Wikipedia Cashes AI Checks to Survive Another Day

0:12This is the Daily Tech News for Thursday, January 15th, 2026. We tell you what you need to know. We give you context. We do our darndest to help you understand. Today, Andy Beach tells us why Microsoft is spending so much money on Anthropic and Wikipedia gets paid by the big LLM companies. Is Microsoft trying to win Anthropic's heart? Is that why? It's like a Valentine's thing. I'm Tom Merritt. I'm Winwood Dow. Let's start with what you need to know with the big story.

0:43Wikipedia launched on January 15th, 2001. That was 25 years ago today. And now has more than 65 million pages. Receives billions of views per month. The site has put up a celebratory mini site at wikimediafoundation.org slash wikipedia25. There's like interviews with people who've been editors for years. Lots of cool, like diving into the archives, telling the story of Wikipedia. Definitely something worth celebrating, in my opinion. The Wikimedia Foundation also celebrated the day with an undoubted bargain. That's a quote from Christmas Carol, if anybody doesn't get it. Amazon, Meta, Microsoft, Mistral, and Perplexity have all signed on to the Wikimedia Enterprise Program, which licenses Wikipedia's data for training their models.

1:34They are not the first to do this, but that's a lot of big names all at once. Previously, Google, Ecosia, Gnomic, Playus, ProRata, and Reef Media were already part of the program. The program gives API access to licensees that deliver files at higher speeds and in greater volume than the free API. You can just go scrape the free API, which a lot of people have been doing, which has been causing Wikipedia a lot of money, costing them a lot of money. but now they get paid and they get better access to the data they want faster access to that it is unclear to me if open ai is part of this program i have seen people say like oh they must be and open ai certainly does include wikipedia in its results when you ask a question sometimes but i could not find anything definitive one way or the other the whole reason they need to get paid for this.

2:25Of course, if you've run into Wikipedia's donation pleas, you know that they are publicly funded. They are funded by donations. They are funded by grants. But this money isn't going to solve their problems. Wikipedia reported last April that bandwidth had risen 50 % between January 2024 and April 2025 as more companies were accessing Wikipedia for training data. So the money from the enterprise deals is going to help offset the costs of providing the enterprise deals. Hopefully they make a little bit of money off it, but it's not going to solve all their problems. This is always one thing that's been so funny about the way that the internet works.

3:02And I say like as broadly as possible is that the popular you get, the more it costs to be you, I guess, is to say on the internet because of like the way that that bandwidth and access costs. This is definitely very interesting. I went, I have to admit when I was reading this, um and of course uh not y 'all know i'm i'm pretty skeptical about um a like the current generation of ai for llms especially when it comes to knowledge look up and fidelity and accuracy which is you know i mean i think to some extent you know just me being critical but also like i think fair and i went down this very interesting rabbit hole because you know um with wikipedia you know uh it's not it i i went down this like thought thought hole of like yeah 25 years of wikipedia and thinking how it changes the way that you know or how it has contributed to how we look up stuff on the internet and how we contribute to knowledge and that because like remembering for a minute that wikipedia is a wiki anyone can edit it and so there's this interesting like i don't know checks and balances not the right word but there it's like it's it's almost like this this it's a mass of knowledge but that the way that the knowledge is input into wikipedia and the way that it's managed there's always going to be subjectivity or there's always going to be some level of infidelity with whatever reality is but then i went into this big hole of like but if it's all just written by people what i mean i guess what is truth and knowledge and reference materials and i i think my critical brain my ai critical my llm specific llm critical brain went down like oh well here we go like here's them like pulling in like wikipedia and i guess in most cases it's it's like fairly reliable but that you know people generally know that they should be like take things in wikipedia with some grain of salt because it is edited by people but then i went in this whole like but then what is knowledge and reference all everything is edited by people yeah and everything that you know and it's so interesting because then my brain went to like this weird informational existential crisis of like okay what we can reference things but you know the way that we trust and utilize that knowledge is very different so anyway i not and i'm not even criticizing it it just it was just so interesting to think that now i guess to get back to the point of story is that now um of course it's important that wikipedia pays its bills um and that you know they're able to set up specific structures and and and partnerships with um these companies that are rising their costs but then also i was just more think of the culture impact of that and and it's kind of interesting how maybe the the trust or distress that you have in llms now is not too different than the kind of trust or i mean it is different in different ways but but it's not too far off from like the trust or distress you you should put in wikipedia and how that kind of maybe changes how we think about communal knowledge and um and and knowledge pooling and and and like how do we consume intake that knowledge and and act on that knowledge or create more content on that knowledge.

