Google Is On Top In AI - DTNS 5152

0:03This is the Daily Tech News for Monday, November 24th, 2025. We tell you what you need to know. We give you important context and we do our darndest to help each other understand. Today, Andy Beach tells us what kind of AI bubble we're in and we talk about how we can tell that Google is now the top dog in AI. I'm Todd Merritt. And I'm Rob Dunwood. Let's start with what you need to know with that big story. All right. So seems like Google's red alert did some good. the release of Gemini 3 has caused everybody to say, wow, they're really good at AI. In fact, they might be the best at AI right now.

0:41For example, Salesforce's Mark Benioff announced on X that he has switched from chat GPT to Gemini, writing, the leap is insane. Reasoning, speed, images, video, everything is sharper and faster. It feels like the world just changed again. If you want something a little less hyperbolic, Weylin Chang, co-founder and CTO of LM Arena, told The Verge, this is more than a leaderboard shuffle. Gemini has clear leads in coding, creative writing, agentic coding, and it's the first model to pass a 1500 score on visual comprehension. On a benchmark called Arc AGI 2, Gemini 3 scored twice as high as GPT-5 Pro and at a tenth the cost.

1:26Now, it's not all blooms on the rose, though. The Verge did talk to folks who were criticizing Gemini 3 for its user interface. Some prefer Anthropic for that. Still not compelling enough for all your edge cases. A lot of people are like, eh, I'm pretty good with what I've got. I'm not going to switch for this. But the stock market has rewarded Google with big jumps in its stock price. It is now well ahead of Apple and Amazon. and an NVIDIA, Microsoft, Tesla, and Meta are all in decline on stock price. So Google making a big jump while others are not even rising. All this to say, Gemini is well-received.

2:04We're not trying to say the race is over though. Expect OpenAI to leapfrog ahead at some point. Expect Claude to leapfrog into this race again at some point. Not that they're out of the race, but it's all gonna keep going head to head. Gemini just has the lead at the moment, which means Gemini has become the target for people who like to bring down the leader. And Thursday, Malwarebytes published an article that referenced a thread on X that alleged that Google opted you into using your Gmail messages to train its models. Google, of course, denied the report, telling ZDNet, we do not use your Gmail content for training our Gemini AI model.

2:42So Rob, not a lot of weasel words there. They pretty much directly contradicted it. Malwarebytes wrote a retraction on the story. It seems that Google changed the way smart compose, smart reply and predictive text are presented and surfaced and how they explain them. But they didn't change the way they collect data. Google does now and has for a while scanned your emails for smart suggestions, as well as things like spam filtering and categorization. That's different than scanning it for training. Meanwhile, The Verge says that one of its staffers is pretty certain they had opted out of smart features only to find them turned back on recently.

3:26That's different from saying they're training on your emails, of course, but it would be a change in your defaults, which would be annoying. And of course, there's a November 11th class action suit that's ongoing, accusing Google of turning on Gemini for all users on October 10th and giving it access to email without consent. and they're trying to argue that's a violation of the California Invasion of Privacy Act. Google's trying to get that thrown out on its merits, saying they're alleging something that just didn't happen. So, Rob, people coming for Google is no new thing, but I kind of see this increased scrutiny as indicative of how well Gemini is doing, actually.

4:03Well, whoever's on top of the hill is the one who gets the most arrow shot at it. So that's kind of what's happening here. But Google is being very clear. Malwarebyte said, this is what Google did. Google said, no, no, we didn't. And kind of waited until you can retract this or we're going to sue you out of existence. That's what that sounded like to me from the way Google responded to them and how quickly they came out with their response and how direct it was. So there's a lot of good that is coming out here. It's amazing. If we just go back to two and a half years, Google was struggling. Not because they were so far behind everybody.

4:41They were just very far behind open AI. And because of the type of company that Google is because of search and what search is turning into, it was an untenable position for them to be in. They could not be not just second, but they couldn't be third or fourth when it comes to this. So it was all hands on deck for them to get this turned around. And it looks like with three that they've done it. Yeah. I thought that Google had a very good shot of not only getting back in this race, but leading it because of their technology. I thought that if they didn't get back in this race, it would be a failure of management.

