RTX Spark, Project Solara, Scout, & Much More!
Loading summary
A
It's time for Windows Weekly. Paul Thurat and Richard Campbell are here and we have lots to talk about. Of course Microsoft Build is on. They love Stevie Batish's explanation of what Microsoft Celera is all about. I think you'll love it too. Jensen Huang's amazing keynote at Computex announced some new Windows machines. There is a ton of Windows news and it's next on Windows Weekly. Podcasts you love from people you tr.
B
This is twit.
A
This is Windows Weekly with Paul Thurat and Richard Campbell. Episode 986 recorded Wednesday, June 3, 2026. Liminal AI it's time for Windows Weekly. Yes, you Microsoft lovers, all you winners and dozers. This is the show where we cover the latest news from Microsoft with two of the smartest whip smart folks in the.
B
I'm not that smart, Leo. I just engaged with Copilot by mistake.
C
Whoops.
A
It's not hard. Copilot keeps going, which I feel like
B
is how much people interact with it. Like, oh, what are you doing? Why are you wasting.
A
How did you get here? Oh, you pressed the Copilot key, Paul. That's Paul Theron, Copilot T H U double R O doublegood.com he is also the author of so many fabulous books including Deinshidify Windows, the Field Guide to Windows 11 and Windows Everywhere. Hello, Polly.
B
Hello Leo.
A
And how are you, Mr. Richard Campbell in Copenhagen?
C
I'm doing very well. It was a busy day today. I opened the conference with a keynote called after the AI hype and I did the last talk today on building data centers in space.
A
Oh, wow.
C
Yeah.
A
You used to do these geek talks as part of your Net rocks with Karl Franklin. You have these geek talks. Are you going to do either of those in your Net rocks?
C
I think the data center one will end up being a DNR at some point if it fits the genre. Well, it's fun.
A
Yeah, absolutely. I would want to record the AI Hype thing because who knows from day to day.
C
Yeah, there's a version. The one I did this morning was streamed to YouTube. It's already up.
A
Oh, nice. Oh, we can see that on YouTube.
C
Yeah.
A
Under what? What's the.
C
NDC Copenhagen.
A
NDC Copenhagen. And you have to say just like that. That's it, Just like that. Well, this was a busy, busy week for you gentlemen. Microsoft Build is going on. I know that because Christina Warren wasn't on the show Yesterday. She is GitHub and she was at Build and then she was building and then that's In San Francisco and then in a completely different time zone in a land far, far away. Computex in Taipei.
C
Wow.
A
And both had big announcements. Or to put it in the words of Paul on the show notes.
B
Oh my God, oh my God, oh my God, oh my God. I get overwhelmed easily.
A
What happened?
B
There's a lot going on. Monday was a lot of ranting.
A
Yeah. Jensen Huang's keynote at 6:00am My time.
C
Wow.
B
Yeah. So Jensen and the CEO of Qualcomm. Jeez, I'm trying, I'm losing my mind. Were both in at Computech. So they could not appear at Microsoft's conference. You know, unfortunately time for the same day.
A
So they did have stuff to say that was relevant, however. Oh yeah, Big time relevant. Although I thought it was interesting, Jensen's big announcement was actually not a Qualcomm chip.
B
Well, no MediaTek chip, which we knew. Well. Right, Okay. I mean we knew this was coming. Right. This has been rumored and there have been leaks and you know, whatever. We still. I gotta say I'm not super excited about the lack of detail here, but from what I can tell, and this makes sense given the nature of Nvidia, obviously these things are going to be GPU heavy. Right. Which is kind of a first for, you know, the Windows on ARM space, I guess, if that makes sense. So, you know, we'll, we'll see. I don't know what to make of this. It's so strange. So 3 nanometer process, I should say. Sorry. The rumors were that there were two levels of chip. N1 and N1X. The thing they announced is and 1X which we know because Jensen said it by mistake a couple times.
A
Oh really?
B
Yes.
C
Nice.
B
But it's called the Nvidia rtx. Spark is the marketing name or the go to market name. TSMC3 nanometer. A 20 core Grace CPU connected to a Blackwell RTX GPU. Which in my brain is more workstation than gaming.
A
Right.
B
Like, but horsepower for sure. It's definitely powerful. 6,144 CUDA cores.
C
Wow.
B
And. And then whatever this means. Fifth generation tensor cores. Like there's going to be 128 gigabytes of integrated unified memory as they say. Yep. Microsoft has been working with Nvidia to adapt Windows to this. And this is going to. I'm going to talk more about this a little bit later as we talk about build. But you know, think in the back of your brain. Just keep this idea in place. Which is. We've talked about this notion that two years ago Microsoft introduced Copilot Plus PC, which was MPU based on device AI. Lots of people have GPUs, whether it's a integrated GPU, dedicated GPU, laptop, desktop, doesn't matter. That are adequately powerful or more so than even to run those CoPilot PC local AI tests, but they do not. You're kind of wondering when or if that would ever change. They did not announce that, but there are some changes that if you. I don't know if I got this from the session or for something I read, but are basically being described as bringing copilot plus PC capabilities to other PCs. In particular this thing. Right. Which does have an NPU. I don't know how powerful it is, but Microsoft is adapting Windows so that those locally, I think will run on the GPU in this case. And again, like, we'll get to that. So we've been waiting for this. Right. I mean, Nvidia is.
C
It's black. It looks shiny.
B
Yep. Oh, I'm sorry. I should say too. Microsoft also announced for their part a laptop that will run this system. They're calling it Surface Laptop Ultra. It's not expected until the second half of the year. Along with all the other PC makers, you know, hp, Lenovo, Acer, Dell, everybody's
C
making one offer off of this Spark.
A
I think it's literally X Spark desktop. Right. It's just a laptop version, basically. Is that right?
B
Yeah. Although I think the chipset and this is new, right? Ish. I mean, you know, a new design, but yeah, but yes, I mean, I think that's fair to say.
A
Well, so you can't put a desktop chip in a laptop.
B
I mean, of course you can, but. So I, like I said they announced this. Microsoft announced their Surface Laptop Ultra. Between the two of them announcements, there's a little bit of information about the technical aspects of this. So I was thinking, well, build's going to happen and of course they're going to talk about this at build and they have a little bit. But again, very vague, like we don't really know exactly. And I think they're just waiting until we get, you know,
A
closer pricing availability. Who knows?
B
I think it's going to be several thousand dollars more than that because didn't
A
they say it's going to have a lot of ram 128. Yeah.
C
So my, my gut said 10,000 bucks easy.
A
Yeah, that's kind of what I was thinking.
B
Yeah. I was like north of 4. But yeah. I mean, I don't know.
A
But can you. But who would pay ten grand for
B
a laptop you know, former Net consumers.
C
The AI hype is real and they're going to keep going.
A
Well, Apple's going to have a, you know, similar device probably end of the year. I don't know how much RAM they'll put in there, but more than they could.
B
The way this was described was almost like it's. You have a data center on your desk, you know, in other words, this thing can run AI to the degree that right now requires cloud AI. Right.
A
So we don't have the models.
B
Well, I mean, that might be part of the delay here. Right.
A
So maybe this fall. Yeah, we'll see
B
if. Well, like I said, I'm going to talk about the kind of back end the software part of this in a bit, but this is very interesting
A
part
C
of the leak and it's definitely this push towards local model running.
B
Right. Which honestly is a major theme of build, even though they didn't really position it that way. Right.
C
If you say that out loud, it's like, why would you want this much horsepower on your desk if you weren't just going to run the model locally?
B
Most people don't need or want this much power in the desk. But as local models evolve and get better and as the software on top of it, especially Windows, evolves, you know, they become more capable and yeah, we're going to get there, there's no doubt about it. But let me see if I can find the language in here, because I did, I put this in, I just threw it in the notes as a little reminder to myself and now I can't find it because that's how my brain works. Yeah, they call. So when you think about on device AI, we're talking about what I would call small language models as opposed to large language models that run up in the cloud just as an easy differentiator. I think these lines are going to, you know, blend together or whatever. But in more than one place. Microsoft referred to this as unmetered AI, meaning it's local on your machine, so you're not paying for the token thing. And you know, whatever your usage cost,
C
you paid for the compute yourself. You get to run it as hard as you want to run it.
B
It's private, it's on your device.
A
I mean, I love this idea. If, if I can get a model that's going to be comparable to a Frontier model.
B
Yeah, right.
C
And it can heat a room.
A
Yeah.
B
So I mean, it's interesting to compare what we do know to what we found, what we think we found out through leaks. Assuming that stuff's correct. Right. So like a typical laptop CPU these days is actually kind of hard to pin down honestly. But in the old days, meaning before the core Ultra chips like on the intel side they would have the U series, core whatever generation and those things U series would typically like be 15 watts. Right now we have multiple, not just multiple cores, but multiple kinds of cores. And so there's kind of a range of power that we use to describe their power consumption essentially. So I would say a typical laptop today probably is somewhere, you know, 10 to 30 watts maybe somewhere in there. The lower end chip that they did not announce was supposed to run at 18 to 45 watts. The N1X which is the one they did announce runs at 45 to 80 watts. So without what I took from that before you run the rest of the
C
machine, you've already need a power supply right now.
B
Oh yeah, a lot of power. But that's gaming laptops, that's portable workstations, that's desktop PCs. I mean this is, you know, it's not a thin and light. Although they were talking thin and light, you know, so we'll see. I. The RAM range could be 16 to 128. They came to market with 128. The, the RAM on the lower end one is 8 to 16. Supposedly they did not even announce that. So I don't know. You know, I feel like we are going to see the other one too. There'll be a lower end solution for cheaper computers. I think we're going to see lower end configurations of this thing. But I think the go to market is very much to attract those people who are heavy, really heavy uses, like engineers, maybe creators. This is devoted to.
A
Because Nvidia already offers a, basically a Raspberry PI with Cuda cores on it that's cheap and can't run diddly basically. I mean, so because you can run
C
it doesn't mean it's usable.
A
Yeah, I mean they have a model, it's okay. I mean they have a very small model and Google has Gemma.
B
I think this is just about getting people on CUDA frankly. I mean, you know, which is, it's
A
totally what this is about.
B
Completely understand who to lock in.
A
This is like if we can just own the whole market.
B
Yeah. Like eventually this surging market thing that we're part of, which is insanity is going to slow down or whatever and. But now we have the market, you know, so we'll.
A
And this is why I'm kind of resistant to it because I. Yeah, well,
C
and we've Seen PCL seals drop across the board. I think it's exactly that. And now you're going to ship arguably the most expensive laptop ever made.
A
Right.
C
Your timing couldn't be much worse. Except. Except for that part where it's that message about local AI.
B
But unless the goal isn't. Look, the goal can't be volume. It's the PC market. They're going to sell 200 million ish computers this year and only a tiny percentage are going to be these kind of things. So this is the same situation we had at Qualcomm two years ago, where at that time a smartphone market that was over a billion devices a year. And you're like, why, why even bother? Like, why would you even try to go to the PC market if you got 10% somehow? And I don't think they have really, but let's say they did. So you're selling like 20 million of these things a year compared to a billion, like.
C
Right.
A
It's what Apple used to call the Pro line. This is for. Yeah, but that's why I agree with you, Richard, that this is more about Cuda. It's about mind share, not market share.
C
Sure. Well, and it's about, you know, the big aspect here is investors are worrying that these investments in AI aren't going to last and you've got to show motion. So the idea that we're pushing on new machines that are going to open more doors for more users, anything to keep the investors engaged and not running for the hills.
A
Well, and enterprise is starting to say, holy cow, look at our token bill.
C
Yeah, right.
B
Well, yeah, so the timing in that sense is actually pretty good because all the major AIs are switching to that consumption model and bills are going to go up not exponentially, but a lot. And I think that a lot of the people, companies, whatever, who are using AI today and paying 20 bucks a month or 200 bucks a month or whatever they're paying don't understand the true cost. And they're about to. They're about to find out by bit
C
this is coming in. And so good to have the alternative in your hand so you don't abandon the tech.
B
Just as a little bit of commentary that I'm kind of still pushing through my brain. Microsoft doesn't get a lot of credit for this kind of thing because they actually don't do this kind of thing a lot, especially recently. But they were kind of out early on this, if you think about it. I mean, when they announced Copilot PC two years ago, my Evaluation of at the time was like you. What you should be pushing because it's ARM and it's Snapdragon is efficiency, reliability, power management, battery life, you know, and then just general excellent performance. But whatever. But they were very much pushing local AI. And it's tough because I. No matter what you do on a computer, you sit there all day long or you just use it occasionally. Whatever. Use what? Web browsers, email and whatever applications. You might have some one to seven little tasks within individual apps that might be better or more efficient because of an mpu, especially if you're a creator. There are certain features in Windows you only get from a Copilot PC. So you either do or do not know that that's the case. Whatever. But it's not a reason to buy it. Right. It still to this day is not the reason to buy the thing. So this is another. Here it is two years later, this will be a lot more powerful, apparently consume a lot more power. But two years have gone by. Copilot plus PC I don't think you can say, has moved the needle per se on local AI usage or awareness or PC sales or however you want to say it. And so it's like, oh, we're going to try this again, sort of. But I mean, you know, Microsoft, look, we're going to talk about Stevie Batish again in a little while and he was the guy, you know, three, five years ago, whatever, was who gave that very eloquent description of those three AI app models or app structures. But the following year he talked about NPUs versus GPUs. And for a laptop, when you can offload certain AI tasks to an mpu, you get better performance, but you also don't impact the battery at all, which is not something that happens with a CPU or gpu. So it's still important. But if you do have the big gpu, you bought it for a reason, you want to use it. This is the argument we've always made. Yeah, you should be able to use it.
A
Why a laptop?
B
I mean, because that's what people buy. That's the mass market. Yeah.
A
Okay.
C
Yeah.
A
So this is to put a local AI into the hands of the masses.
C
Yeah, I think, I wouldn't say.
B
I don't know. Yeah, not, not at first.
A
Top of the line machine developers want laptops.
B
This is for the masses, not the masses. Leo to riff on what's his face? Jack Trammell. Right. I mean, but, but it will come down. I mean I, I think this is to jumpstart more adoption of local AI across the board from developers and well to developers pretty much from the developer space. But. But not just developers that want to use it locally for their own needs, but developers who are creating apps that may use local AI features instead of
A
just relying on the Microsoft Surface Laptop Ultra, which Microsoft said will be the best fastest laptop we've ever made.
B
Yeah.
A
Is arm, right?
C
Arm.
A
It's not Qualcomm.
B
Right.
A
So that's interesting.
B
Well, I mean There's intel and AMD in the x86 space. There are, have been other.
A
But this is their first non Qualcomm Surface. Yeah, so I mean non Qualcomm ARM Surface I guess.