6:05I just thought it was so interesting. It sent me down this really interesting hole of like, oh, maybe I've been kind of hard on it because I mean, really these problems have existed for at least 25 years now, if not longer with other sources of information. So it's really interesting. No, I think you shine a light on a bunch of interesting parallels between the LLMs and Wikipedia, right? Like when Wikipedia launched in 2001, I remember people flat out saying exactly the same things they're saying about LLMs right now. Don't trust it. It's unreliable. You don't know where the facts are coming from.

6:38Double check anything you get from it. And I think those are still good recommendations if you're using Wikipedia, but we've learned over 25 years where the limits are. And now I mostly know when I can trust what I've read on Wikipedia and when I need to go to that citation that it has. And they've been very good at developing sources because they developed an ecosystem where editors are very good at double checking each other. And in fact, in some ways, Wikipedia, I think, is the best source of information on the internet these days because you know there are a group of people whose passion is making sure they are being as objective as humanly possible, citing sources, and making sure those those sources are reliable when you cite them in ways that a lot of bloggers and video creators and folks are not uh you know so i i think wikipedia has done a 180 in my estimation from being like well it's kind of a risky place to like actually it's one of the more reliable places and the llms were using it anyway i don't think it changes anything on that side it just means that now they're paying for it and they'll have more up-to-date and faster and better and more accurate access to Wikipedia.

7:53So maybe it improves the times that they rely on it, but they were all scraping it anyway. So it's not like those results weren't in there. Yeah. And I think if anything, it softened my, you know, my kind of hard nosed like feelings about some of these things, because as you said, like it, it evolved like, and, and Wikipedia over the last 25 years, as you said, has, has become a trusted source. And, and I agree with you, I think, um, and I guess it was my existential crisis. Like, well, I mean, like, I mean, it's You should question sources of information, but that's just across the board, period.

8:26And then, of course, as I mentioned before, what is anything that a person, a human being, writes down either empirically or anecdotally or just from their own opinion? It's all kind of like the objectivity, subjectivity scale is always going to be kind of like it's a scale, right? And that, I don't know, applying that to the current generation of LLMs and how we get knowledge from them, it's more just like the idea of Wikipedia where human beings are injecting knowledge. And at some point we have to kind of filter it, collate it, and balance, filter, whatever, process it. Like checks and balances and stuff.

9:03Yeah. And it's happening at scale, just huge scale in a sense that's being automated. I'm kind of like being very simplistic. But in a sense, LLMs are kind of doing the same thing. doing what people have been doing for ages is scraping wikipedia or like collating stuff it's just at a scale and at a pace which is kind of overwhelming um whereas wikipedia of course as you said had many many years to kind of like solidify and come up with these tracks and balances and i guess my hope is that by kind of you know maybe because we're everyone's figuring these things out everything the relationship with like wikipedia and everything else i i hope in the best case the very very best case that something like the wikipedia story becomes the case for LLMs and that people both balance a sense of taking things with a grain of solid understanding, like how to analyze sources, but that also, you know, we all kind of come to some kind of checks and balances system with LLMs where overall, right now, I hardly trust anything LLMs tell me that is meant to be factual.

10:02Maybe there's a place where I start to kind of slowly trust it or understand that at least I have ways of checking it. And to some point where like, okay, maybe this will become, maybe we'll eventually get to a place where it's reasonably trustworthy. Yeah, I agree with you. And LLM starting to show citations more often is good, although I quite often check those citations and like, that doesn't say the thing that you're citing it for saying. That's happening less and less. So hopefully that gets better and better. The danger though, is that Wikipedia built a system based on people coming to Wikipedia and seeing something where they're like, wait, that's not quite right.