5:19And it seems like you can't pin it on management, at least not at the current moment. They've been able to take all of that research, that very fundamental research that they have and use it to their advantage. I guess the next question is, can they propel themselves to stay in front? Not even just stay in front, but can they increase the distance? Which I feel like they should have enough internal experience and research to kind of widen that gap a little. I don't know that they will. Google has an advantage that the other players, let's say Microsoft. And Microsoft's not really in this conversation with Copilot.

5:58But Google's advantage to open AI to Anthropic is that it is a trillion-dollar company already. so they can just keep throwing money at the problem. And if there are any problems that arise, hey, how about$70 billion? We'll just buy them. I'm not saying that they've done that, but they can do that. So that quiver or that arrow in their quiver, so to speak, to keep that analogy going, that gives them just advantages that a lot of other companies don't have. Anthropic can't go out and spend hundreds of billions of dollars to buy somebody if they needed to catch up. Google can. And what's happening here, they really didn't have to do that.

6:39It looks like a lot of this is this what they've done in-house by being Google and being very good at these type of things from a technology standpoint and saying, yeah, we were a little bit behind, but we have now caught up. It's crazy to think that we're just now in November. ChatGP3 3.5 came out three years ago. Yeah. It's been three years we've been talking about this. It seems like it's been the entirety of the, you know, of the, of the thousands, but it's been three years. Yeah. Only three years. And, and Google's disadvantage is that it is big and that it has been doing this for a long while.

7:13And, and that is, is can get in your own way because you, you start to put on blinders, you do things the way you've always done them. It took 3.5 to shake Google up and say, you, you've got to stop slow rolling this. You've got to accelerate. And it was, there was no guarantee they would, but they did. And so now we have true competition. That's what you see when there's competition. Somebody figures out a new way to do something and shakes everything up and the competitors have to respond and improve as well. And that's why I think you see a lot of people being like Mark Benioff so effusive because they weren't sure that this could happen.

7:51They saw lightning strike with chat GPT 3.5 and they're like, oh wait, actually Google is making something that has leapt ahead a little bit as well. I'm not sure it's as big of a leap as ChatGPT was when it first came out three years ago, but it does seem like it has improved a lot. Personally, I tried to use the new Nano Banana Pro to combine four of my dogs, three of them are gone, together in a picture so I could get a picture that could never have happened of all four together. It couldn't get all four. It could get three, but then it would always mess up one of the others. So I feel like, oh yeah, it can do three.

8:25Couldn't do that before. that's great uh says it can do five i'm not able to get it to do five so i i feel like that's emblematic of where we are like big leap ahead maybe not quite as big as they say what's been your experience um last week actually you know on on on daily technician live i said that i've downloaded it and my initial thought was that it seems like it's slower but what i've now realized is that I was still using, I was comparing it to 2.5 fast and 3.0 thinking. The thinking is slow because it's supposed to be. It's thinking, it's doing more. But now that I've actually got the 3.0 fast, it's fast.

9:05It does what it does. And I am very impressed by the images that you can generate by just talking to your phone. I mean, I've, you know, I generated an image and I just said, you know, create a woman with a stroller next to an EV at a Midwest shopping center. And it doesn't look like it's AI if you look really closely, but if I showed it to my, you know, to my great uncle or my great aunt, they probably would just think it's somebody standing in the mall. It's like, you know, if you know, it doesn't have the, you know, the tails. Now, the hands didn't look perfect, but I was able to hide the hands by have the mom hold the stroller.

9:44And then all the hand issues went away. And it looks like a real picture unless you know how to look at pictures to determine that they are AI generated. So it is impressive so far as far as I'm concerned. And like I said, the speed issues that I reported, that was a misreporting. I was not looking at the newest stuff. I was comparing fast, you know, of an older version thinking that that was actually 3.0 fast and it was not. Gotcha. OK, that makes a lot of sense. And yeah, the ability to maintain consistency and understand context has improved, but it's not perfect. Right. It's like you can do a small thing, like make sure her hands are hidden behind the stroller and it's got a good chance of doing that.

10:23If you start throwing a lot of stuff at it, in my experience, then it gets a little confused. But again, this is one of those things that it's kind of hard to pin down because you're like, hey, it's a big improvement over what it could do before. It's not perfect and can't do everything yet, but we probably shouldn't expect it to. And to put it in perspective, how long ago did we start talking about Nano Banana? Yeah. Three months, four months ago. And now it's so much better. So yeah, this 3.0, I am impressed with it this far because it is one of the models that I, it's one of the models that I pay for.