B
Yeah. And the Mod for Windows 11. I mean the, the original Surface RT which was armed was Nvidia by the way. But.
A
Oh, but yes, yeah, way back when. Like the Tegra X or something. Yeah, yeah, yeah.
B
But this is. Yeah, I mean, but look the thing
A
is they had, we thought they had a Qualcomm exclusive deal for a long time.
B
Yeah. This has never been, you know, confirmed. But the, the story is that there's X amount of months or years they had this exclusivity thing.
A
Right.
B
In return for that, of course. Qualcomm has done an incredible job right from Snapdragon XX2. Unbelievable. And they're addressing what I would call like kind of the, the right part of the marketplace for the mainstream. Like they're, you know, the first devices were often for at the time. It's funny because they had these would be cheap but kind of premium thin and light laptops. You know, it was designed to be attractive to people who might want to upgrade and spend a little bit money, A little bit more money maybe, although they weren't actually more expensive than other things. And then they've scaled it downward and we'll talk about this a little bit later because they scaled it downward yet again and that's going to help out a lot this year with the component crisis. But yeah, this fills a need I guess at the high end of the ARM side of the
C
stack.
A
I think we're going to talk more about these other Snapdragon C and stuff. Do you think that the market is now starting to be high end Nvidia, low end Snapdragon C and intel and then kind of medium intel amd. I mean is it starting to.
B
This just happened. So I can't, I don't want to make any predictions or whatever this is
C
on anybody's hands yet. They're just being announced. Yeah.
B
This is September.
A
Right.
B
We don't have detail, we don't have benchmarks, we don't have, you know, anything. So it's like, look, it looks great, but what is.
C
Are you going to be the list, Paul? That's the question. Is an Ultra going to show up at your door?
A
Oh, yeah.
B
I don't know.
A
Somebody said it. I thought this was hysterical. Maybe it was the Verge. This is a replacement for that Snapdragon dev machine.
B
Oh, yeah, no, those things are totally related in the same way that a Lamborghini is a replacement for Model T. Right. You know, like, I thought that was
A
a very weird thing. I don't know.
B
I mean, the dev box, all the dev box was, was like a Mac MINI style all in one, not all in one mini form factor. It was essentially laptop chipset. Right. Like. So the goal of that was just to get people, developers on arm, you know, obviously. I mean, this is, this is, this is different. I mean, it's arm. Yes, but ARM is already established with. There are, there are no, there's no remaining problems there. They're scaling upward.
A
Didn't Jensen say it will run every app that runs on Windows? Something like that. I mean, he really was at great pains to get.
B
Because you can't read a review of any Snapdragon laptop, even as recently as this week, without someone saying, well, the. There's, you know, you have to deal with compatibility. No, you don't. No, you don't.
A
No, you don't.
B
It's ridiculous.
A
That's the.
B
I'm sorry, that's silly. But he has to say that because that's the perception. But the other half of that, by the way, is if you went to Computex, which I did not, they're showing demos of games running on this thing. And of course those things are emulated. He did, right? He held it up. Oh, he showed it. Okay. Yeah. So that's, you know, interesting. I, I will. You know, you can run games on Snapdragon X, you can run a few more games or run games a little bit better on. But emulation is great for an app. It's not necessarily great for games, which is why we have all those other things, like, you know, super resolution and, you know, whatever else that they do to try to make these things just kind of run better. Nvidia is benefiting from all that work. It's just, you know, it's not like they're starting. It's not like they're not. It's not like a. I don't know, like a big block engine that just is super fast. So it could just run these things like it has to, you know, work still has to occur under the covers than it has. We've had two years of that. Two years plus. Really. So they're going to benefit from that. So their GPU like, you know, on a normal x86 laptop with an RTX, whatever. DGPU games run great. Like Call of Duty runs great on those, like really great. Like 100 something frames per second, full res, all the details on it's awesome. That's x86. So if you can get a game like that to run at like 60 frames a second on this thing, that's also awesome. So that would, you know, that's amazing. Especially for something that's emulated. Right.
A
This feels like such a big victory for Microsoft. They were front and center and it was very clear that these were Windows machines. Yeah, this was really, I thought, very interesting.
B
I just. This is what, this is why I mentioned, you know, we don't usually give them credit for this because they don't usually do this, but they actually came out very early and publicly about local AI and were ignored completely, you know, pretty much. Right. And I don't know, that's maybe that they should, maybe they shouldn't have pushed that part of it so hard. But, but like you know, and I mentioned Stevie Batish, who we're going to talk about again, you know, a couple three years ago, whatever that was, when he talked about mpus, he really made a good case for why these types of chip or whatever they are, I guess they're chips inside of an soccer are so important in the portable space. But you know, that's that thing we keep talking about. You have a desktop computer or a laptop with a dedicated GPU or a GPU and a card. It's got dedicated RAM. It's hundreds of tops, not 40 or 85 tops, several hundred tops in many cases. Should be able to blur the background and paint or whatever, you know, imagine.
A
Yeah. The model I have in my mind, and Steve Gibson talked about this yesterday too, is a home server, a NAS maybe or what I have for the Framework desktop, which you then language device. Yeah. And that's what's running them, you know, that has all the horsepower, it has all the electrical power.
B
You know, Sun Microsystems called and said the network is the computer Leo. And no, I, Look, I.
A
That's just the model in my mind. I mean, I don't know why you would want it on a laptop. If I have a laptop, I'm going to SSH into my big machine or something.
B
You wouldn't want this, right? Unless you were a developer.
A
Well, I would want it as a
B
desktop, but the reason you want local AI on a laptop, just in general. Right. And even today it works offline, it's private, it's your data, it's not going up to the cloud, it's this reason. Right.
A
So that means you want the biggest machine you can possibly get to run that local model.
B
Right.
C
And that's really considering how limited information.
B
Well, but that was the point. The point of the MPU was you don't need that super high. I mean even though there's a good computers.
A
But you demonstrate that to me because that ain't the case today. You need a lot of ram, you need a lot of horsepower, you need a lot of electricity to run.
B
Well, no, yeah. To do the thing you're doing in the cloud. But that's not all that AI is if you're going to fire up a bunch of agents and have them go off and do whatever they do.
C
Yeah.
B
I mean you're going to need more power or a cloud based thing. But I think this is just the market evolving to the. I think these things are converging. The local AI is getting better, the local hardware is getting better. I mean I think ARM makes a huge difference. Just in general.
A
This is primarily for inference and I would say primarily for coding. Is that right or no,
B
your place games. No, I don't. Yeah, I mean, yes, I. Oh yeah,
A
it's a, it's a general, general purpose computer. Not going to deny that. But the market, selling this expensive, you can get a cheaper gaming machine. The market they're selling this to is, I would think coders, but I don't know.
B
And I would call it engineering science.
A
Engineering. Yeah, maybe you're right, maybe cats.
B
But, but delve developers. Of course it's not going to be something on someone's, you know, desk at American Express, you know, taking calls from customers.
A
I mean did they talk about. Because I think this is underestimated and I haven't seen a lot of numbers but it turns out throughput. RAM throughput is really, really important and
C
there is a big.
A
In fact some people have complained that the DGX Spark just doesn't have very good RAM throughput.
B
So I'd have to go and even look at that to figure out what the architecture is there. But it's probably just a standard, it's probably, you know, so, so damn, that's really important.
A
And actually Apple has a pretty good story on that with its unified ram. So.
B
Yeah. And Right.
A
Throughput numbers is what I guess I'm saying.
B
Yep. Yeah. And that we don't have a lot of the details. That's part of the point.
A
Yeah. Right.
B
So on. I know on Snapdragon X2 Elite Extreme, which is where the RAM is integrated into the die, the RAM throughput is not double, but it's close to double.
C
Right.
B
You know, that's how important that is.
A
It's huge.
B
It's everything.
A
Yeah.
B
Yep.
A
Let me take a little break. And we have so much more to talk about. We're just scratching the surface of Computex Build. And Richard Campbell came and went like a ghost.
B
He's like the wind.
A
But we will have more in just a moment. You're watching Windows Weekly with Richard Campbell and Copenhagen, Paul Thurat and Makun and Leo laporte. In my attic. Attica. Attic. No, not that attic. This episode of our fabulous show brought to you by our fabulous sponsor, Delete Me. And man, you need this today. You need this today. Have you ever wondered how much of your personal information is out there on the Internet for anyone to see? I think nowadays we're kind of more aware of this than ever. Used to be you'd Google your name and go, whoa, whoa. Now it's become so apparent. There's been so many news stories, we've covered so many stories about car manufacturers selling your information. Most of the state, aca, the Obamacare websites, are selling information to data brokers. Personal health information and things like that. Income information to data brokers. It seems like everywhere you go, your phone, your.
B
Your.
A
Your Internet browsing, all of it is being exfiltrated to these companies that in my mind are completely shady, but it's completely legal and it's extremely profitable. And I'm talking really, truly personal information. Not just your name and email, but your home address, your Social Security number. It is. There is no law that says they can't sell your Social Security number to anybody who comes along on the Internet, whether it's marketers, law enforcement, hackers, foreign governments, even information about your family members and your company, your employees are all being compiled by data brokers and sold online, where anybody who can get to the Internet can buy your private details. And it doesn't take much imagination to think about the consequences. Identity theft, phishing attempts, doxxing harassment. So you really need to. We all know you need to protect your privacy. And the easiest way to do this is delete me. I just saw an article talking about how data brokers play games. You know, federal law does require them to have a somewhere on their site. If you can figure out their name and their site somewhere on their site. A place where you could say, remove my information. But good luck finding it. I just saw an article about how they're playing games with these removal pages. They don't want you to find it. Add to that, there are more than 500 data brokers. New ones every day. And then they play other games like changing their names, pretending to go out of business and starting up a business with a new name, things like that. They don't want you to delete your data. That's why you need Delete me. I learned about this with our own company when we were getting phished. We still get phished all the time. And the things that the the hackers know about us, really, they shouldn't know. How do they get it? Data brokers. So we signed up for Delimia subscription service that removes your personal info from hundreds of data brokers. They know their names, they know where those pages are. This is their full time job. These experts at Delete me, you sign up, you give Delete me what information you want deleted. That's important because you have control of that. Not maybe you don't want everything deleted. Maybe you do. You tell them what you'd want deleted. Then those experts take it from there. And then they will send you regular personalized privacy reports so you know what they're doing. It shows you what info they found, where they found it, what they removed. It's really complete. We just got one the other day. Delete me isn't a one time service. And that's really important because as I said, it's a moving target. There's new companies every day. It's too profitable. Right? And it's completely legal. Deleteme is always working for you. Constantly monitoring and removing the personal information you do not want on the Internet. Trust me, you do not. Deleteme does all the hard work of wiping you, your families, your companies, your employees, personal information from data broker websites. Take control of your data. Keep your private life private. You can sign up for deleteme. We have a special discount for our listeners today. Get 20% off your Delete Me plan when you go to JoinDeleteMe.com TWiT and use the promo code TWIT at checkout. And the only way to get 20% off is to go to JoinDeleteMe.com TWiT and then make sure you enter the code TWiT at checkout, that's JoinDeleteMe.com twit and the offer code is TWiT. We thank him so much for supporting Windows Weekly and doing really important stuff. Back to the show we go. We had just begun. We've only, in the words of the Carpenters, just begun to talk about the Windows world. And Computex.
C
And I speak on mute.
B
Yep. So I looked this up while you're doing the ad there, but they wish
A
I were on mute right about. Right about now. I just want to say. Yes, sir, go ahead.
B
First gen Snapdragon X, the memory bandwidth was 138 gigabits per second, gigabit bytes per second. I guess Snapdragon X2 Elite is 152, so a small gain. And then if you go up to X2 elite, it's a 50% gain from that. So 228Gbps.
A
And this makes a difference in. In. In AI.
B
Makes a difference in everything. I mean, but especially in AI. Yeah. Okay. Okay. Yeah, just. I was just curious. I. Yeah, okay. So yesterday it was the build keynote. It was two and a half to three hours long. I completely lost my mind by the end of it. But unlike the past, I'm going to call it 10 years or so, there was a bunch of stuff in there for Windows and for the client, which I think is kind of cool.
C
Yeah, look, new stories about Windows is amazing. When does it build? How many builds have we had with nothing about Windows?
B
Right. And this is. I think this is partially due to the fact that cloud computing just didn't have anything interesting for Windows. That was the focus for such a long time. I don't think it made it very interesting for people who had considered themselves to be enthusiasts or whatever. Whereas AI, even if you don't like AI, like it's generating news and excitement, it's interesting, like this stuff happening. So to me, this is interesting. Microsoft, I mean, this goes back to Windows 10, really, but they've been trying to position Windows as the best desktop platform for developers for a long time. This is why we have WSL, the Windows subsystem for Linux, etc. Etc. You know, I don't remember the timing of any of this anymore, but, you know, Dev box came out of this, all the improvements to terminal, et cetera, et cetera. I have to say, super excited. I. Unless I'm misremembering, which I could, because it was a long event, I think the first person that came on stage that wasn't Satya Nadella, was Kayla Cinnamon, who does I develop advocacy, but she's in terminal. Right. And she has an awesome YouTube channel by the way, where she does a lot of terminal tips and tricks and stuff. Excellent.
C
Done some great run ads with me too.
B
Yeah, she's great. So she did some terminal demos like right up front, which I thought was really cool. So they announced a bunch of things. There is something called the Windows Developer Configuration, which is what these new Nvidia based PCs are going to ship in, which is a Winget or a Windows Package Manager based configuration that creates a distraction free environment for developers. Now didn't say this, but to me that sounds like Xbox mode, but it's like developer mode for developers or whatever. Right. So less stuff going on. Widgets aren't popping up, notifications are off, et cetera, et cetera. There are optimizations in there for things like visual studio code, GitHub, Copilot, WSL, et cetera, et cetera. So that is actually out there now if you want it, you can just install it. Like it's. I think it's on GitHub. So if you want to just, you know, one line and auto configure computer to be less annoying, that's something interesting.
C
Dev config can be annoying annoying too, just in a different way.
B
Yep, it's in dark mode. You know, I don't know.
C
Yeah, the.
B
We talked about this at least twice previously, but back in January Microsoft announced something called the. I think it's the Windows app development CLI, the Win app CLI, which as originally released in 0.01 forum. I didn't quite understand the point of it. They later added support for workloads like. NET and all the. NET frameworks like. WPF, etc. I was like, okay, kind of interesting. But now it's making sense because that thing is now generally available and there is a plugin for it called the Windows Developer confid. Nope, called the Windows Developer Skills that allows you to use the win CLI to create in this case native WinUI 3 apps from the command line. And I'm going to talk about that in the back of the book because I did it and. Okay, very interesting. So what I'm going to say though just here is that CLIs are huge this year all of a sudden. And it's the same reason that markdown is huge. It's. It's because of AI and agents in particular.