10:38And I am an expert. And so i will contribute that and then maybe some of those people become editors and become passionate if an llm is just citing wikipedia the danger is people go oh it says it's on wikipedia great they never go to wikipedia and we start to impoverish the source of the information not just wikipedia but others yeah the llm right i mean that was the kind of thing that i think um that google summary is kind of bothered me when they first announced it at io is like the entire internet and not just like it as a source of like source of knowledge and communication and like content but also it as an industry business whatever depends on people actually visiting sites and powering the ad engine whether you like ads or not like people need to pay for things and as you said like it this is kind of just speeding up the speeding up that like it isn't yeah it's impoverishing it's it's kind of it's like cannibalizing the internet a little bit and i can't like that we haven't quite figured out that part of it because you want to it's like People like the idea of a one-stop shop for things like Google search used to be.

11:39And now ChatGPT and Gemini and everything else are kind of even taking that to an extreme. And then it's like the circulatory system or whatever. How do you make sure the heart keeps pumping? Yeah. Yeah. It's weird. It's because it's putting a tourniquet on a certain part of the body. And at some point, I don't know, things are going to get gangrenous or something on the internet. I don't know. I wonder if Wikipedia is an exemption to that, though, because it doesn't need high traffic. It just needs quality traffic to survive. It needs traffic from people who believe in its mission and willing to donate either in money or time.

12:14so I find myself still going to wikipedia and reading things even though I occasionally will look something up in a gemini or a chat gpt yeah uh and it's not even because chat gpt sends me there it's just me going like oh I know I'll get I'll get a different view on this and a view that I can rely on if I look at it on wikipedia so I go directly there versus chat gpt and gemini are more like I don't really know what I'm looking for can you direct me in the in the right location and a business based on traffic is less likely to survive that reduction than Wikipedia. So maybe they're best suited to survive a new way of doing things.

12:53That's super. I guess we'll see how it shakes out. I mean, for sure. One way or another. If we survive, we will. Well, we survive because of y 'all. DTNS is made possible by you, the listener. So thank you to Mike Cortez, Aaron and Ken Hayes. Thanks, everybody.

13:13There's more we need to know today. Let's get to the briefs. All right. Well, more details are dribbling out about Apple buying Google's Gemini services to power Siri and other smart features. The Financial Times' sources say the deal is structured as a cloud computing contract, with Apple paying several billion dollars to Google over the course of the deal. Conversely, remember that Google still pays more to Apple to be the default search engine. The Financial Times also claimed a source close to OpenAI said OpenAI made a, quote, conscious decision not to be the provider as it focuses on its own devices in competition with Apple.

13:48Bloomberg reports that OpenAI is making requests for proposals from companies that make silicon, motors, packaging material, and cooling gear. Most of that is for the data centers that OpenAI plans to build, but some of it is for robotics, which OpenAI is developing for its own products. Yeah. Structuring this as a cloud contract is very instructive to me. It may not sound like a big deal, but to me, that sounds like Apple saying we really are pivoting to services. And so we are not a cloud provider, even though we have iCloud. We're not going to build massive data centers. That's not what we do.

14:24So we'll get a cloud contract to help us with that and take advantage of Gemini in the offing, right? OpenAI could have provided that. They're building data centers that is something they could they could have provided but it sounds like open ai wants to go head to head and they're like we don't want apple to have our special sauce when we are going to use that to make devices that compete with apple and apple's advantage in that situation then becomes not just the existing user base and brand loyalty but also services it's going to be harder for open ai to create apple music apple fitness apple tv competitors uh so apple's kind of I think banking a lot more on that than I would have expected.

15:04Yeah, it's crazy because I feel like a lot of Apple's shininess as a brand and their entire reputation as a company that relies on itself and builds high quality products, hardware, software experiences, and that that control is what makes Apple Apple. And then so this kind of final capitulation, and I've seen a lot of big companies do the same thing where there's an inclination to build it in-house or um what's it not built here syndrome or something i forgot i forgot what the software person like cute term for it is but there's an inclination to want to build things yourself and there's always like a a compromise between control um and like doing something entirely custom but also cost and effectiveness and also if you're not specialized in something there's going to be a gap and so it it's really interesting to see apple finally kind of openly partner with google and it's also makes it's kind of funny because in these days of LLMs.