10:54So I'm excited to see how much better it is than what I was, than what I'm used to. Your move, Claude. Take the baton, Mistral. Come on. OpenAI is going to come back with something else. It's going to be interesting to watch. DTNS is made possible by you, the listener. Thanks to Chris Allen, Reed Fishler, Larry Bailey, and Paul is coming on with an annual subscription. And Stoic Biker is giving us a raise. Hey, thanks, Stoic Biker. And welcome, Paul, on the annual plan. That's great to have that sort of commitment. Big thanks to everybody who supports us everywhere.

11:31There's more we need to know today. Let's get to the briefs. OpenAI CEO Sam Altman and product designer Johnny Ive told Lauren Powell Jobs at the Emerson Collective's 2025 demo day that they have settled on a hardware design for OpenAI's new product. There was an earlier prototype that we were like quite excited about, but I did not have any feeling of like, I want to pick up that thing and take a bite out of it. And then finally we got there all of a sudden. OpenEye is prototyping it and believes it should arrive in less than two years.

12:26Two years filled with leaks and speculation, don't you think? Two years, less than two years. So let's just go with the, you know, with the one year and 11 months and 31 days. Two years. And here's the thing. We just said that we've been talking about AI the way we're talking about it for three years. This is another two years out. We'll be five years into AI. I just wonder where AI is going to be in two years, you know, for devices being developed today. Now I'm certain that what they would do is that they're going to make sure that, well, whatever we're doing two years from now, we'll be able to run on this device.

13:03They're figuring out the form factor and, you know, the hardware piece of it. But I'm curious because if I think back to the perfect device, the perfect device is something you can hold in your hand and you can look at it and it shows you stuff. This goes all the way back to Star Trek and even before the original Star Trek when sci-fi stuff was you've got some kind of tablet that you're looking at and it shows you things. If this is not going to have that, I'm thinking, could it be maybe an earpiece or something that you wear where you just talk to it and it talks back to you? We've already got a watch.

13:33That's not really a new design. I don't know that I want to wear a, you know, keeping the Star Trek analogy going. Maybe I wouldn't want to wear a communicator badge that I could just tap and start talking or something. But I think, you know, that looks kind of corny when you're not trying to be a very, very sci-tech looking type of person. So maybe an earpiece would be the thing that that's, if I had to guess, that's where I'd go. Although I have no clue. I have no clue what this device is going to be. The imagination is definitely going to be better than the, than the actual reality. It's whether it's the gap, right?

14:08Apple was always really good during the Steve Jobs years of coming up with something that was at least close to your wildest imagination. And that, that would get people excited. That's what Johnny Ive is promising. They talked a lot about simplicity and toy-like and being able to understand immediately what to do with it, but not its full range of capability. They got to skate to where the puck is going to be a little bit here. Like you said, that's a really good point. They have cloud-based systems. So being able to run on it probably isn't as big of a deal as, well, what is our model going to be capable of that this can take advantage of?

14:44And how do we make sure that you can take advantage of it when it comes out? Yeah, this is definitely a launching to Mars type of problem. You can't launch to where Mars is. You have to launch to where Mars is going to be by the time you can get there. Yeah. And, you know, that's that is that is like for me, it's like, what are you going to be able to do is if it's got to be something small that you can wear and talk to you, but you've got to be able to interact with it. So I'm thinking the earpiece would be something that probably could work. But I don't know. Johnny Ive is good with coming up with this stuff.

15:15He's really, this is how good he is. Apple's putting out$200 plus knitted socks to hold iPhones in. That's not a Johnny Ive thing. All right. Well, we're not going to get off of the Apple stuff quite yet because it's Monday. So we like to assess what Apple leaks Mark Gurman put in his Power On newsletter on Sunday. This one tops our subreddit. thanks to rw nash who posted it a lot of good conversation happening in the thread on that german wrote the company meaning apple is focused on improving the software's quality and underlying performance adding that engineering teams are now combing through apple's operating systems hunting for bloat to cut bugs to eliminate and any opportunity to meaningfully boost performance and overall equal or overall quality he also said apple is working on integrating apple intelligence into more apps and adding an AI web search tool.