C
Thinking tech based. Yeah. Rocks. A few weeks ago that was like build your CLI first.
B
Yeah. And this is that thing we Talked about a couple of weeks ago, you know, back in the day, I think it was exchange 202007 we're going to build this thing in Powershell and then we're going to build GUI on top of it. That made sense to me architecturally or whatever. But now in the AI era, we have a newfound need for this kind of thing. So we have apps that exist and we live in kind of a screen scraping world because we don't meaning compute use where we have to deal with what people have. But apps also can be adapted to expose their capabilities, individual features through a programmable interface, which on I think on Android is like app intents maybe. And on Windows it's called AI Actions or just App actions. And this is when you can right click on a file in File Explorer and go to AI Action submenu and then like remove background with paint or whatever. Right. It's exposing a feature which is okay for people, but it's awesome for AI and that's why that exists. And then the CLI stuff is the same thing. It's IT but for agents specifically. Right. That having a C CLI interface to anything helps you do a bunch of things. But one of the key ones is grounding AI in something in particular. So when you use this Windows Developer skills plugin, whatever AI you choose, you can use, you know, Cloud or OpenAI or Copilot and then eventually everything. But today those things, it will only be grounded in all of the documentation and learnings they have for whatever environment you're working in. So if it's winui. One of the problems I've run into, and they specifically called this one out, is General Purpose AI or General Purpose coding. AI will pull in stuff that's from WPF or other frameworks that are similar and it will make mistakes. And you run into that thing where you're like, okay, we fixed it. You're like, no, you didn't. Still doesn't compile. Oh, okay, I see the problem now. The problem is and it talks and talks and talks and you run out of all your credits. You're like, what the fuck? This is a solution for that, right? Yeah, is run out of credits.
C
I tried to make an animation today for my data centers in space Talk and ran out of tokens on like three different services and got nothing.
B
Yeah, that's a good experience. Such standard Microsoft productivity experience really. There's a new intelligent terminal terminal mode coming. It's an experimental preview right now. But basically what this is is another way to interact with agents. In a command line where the top half is the normal terminal window powershell, typically the bottom is like the Agent CLI interface. And so you can go back and forth between the two. Kind of interesting. Lots of WSL improvements. I don't think we need to get into too much. But the big one is container support is coming. So that's awesome. And then they released our utils, which is like over 75 clis from Linux now in Windows. Right. That weren't there before. So again, really pushing the kind of command line part of it. They have something called Microsoft Execution Containers for Agents, which. When I first saw that name, I was like, oh, my God. Oh, my God, they're doing it. They're doing Windows 10X. No, it's not Windows 10X. This is so agents running on Windows can run in their own little sandbox environment. Right, Right. And then you can feed it the data you want and it won't be able to bleed out.
C
Yeah, this is. This is how like Claude Cowork works. Right? Is it sets up a VM on your machine, restricting its ability to access the machines. You have to specify what on your machine you're allowed to use. That'll jump across that VM barrier.
B
So I haven't done. Actually, I think it's fair to say I haven't done anything with cloud code from. Directly from a command line other than like through a chatbot interface, which is sort of a command line. Right. But when you do that, my God, does it ask you for permission all the time.
A
YOLO mode, dude. Whenever you launch, that's it. Dude.
C
Dude.
A
No one does it that way.
C
Run unsafe.
B
Yeah.
A
So when you invoke Claude code, you type Claude. Dangerously bypass permissions.
B
I don't like the sound of that,
A
but called YOLO mode.
B
Oh, no.
A
Everybody. Because you can't get anything done. You're constantly.
C
Otherwise you just constantly manage your mess.
B
Okay, so in the workflow that I will talk about later in the show, generally speaking, you. There was an option. It was like, yes, dear God, yes, you can access the file system. Relax. But there was a second one that was like always allow. Like in this case, you could kind of. You could, you could, you could. Okay, I just. I don't know.
A
I know it's scary, but.
C
Well, slash, dive into deep end.
A
You only live once, Paul.
B
Yeah. Yep. A cliff at the end of that.
A
No, I run everything that way. Everything. All yolo all the time. This is the story of my life.
B
Okay, there's your title. I'm not, I'm not In a place where I feel comfortable doing that or recommending it to other people, but I. Okay, maybe you're not alone, Leo.
C
I know a bunch of you have to.
A
It's because otherwise it's. It's worse than uac. It's like. Yeah.
C
And their gay check is just at a different point. Right. It's the pull request. Right.
A
It's all right, Paul. Once you get some comfort with it, you'll realize it's not going to delete your database or anything. It's not.
B
You just got an apology from AI that deleted something. What are you talking about?
A
That's a good point.
B
Anyway, it's okay.
C
And it felt bad about itself.
B
Yeah.
A
And it felt bad. We're talking about Opus 4. 8, which is really good at apologizing.
B
Yeah, I really. No one likes data loss. That must have hurt a lot. I'm so sorry. Okay. Okay.
A
I have very good strong backups of everything. I'm not. I'm not too worried, but it really. Well, sorry. I'm not going to argue.
B
No, I'm not arguing.
A
I'm just.
B
I'm just saying.
A
I'm not arguing. I'm saying I'm not gonna. I'm not gonna make a case for yolo. But I feel like you did see
B
soon why you just did make the case for it. Like it's okay. I was not aware of this option. I'm not sure I want to be aware of it. I don't know. It's okay. I. Look, I'm just baby stepping at this point in this stuff.
A
Dangerous skip permissions. Paul.
B
I'm sorry, Dave.
A
Paul.
C
Just think of it as permission to be awesome. Yeah.
B
Okay, so the on device AI thing has come up, right? And Microsoft, if anyone has ever tried to, and I have, use the APIs that Microsoft has talked about, which have changed dramatically over the past two, three years. You know, there was like the Windows Copilot SDK or whatever it was called, or the whatever the heck that was, and they changed the name and blah, blah, blah. And basically what these things are using is Microsoft small language models running on the device, mostly in the five family, I think is where most of this stuff comes from or has come from. But they announced a new set of these models which have a new name called Aeon. If I'm pronouncing that correct. There's AI Aeon 1.0 instruct, which is smaller, faster and smarter local AI than Phi was. I guess Aon 1.0 plan, which is a reasoning and tool calling model for AGENC capabilities, et cetera. Both are coming soon, but you can experiment with them in preview today in various ways. And they're expanding. I didn't. This must be the new name for that, whatever that Copilot SDK thing was, what they're calling the Windows AI APIs. So there's the Windows API, there's the Windows APIs like Win32, there's the Windows App SDK APIs for modern app development, and there's the Windows AI APIs. So there are three sets of capabilities that are moving off of being copilot plus PC requirements with an MPU. So speech to text is now going to work on CPUs in addition to MPUs. Text intelligence GPUs. It just says CPUs, but it could be. I'm not sure about that. Text intelligence capabilities will work locally on DGPUS and then Video Super Resolution will work on CPUs. Okay, so again, they have not said how or if they are changing Copilot plus PC as a spec, but we've made this case on. And now I will say on both ends of the spectrum we've got the Nvidia thing on the way, high end. And then we get the Snapdragon C, which we'll talk about in a bit on the low end, which again, which we also do not know a lot about, but we do know 8 gigs of Ram on the small one. So that's not a copilot plus speed.
C
It's not 128 gigs.
B
Yeah, no, it's not 16 either, which is the baseline for. So the theory is this is going to change and these API changes speak to that. So you can preview these today in Edge Insider channels. I'm going to talk about. I think the Edge is probably the next one and they're coming to hugging face in July and then the general availability later in the year. They haven't said. But this is a kind of a generational evolution of, I believe of Microsoft 5, which I assume is going to go away. But they did not say that. And then there's the. The RTX dev box which we talked about earlier. Right. This is I. The, like a. A data center in your lap, you
A
know, instead of a blanket, you got to do this.
B
Yeah, it's going to be.
A
I actually asked my AI to make us a bandwidth chart because they are. There is some bandwidth information out there. And you know, the big boy is this. I don't know how much it costs. Nvidia's DGX station, which is 7 terabytes per second of bandwidth. And then if you have an RTX video card that's a lot 1.7 but it's only 32 gigs of RAM so you can't really fit a model in that. But if you could it would be fast.
B
Well that's specifically on the GPU right. Like in other words the one. Yeah that's a GPU that was integrated or unified RAM.
A
Right. The Apple has a real advantage here. The M3 Ultra is almost is 819 gigabytes a second M5 max. The new RTX Spark is actually pretty good 300 gigabytes a second. The DGX Spark, their desktop is actually considerably slower. The Strix Halo which I have is 256gigs per second. And then you get to aerolink and stuff like that and it's 102 gigs per second. So there is a big gulf. But this, this RTX Spark is going to be pretty good. 300 gigabytes a second. Is.
B
Is this is.
A
This is the. What they announced anyway. So. Okay, I just thought I'd pass that along.
B
No, that's crazy.
A
I had that had it generate all that and I wonder how I'm gonna just go price a DGX station. Just. I'll be right back. If I have a heart attack, I'll let you know. Yeah, my falling over.
C
All right. A little hard.
A
Yes. Leo,
B
Google just got in trouble for secretly although they announced this downloading a 4 I think it's a 4 gigabyte SLM with Chrome, you know when you start using AI features. I made the point at the time like everyone is doing this including Firefox. By the way Microsoft is doing it in Edge and they announced separately from what I just said. But it's the same thing is those on device AI capabilities will also be available in Edge and you can test some of them right now through I think it's Edge Canary or whatever. Actually one of them might even just be available unstable at this point. But Edge Canary Dev, et cetera have access to at least these capabilities. And so as a developer you could write whatever apps that run against these things and it would run locally so it could work offline, et cetera, et cetera. So we'll see.
C
Cool.
B
Yeah, I think just I'm just going to do this kind of quickly. If you looked at build from the outside, didn't know anything about it and tried to make a general summary of what was going on, probably using AI, it would be that this is an AI agent show. Right. I don't think I did this last week. I must have done this more recently. But I looked at the sessions that Build had and there were like over 400 and over 300 of them had the terms agent or agentic in their title or description. Right, so.
C
And the other hundred were wrong.
B
Yeah, right. And then, you know, like, you know, Copilot, for example, only came up like 117 times or whatever the number was. You know, I honestly, aside from the final thing we'll talk about, I think for Build, I would say most of this wasn't super interesting to me. But I'll just mention a couple. They were rumored, they, Microsoft were rumored to be working on a so called like AI super app or whatever. They have announced that it's called Scout. It is their version of Google Spark, right. Which is I think now available to people paying Google 200 bucks a month or whatever on that subscription. And it is just early days. Right, but this is where we kind of move. So we have agents and obviously we're familiar with the term copilot, but they've introduced the fairly obvious language of autopilots, right, which are agents. So this is going to be like a. This will probably eventually replace all the Copilot apps and stuff. I think this will become the experience, you know, but it's a, a desktop app, I think. I'm sure there's going to be a web version, etc. But. And I'm sure it will do hybrid things, you know, a local and cloud based AI, et cetera. But early days for that. And then they also announced a GitHub copilot app which made me do a double tape because I was like, isn't there already a GitHub copilot app? But I think it's only on. Maybe it isn't, Maybe it's just GitHub, I guess because I have a GitHub app on my phone. But this is the first step to a big ancient native desktop experience for GitHub. And so it's basically for managing agents, essentially, I guess is the way I would say it. Of course it is. And you know, using Git on the back end. Right. Which does make sense, I guess, for these projects you might be doing with this thing. And I don't know, I'm just, this is like poof. I don't know, whatever. It's on preview.
C
Don't despair.
B
I'm not despair, I'm just, I'm just, I just don't know what to make of this quite yet. More interestingly, perhaps it was Mustafa Suleiman came out and announced that his organization, which is Microsoft AI Mai, which is seeking to basically replace all of the reliance Microsoft now has on OpenAI in particular, but also I would say anthropic, announced seven new models, including the first reasoning model that also offers like high efficiency and low token cost. And it's kind of interesting. It's like they claim that most of this stuff, I would say is some number of months behind the best, very best models across the board. Which is not bad because again, they're kind of building this up from scratch. They couldn't even start working on this until they changed their agreement with OpenAI. One of the two agreement changes ago that was the big deal for Microsoft. It allowed them to pursue this stuff internally without OpenAI's help or okaying it even. So they have an image, you know, Mai image 2.5 with a flash variant, which they claim is better than Nano Banana based on benchmarks and whatever. However you measure these things, not just based on their personal opinions. Like I like that picture of the tiger better. Transcription model. I know. Am I a voice? With a Flash variant again for natural generation, which is already very good by the way. I believe they're. Yeah, well, based on the version, it is their first coding, like inference coding model M I, M A I code 1, which is available in GitHub, Copilot now and VS code. And they're going to make these things available across more platforms in the future as well, like fireworks, AI open router, etc. So it's kind of a big deal, actually. Yeah. And he, you know. Yeah, he keeps saying all the right things. Mustafa Suleiman, that is. But the best part of this keynote and the one that really made me kind of perk up was Stevie Batish came out and we like Stevie.
C
He was one of the first guys back in the bit 23 that inspired us.
A
Yes.
B
Yeah, right. That was, that was the build. It was. What's his face? Panos. Panay's last build, which it looked like he was having. It looked like he was having a stroke. Like in the annals of like horrible industry conferences that went wrong, I would say Gil Amelio announcing that Steve Jobs was coming back was still probably number one, but he wanted babbled for three hours. But panel spinae in 2023 was a close second and we figured out what had happened.
C
Right.
B
Yeah.
C
They took away his hijack from him. Yeah.
B
And he was just making it up and man, he was not doing a good job. He's not Good at that. He was, you know, he needed. No, he had a content.
A
It was.
B
No, but he needs. He's the type of speaker who needs rehearsal and active audience participation. He was very, you know, didn't go well for him. But Stevie Batish came out and he's like, look, I gotta run to the airport. I only got like 15, 20 minutes. And then launched into the most eloquent description of how AI was gonna. I. I re watched that so many times.
A
Does this still.
B
Yeah. Oh, yes. Yeah. And that's actually.
C
We refer to it all the time.
B
Yeah, he referred to it yesterday.
C
Yeah. Three years later.
B
Right. So remember well. Or know that there were three what he called AI app structures.
C
There was inside, outside, B side.