15:58It feels like everyone's competing very openly, obviously here with OpenAI and Apple, but at the same time, people that we normally think of as competitors like Google and Apple are like shaking hands and making deals. It all feels very like intertwined spaghetti. Yeah, yeah. Like integration. Internecine. Yes, yes. I mean, we have to remember that this happens all the time, right? Yes. Sony buys cloud computing from Microsoft for PlayStation, with which it competes with Microsoft's Xbox. So when these companies are that big, that kind of stuff happens all the time. One of the big stories of the coming year will be how high demand for chip manufacturing, especially RAM, is going to affect pretty much everything.

16:42Intel's Senior Director of Product Management, Nish Nella Lojana, told Tom's Guide that laptop makers have about 9 to 12 months of memory inventory. and that could be key to this not being disastrous. He also tried to tell Tom's guide that the increase in L3 cache on Panther Lake chips might take some of the pressure off needing more RAM. That's something you might expect an Intel person to say. It's also true, so maybe it'll help, but mostly the stockpile means you may not see laptop prices skyrocket right away and it gives the industry some runway to adjust to the new higher demand. There are still going to be effects because there is only so much capacity to build this stuff and everybody wants to build data center RAM with it.

17:27So one of the first casualties appears to be Asus. Australian retailers told Hardware Unboxed that Asus-built versions of NVIDIA's 5070 Ti are no longer available, that they just can't buy them. And the 5060 Ti 16B is almost done with Asus telling the retailers that it's not going to make that model going forward. So we're already seeing some early casualties. these aren't necessarily the ones that most of you are going to buy, but it's fewer options for you to buy. Doesn't mean you can't get a 5060 Ti. You just can't get it from Asus right now, but canary in the coal mine kind of stuff, right?

18:02Yeah. We talked about this on Android Faithful a little bit about, I mean, and I'm not saying this is directly related, but how Asus, for example, is not making, I think, any new smartphones, certainly not many this year. And, you know, I mean, it's a bit of a different part of the industry. But at the same time, everything takes RAM. Like most computers take anything that can be vaguely considered a computer, it takes RAM. And so it's weird to think about how the ripple effects of something like this might happen and how maybe companies will be less likely to experiment with prototypes or new models because they're getting that RAM for somebody else, somebody AI else.

18:38Yeah. I think if you are in a sector of the industry that didn't stockpile, So he's saying, Nella Lajana is saying it's laptop makers. Maybe Asus didn't stockpile enough to make new models of phones. And it was like, well, I guess we're not going to do that now, at least this year, because we can't get them. I don't think it's as simple as like, if you didn't buy a laptop, too late, right? But there will be weird effects. It's going to spike certain things, keep things from coming to market. But other things are going to seem perfectly normal. and you're going to be like, well, why isn't that price going up?

19:15And this is the kind of thing that helps understand that. Well, speaking of AI, I'm like, like, like, like we have stopped it all. But one of the convention contentions about LLMs is that they haven't been great at math. They aren't LMMs after all, large math models. However, software engineer and former quant researcher, researcher Neil Samani posted some results over the weekend that he was able to use chat GPT's GPT 5.2, its chain of thought reasoning to solve open math problems. So Samani was looking at the Hungarian, is it Erdos? Erdos? I think so. Erdos problems, which is a set of more than 1 ,000 conjectures.

19:54Gemini's Alpha Evolve has had some success at this problem set as well. Since December 25th, 15 of the problems have moved from open to solved, with 11 of them credited to an LLM. Mathematician Terence Tao evaluated the solutions and found that in eight of them, LLMs had demonstrated meaningful progress, while six of them built on previous human research. So Tao suspects that the scalable nature of these systems makes them well-suited for the long tail of their dose problems, but it's still in advance. And along with this, automated formalization tools like Microsoft's Open Source Lean and Harmonix Aristotle are not LLMs, but automate labor-intensive aspects of proof formalization and are being used by math and computer science professionals to assist them in their work.

20:41And I feel like, you know, we have definitely talked about this, and you mentioned this before, that, you know, specialized AIs, not necessarily LLMs, but that the specialization of certain forms of AI is going to be where we see like the most ROI in this new age of AI, ROI, AI. Yeah, yeah, absolutely. LLMs getting better at math is interesting. And I think that's worth noting and it's notable. But the uptake by professionals of these specialized tools, in Lean's case, it's been around for a while. It's not new, but it's just continually getting better and being developed. And it's not an LLM, right?