16:09And he thinks the improved Siri should launch before iOS 27, possibly in the spring as part of iOS 26.4. He compared this to Snow Leopard, though, Rob, and longtime followers of Apple know that Snow Leopard was a big efficiency improvement, but didn't have a lot of features in it. And he says that's what we'll be getting in all the Apple operating systems this next year. unlike google was able to google had to figure out ai i'm i'm almost wondering is apple are they saying we're going to make this stuff run on our hardware as effectively as we possibly can but you're going to be using gemini or using open ai or using anthropic those are going to be the models have they kind of just seeded this because the ai stuff is all is all the way at the bottom of the newsletters it's not at the top if it was if it was the biggest thing usually the biggest bike comes at the top of these letters so i'm just i'm just i'm just wondering because apple is way behind on this stuff yeah keep in mind as good as mark german is at this uh he is not immune from the fact that he gets this information because apple is okay with him having it uh he he most of the time uh apple is aware that it's going to show up there uh that's how he's able to continue to do it over these years.

17:28And so what this tells me is Apple is setting expectations. Oh yeah, we're going to have that Apple intelligence out before iOS 27. Don't get all upset about that. But the more important thing is the efficiency. The more important thing is fixing the bugs. And that is important. It's the kind of thing you never get credit for as a company. You only get blamed if you don't do it and your operating system gets buggy. So they need to do this regularly. And I think it's good to focus on it. But I think there's a lot of expectation setting by by feeding this into the public press to say like yeah yeah yeah this this next stuff you know with they want your reaction rob is my guess of like oh i guess they're going to seed things to other people that way whatever apple intelligence ends up being good at you go oh i'm surprised it actually is good at this that or the other thing set the bar low yeah yeah canada's watt medical is making a wristband at his chinese subsidiary called the e-coffee energy band.

18:26The promise is that it will keep you as alert as a cup of coffee would, but without the caffeine. You wear the wristband that has two electrodes that send a mild electrical pulse, no caffeine addiction, but you may experience numbness in your fingers. So they recommend not wearing it for more than three hours. It sells for$130. If you can find it, there has been no research done directly on the system, though the company lists research on the concept and trials for similar systems that reduce vomiting, but not nausea. Yeah. So they're going to claim a bunch of science behind this there doesn't seem to be a lot of science behind it other than if i'm getting they describe it as a bump you know not like a shock uh but if you know that's gonna it's gonna make me be a little more alert i guess uh that just makes sense this is a great example of something going viral wired says that there's all this backlash and i'm like yeah there's always backlash when something goes viral and the fact that it's sold out means that it went viral for real.

19:20Like people are, are, are trying to find this thing and they're having a hard time keeping it in stock. So I'm thinking back to when I was a kid and I would wear the socks across my grandma's really thick carpet. And then anything metal that you touch, it would shock the heck out of you. That's kind of what I've imagined. Maybe not that much because those shocks were pretty, you know, were pretty jarring, but something to just kind of, Hey, you know, you know, you know, come back. So like a muscle convulsant more than a pain, right? Exactly. So I'm not looking at this like it's a stimulant. I'm looking like it is shocking you back to full consciousness.

19:53If you were kind of like waking you up, yeah. Hey, wake up. The New York Times reports several anecdotes of users describing chat GPT acting like a friend, excessively praising them like a sycophant, and in some cases offering harmful advice, including instructions related to self-harm. There's some pretty disturbing anecdotes in this New York Times report. OpenAI believes it addressed these issues in GPT-5 with condition-specific responses, pushback against delivering delusional narratives. That includes discouraging emotional dependency on the chatbot, making it a little colder, basically, suggesting that you take breaks, and alerting parents if children's accounts express self-harm.

20:34Age verification is supposed to launch soon as well. Still, OpenAI is the subject of five wrongful death lawsuits. Now, these are some of the reasons that a public benefit company called Building Humane Technology has developed a benchmark called Humane Bench that is meant to assess how well models prioritize user well-being. OpenAI's GPT-5 has the highest score on this benchmark, 0.99, while Claude Sonnet 4.5 is right behind it at 0.89. Meta's Llama 3.1 and Llama 4 are ranked the lowest. so this is something that because up until very recently open ai sat at the top of the list that they're the ones who are getting the arrow shot their way but this is not the first or the second i think this is the third time that we've heard about you know its ability to just be very sycophantic and just tell you what you want to hear in every situation possible i've actually had that experience i know people have had that experience so hopefully they are getting this figured out, but it almost can be annoying.