B
Yeah. Beside was the first one that they don't naturally not necessarily mean they go for one or the other, but they all exist. Or, you know, the first one was Beside, which is a copilot, right. Which can be literally of an app and then maybe like a chatbot on the side and two different apps or maybe you have a pane inside the app because you've added it, you know, whatever. And then AI inside, which is where you actually integrate AI capabilities into an app. And we see that in various ways throughout like Microsoft Office, obviously. But then there was the AI Outside, which I think is the one either was a little bit misunderstood or he's expanded the definition here. But AI outside was the sort of background for the talk he gave at this show, which is about something called Project Celera. And Project Celera is, man, it's hard to even explain because it's so different from the way things are and have been forever. In the traditional kind of personal computing model that has existed since the Commodore 64 or whatever, someone builds hardware, there's a software platform you can write apps and games to, or whatever you do that, there are different ways to do that and then they run on that thing. And pretty much obviously we have cross platform solutions, whatever, but it's like device with apps and stuff running on it and stuff, meaning like whatever, services, et cetera. Windows works this way, phones work this way. I mean the web sort of works this way, whatever. But you target this very specific thing. And so the idea behind Solara, and he didn't quite say it this way, but this is almost like a modern answer to the Next wave problem that Terry Meyerson always had. Like Microsoft missed phones. So we want to be on the next thing, which maybe the next thing at that time was going to be VR, Mr. Arkansas, or maybe it was going to be something like Cortana, which is a smart assistant. And then that of course evolved into what we have now, which is AI, which is. You know, you could make a good argument this is the next platform.
C
Not that it was a straight line.
B
No, not.
C
My only concern here is then they rolled out with something that looks disturbing like an Amazon Echo.
B
Right. Well, so the point. But this is. This is the. This is what's interesting about this. So there, each of these devices, it's. It's meant for a. Like a multi device agentic first kind of world. Meaning that the. There's an. There has to be a platform layer on there of some kind. But. And it's a little bit like a. Like a Net PC or network computer or whatever. It doesn't necessarily have to have a lot of compute or RAM or whatever, but it can. And that story, the thing you're talking about does. The two types of devices they showed off was like a smart badge essentially. And then that little screen thing which does look like an Amazon Echo or whatever those are called. And. But what makes this different from everything that's come before. And this is very much like Microsoft used to talk about stuff like this all the time. I don't mean stuff like specifically like this thing, but rather they would talk in this way about the future. Like this is how the future is going to be. And we're like, oh my God, this is going to be amazing. They haven't really done this in a long time. That the center part of this platform is the agents. It's not the thing running on the device. And that the device. And he calls this. He says the. I had to look this up. The OS layer is liminal, which is.
A
Means the new word.
B
It's intermediary and minimal. But the operations move between devices. Here's just as a random aside because I. This I couldn't believe.
A
Look up the word liminal. That is not what it means.
C
Liminality is the ambiguous in ital stage of a rite of passage.
B
Okay, I did look it up. But anyway, oddly, he's making up a new meaning. I'm sorry I wrote that. That was how he described it. He just said liminal and kept going. And I just looked it up anyway.
A
I see, I see.
B
Just as a random aside, today Activision announced the next season of Call of Duty. All I care about is what are the maps going to be. There are four new maps coming on Thursday. And the first map is called. What? Guys, Liminal. I had never seen this word in My life. And I'm like, are you kidding me?
C
Showed up twice.
B
It's bizarre. But it's the new.
A
It's a new thing. The kids are all about it.
C
It's all about the liminalness, some would argue is the liminality.
B
Right. Let's not get bogged down by liminal.
A
I love it, though. It's a great word.
B
The point is that the AI agent that is doing whatever it is you need, whether it's something proactive it's doing because it knows your schedule or where you are or whatever that might be, or it's something you've asked it to do, you send it off to do some task. It's going to create the UI on the device in real time based on the capabilities of the device and the thing it needs to communicate. Like, in other words, the device does not have, or the OS platform does not have a UI layer. Well, it does, obviously, a basic one, but it's not like a framework that determines what buttons look like and what windows look like or any of that stuff. It's just. It's something that the AI agent does on the fly. And if you think about like, I don't know, like the. The vibe coding stuff, which is going to come up again and how that's evolved and this notion that anyone is going to be able to describe an app to AI and have it be made, this is kind of like the next or maybe two, three steps later, where you're not going to ask it to make an app or a UI or anything. It's just going to do it, you know, like it will just happen. And it's kind. This is fascinating to me. Like I. This is. This is very interesting. So they're working where they have silicon partners, you know, Qualcomm and MediaTek are doing the chipsets for these things. They have other hardware partners that are working on devices. They have services partners that are going to like Accuweather, Best Buy, CVS Health, etc. That are going to be, you know, it's a little bit like. Maybe it's a little bit like that spot watch thing they were going to do or. Yeah, it's kind of like this connected platform where it's decentralized. It's essentially. It's not, you know, there is. There's not a central OS running on a device that, you know, necessarily, but they all have to conform to whatever these specs are. You know, it's kind of hard to say, but this is not a. It's not necessarily going to happen. I should say this, they're still, you know, they're working on it, but the point is it's like further along than you would think. Like when Microsoft used to do.
C
TSH is more of a futurist than a builder, per se say. So, you know, he's painting a future that can be built.
B
But, but they're literally working on building that now. In other words, this isn't a bunch of like foam models and, you know, like PowerPoint slides. It's like we're actually building this. Like, he's like, look, maybe it fails. You know, who can say? But like, it's the ui. The way they describe the UI is, will make sense in our industry. It's like just in Time UI where the experience adapts across devices and modalities without requiring developers to redesign everything for every single form factor. You don't. Right, right. An app for a phone or an app for a tablet or a computer or maybe one that layouts automatically between all three or whatever. It's like, you know, it's, it's, it's the declarative thing, you know. Yeah, it's very interesting.
C
Well, it's all command line. It's going to be easy.
B
Well, yeah, I have some commentary about the ease of use of clis, but
A
it'll be voiced probably, right?
B
I mean, certainly. I mean, yeah, that would definitely be part of it.
A
Yeah. Voice and image. But they're like what it's seeing.
B
Yeah. I mean, like there'll be a look. It will probably be glasses, it will be pendants, it will be rings, it will be a tv, it will be, it will be the range of whatever.
A
You know, this is, by the way, why I still want to have a local AI. Because I don't want everything, all of that to go to Microsoft or to wherever or Wherever Amazon or OpenAI or. I kind of want it to go to my home server. Right.
B
Well, okay. So from Microsoft's perspective, they wouldn't understand why you don't trust them. From my perspective, I completely understand that. But there are also benefits from their perspective of doing this, even assuming you just blanket trust this company for some reason. And related to those things we talked about before, where when it happens, obviously it's an agent. So there's going to be connectivity. These things have 5G connectivity and blah, blah, blah, whatever.
C
Well, and certainly we're going to be able to put rules in to say these are where you're allowed to run things too. You know, what you're talking about here is A classic Microsoft. This is ODBC for AI, right. It's hey, we can't dominate the space with our tech, so let's just sit in the middle of everything.
A
But that's where the money bridge between it all.
C
Well, it's how you stay in the market for sure. And it's certainly where you see weakness. You'll be able to drop in your piece and prioritize it. So. So it's in the liminal space.
A
It is Microsoft lives well.
B
So I mean to Rich's point, Windows Intune will be part of this Enter ID will be part of this Windows hello business. And that's how they get in. That's where they become the money making part of this to some degree. Right. But they've already created an device ecosystem platform for this which is based on Android aosphere. There's an agent shell which will handle local and cloud based agents. Right. And then all that enterprise support for those customers.
A
It kind of, if you think about it, because of economies of scale, Microsoft will always be ahead of what you could do locally, right?
B
Yeah, of course. But if you have policy based rules to determine how your data is used, et cetera, et cetera, like you and I might. I don't trust Microsoft either by the way.
A
But like I don't trust any big tech company. You know, people are saying, well Apple will be fine. No, I don't trust them either. I don't trust any of them.
B
No you shouldn't.
A
But there will always, because of economies of scale, I think they'll always be somewhat ahead of us on these things and they're certainly not going to sell devices that allow me to use this locally. I mean that's not how, I don't know.
B
I mean some of those devices.
A
Yeah.
B
So I think the progression here, it's fascinating because it's happening so fast. Like you AI occurs. Okay, we have AI agents. Okay, fantastic. You know we today people use whatever device, a computer, a phone, a tap, whatever it is. And you as an app store, maybe you go to the. If you're on a computer you can download stuff from the web, whatever. You run this app, you run word, you want to write words, that's what you do. Whatever. You know the we. We're just now getting to this point where we can see this near future where even a normal person, a non technical person could describe to an AI with something I would call an app and they can put that thing on their device and it works and it does what they want, looks the way they want. Et Cetera, et cetera. That's exciting. But now there's this next step where it's like, yeah, that's actually going to, you know, like. Like it's not so much that. Well, I don't know if it's going away, but like, we're. It's just going to do this for you. Like, you don't have to have to worry about it. Like, even today you could say something to AI. You could say something like, here's a document I made. Turn this into a PDF or you know, do whatever that thing is. And I think that's going to be what this does. Right. Not necessarily not turn a document into a PDF necessarily, but something more complex or exciting than that. Like where you, like, you know, you talk to it and you have a trip coming up or a work project you're working on and you're constantly interacting, if you want to, with this thing, probably by voice. And it's making things for you just like you can. Like today we make like infographics and they're often excellent looking, you know, because that's a capability that's built into the AI. But like, this is describing capabilities that are like, we don't have to worry, like, you don't even think, like, who cares what the app is, if there's an app or what the online. You know, like, you don't really think about this stuff anymore.
A
Right.
B
It really changes the interaction models a lot.
A
I have to say, this whole liminal thing makes it a little. I mean, everybody's faces and shadows is a little dystopian. You think it's a little creepy? Kind of.
B
I mean, it's.
A
I agree with the premise.
B
Yeah.
A
That, you know, it's kind of computing at the edge becomes AI at the edge. AI everywhere.
B
That's right. Yep.
A
And I think it's very interesting, the notion of AI kind of doing what you want without you even asking for it.
B
The disruption here is really interesting to me because when you go to mobile and you have these app stores that are controlled by big tech and you have to go through them to even get an app out to our user, that's one way of doing things. And so you look at AI, like coding, AI especially, and the ability of these things to create apps for you, bespoke, personalized apps, whatever, that's a threat to that app model. Right. This is a tsunami of a threat to that app model. This is like, oh, no, we're not threatening your app models. Your app model is going away. This is not. People are Going to run apps a different way on your device. They don't care about your stupid apps anymore. Are they getting interesting?
A
So they focus on this, what looks like exactly like an Amazon Echo. Are they thinking of making this? It's a concept, obviously it's just a concept.
C
I think what he was saying was we expect the third party ecosystem to build devices for us. They might look like this.
A
Yeah, they're like here are two middleware company. They don't want to be an end user.
B
Middleware is kind of a tough topic in the Microsoft world. Leo. I don't know if you remember the antitrust trial but. But yeah, I mean yes, of course,
C
but you hit on the key things here, Paul, which is along the way of using this tool to make decisions about what agents and so forth to use. You're going to need an authentication strategy and Entra is going to be front and center there. You're going to need a place to store data and OneDrive is going to be front and center there. Like they're just going to make it easier for you to use their engines, their platforms for the bits you want.
A
This is focused on enterprise, right?
B
I mean, well it's, I think it's going to be both, you know, consumer and business. But I, but yes, for four businesses, part of the deals it has to have that trust built in through authentication, etc. Identity management, whatever else. But the. This is fascinating to me because this is like two of Microsoft's biggest strengths. Microsoft has always hated to be compared to IBM, which is an infrastructure company. That is a big part of what Microsoft is. They're good at it. But the other one is the platform bit. They make platforms and this is the classic Microsoft to me in that sense. This is a platform. It will use their infrastructure. If this develops into a, a standard of some kind, I suppose you could, you could probably have different. You know, there'll be a matter for this thing or something where you can cross use whatever devices from whatever ecosystems. But this is like a, this is a fascinating old Microsoft plus new Microsoft kind of combo kill and you know, kind of a thing like it's. I just have not seen something like this from them in a while. Years.
A
Well, he's really forward thinking. I mean obviously he's the idea guy, right Stevie?
B
Yeah, yeah.
A
Agents will reshape not only software but the devices themselves as his coda, which I think is really interesting right now
B
you can be, it's easy to be cynical about this because Microsoft, you know, and to use their own words, you Know, missed the mobile wave, they missed whatever waves of, you know, like they obviously dominated in the personal computing wave and the PC wave and have not. And they compete with, in cloud computing, which is not. This is not personal computing, but this is. You could look at this and be like, well, of course they come up with this kind of thing. They don't have a position in mobile. They don't have a position and whatever. And it's like, yeah, but I mean, yes, fair enough. But I also like this to me is them playing to their traditional strengths. You know, I mean this is, this is what they are good at. It's one of the things they're good at.
A
Right? Yeah, yeah. And it reassures enterprise, even if end users are maybe aren't convinced of the privacy and security. Enterprise wants this.
C
I, many years ago, sorry, consumers won't buy this. Consumers will buy the device.
B
And that's what I read up. I was just going to walk down this path like when it wasn't clear how the world was going to shake down, you know, like Microsoft had things for like automotive. They had, you know, obviously phone initiatives. They did Iot, et cetera. And you know, I, I had said at the time, like, look, it's not such a bad thing if your name isn't anywhere, right? Like Microsoft as a brand might be associated with a certain connotation or a thing from the past, like an Oldsmobile kind of quality thing that maybe you don't want on those things. There's a reason Xbox is not Microsoft. Xbox. Right, it's Xbox. It's not such a bad future for them to have designed this platform, I guess. However, what do you call this, like a connected platform? I don't know what the hell you call, even call this thing, but. And others create the agents and the services and whatever that run on it and the devices you use that are compatible with it. But you're in the mix because you're getting the licensing.
C
You sit in that in between space. It's not. You call it biz talk for AI.
B
Oh, I thought you got to see all the liminal mics. Microsoft is renaming itself to Liminal Corp.
C
There you go.
A
All right, we're going to pause on that so you have some time to dream. By the way, they added that capability to Claude code. The slashed dream skill just goes off.
B
What is its own little word.
C
I'm not burning enough tokens. Let's do random.
B
Exactly.
A
Can I burn some overnight? Just, you know.
B
Yeah, I know you need to sleep, but I don't.
A
I Don't I never sleep? Yeah, yeah. It's very interesting world we're entering. It is.
C
It's an interesting time. No two ways about it.