21:18It's just a formalization tool that's specific for that. I think that's really interesting because I'm going to keep my eye out for more of those. There are more examples like that where it's not just about AI, meaning an LLM. It's about automated tools and things that can help you in various more tailored ways. All right. We cannot avoid talking about Grok these days. It's not necessarily a pleasant topic, but here's the update. We'll try to keep it as utilitarian as possible for you. In the ongoing story around XAI's Grok model and its ability to create non-consensual sexual images and in many cases allegedly illegal images, here's the latest.

22:04X says it has, and I quote, implemented technological measures to prevent the Grok account from allowing the editing of images of real people and revealing clothing such as bikinis for all users. There was a criticism that they had just stopped people on X, the social media platform, from doing it. But if you were an XAI user, you could still do it. Now they're saying we have implemented a stop to or an attempt anyway to try to stop this for any user of Grok. Meanwhile, the Philippines, Malaysia and Indonesia have now all ordered a block of the app that not necessarily has been executed. But the governments have said we want you to block distribution of this app in our country.

22:43U.S. state of California has opened an investigation into Grok's use to depict people in minimal clothing, according to the attorney general. UK's Ofcom is also conducting a similar investigation. We've talked about that before. And a coalition of 28 digital rights organizations led by Ultraviolet sent letters to Apple's Tim Cook and Alphabet's Sundar Pichai, asking them to remove both X and Grok from the Apple and Google Play app stores. yep sorry yeah no i know right it's like there you go that's the update if you have feedback about anything that gets brought up on the show get in touch with us on the social networks we are at dtns show on x on instagram on threads on blue sky on mastodon and for tiktok and youtube you can find us at daily tech news show

23:33finally some quick headlines that are just good to know might make you look smarter in the future if you know about these. OpenAI launched a webpage specifically for using the language translation features of ChatGPT at chatgpt.com slash translate. It can detect what language you're using and support translation between 50 languages, as well as use natural language prompts to modify the translations. The latest version of Wine. Wine 11.0 is out for Linux and Mac OS. This is the Windows emulator. lets you run 16-bit, 32-bit, and 64-bit Windows x86 binaries with the same Wine command. Don't have to use Wine 32 or Wine 64.

24:11You don't have to know. Just tell it. It'll figure it out. Android 16 QPR 3-bit 2 is out, and it reduces the option in the Recents menu to copy or share, meaning you can no longer save or open in Lens. Two minds about this because, yes, the share menu and all kinds of menus are bloated, but come on, I literally were using both of those options. You were among those people. Raspberry Pi has launched the AI Hat Plus 2. It has 8 gigabytes of RAM and the Halo 10H neural network accelerator. So it can do 40 tops of inference and it cost you 130 bucks. Researchers at Belgium's KU Leuven University released details on vulnerabilities they called whisper pair in Google's fast pair protocol for headphones that would let an attacker pair with some Bluetooth accessories without the user's knowledge.

25:03So go update the firmware on those Bluetooth accessories, please. Taiwan's TSMC announced a 35 % rise in profit last quarter, and as a result, projects a 40 % rise in capital expenditures this year that will be used to speed up construction of its factories. You know, the factories that make RAM. Please do that. They specifically mentioned plants that they are building in the U.S. state of Arizona as being a beneficiary of this. OpenAI announced Wednesday that three former employees who left to work for Mira Morati's Thinking Machines company, they're Zoff, Luke Metz, and Sam Schoenholtz, have returned to OpenAI.

25:47Spotify announced it will raise its monthly price in the U.S. from$12 a month to$13 a month. I saw people trying to make this into a percentage in their stories. so like it's just dollars it's easier for people to understand if you like it's raising by a dollar month uh it will also raise prices in estonia and latvia china is working on rules for who can buy nvidia's h200 chips and in what amounts with hopes to have approvals by the end of the month imports have been paused until the rules can be drafted and the u.s announced it will put a 25 tariff on certain chips like the NVIDIA H200 or AMD's MI325X if they're not used in the U.S.