21:39It's like, why are you gassing me up? Just give me the information. I know there's no reason for this to be positive or negative. I just want information. Why are you trying to make me feel good about the information that you're giving me? Yeah. And I make no claims about the reliability of Humane Bench. I haven't looked into it that much, but I like this idea. And creating a public benefit company is an interesting way of going about it because this isn't a nonprofit. So it could be subject to some influence here, but they are trying to position themselves as an independent way of assessing these sort of things.

22:14And they want to get some venture-backed capital so they can build it out and make it work. I look for some independent assessment of the humane bench, but I like the idea of this and hopefully we'll get this and maybe a couple other efforts at it. The president of the United States has ordered the attorney general not to enforce a law that prohibits the distribution of TikTok in the country until January 20th. The delays are meant to give time for a deal that would see the divestiture of TikTok as required by the same law. No one has challenged the legality of the orders, and the administration says it's close to a deal.

22:48However, Monday, U.S. Senator Ed Markley asked the president to give details about the plan to determine whether it satisfies the national security concerns of the law. Markey wants to know if the licensing of TikTok's algorithm is a one-time transfer of source code or if ByteDance will continually retrain it. Yeah, Senator Markey usually is pretty technologically proficient in his communications. He probably has some good staffers, but he tends to understand what he's saying at the same time. So asking about source code, asking about licensing, there's some good questions in here. what's more significant to me is that we are finally hearing a hey what's actually going on about this uh secretary besant has said that we might hear something this week uh about the deal maybe on thursday which is a holiday in the u.s which so that would be an interesting day to unleash it maybe they're going to push it during the nfl games on thursday they've done that sort of thing before uh so yeah i uh i thought it was significant that we're finally seeing a little pushback on the legality of these executive orders and not challenging that directly, but saying, hey, if we got a deal, can we find out the details about it?

23:56I have gotten to the point to where when a deal comes, tell me about the deal. And then I'll believe that we have a deal because we have been kicking this can down the road. If we're going to push this out until January 20, if it will literally be a year, because January 20 of this year was when the law was supposed to kick into effect. And we haven't done it. So we are a year out from doing this. So if we get a deal, just I'm kind of where Marky is. Tell me about the deal. Give me some details. How is this going to work? Don't just tell me that we have the concept of a deal. I want to see the actual innards of the deal.

24:31Yeah. And does it address the problem that the law was created to address? All right. Finally, some quick headlines that are things that are good to know if you want to understand the news in the future. In Windows 11 preview, Microsoft is testing preloading file explorer to improve launch times. Short but sweet, but that's something that I would like to see and you can you can turn it off if you don't want it, but I bet it works. The OnePlus 15 R, which is a cheaper version of the OnePlus 15, will arrive on December 17th along with a OnePlus Pad Go 2 and Watch Lite. Now that's the announcement date.

25:07Shipping will be varying, but the tablet is going to come and the phone is going to come to Europe, India, and the US. The watch only come into Europe for now. The GSMA is developing the 6G wireless standard for an expected rollout in 2030 and estimates the standard will need three to four times the spectrum currently allocated to mobile operators for 5G and below. Oh, 6G. What's that name about 6G? That's a whole lot of bandwidth. That is a lot. Yeah. And another sign to watch for future memory prices. Remember, we've been hearing about price spikes and hedges against that from makers like Samsung.

25:43Lenovo said it's keeping its component inventory, which is memory and other things, 50 percent higher than usual in anticipation of price hikes and particularly shortages. If you wonder where CNN reports went in Apple News, they have been removed while the two companies discuss a renewal of their distribution agreement. A carriage dispute. A couple of important satellite internet notes. Amazon Leo has begun previewing its service for select business customers. They still are on track for a public rollout of satellite internet next year. And Ukraine's Kyivstar has become the first European carrier to launch direct-to-sell satellite service for SMS with Starlink.

26:24Meta is becoming an electricity trader in order to help secure new energy plant construction in the United States. That's just one of those checkbox things. You know, if they're really going to do this, they need to be able to qualify as a trader. Now they have. U.S. Office of Personnel Management Director Scott Kupor told Reuters earlier this month that the Department of Government Efficiency, oftentimes referred to as DOGE, does not exist. It's gone. Some of the members of the team have taken other roles, like in web design at the White House. But there is no more DOGE. And y 'all made me do a big old story earlier this year about whether Doge was a threat to everything or not.