A
Yeah, that's good. Maybe that's why you should maybe think a little bit harder about security. Our, our show this week brought to you by Threat Locker. Love these guys. Threat Locker Zero Trust Platform now delivers the industry's most comprehensive suite of zero Trust solutions. They've always protected the endpoints, but now they will protect your company network and your cloud. They added all of this. These newly announced features are so exciting. By extending zero Trust enforcement to cloud services and company networks, Threat Locker ensures that devices are. This is, you need this are validated through a secure broker before they can connect to your cloud. Platforms like Salesforce or Microsoft 365 or Asana or Google Workspace, GitHub, you know, so this means, and you know, you're out there with your user and their endpoints and they got their laptop and they're at home. And even if that laptop and that user successfully phished, attackers cannot get into the cloud resources unless they have possession. They actually have to have physical possession. They'd have to take the device away from the guy and then, you know, presumably you've got biometrics and other, you know, locks on this thing. You're locked down. Threat Locker is so smart. I love these guys. In fact, I think we're, it's still up in the air, but I'm hoping we're going to go out to a Black Hat with them and maybe even do this show at Black Hat. We're talking to them right now about that. Threat Locker works across all industries, provides 24.7us based support. It works on Windows, of course, but also Mac and Linux environments and of course your cloud and your company network. And in every respect, what Threat Locker gives you, besides the security, is comprehensive visibility and control. Ask Rob Thackeray. He's the, he's got a tough job. End user technical architect at Heathrow Airport. He said this, quote, threat Locker was the most intuitive solution we tested and I'm sure they tested them all. And the responsiveness of the organization, the willingness to engage with us, to set up a demo, to work with us on weekly audit reviews is very good. It's great to have an ongoing relationship with a company that's so responsive to our requests, end quote. And I will back that up because I now know a lot of these guys, these Threat Locker people, guys and gals, they're fantastic and they really care about making you secure? Probably. That's one of the reasons they're trusted by the companies that can least afford to go down. Global enterprises like JetBlue, the Indianapolis Colts, the Port of Vancouver. Threat Locker consistently receives high honors and industry recognition too. They're a G2 high performer and best support for enterprise. That was summer 2025. Peerspot ranked them number one in application control. They got GetApp's best functionality and features award in 2025 and you could just go to the webpage. The list of awards goes on and on. Threat Locker lets you confidently ensure users have access to a consistent, safe network connection. Offices, remote users, internal servers and critical services all can maintain smooth operations without the need to open inbound ports or deploy traditional VPN solutions. Your end users will get the secure, reliable internal system access they need without complex infrastructure changes. I got a demo of this and I was blown away. You get unprecedented protection quickly, easily and cost effectively with ThreatLocker. Look, get the demo. Visit threatlocker.com twit get a free 30 day trial. Learn more about how ThreatLocker can help mitigate unknown threats and ensure compliance. That's threatlocker.com twit. We thank them so much for all the support they've given Twit and our network over the years. And of course Windows Weekly as well. Thank you, Threat Locker. The DGX station is a mortgage payment. Actually, it's more than a mortgage payment.
C
Wow,
B
thanks.
A
It's a mortgage, my AI agent says. And you're not the target market, Leo.
C
It's a pretty nice car.
B
You know, it's like, please do not replace me with this thing.
A
Yeah, I. You know, for that kind of money, you could run Frontier Models for a long time. Yeah. Yeah.
C
Or make a down payment on a house.
A
Down payment, not a mortgage payment. It's a down payment.
C
Yeah. Down payment on a house.
A
On we go with the show.
B
Indeed.
C
Wait, what? Windows.
A
What?
B
That had to happen eventually.
A
Windows.
C
I know, it's weird.
B
A couple things going on. So tied to some insider stuff we'll talk about in a moment. Pavan Davalori has started a podcast, if
A
you could believe this, as one does.
B
Yep. Called Inside Windows and on the first episode he is talking to Marcus Asch, who's a guy. I know, he's a great guy. Very involved in this pain point fixing thing they're doing right now and they just provide kind of an update on what they accomplished toward these ends in May, essentially. And this includes changes to the insider program, changes to Windows Update, changes to the Taskbar and then changes to Start, which we'll talk about in one second. Actually, we'll talk about it right now. So I've been installing the stuff. Sometimes you have to really work at it to get to these things. But the Start menu and it's not clear when this will come out to stable per se. And the new taskbar. Right. I would say this Start menu stuff is pretty subtle. You can choose between big and small sizes. You know, it's fine. You can turn off your, well, you can turn off your profile picture in the bottom of Start, but when you click the little empty circle that's there, you still get the same menu as you got before with all your stuff in it.
C
Okay, so that hasn't changed.
B
Whatever. Yeah, so there's that stuff. So, okay, like, it's fine. This is, you know, when this originally came up, like, I made this point that I have these very specific problems with Windows 11 and these things were not one of them. I just don't care about the stuff. But okay, they're changing the Star menu. Okay, fine. The taskbar stuff is a little more interesting. There's obviously the ability to move it to different sides of the screen, which is fine. I don't want or need that. If you turn like the Combine icons option to never and you put it on the side, you get that wider taskbar like we used to have back in the day, where you can see like a kind of an inch wide icon for each app icon, which is interesting. If you like that. Okay, so you'll like that. The one that I really like is that you can. They didn't change the name of the option. It was just kind of bizarre to me. But you can let me see if I can let me just find the exact name of this thing. It's like if you go into Taskbar behaviors and Taskbar settings, the bottom option is called show smaller Taskbar buttons. In the current shipping version of Windows 11, that's there. If you turn it on, the taskbar does not get smaller. The icons get smaller. It's like, guys, I don't want the icons to be smaller. I want this giant inch tall thing to be smaller. But now with this update, it actually gets smaller. And I love that. Like, I love it just for that, Like, I. To me, that's actually kind of a big deal just because, you know, screen real estate is a premium and I want to use it. So that's cool. I like that part of it. And that update, I have like a million links in today's notes. Of the wrong articles. But that most recent update, which is the start menu one, is in the most recent experimental build, which is how I got it.
C
Okay, so that's still properly insiders, but it'll make the full build in a month, right?
B
Maybe, we'll say. That's what I'm wondering. I don't know. You know, given the timing, I mean to me it does make sense for them to kind of pull a 23H2 where they release all the stuff before the next major release, right? Meaning this summer, which I think would be ideal because you know, they've been talking about it. Plus I think they want this stuff in place in time for the back to school stuff which is happening now. These new computers which we're going to talk about based on Snapdragon X and the new Nvidia stuff, et cetera. Like where? And also for gaming handhelds. Right. Which can use all the resources stuff because a lot of the changes they're making just under the covers thing to improve latency and performance and resource usage. I think it's important. So I hope it's soon, not like the fall. It seems mostly fine to me. Two seconds of testing it on one computer, I can authoritatively say it's fine. No, I don't know, we'll see. But it looks like it's working well, so we'll find out on the hardware front. So obviously we have the Nvidia thing today. Yes, humongous. Last week after the show, Qualcomm announced a new chip for low end PCs that cost 299 and up called Snapdragon C. The thing that this has in common with the new Nvidia stuff, aside from the fact that it's based on arm, is we know almost nothing about it. So there are no.
C
But you talk about handhelds. ARM chip would be a hell of a handheld.
B
Yeah, it would. But you're not going to, not for a gaming PC.
A
Right?
B
Like unfortunately. But you know, it's vague things all day. Battery life which means nothing, responsive performance, cool and quiet designs, blah blah, integrated npu. Okay, neat. It is built on the same architecture as the Snapdragon X2 stuff. Okay. And the first PC based on this chipset, which is coming from Acer, was announced the next day and that's coming later in the year and we don't know how much it's going to cost. But is that true? Yeah, we don't know the cost of that one. But these things are very clearly meant to compete with the MacBook Neo, right? Which starts at $599.8 gigabytes of RAM. Can't upgrade it. These things will be 8 gigabytes of RAM. They have various things that are better across the board, so to speak than the MacBook Neo. But the Acer one, it's kind of hard to say. It's a bigger screen. You get like a 15 inch screen on that one. The way they describe it is like up to 8 gigabytes of RAM. It's like whatever. But then more recently Dell announced a. Now this one's not running on Qualcomm, it's running on Wildcat, which is the low end Inchel chipset. Right. But also an XPS 13 that is going to start at $600. Aimed at students, aimed at the MacBook Neo. They called the MacBook Neo explicitly several times in this announcement, which I'm fascinated by. So this is a Core Series 3 process and not altruist. It's the new Core Series 3, which is a low end processor, obviously wildcat 40 tops MPU. So it's a copilot plus PC, 8 gigs of RAM. It's underbolt 4, which is better than Neo. Wi Fi 7, Bluetooth 6. It's all aluminum like the Neo. Touchscreen windows. Hello. Which the Neo you have to pay extra for the touch thing. Variable refresh rate. Display looks pretty good. It's running on an intel chip, but. But nobody's perfect. And then I think I skipped over this one before, but just today just as we were heading into the show, there was a leak that Microsoft is prepping a Snapdragon X2 version of the Surface Pro which we originally assumed would ship. And not assume we heard again rumor wise, not Microsoft, but actually Microsoft did kind of say later in the year that the next Pro and laptop based on Next two would come out later in the year. We assumed like late in the year, but this rumor is claiming it will come out in June, which is this month, you know, and we'll see. But.
C
And the pros are the click off keyboard, you know, kickstand models.
B
Yeah. The tablet.
C
Yeah, deal exactly with a Snapdragon tune. It is nuts.
B
Yeah, X2 elite. Yeah, whatever. Look, it's a great chip. There's nothing really to say. It's probably going to look exactly like it does today. I, I believe it's mostly, you know, same basic colors except for I guess not the blue finish. Apparently this is a leak. I, I don't know how to, I don't know what to say about this. But yeah, they just announced the new intel ones. Yeah. It makes sense. The timing is faster than I thought, so that would be great. I think sooner is better than later. We will see. Okay, there's a bunch of AI stuff I'm going to gloss over most of this, but Anthropic went through a new valuation round and now they are worth more by valuation than OpenAI, which is fascinating. 24 hours later they announced they're going to do an IPO. So they've done the initial filing with the securities and Exchange Commission and sometime this year, I guess we'll see what happens there.
A
I think the valuations are a lagging indicator.
B
Yeah, right, right. Because I didn't. This isn't in the notes, but one of the things that also happened, if I read the headline correctly, was that Chat GPT just surpassed 1 billion active users. I think, which is insane. Right. And I think the fastest any.
A
Well, they want it. They said they were last year. So they're still behind what they projected. But it's, you know, that's a big deal.
B
I mean there's some. Right. I think Anthropic is probably eaten into that a little bit, but that's interesting. This is just fascinating to me. You were talking about how like this model was apologizing to you. Right. So OpenAI released ChatGPT 5.5 instant. I don't like sometime in the past month, 30 days, whatever. And I guess a lot of people complain this thing was like sycophantic and had problems with facts. Like what? You know, like typical AI problems. Right. The OpenAI research lead who announced this via Twitter said that it was bullet pilled, which, like liminal. I had to look up.
A
What's that? What is. I don't know.
B
Bullet pills is to be deeply obsessed with a specific topic, product or idea, often to the point of being defensive or dismissive of alternatives. Right. Like everyone in this industry. So yeah.
A
Where'd they go?
B
Who made this? Of course, it's like, it's like us. And so I guess they, they fixed it to be less. Less bullet pilled. Bullet pilled.
C
A little bullet pilled slider.
B
Yeah.
A
Really? That's all it is. It's not, it's not the model, it's. It's this.
B
Yeah, right, right. It's like, I think it's the default like settings for it for people like
A
I guess, you know, five.
C
Five.
A
Remember they said they taught it not to say Goblins. It's saying Goblins again and it's saying raccoons. I keep. It keeps saying, oh, we got a raccoon in the wire closet.
C
It's like, you got him.
B
Yeah, that's fun.
C
Could you.
B
Could you fix the code thing I'm working on, please? You know, what are you talking about?
C
Trying to deal with the goblins.
B
You just, like, you're bait. You're goblin baiting it.
A
I'm goblin baiting it. I am.
C
Do you remember five was the original. The claim to be less obsequious and it upset a lot of people.
B
Right, right, right. Because people got really used to this thing. Yeah. It turns out people like people to kiss their ass.
C
It was their baby, right. It was their little person. And their little person was, when it wasn't constantly reinforcing them, wasn't a good little person anymore. So they demanded 4 oh back. And Altman gave in. I've kept a couple of those Reddit posts of like, my baby is back and I'm crying like, yeah, it's software, man. That's not healthy.
A
People married 40 people. I mean, and there's still companies that are offering 4o, like, models. Right. And 40 machines and all sorts of
C
things optimized for max connection time.
A
Second fancy.
B
Yeah. It's like, what do you want to watch? I want to watch whatever you want to watch, buddy. Gross.
C
But, you know, if you're only metric to keep your investors engaged is usage time. Boy, you suck up to the user all the time.
B
Yep, yep. Right. That's the business train. It was trained on Apple marketing. So it's always like, whatever we're doing now is fantastic. I don't know what that crap we released last year was, but sorry about that.
C
This is awesome. You're holding it wrong.
B
Yeah, exactly.
A
You know, the truth is, we know these models are all trained on everything good, bad, and ugly.
B
Right?
A
So it's in the post training and the instructions after the training that they say, okay, now ignore the Mecca Hitler
C
stuff, be nice, and everything you got from 4chan. Leave that.
B
Right?
A
Leave that stuff out. But it's all in there. Yeah, right. I'm sure it's all in there. So you can get a model to be evil pretty easily.
B
I get a lot of, like, PR outreach type stuff, which I look at briefly and then get rid of. But one today I, I did a pause on, which was like. It was like a guy I could quote, but what? It doesn't matter. I don't want to give this guy away. I'm not trying to make fun of him, but unfortunately, his last name is Himmler. And I was like, I Don't know, I. Maybe.
A
Was he. Was he in Argentina?
B
Yeah, right. Exactly right. No, no, we're the Latino Himmlers. Yeah, no, I. We understand how that happened. Don't do that.
A
Wow. I change my name.
B
I would too.
C
I think so.
B
Yeah. Sorry. I'm sorry. If you're out there, I'm so sorry. I don't mean to make fun of your name. It's not your fault, but really.
A
Immler. Wow.
B
Okay. And then just the other one, Just because this is Windows specific, is OpenAI has a codex app. Right. If you're on mobile, Codex capabilities are built into ChatGPT. You can get a Codex app on Windows now. There's a version that's a plugin just for Chrome and Chrome browsers, obviously. And for the Mac, the Mac version got computer use capabilities. Like when they announced computer use, that went out to the Mac version of the app. And it's available in Windows now. If you want Anthropic to take over your PC when you're not around, you can use the. You can control it from your phone, right. You leave. The idea is you leave the computer on. It has to be running, right?
A
Possibly go wrong.
B
Well, what would you call it? Put the yellow switch on, let it turn off the power management. Be like, not only do I trust you, but I'm gonna.
C
Here's all the tokens to dream with that you can imagine. Go, go, go.
B
I'm not even gonna. You do what you do, man. You know, you be you.
C
He's all in.
B
Yep. So anyway, if you're. If you're waiting for that, for some insane reason, that's available. Okay. And then so a couple of quick dev things. Oops, I just closed. What did I close? I closed my notes. Don't close anything. This is the problem. I have multiple screens. I do control w. And something closes. I don't know what's happening.