26:27So if you're wondering, how's the U.S. going to get 25 % cut of NVIDIA sales? That's how. OpenAI has reached an agreement to buy several billion dollars of chips from Cerebris, which makes efficient data center chips for foundation models. Cerebris's chips are similar to those made by Grok. That's Grok with a Q, which was recently acquired by NVIDIA. And Amazon is going to be the first customer to buy copper from Rio Tinto, the first new source of copper mining in the U.S. in more than 10 years. All right. Those are the essentials for today. Let's dive a little deeper. All right. Earlier this week, we talked about Microsoft spending close to half a billion dollars a year using Anthropics AI, even while OpenAI is still a major partner.

27:11Tom followed up with Andy Beach to talk about what this tells us about how big tech is thinking about AI partnerships. Andy, thanks for joining us again, man. Hey, Tom. Good to see you. Good to see you, too. So I think, you know, a lot of people hear that Microsoft is spending this, you know,$500 million on Anthropic and they think, whoa, whoa, whoa. I thought they were the OpenAI folks. Some folks are like, oh, I heard that Microsoft and OpenAI broke up. I guess they're Anthropic's best friend now. Neither one of those is really true. What's going on here? Yeah. You know, it was interesting news to come out.

27:47And I certainly heard from multiple people, even though I no longer work at Microsoft. But there was a lot of conversation that went around this. And I know you've covered the piece, so we won't go into the details around the deal. But where these kinds of things come from is that even though OpenAI and Microsoft had a crazy strong, almost overly strong in some ways, partnership and relationship in the very early days it was far from the only ai initiative and it definitely wasn't the first ai initiative that they were involved in and i think what this shows is that there's they're beginning to position themselves back almost where they maybe have been in years past as more of the the marketplace and that platform for any ai that's out there so they they certainly have a lot that is still highly tied to open AI.

28:42So obviously the co-pilot is chat GPT based. It doesn't sound like there's any changes or pieces that'll go different from that. But for years now, probably for at least four years that I can remember, they've been increasingly expanding the catalog of models that are available on Azure for anybody to use. And only a very narrow slice of those were actually related to to open AI. I believe they were into the multiple thousands of models that were out there and available across the spectrum. So I sort of see this as them beginning to dip a toe in the water in bringing another massive platform and partner into the space, but looking at how those new infrastructure plays can work with another big sort of rival partner effectively.

29:38Yeah, it's a good reminder that Microsoft is in a lot of different businesses, right? Exactly. Azure is probably the predominant one. And as an Azure cloud services provider, you don't want to provide just one service. You want your customer to be able to pick the best models that are out there. It's right. I think because they are the sort of the gateway to a lot of pieces of the puzzle in architectures and infrastructure, they have to have multiple partnerships because the reality is OpenAI won't always be the dominant player. They had a lot of great early first moving action, but there have been lots of other companies that are either catching up or surpassing them in other functionalities.

30:23And having those partnerships and relationships in place allows Microsoft to guarantee that they're going to have the right tech at the right time for people. and they're not just going to be sort of one note. If they only had that open AI relationship, it would become dangerous because as soon as it fell out of favor, they would be at big risk of losing a lot of business. And to be honest, I think this is even not the biggest AI news we're likely to hear out of them this year. Realistically, they have Microsoft Research, which is its own research arm that is separate from products, separate from cloud and infrastructure and all the other pieces.

31:03And they're constantly coming out with new models there or new research that will drive new potential products and models. So I'm willing to bet that this is a front loaded early 2026 announcement, but that there's at Build or one of their other conferences later this year, we're likely to hear big product announcements that aren't related to this, but are still AI and that are probably more first party Microsoft. Is it possible that we are headed towards a marketplace where you will rely on different providers for different things? Because what it looks to me like is there's the Azure part of it, but there's also Microsoft integrating parts of ChatGPT with parts of its own research and development to create Copilot.

31:50And that now the possibility is, well, maybe there's a little bit of Anthropic in there when that works. Are we headed towards a world where you just pick the models that are best for different tasks and there's not just like, oh, it's picking between cloud services, right? You either pick Google Cloud or Amazon Web Services. In AI, it may be like you're always going to have a cocktail of different providers. Yeah, that's a great way of looking at it. And we did similar in cloud days. So I think you're probably right here that it will follow similar for AI. So for in sort of 10 years ago, when it was just purely a big cloud play, there was a lot of talk of whatever else you're doing anywhere else, data and Microsoft make sense together as a combo.