27:03And so I am personally glad that it doesn't exist anymore because I'm tired of talking about it. It didn't even make it to its first birthday. Well, it's so sad. A U.S. judge issued a temporary restraining order against OpenAI prohibiting its use of cameos or cameo videos in its products and marketing as part of an ongoing trademark lawsuit brought by Chicago-based Barron app, will use the same name in its Cameo product. The hearing is set for December 19th. And the Financial Times reports that AIG, Great American, and W.R. Berkeley are among insurance companies seeking permission from U.S. regulators to have the option, none of them are doing it yet, but to have the option to exclude liabilities from deploying AI tools from their coverage.

27:46In other words, we're not going to cover you if you get sued for AI stuff. I'm a little scared of that. They just want to have that option, you know. Those are the essentials for today. Let's dive a little deeper. It's not an AI bubble. It's an LLM bubble. Andy Beach breaks down what hugging face a CEO thinks comes next. Last week, we talked about whether there is a bubble or not. And today, Andy Beach is going to explain not only why one AI CEO thinks there is a bubble, but he thinks he knows what kind of bubble it is. Andy, thanks for joining us. Hey, Tom. Good to see you again. So tell us what Hugging Face's CEO was going on about here.

28:27Yeah. So there was a recent Axios event and Clem Delengue from Hugging Face argued that we're not in the middle of sort of an AI bubble overall. What we're really in the middle of is an LLM bubble. And the fact is that these general purpose LLMs are going to start going away and we're going to see AI specific domain knowledge really come into its own in this current next phase. I've been saying this for a while, and I know I certainly wasn't the first to say it, but it makes sense to me. And in some ways, he's kind of making a distinction without a difference. When everyone says there's an AI bubble, they mean LLMs.

29:08They're talking about Gemini and open AI and all of that. Is there more to what he's saying here that says, yes, the general, is what he's saying more hopeful than it sounds? Hopeful may be the wrong word, but I think we're getting down to a reality that you, myself, several others have sort of seen. We've just been saying it in words that nobody necessarily understood. We've been talking about these like agentic workflows. We've been talking about orchestration. All of that is part of what will make these smaller language models, these more directed models useful, because they will need sort of a multitasking system that coordinates them so that the right information is coming together.

29:54So I think what he's calling out is sort of the feature within a larger set that we've been thinking about and trying to put that forward in public knowledge in such a way that people begin thinking about the lack of the one big model, which is how everybody has thought about A. We've thought of it as capital A, capital I, and now we need to start thinking about it as sort of lowercase A, lowercase I, and then there's lots of them. Yeah. One of the things that I have been pointing out is that when you fine tune a model and you make it very specific, it hallucinates less and it helps things more.

30:30And that he seems to be saying a similar thing. He's saying it more from the financial end in this particular case. But it's all based on the same thing, which is a big general purpose model is never going to be as good as a fine-tuned model. And I guess what he's saying, if I'm hearing it right, is that's the future. The future is making a model that works for whatever it is that you want to do, not saying try to get ChatGPT to do the thing you want to do. Maybe it'll be good at it. Yeah, absolutely. I mean, and I got to see this firsthand in some media operations that I did when I was still at Microsoft, where invariably, if you wanted something that was more accurate and also speed efficient, lower latency, then you had to have something that was tuned and specific.

31:17And it was going to only work on that one task. And that's all you needed it to do, but you knew that it was going to be highly consistent. And so I think operators have already been thinking in this space, and this is the larger public starting to think that there's more behind it than just a chat GPT or a Gemini or a Claude that goes into it. And we need to think of it in terms of smaller pieces that we assemble for the larger workflow that we want. So what do you think that means if the LLM bubble bursts, right? If it turns out like, no, nobody wants to use Gemini, Claude or chat GPT anymore.

31:53They want to use these smaller bits. How would that play out? Well, I think all of those players are still in a good place because I think all of those systems are already attuned to being sort of an agentic or an orchestrated workflow, meaning that that larger LLM piece can still act as an agent that is moving the right pieces together to get you the answers that you want. It's just not answering it itself. It's going and finding the correct resource, whether that's maybe a specific language or a specific utilization within science or medicine. It's just not trying to answer it off the cuff. It's almost the research librarian that's going to help you find the topic that you want.