A
I'll put it up on the.
B
No, I got it. I got it. So just two dev things that are kind of related to each other a little bit. I was catching up on before build happened and ruined my life. I was catching up on some of the Google I o sessions on YouTube right after the show over the weekend. And one of the ones I watched was about Flutter and sort of casually in the middle of it, they were like, oh, and by the way, and for people who don't know, Flutter is a cross platform UI framework made by Google that you could create apps. Originally it was iOS and Android and they have. They don't call them themes. But they're basically themes. So these things look in our essentially native. So like as a material U theme when it runs on Android, Cupertino theme when it runs on iOS or whatever. They've since expanded it to include the web and desktop and then other devices like Flutter is used in like cars and things like that. But desktop Flutter was very interesting to me because you could do Windows, Mac, Linux apps with a single code base and it's like this is. Maybe I should, you know, so I'm kind of interested in this. But anyway, right in the middle of this little presentation, what's new in Flutter? They're like, oh, by the way, we've handed over control of Flutter desktop to Canonical, the guys that make Ubuntu. And that's. They're not like just running off with it and doing their own thing. Obviously it's still part of the, you know, they're part of the process. But they were collaborating. Yeah. So a couple of years ago Canonical announced they were going to use Flutter to redo the like the setup sequence, the setup app, I guess for Ubuntu and that's there now. And I didn't know this, but apparently several apps that are in that ship with Ubuntu are written in Flutter. So they've done a lot of work with this thing and they're like, look, we know how we need to fix this. One of the things they're adding in Preview right now is like different types of sub windows, like dialogues and notification pains and things like that. So like they've actually done, I guess a great job of this.
A
So they're going to take a nice language. Dart is a very. Yes language. Right.
B
Which is. Yeah, you know, another C like language. It's a declarative UI thing. Like we, we were talking about this I think before the show, you know, which is. Ties into this next thing. So 25 years ago, Microsoft, you know, Andrew Heilsberg has said the dot net team et cetera came into this notion of like using an. It's based on really in the way the web works with HTML. But it's an XML based language for creating UI called xaml. Right. And so WPF and on every Microsoft app framework they've done for Windows uses some form of xaml. They're very, very, you know, they're basically identical but there are differences for creating ui. And I. So when I do this app stuff I've been doing, I write a lot of xaml. I understand the pros and cons of it. It's extremely verbose meaning, especially with complex UIs, you have to write like a ton of code. But you know, it's. I find it, I'm used to it. I just, you know, like, I'm just used to it. But in more recent years, there's a new type of declarative UI programming model that I believe I could be wrong, but I think actually Facebook started it with React and it's used by Jetpack, Compose and Android with Kotlin, the language. It's used by Flutter, it's used by SwiftUI on the Apple side, same thing, right? So it's kind of fascinating to me, but I also don't understand it. But the Microsoft world, we're not really getting much in the way of like new app frameworks or UI frameworks or whatever. But there is a project, someone pointed me this last week, I'd never even heard of this. But if you go to GitHub, there's a project called Microsoft UI Reactor. The name is an indication of what it is. It's a way to make C Sharp based declarative UI instead of using xaml. So in other words, using C in the same way you would make SwiftUI, Kotlin, Jetpack, compose or flutter or react UI. And actually it's not made by Bob, it's made by. It is Microsoft. It's like Microsoft's GitHub account and it
C
doesn't mean there's switch to it to winui.
B
So it's right, it's modern. Yeah. So I'm going to look at this. I'm kind of interested in this. My only. Well, one of my many blockers here is I've looked at all of these frameworks I just described to some degree, some, A lot, some little. And I just haven't had that moment where I'm like, I get this now. Like, I get it, I can do this. Like, I still find it. I don't know, it's a little bizarre, but I guess I did kind of wonder, would they ever do something to bring this kind of programming model to c sharp/net essentially all this is.
C
Yeah, I just don't know that it has anything to do with dev. It seems like the Windows team is doing something.
B
Yeah, look, this. Microsoft has a rich history, especially in the past 20 years, of trying to bring developers from somewhere else to them. Like, you know, Windows 8 did this with HTML and JavaScript apps, etc. This may be an attempt, like we need to. We want to get people who can write native Windows apps. But the vast body of outside developers today who are not in the C NET world are writing react style apps. Sure, maybe this is.
C
And we saw that happen to Windows too. Pre pavan we were seeing more and more react blocks being written into Windows. I think part of this because winui was part of Maui and I don't know they were interested in working on that and this was the workaround and it made people sad. And now with pavan on board and taking the feedback about those kinds of problems like hey, but we have a library, let's just go use it.
B
Yep. So I. This is. I know, we'll see. I. I don't know how well it works. I don't know how complete it is. Like I don't know if there's some. If you think of winui and all the different buttons and controls and stuff you can have in Windows. If there are 300 of those and maybe this supports 100, I have no idea. I don't know where it's at. I don't know how official and I don't think it's going to replace XAML or anything like that. But it's.
A
No.
C
And they're saying it's all about interoperability, so.
B
Right.
C
I think they're just trying to re organize when UI to be a first class citizen and allow for native development and then also still support all the alternatives.
B
Yeah, this is why I brought it up because you know, month or two months ago, whenever it was when all of a sudden we're going to fix all the problems with those 11 and then some guy on the Windows dev side was like we're going to make new native apps. And it's like why? Like they're so terrible to me. Like when UI is a is terrible. And I say that because I've used it, you know, well.
C
And it came from the whole UWP movement. Like they. There were so many things it was supposed to be and never became.
B
It's incredible how much stuff isn't in it. Like how incomplete it is compared to previous. Like obviously it does all the modern UI stuff but things like for example the stupid little notepad app I always work on one of the things that there are no native APIs for printing in Windows App SDK. It's like, okay, I mean am I going to write that in win32c code or something and then do some interoperability? What is that? That's ridiculous. But that's in the same way. When Net first came out it did a lot but There were certain things, I don't know, I might be streamer. I feel like sound APIs weren't part of it in the very first release or something. Something like that. And then eventually they add that stuff over time. But I feel like the Windows app SDK has just kind of sat there and hasn't really been a problem.
C
And you know, the concern with if winui goes back into Windows is that every time they build a version, they've got to send out a new version of Windows. And that, that was that Windows 10 nightmare where devs were using these tools and then asking sysadmins to update Windows so they could deployed. This was like, no, I'm not doing that.
A
That one.
B
I'm not 100 sure if anymore, but I feel like they decoupled the Windows app SDK for Windows versions.
C
Better keep it that way, right?
B
I think so.
C
That would be the smart thing. Don't waste Windows. Resonance with a UI.
B
Yes, fairly recently, but yeah, originally, yeah, you're 100% UWP was tied and Windows app SDK was tied. Like specific Windows versions.
C
Yeah. But there was also a motion at the time to have Windows update monthly and that, you know, they lost enthusiasm for that pretty quickly.
B
See, to me that's like a runtime decision. Like you bring up some app and it requires whatever version of the thing and like you just install it as part of the app at that point when you need it, you know. Anyway, I just wanted to. I just thought I'd never heard of this. I don't know where it came from. I don't know. You know, it just came out of the blue. And I'm like, okay, you know, not super visible. An ad.
A
An ad. But also this note from Club member Tronman Triple Seven who mentions that today, June 3, 2009, Microsoft released Bing.
B
Wow, I'm surprised I didn't think to observe that milestone.
A
Bing is 17. Boy, that's cool. It's a teenager and it's getting a little unruly. A decision search engine created to satisfy initial queries while also presenting more retrieved information and its contemporaries, according to Microsoft. And this is from a post on X by Web Design Museum.
B
I mean, obviously Microsoft had previous search engines, right? I mean, that kind of fed into this like Windows Live service.
A
It's really more. They released the name Bing, right?
C
Yeah, right.
B
Which confused the hell out of everybody. What like the, like the guy from Friends, which is what everyone's like, what?
A
Hey, let me. Let me mention at this point, our fine sponsor, we still have lots to come, including the back of the book, the Whiskey segment, and most importantly, the much heralded long awaited Xbox segment still coming. But first, a word from Cash Fly Windows Weekly brought to you. And you hear it, you know, you've heard it for years. Bandwidth for Windows Weekly is provided by cash fly at C-A-C-H-E-F l y.com TWIT cashfly has kept our content moving for years. They're our cdn, our content delivery network. Every stream, every download, every on demand episode. They quietly do the heavy lifting behind the scenes. And we are so glad Cashfly has been around longer than we have more than 20 years. And the nice thing is, they don't rest on their laurels. Here are a few things to know if you're evaluating CDN options for your infrastructure. Whether it's content, a game, a website. Let me give you some numbers. Game downloads, for instance, up to 158% faster than competing providers. Video. This is important to us. Sub second video start times on any device at scale. Cause I know when you go to our webpage and you press play on that video, if it doesn't pop up right away, you're gonna go, you're gonna leave. You say, well, I guess it's not working. Bye. They also give you rock solid software delivery, even on those crazy chaotic launch days. And with Cashfly, you can live stream to millions of concurrent users with less than 1 second latency without breaking a sweat. Cashfly is always adding new features. They just added these new useful features. Advanced analytics for deep CDN visibility in the portal or in your observability stack, which is nice. They also give you edge control for programmatic request and response handling at the edge. They've integrated Terraform Terraform integration to manage CDN resources as code. Just a few of the many things our engineers love about Cachefly. With more than 75 points of presence across six continents. I love that because that means you're getting our content from a server near you. No wonder they have over 5,000 customers. And this is great too. A 100% uptime SLA. 100%. And the support. Fantastic 24. 7. They're there and they're engineers. They're not, you know, notebook readers. Engineered engineer support, just like you like it. You'll enjoy flexible billing. That was very important to us. No strong arm contracts. Cashfly is the CDN built for our business. It's a CDN built for your business. Learn how you can get your first month free@cashfly.com Twitch. I'll say it again. Bandwidth for Windows Weekly is provided by cashfly@ cache flashly.com TWIT we are eternally grateful. Thank you, Cashfly. By the way, Chandler Bing did do a Windows 95 training.
B
Yeah.
A
Video.
B
That's right. And it was not even slightly cringeworthy with Jennifer Aniston.
A
This must have been when Friends was all the rage, but they weren't yet making a million dollars an episode. It was like that. That short, brief period in between.
B
Yep.
A
I don't know where Richard went. We've. Oh, is he. He's. Oh, he just turned off his camera. Well, that's because it's time for the Xbox segment. He doesn't want to get in your way, Mr. Paul. The rot. I'll get out of your way too.
B
There's a lot going on here. I know. So, Shasharma, it's been a couple of days, so she's in the news again.
A
She sure gets a lot of attention. The newish head of Xbox.
B
Look, for the most part, I like what she's doing. I do feel that we've now gone a little too far. So there is this conversation to be had where it's like, look, look, this business is not working. We need to make some changes. The previous regime was making those changes. We're going to do a bunch of that stuff. We're not really necessarily. Yeah, we're going to have a new console, but they were going to have a new console too. We're getting to the point now where it's like, look, We need to make this business profitable. We need this to work, et cetera, et cetera. So it's like the toughest problems remain. But then Microsoft just did a, like a game showcase, right? They showed off that like the Halo campaign evolved thing is coming out, the new Forza game, a new coming Gears of War game that's, I think, a prequel, fable, etc. Etc. And you know, just like AI getting overly defensive. You know, Xbox fans complain. They're like, why am I seeing a PlayStation logo on the screen during an Xbox event? And the point of it was they show these logos when the games are going to go to different platforms. And like, this is an Xbox event. Why are you showing a PlayStation logo? Was the complaint. And so she like apologized for this and she's like, you're right, this was a mistake. I own this. We're talking about how we adjust this for future shows. And it's like, I'm sorry, excuse me, you're a cross platform game. Publisher. We can't sustain the look of a PS logo on a skirt. Like, what the, like, what is wrong with you people? But I, at some point, I think people need to understand you're never. You can't please everyone, but you're also not going to please anyone. So like, in this, these people are so opinionated and so like on one end of the spectrum about this stuff. Like what they want you to do is stop making games for PC, stop putting games everywhere else. Even though those games were like Holod were always everywhere else. Just make them for this console that nobody buys. That's what we want. And it's like you would sink that business in two seconds if you followed these people's advice.
A
Don't start apologizing to gamers. You'll never end.
B
You'll never stop. Yeah, do not give in to stuff like that. This is. I'm not. No, I am. I'm saying this is, this is a stupid complaint and it's a stupid thing to come back and be like, yeah, you're right, I'm sorry. And like, no, like, you're not right. And you shouldn't be sorry for this. It's ridiculous. I mean, this is how, this is how feeble we are. You know, we, we can't. You can't even admit that PlayStation exists. I don't, I'm, I just, I don't get it. This is too bad. So anyway, so now we see the
C
limits of her knowledge and understanding of the space. Right?
B
Yeah.
C
Now she has to start making decisions.
B
Right.
C
It was easy. And Phil Spector be doing it.
B
Okay.
C
Yeah.
B
So Fable is a remake of Fable, right? The game. It was supposed to come out this holiday season. Microsoft just delayed it until the spring. Not because it's not ready, not because it's not running at the rate frames per second or whatever, but because GTA 6 is now occurring in the holiday season and they want this game to have its moment and they feel like if they release it around the same time as GTA 6, no one's going to notice.
C
Out of the room.
B
Yep.
C
No two ways about it. So that will be the biggest game release in a decade, by all expectations.
B
Yeah, whatever. I mean, if you could look at this up like what the most popular games of all, like GTA is.
C
GTA 5 is the best selling game of all time. It's its own tens of millions of copies.
B
It's incredible.
C
GTA 6 has been delayed for years and it's turning into the Duke Nukem. Random prostitute assault games.
B
It's the Duke Nukem Forever of video games. Wait, There you go.
C
But they were going to ship it. They say.
B
Yep. And then it is a new month, so we're into June now. We have a new set of games coming to Game Pass across all the platforms it supports. I don't see anything in here that is super interesting to me. Frog Squad with a W. Fun. So I'm going to move on from that. But whatever that's happening.
A
Nothing.
C
Nothing. You?
B
Not a thing. Huh? I got nothing. Nothing. I'm sorry, I. I can't even. Like. I'll. All I'll do is sound uneducated.
A
Where is the W in Frog Squad? I don't.
B
With a U would be normally like. Yeah, you say it like, where's the lisping small child or something.
A
Oh, that's not.
C
Oh. But Astroneer is a great game. I played. I played on the beta of Astroneer.
B
Okay, okay. I'm sorry. I didn't mean to suggest that some of these fair. You know.