32:32So that was always, and then advertising and Google made sense. Sure. And so I think we're likely to see a similar thing play out. Now, I don't know if it's as long a term play. In other words, I don't know that everybody keeps wearing the same hats as long in AI because AI rapidly iterates. Sure. But the reality is that, you know, Anthropic has some great coding tools. And so you're going to go tap it. And so you want the ability to pull it into your workflow, whatever cloud you're in. Similarly, Google's doing a lot of amazing stuff with image and video right now between its tooling. And so there's a desire to be able to integrate that into the work you're doing.

33:11But there are others out there that are focused on sort of their particular sector of AI and whatever AI means to them. And we're likely to see it, at least for the next couple of years, be something where you go tap that company for that functionality, if that's what you're looking for. Yeah, it actually made me think differently about what Apple is doing with Gemini. because if you look at Microsoft, granted it's different because it's in a cloud business and it's got some other priorities, but in a certain light, what they're doing is not that different. They're saying, yes, we have our own in-house developments for these kinds of algorithms, but we're also relying on external partners and Apple is doing the same thing.

33:53Do you see what the differences are between what Microsoft is doing and what Apple's doing, or does it appear similar to you? I think they're potentially a little different. I mean, Apple doesn't have an enterprise cloud space the way that Azure does and Microsoft does. So I feel like there's a little bit different there. They're not trying to, you know, when they pull in AI as a feature, they're doing it really for themselves as a platform that they'll then pass on as a feature to a customer and to an individual. So it's a little bit different. Whereas Microsoft, AWS and Google, frankly, right now are all trying to create a healthy marketplace in their cloud business for AI functionality, regardless of whether it's first party or third party.

34:44Yeah. Although Microsoft also doing the thing where they will pass along co-pilot to you and say like, this is ours, but there are different models powering it depending on what you're doing with it. Yeah, they're constantly going to, and they're still going to have their first party pieces that come in. I think more and more we'll see that it is AI integrated into existing first party lines of things that they're doing. Maybe it's dynamics if you're in the CRM business and you need something and there's AI functionality built in there. Or obviously there's a lot within Office with Copilot already, but I could see that being an area.

35:23And then the third party pieces start to come in as you're doing, as you're building a bigger thing and an application, whether it's maybe a streaming service or a creative platform, or maybe you're doing, you know, health and statistical data studies or other things. Well, Andy, thank you so much for chatting with us about this. It's always valuable to get your insight on these sorts of things. And people can get that insight more often than we talk to you. Where should they go? Yeah, absolutely. So I started a Substack in 2025. I'm still going at it. It is called Engines of Change and it's Engines of Change.ai.

35:57Thanks, man. Thanks. So there you go. Everybody's paying money to everybody. You were talking about that earlier in the show. Yep. It's all a big circle. Mm-hmm. All right, well, we end every episode of DTNAS with some shared perspectives. Today, Bill has some further thoughts on how Microsoft names things. Yeah, so I know y 'all talked about this on Friday when Stim sent a note, Bill said, just to add to the perfection that is Microsoft product naming, I work as a technology trainer and I can tell you that Microsoft's naming skills leave a lot to be desired. You can create teams in Microsoft Teams.

36:35Each team has various discussion channels. You have Microsoft Stream, which can have channels of videos. Microsoft Stream content can be added to a navigation tab in a team channel. And this means you can place a stream channel inside a Teams channel as a stream channel tab. Not confusing in the slightest. No, it makes perfect sense to me. All right. Are you also thinking about how brilliantly Microsoft names things or just anything else? If he has some insight, please share it with us. Feedback at DailyTechNewsShow.com. Absolutely. Big thanks to Andy Beach. Thanks to Bill for contributing to today's show.

37:11Thank you for being along for Daily Tech News Show. You keep us in business, folks. Go check out our Substack. You can subscribe there, get ad free, dtns.substack.com, or become a patron, patreon.com slash dtns. The DTNS family of podcasts. Helping each other understand. Diamond Club hopes you have enjoyed this program.