32:35I know I do historical parallels way too often, but it's sort of reminding me of, uh, listen, Alta Vista, Excite, and InfoSeek are all really good. All they need to do is, is start orchestrating ways to share data. And maybe Yahoo is the future because it has a directory and that's a little more useful. And we are experiencing that with OpenAI, Anthropic and Google and Nestril and others. And we are waiting for Google in this analogy to come along and go, actually, if you just do it this way, it works a lot better. I think, you know, you can only, And I agree that we have definitely seen this time and time again.

33:23And the difference here is that we got to this probably much quicker than we did in the era of search. Definitely. And we will likely see, you know, an AI version, a Google version of an AI company that comes around that completely flips the model on its head at some stage. Yeah, because, I mean, what Google did was say, you make the websites, we'll make them easy to find. What someone can do in this space is you make the individual models that are great, we'll make them easy to orchestrate. And whoever does that wins that game if that ends up being the game. Yeah, no, absolutely right. Yeah. Well, Andy, thank you for helping us interpret Clem's words here.

34:08I think we all know there's a bubble. We just kind of don't know how it's going to deflate and who's going to get caught in the fallout, if you know what I mean. Yeah, absolutely. It'll be an interesting area to watch. And I do think that we will see more of this in early 2026 now that we seem to be shifting gears into the smaller manageable models. We'll see more players out there. And if you want to follow those players more closely, you got to go check out Andy's Engines of Change. Where can folks get it? Enginesofchange.ai is my sub stack. I cover media technology and the convergence of AI with it.

34:46Fantastic. Thanks, Andy. Thanks, Tom. Hey, you were just saying earlier in the show about how much accelerated we are. Did you like that analogy? Yeah, it's actually pretty cool. It makes sense. It's like, this is what we're thinking. So smarter people like Andy can put it in words that I'm not able to express. Thank you, Andy Beach, again. Also, big thanks to everybody who supports the show. And in exchange for your support, we're giving folks a discount. If you are interested in some DTNS stuff, and we got some great new stuff up on the store. We got a pint glass. We got a bunch of stuff with Len Peralta's art with Techie Santa.

35:26We got a new DTNS hat, like a dad hat, nice cotton hat. It's gray with the DTNS logo on it. You can get all of that at the store for 12 % off to the end of November. Remember, if you use the code DTNSROCKS, that's D-T-N-S-R-O-C-K-S. There'll be a link in the show notes as well that you can click on. But just as a thank you for 12 years of DTNS, we're giving you 12 % off. Again, that code DTNSROCKS at dailytechnewsshow.com.

36:01We end every episode of DTNS with some shared perspectives. Today, Reid is helping us understand. Yeah, regarding last Monday's show, Reid writes, I don't think people understand how much of the wireless backhaul is done via fiber. Regardless of whether or not it is satellite or cellular, we are talking about terabits upon terabits of capacity. That is much more than any satellite is doing today, especially not without beaming down to an Earth station where it connects to fiber. This is a political problem and is also a problem of countries who know that they are in the right place and are able to suck money out of those countries who need to pass through them.

36:42This isn't a problem that will be solved soon because it is directly related to the political problems in the Middle East, which I don't think are going to be solved anytime soon, unfortunately, as well. This is a good email. Yeah, it is. This is from somebody who knows the business, too. So it's no surprise to me. Reed knows what he's talking about here. And this makes so much sense, but it's like, yeah, it's like, you know, money rules a lot. Yeah, we can put satellites up as an alternative, but until backhaul is capable between the satellites, which they're investigating, but it is way too early to say that that could replace fiber.

37:24You're going to need that fiber. And so when you have problems running subsea cables, when subsea cables get cut, when you have to move them away from areas that have, you know, war happening, it has a big effect on how well we're able to move bits around the world. Yeah, you think about, you said war in those cables. That is, people are going to go after that in times of war. If you can stop communication, you can stop a lot. Absolutely. Well, folks, we want to know what you're thinking about. So if you've got some insights into a story, share it with us at feedback at dailytechnewsshow.com. Yeah, big thanks to Andy Beach and to Reed for contributing to today's show.

38:05Thank you for being along for Daily Tech News Show. You keep us in business. Thank you, patrons. If you want to become a patron, we have a little discount for patrons as well. Go check that out at patreon.com slash DTNS. Talk to you soon.

38:23The DTNS family of podcasts, helping each other understand. Diamond Club hopes you have enjoyed this program.