a massive improvement in what I've been able to create and work with on this. So, um, And I think we're starting to see Claude Mythos spill into the, the Claude capabilities now, but they definitely, uh, did that after pulling back some of the cybersecurity issues that were in there. But we'll see if we ever get this full model out there. I'm sure it's getting close to that spot. Uh, China has been interesting in making a lot of moves. Kimi, Deepsea, Quinn, GLM 5.1 are all making massive improvements. Shouldn't be a surprise to anybody. We still have a compute and we still have about a 6-month lead on them, but they're moving fast and bringing it together. This matters because it is the fastest growing segment of AI. And while it may not be able to, you know, hit here in the US as most people will use US models worldwide. These Chinese models will start to dominate because they're cost-effective and they're, they're killing us on cost and accessibility. So, and you can download these and build your private models and all that kind of good stuff. So, um, it'd be interesting to see where this goes, but I think, you know, worldwide, um, there is competition out there from China without a doubt. And the government now is waking up and starting to come in. Uh, you can expect an executive order coming, uh, FDA-style pre-release safety requirements planned for most models. Google, Microsoft, and XAI have already agreed to give the government early access. We'll see if Anthropic does it or OpenAI, but they may be forced into it. Um, there's already been a lot of pre-release model evaluations from the Department of Commerce, and, and NSA is probably going to be the one that's leading all this as well. Um, but you know, this is an era of, of freely putting AI models out there become way too powerful for the government not to pay attention to it. So I would expect that we're going to have the government having to pre-clear what's available. Uh, hopefully this doesn't slow down progress and and slow down commerce on this. But we'll see how that plays out as we go forward. And then finally, you know, what this tells us this week, um, AI is a platform war. That's not new. But wow, um, what's cool from a consumer standpoint of this is we're getting really amazing capabilities, um, and we're able to pull those in to our businesses and make massive impact on our business when we do these things. Safety is real, and the government's starting to wake up and get in there. The money in this is unfathomably enormous. I mean, we're talking trillions, which is hard to even really get our heads around how much money is out there on this. But the fact that, you know, Anthropic's almost a $1 trillion valuation and their first profit signalize this is starting to mature and become very high stakes. Uh, and China is winning open source, and, you know, they may, may be formidable across the, across the world. If not, you know, competing here in the U.S., they're definitely competing in other countries, um, that, you know, need cheaper models and don't have the same kind of income to afford $100 a month model. So interesting week as always, um, a lot going on. If you're not in here and you're not testing this stuff, stop falling behind. And I, and I don't mean that as a warning. I don't— I'm just saying that this has been a godsend for my business. I've been able to make a lot of money off of, uh, being able to put this into my business and service my clients. Not just— I mean, it does better work when I, when I put it in there, and it does it faster and allows me to service more clients, um, in my own business. So, you know, don't hesitate, get in there and start playing now.