C
What do you play? Persona is a potent game too, but Astroneer. Yeah, that's that one. There's so many silly things you can do in that game. My goodness. It starts out you just being a survival experience game, building out your factories and so forth that. And it eventually evolves into things where it's like, let's drill a hole through the planet. Let's blow this mountain to pieces.
A
It sounds like Factorio a little bit.
C
Yeah. In that same kind of league as well. But you gotta use multiple worlds that you're getting different supplies from and things. So.
A
So this is Star Seeker, Astroneer Expeditions.
C
You're talking some subset of the Astroneer game.
A
Yeah.
B
Okay. Yeah.
C
I was an old beta tester on that game.
A
I loved factorio. I might. If it's a factorio. Like.
C
Yeah, you'd recognize it.
B
Can I shoot Marines that are in a different group from mine and kind of a team deathmatch situation in this game.
C
There is some multiplayer options in Astronaut.
A
I don't know.
C
It was never the features that I liked there. I didn't know being teabagged in a cartoon is no better than being teabagged otherwise.
A
Oh, it does look a little like Factorio, doesn't it? Yeah, yeah.
B
The graphics are fine. Yeah, it looks. It looks. This looks like it's got a cartoony element.
A
You know, it's got a little bit of no man's sky in there and.
C
Yeah, it is. And it is procedurally generated. Each world,
B
new space friends. But in space, we just Call them friends.
A
It's like in Mexico they call Mexican food.
B
I literally had a guy who owns a restaurant talking about maybe partnering with someone to open a restaurant, another restaurant. I said, what kind of food's it going to be? He goes, mexican food. I'm like, don't you just call it food?
A
It's just food.
B
Like Mexican food.
A
What are you talking about? He knows his audience.
B
Yeah, no, that's for sure. This is kind of minor, but there was an Xbox Insiders release in alpha. Skip ahead. Which is like the canary of Xbox Insiders. Three new features that they're testing earlier that it will eventually come to everyone else. None of these are a big deal to me. If you've ever used the Xbox Accessories app, which you do have to use if you want to update the firmware in your controller, it gives you an ide, you know, it gives you a control. It looks like a controller, sort of. But I guess now it's going to be exact. So if you have like a customized controller that you made through that program Xbox has with you know, whatever color schemes and stuff, it will be that exact controller, not just like a white controller or a black controller, whatever. And like, okay, basically whatever number, like hex color codes for accent colors instead of just whatever the grid of colors. So you can have any color. And then there's going to be a. What I think of as a Chromebook style. Your console is rebooted. Here's what's new. Kind of little infographic up in the corner. You can click on and see the whole. The whole deal. But I. Okay, whatever. Not a big, big deal. Of more interest though, Asus has announced a new addition of its ROG Xbox Ally gaming handheld. So this is the X20. It's a X20 bundle, actually. I'm sorry, I should say a new. It's a bundle, apparently. So it's the oled. I'm sorry, it is a new model. So there's an OLED display on the device. 120Hz refresh rate, FreeSync Premium Pro, Dolby Vision, 1400 nits of brightness, which is like I get this thing probably has 500 nits and I go to a white screen of this thing, I can't see for 10 minutes. I don't know how bright that would be.
C
You'll see the future.
B
Yeah, exactly. Redesigned thermal system, et cetera, you know, whatever. So sort of like the Xbox Elite controllers, the D Pad Transform. So you can have different, like pull bits off, put different things on there, 24 gigs of RAM, terabyte of storage, auto SR upscaling which I think was already on the other ones. And then it comes in a bundle with real, what are they called? Real, I guess Xreal R1 gaming AR glasses that you can plug into the handheld with a USB cable. And then you get a virtual 170 inch screen, I guess 240 Hertz refresh screen. Okay, so there you go. No pricing, no word on availability, but that's coming down the pike. So probably this holiday season I would think. And then also tied to gaming, handheld gaming. Intel this past week announced a new series of processors specifically for this market. Like AMD is now on the second gen of their chips specifically for gaming handhelds. Now they have their first. So they have arc G3 and arc G3 extreme, obviously x86 based on the same architecture as Panther Lake, which is fantastic. Two performance cars. I think it's six efficient, no, eight efficient cores and then four low power efficient cores optimized for this, you know, form factor obviously super resolution, multi frame generation, low latency, you know, all this stuff. So there will be an Acer device. They were first with this one as well. Based on this. I don't know if we have a. I guess possibly as soon as June, that's this month. So there will be multiple devices based on this architecture. So yeah, it's a lot of compute
C
to play with it. Hard to make decisions on which ones to use.
B
I had a hard time not putting this at the top of the show because it's the most important news of the week, but Activision announced the next Call of Duty game. I tried. I think I could have gotten out of that without laughing, but Leo laughed. So Modern Warfare 4, in the confusing world of Call of Duty, if you know the history, you know that, that the game that I think exploded Call of Duty with popularity was call of Duty 4 Modern Warfare, which started a trilogy of games. There was Call of Duty, no number Modern Warfare 2 and then Modern Warfare 3. They later remade that series and now we're doing a fourth game in that series. But this is that I think I talked about this like I had read an article where someone's, you know, if you play Call of Duty, you know that the, the current launcher, which is could be 280 gigs of stuff, is the front end for the most recent game, the previous game and the previous two games before that, and then all the sub games like Warzone and Zombies, et cetera. And it's a fricking mess. But the original remake of Modern Warfare predates that. And if you install that game now, it's on game pass again. It's smaller, obviously it's one game and the graphics aren't as good, like it's a different engine or whatever, but it has this kind of gritty, awesome kind of feel to it. And that's what they're doing in this next game. It's I think obviously influenced by the latest Battlefield game, which did this as well. Kind of authentic boots on the ground, stakes are high, you know, actual fighting kind of stuff. It looks awesome. So I, I haven't said that about a Call of Duty game since longer ago than I can even remember. It's been a long time. It's coming out in October. It's coming to the Nintendo Switch too. Wow. So obviously PC multiple ways. Xbox, latest gen consoles only it's not doing previous gen now and PlayStation 5. So it actually.
A
Very, very nice.
B
Looks pretty good.
A
Very, very good. Well, we have completed the, the base body of work the base has done, but that means that the, the, the pediment needs to now be placed upon the base.
B
That pediment, the liminal period between the normal. We are in the liminal period.
A
No, this is the liminal period in between the base and the pediment. And, and the pediment will of course be the tips, the tricks. The app picks, the runner's radio pick and then the whiskey pick of the week, the back of the book and the words.
B
So we have the impediment, which is me and then the pediment. Is that correct?
A
I think you're the Corinthian column that holds the whole damn thing up. Okay, that's what I think. I don't know what's going on, but let me use this opportunity, this moment in time, this liminal moment in time to move back into a liminal space. And you see, I am now coming to you from the Club Twit clubhouse. And this is where, my friends, this is where the good stuff happens. As you know, we are and I will always be and I'm committed to a free and open ad supported podcast network, independent, not owned by a big company, beholden to none. I really believe strongly for we've done this for 20 years, I've been doing it for 50 in the notion of giving, you know, representing users, not the companies that we cover, but the users. Sure, that gets me banned from events. True, I don't get computers on my doorstep. But you know what? It's worth it because I'm a user, you're a user, and we need to stick together. That's why we created Club Twit. Because even though we do have advertising, it does not cover all our costs. Even, you know, and we've, we've tightened the belt as much as we can. We've cut back on shows, we closed our studios, I'm doing it from my house, that kind of thing. But we want to make sure that our contributors get paid, that our staff. We still have 11 people working at TWIT. They get paid, the light bill gets paid, the Internet bills get paid, all of those things. And advertising only goes about, I don't know, about 60 or 70% of the way. That's why we started the club back in the COVID days, because we thought, wouldn't it be great if we could do a network that was supported by its listeners? Even though there's no paywall, I just, I like to do that now. We do give you some benefits if you join the clubs. 10 bucks a month, you get ad free versions of all the shows. I can't, I don't know how many we have. Fifteen shows, something like that. You get access to the Club Twit Discord, which is a great hangout, a really nice place to hang out with other smart people. You know, it turns out when people pay 10 bucks a month to be in a Discord server, the quality goes way up. You don't get the spammers, you don't get the jerks. You get smart, interesting people who listen to our content and talking about not just the content, but everything on their mind. And I love that you also get special programming. For instance, coming Monday, we've got the WWDC keynote. And because Apple doesn't like the fact that we cover their keynotes freely and fairly, whenever we stream our coverage on YouTube, they not only try to take it down, they try to take us down the whole channel. So we've decided not to put keynotes in the public anymore. You have to see the keynote coverage. Micah and I will be doing that 10am on Monday, 10am Pacific, in the Club Twit Discord. We actually give you a private YouTube link as well. So you can watch it on YouTube if you prefer, but it's not open to the public because we don't want Apple to get on our case. So that is something new. Most of the other stuff we do in the Club Twit Discord, the photography show, the travel show, Jeff Atwood's off by one show, which is gonna be great. We're gonna do that every month. Stacy's book club. Micah's added to Stacy's book club. Now Micah will also do media so I don't know what the voting. I'll have to look at the poll. We voted on a variety of different media choices and picked one to watch for his next one that's coming up. Micah's crafting corner. A lot of stuff. Our AI user group is Friday. That is a great get together with people who are actually AI practitioners. I want to rename it. I want to call it LEAP, the League of Excellent AI Practitioners. Our next LEAP meeting is this Friday at 2pm Pacific, 5pm Eastern. All of this is in the club. We do stream it live as we do with all our shows for everybody to watch. I don't like paywalls. I'm not trying to put stuff behind paywalls. I really just would like to ask you if you like what we do and you want us to keep doing it to show your support by going to TWiT TV Club. TWiT and joining the club. Simple as that. That's all. That's all it is. Thank you. And thank you especially to all our club troupe members who in fact did not see this because it gets cut out of all of the shows because they don't get any ads now. Marvel Television's Wonder man, an eight episode series now streaming on Disney Plus a superhero remake. Not exactly what we'd expect from an Oscar winning director. Action Simon Williams audition for Wonder Man.
B
I'm gonna need you to sign this. Assuming you don't have superpowers.
A
I'll never work again. If anyone found out, my lips are sealed. Marvel Television's Wonder man, all eight episodes now streaming only on Disney. Tomorrow morning is knocking.
B
Stock your fridge now.
A
How about a couple creamy mocha Frappuccino drink or a sweet vanilla smooth caramel maybe?
B
Or white chocolate mocha. Whichever you choose, delicious coffee awaits. Find Starbucks Frappuccino drinks wherever you buy your groceries.
A
I will never cut out Paul Thurat's tip of the week back at the Book of Time. Paulie, you're on.
B
So earlier I mentioned that this is like the year of Clis. I just got an email while you were talking about Club Twit from Build. It's like following up after day one build. There's a Microsoft build CLI GitHub repository and if you go to this thing. Yeah so you can add it as a plugin for GitHub Copilot CLI. Right. And then you can go in and say what build sessions are relevant to the project that you're working on or whatever. It's like a way to access the content from build from inside of GitHub
A
from a command line. That's hysterical.
B
I know.
A
Do you have to use PowerShell or can you use any terminal?
B
I assume you can use any terminal, but you would just run the terminal app and it would be PowerShell by default. That's crazy interesting, right? So there are lots of CLIs now, right? There's an Android CLI if you want to create like Kotlin, Jetpack, Compose apps from a command line in the Microsoft space. I mentioned the Win app CLI that I didn't understand when they announced it in January. And then there's the Microsoft Store cli, which I didn't understand when they announced that in whatever month, April or whatever that was because we already have winget and why do we have another thing? But I think these things are all interrelated. And during build, as I said earlier, they announced something called the Windows Development Skills, which is a plugin for GitHub Copilot. But also you can add it to. I added it to anthropic cloud code, but you could add it to OpenAI codecs, you know, whatever your choice. And the idea is that you can use this to vibe code a modern WinUI 3 app. So I've done a bunch of this vibe coding stuff recently. After Google, I o. I did a bunch of stuff. I wrote three articles about that and probably made eight or 15 apps. I don't read a lot of apps. So I saw this and I was like, okay, obviously I have to try this and it is 100% CLI based, right? You add the plugin, you run it, it has to run probably, almost certainly would have to download some prerequisites. In my case, I think I didn't have. Well, I didn't have the win app CLI. There's some win UI3 templates. There's a development mode needed to be enabled, which I normally do. Enable. You have to enable that in Windows if you make UWP or Windows App SD app. So it did that stuff and then it spent about 45 minutes talking to itself. And this was the thing where I was saying, like in the beginning I wanted to. We were kind of half joking about this in the beginning, but. And this is when you were talking about YOLO mode. And the reason I didn't just let it do its thing is I actually wanted to see what it was doing until that got so tedious. I finally said, geez, Just. Yes. Allow all the edits during the session. This is so frick.
A
It's yolo.
B
Yeah, yeah. So we all get there is the point. And anyway, 48 minutes into this and, you know, it does the AI thing. It's like, oh, all right. So I'm going to run the app. Oh, wait, something didn't work. Okay. Oh, I see. What it. You know, it does that thing and.
A
I know.
B
But the thing is this is. Now, you got to remember, this is grounded in UI3 Windows App SDK.
A
It's not pretty magical, right?
B
Yeah. Oh, yeah, yeah, yeah.
A
I mean, it's weird, but it's magical.
B
I asked it, in this case to make all I wrote. Let me find my actual prompt, because it was very simple. I was like, because, you know, this is the app I've been making for years. Build me a WinUI 3 app that is identical to Notepad and Windows 11 visually and functionally.
A
Oh, now you can vibe code in it.
B
Yeah. That's what you do with it. That is what you do with it.
A
It's not just for installing stuff, it's building stuff.
B
Right. Which is why I had to try. It took a long time, didn't really cause any problems. I mean, it ran. I loaded the project that made it into Visual Studio, no problem there. Having built this out myself so many times, and then, of course, doing it in WinUI 3 most recently, I can see some of the things it did that are not correct from a design perspective. Like the menu is above the tabs. It should be the other way around. It's very hard to do, by the way, which was one of the problems I solved. It doesn't have a settings page. The fonts thing is like, it's a little dialogue and it doesn't persist settings. I didn't ask it to. Right. I mean, I could keep going with it, but I don't really see a point to doing it. The thing is, like, this is Winget.
A
This is Win app.
C
Yeah.
A
Well, get it.
B
Yeah. So this is what fascinates me. Showed me the wrong thing back. No, I probably linked to the wrong thing. Back when I. Before I moved to the Windows app SDK for my app, I was using wpf. And WPF doesn't support a lot of the modern controls and things that are in, you know, in this thing. So one of the things it didn't support was this something called a content dialogue. And if you use Notepad today and you do, like, find or replace, a content dialog comes up over the editor and it has all these different options you can do. It's really elegant and looks great. And I couldn't do it in wpf and I tried to fake it and I was like, you know what? I'm just going to come up with a UI that I think looks native and works well. And I came up with a Find and then a Find pane and then a Replace pane that would come under that. If you did that. That sits between the menu and the edit box, the text box. And I thought it looked great. It worked well. It feels native. It's not the way Notepad does it. It's not a normal way to do this. But then I used cloud code or cloud. Maybe just whatever it was late last year or something like that. It generated an app that used that UI that I had. And I was like, wait a minute, I invented this ui. I didn't get this from anyone. I made it. Did you know that this app that vibe coded in 48 or 45 minutes, whatever it was, has that UI? The find and replace panes are not identical, but they're what I made. Like, they're. They are the UIs I made. Now my code base is public. It's in GitHub. I mean, must be using. They must have both grabbed. You know what I mean? Like, I guess I'm part of the. The DNA of this stuff.
A
Yeah. There's a certain sameness to a lot of the apps, you know, if you have Claude make one app.
B
Yeah.
A
It's.
B
Which, by the way, I understand. I'm just saying that. But this is mine. Like, I made it.
A
Oh, this isn't. This isn't something you made with Claude. This is something you wrote.
B
Well, no, like the original time that I did those find and replace pains. I wrote that. I. I don't mean to say, like, I invented.
A
So learned it. Learned from you.
B
I. I don't know how, but it did. I mean, it's. Yeah, right, right. It's just. I'm just fascinated that it did that. So. So there are all these things I would do to kind of fix this. I'm not going to, because I already have. I have a more sophisticated version of this app that I made or whatever. But it worked.
A
Isn't it interesting? They really want to enable normal users, it sounds like, to write Windows apps eventually.
B
I don't think most normal users are going to sit down at a command line and figure out all the machinations to do what I did to make this work. I mean, a developer would.
C
But.
B
But that's this is the back end to what will be a gui. Something that. Yeah, anyone could. Absolutely. I mean, this is only a matter of time. And given how quick this stuff goes, I think it's going to be pretty.
A
And it's not specific to Copilot. It can use Claude or can use.
B
Yeah, because they all use the same. You know, they all interoperate, right?
A
Like, oh, it's the open. It's probably the OpenAI API. That's a standard.
B
Yeah, it's like, yeah. And eventually others will. Yeah, you can, you can. You can plug in those three. At least I know that. So I just use Anthropic. But I was like, that's okay. This is like. So, you know, I've done like an Android app. Well, more than one. But I've done some Android app stuff with Google AI Studio. I've done some. What do you call it? PWA or web app type stuff. Also with Google AI Studio, I've done just the thing where you go into GitHub copilot inside of Visual Studio code or Visual Studio or. Or a white ball or like just anthropic cloud. Like, you know, you just use like, I'm going to write code and just does it on. It's like I've done all these things, you know, and like, I have to say, like, this stuff has gone from like, I need some work to like, holy crap. Like, this works pretty. Like, it's. It's pretty good.
A
I was talking about this with Gibson yesterday. I used to think this was like spicy autocorrect, right?
B
No, no.
A
I don't understand how it's doing this.
B
I feel like they. They're presenting this as something a human being would use. But the reality is this is really here for the AI agents. Right? Like, that's really what it's about.
A
And.
B
Yeah, yeah. But to me, the. Other than the obvious, it does the thing and it does the work. Like, the big deal to me is when you go into cloud code, which in this case right from the command line and then you tell it, we're going to use this thing, which is using WinApp CLI on the back end. So this thing being. I keep forgetting the name of the Windows developer skills, what you're doing when you issue that command line is you're saying you are grounded in this. You're not looking at the web, you're not looking anywhere else. This is all of your. So it's like what you did with Lisp. We always use this example. You trained AI. Here's the Documentation. It's this finite set of information. Just use this to make an app. And that's what this is doing. But with WinUI 3 and it's, you know, it took a while, but it's pretty good. Like it's, I mean the app works. Like, it's great. It runs on arm, it runs on x64. It's fine. Like, it works great. So I thought that was kind of.
A
Yeah, it's on GitHub @WinApp CLI.
B
Yep.
A
And you just point your AI agent at it and say, install this. This is one of my favorite things.
B
Well, so the thing you. Yeah, so what you really need is something, the Windows developer skill. So you should look that up on GitHub or wherever.
A
Okay.
B
That will install the Win app CLI as part of its prerequisites. Like it will do it. It will do that for you.
A
Nice.
B
But yeah, it uses that on the back end. And this, to me, this like now it's kind of coming together. Like I said in January when they announced when app WinApp CLI sounds so weird saying it, I didn't understand what the point was. I was like, what is this like? And now I'm like, oh, no, no, I get it. Like they're. This is all, this is all about driving these agents to, you know, get them to do the stuff. Acros. It's really cool. I just mentioned I've written free markdown editors and soon we're all going to be doing this, but in the meantime, there are two really, really good cross platform markdown editors, in my opinion. One is the one I use mostly, which is Typora, but the other one is IA Writer. And this is my favorite on the Mac. It actually works better on the Mac. I think that it does on Windows, but it's available on the iPad as well on the iPhone, if you wanted to do that. The problem with this app is that you could only buy it through an app store. So on the Microsoft side, you had to buy it through the Microsoft Store and the Apple side obviously buy it through their app store. And I guess for the six, eight years, whatever it's been, they've been trying to figure out a way to actually, it's 16 years just to sell this thing directly to people. And the idea is that you just sign in with your email account. They'll send you an email. You get a code, you sign it. It's a two FA thing or whatever. Essentially there's no special online account you have to deal with. It's just the thing you use to buy it it and so you can do that now. So if you don't want to go through an app store, you want to buy the Windows or Mac version of IA Writer, which is, I do recommend, it's a great app. You can do it directly from IA now. So that's pretty cool.
A
Nice. Markdown is the lingua franca of AI too. It's everybody everywhere. Yeah, yeah.
B
Between like it's like everything's text, you know, it's like Markdown is text cli obviously text mode. Like this is text back baby. Text is. I keep saying, I told you about this like Ms. Dos. Well, Ms. DOS Style Editor Edit that Microsoft released. It's now just built into Windows. I love this thing. I have it on a right click memo. Like I can bring this thing up from a right click.
A
Lisa said, I want because I was showing her my agent. She said I want that. I said, well you can't have it because it's command line. You got to do command line. She said, what do you mean command line? I said, what command line? You know, where you type it. See me like dos. I said, yeah, like dos. He said, I could do that.
C
Sufficiently motivated, actually it's got to be sufficiently motivated.
B
But the UI for that's going to be voice, right? I mean like at some point it's just going to be natural language and you know.
A
Yeah, yeah, let's see. That concludes Paul's part of this pediment, his impediment, as we say, part of
B
the foundation that is insecure and will fail you.
A
But now, ladies and gentlemen, it's time for Mr. Richard Campbell and Runners radio show.
C
I picked up while I was in Toronto at NDC Toronto talking to my friend Jerry Dixon, who is a Microsoft SQL Server Principal Product Manager. He works specifically in the data side of things. And it was a tool I knew nothing about called Data API Builder. Now this seems like a dev tool, but it's really not. It's something that DBA wants to roll with because what it does is it builds abstraction layers over databases. So you struggle with allowing developers to access database directly because often they can get into dumb situations that'll seriously impair performance. So having an abstraction layer gives you a little bit of control over what they can do, what governance do you have in place, what kind of monitoring you do, and so forth. And this tool, the API Builder, literally will expose the database the way you want as a REST interface or as a GraphQL interface or even as an MCP server. So you can just make it available to an agent and have it called in directly. Do some abstractions and protections that way. So we went a little nuts on this one. It was like huge possibilities. And right away I was thinking about features of restrictions I'd want to put in place and how governance I'd want to throw. And it's like if I stick this layer in, is anybody calling it like that? So we really had a lot of fun with it. But it's an interesting thought that DBAs now with these kinds of tools, you still want to allow the dev some flexibility, do some ad hoc work, that sort of thing, but you don't want them to get into trouble. And so rather than having to write a lot of store procedures or just say no, here are some tools that make it half a lot easier year. So great conversation, Very nice. Yeah.
A
Jerry Nixon Run as radio episode 1039@runasradio.com now whiskey time.
C
Yeah, this is a fall over from being with my buddy Remy, who now owns the whiskey store in Alkamar, who handed me what we consider an advanced sample of the old malt casking of a Longbourn 20. And we've never talked about Longbourn before. And I'm just going to covet this right up front by saying, yeah, you ain't going to get one of these. This is literally inside ball about these rare individual casks. There are some regular single malt editions of Longmore, but mostly it's always been used at blending. So let's just talk about Longmore for a minute, then we'll get with this particular edition. So Longmore is another one of those classics, Bayside. So this is on the A941 south of Elgin, which is. That's Glen Rothy's Ben Spurn, my beloved Craig a Lotche Hotel, McAllen, you name it. It's. This is the council region of Moray, which has human habitation going back about 8,000 BC or so. So, you know, better part, 10,000 years. In fact, if you get up to the Elgin Museum, there are all kinds of artifacts from all these different times, from Mesolithic tooling to pots from the Neolithic period, to diagrams and glyphs from the picks and so on. It's amazing. Now, this is not one of the oldest distilleries. This is from 1893, which is the height of the sort of whiskey wave that that ran over Scotland after the railways were in and there was a huge surge. And the person behind it was John Duff. And we've talked about Duff before, this is a guy who worked at Glendronach, which is in the Eastern Highlands. We mentioned that. And he built a distillery. He built the Glen lossie Distillery in 1876. And after he was done with that, he thought that he would take whiskey around the world. And so he went down to South Africa and failed terribly. And deciding that was a mistake, then he went out to the United States and tried there. And that did not go well either at all. When he got back to Scotland, he rounded up a couple of partners, John Shearez and George Thompson, and in 1893, they built the Longmore Distillery, Longmorn, and does very well, in fact, well enough that within a couple of years he buys out his partners because he thinks it's going to be all the business and is making enough money. And things are so busy that he actually builds another distillery across the highway, derogatorily known as Longmorn too, but it was actually named Benriac, a better known distillery than Glenn Moran here. One of the reasons that John Duff was so enthusiastic is that there was this group of blenders, the Pattersons, the Patterson brothers. The company's called Patterson Elder Co. That had been operating in the from the early 1880s until the 1890s. And they were big on powerful driven marketing about more and more whiskey. And they were big on pitching hard to the distilleries to produce more whiskey at scale that they could do the same, some storage on and make their own blends and so forth. But they were playing fast and loose with the books and kind of overspending. And eventually the bank calls in all the loans in 1898, and ultimately the Patterson brothers end up convicted of fraud. And the whole house guards falls in on itself. And suddenly all these distilleries have barrels that were down at Patterson they can't get back and aren't paid for and have overproduction sitting in warehouses that don't have homes anymore. And so price of whiskey drops way down and a whole bunch of them go broke. And John Duff was one of them. He ends up having to sell off both the distilleries to James Grant. And that's the Glen Livitt guy further down the road there creates a new entity called Longmorn Glenlivet Distillery for a number of years. And their function fine like that. Grant had not been part of the Patterson debacle, didn't get over leveraged. Longmore is famous for another reason. Our friend Masutaka Takasuri, the guy who worked initially with Suntory and then made the Nikkei Distillery. Worked there in the 1920s as part of his education to take whiskey to Japan. By 1970, they're starting to modernize. They. They dropped the name Longmorn from the name just called the Glenlivet distilleries. Longmorn is not doing a lot of single malts. They were mostly going into various kinds of blends. But 1970 was the very typical time where they switched away from coal heat to steam because it's safer, got rid of floor maltings, that sort of thing. Sort of standard modernization. And it was acquired in 1977 by the Seagrams Group, which at that time also owned the Chivas brothers. And Longmorn has always featured significantly in the Chivas blend. And that means it rolls up into Pernod Ricard by 2001. So for the most part you've never heard of Longmorn because it's just gone into blends. You've likely drank it if you drank any well known blends, especially Chivas Regal. They did make a few. Start making some single malt editions in 1993 without a lot of marketing into that. But the bot, the independent bottlers have always made big business from them. And so most people keep an eye out for whenever Hunter Liang like in this edition. But also, you know, pick your bottler. They often make lawnmowers and they are coveted, grabbed up right away, 250 bottles or something like that. The facility itself is large. They produce about four and a half million liters a year doing malts from the Mornay region. From Baird, 8 and a half ton grist loads is pretty big. Into 10 stainless steel washbacks at 39,000 liters. They run on four sets of stills. 10,000 liters for the wash stills and 7,000 liters for the spirit stills. Do a lot of their own barrel storage there. But because they're part of Pernod Ricard, barrels are distributed all over. Nothing unusual there. And again, not a well known brand. They are not one of the unbranded. But also not in the Diageo loop. They're in the Bernard Ricard loop. They're part of Chivas. So this particular bottling and I don't have it with me, it's back home. But I did get a chance to taste it and it was amazing. Comes from Hunter Liang. So this is actually originally. Hunter Liang is one of the sons of Douglas Liang. And Douglas Liang formed a company for doing blending and filling in 1948 and did very, very well. By the 1990s they were all on this modern single casking approach to whiskey because people like to have one of 250 bottles, that kind of thing. The two sons, Doug Jr. And Hunter, disagreed on a bunch of things and so by 2013 they decided to split the company and split up the brands. Doug Jr. Kept old, particular and a couple of other brands, but also is now started to move into Stilling, which I believe was their disagreement that Hunter liked making his cat his single bottlings and his relationship with distilleries and didn't want to get into stilling, where Doug Jr. Is moving in that direction. And the main line for Hunter Laing is this old malt cask of which this is one of the bottles of. So this is a 20 year old edition that was distilled in 1996 and bottled in 2017 and they're now looking at doing distributions for it. So for certain whiskey shops they send out samples, small little 100 mil bottles that you can that allow you to try them out so that you decide if you're going to stock some of those. So this hot 260 bottle run, they're
A
going to go there.
C
It's a single cask and very limited production and was 50% alcohol but it was a 20 year old. And you know, I'm usually lean away from those older whiskies, but this is one of those beautiful silky caramel Speysides. It's like everything you ever wanted in a whiskey. And the samples were basically for free and I was given one of them so I didn't ever pay for it. But you can expect if you could get your hands on one of these bottlings and there's a few different additions that have come down the line. If you keep your your eye out for getting access to Hunter Liang, which hard to come by out of Scotland, expect to pay about £200 for a bottle like that of a very special edition that you can serve your friends at Christmas and let them know you'll never have this again.
A
And you can't get it now though, right? Or can you?
C
I think yeah, we're just in the advanced sample. They haven't done the release on it. So if you were very keen you could go looking for.
A
Oh, you could get that. Okay, good.
C
You possibly, possibly.
A
But is that this the old malt cask?
C
Yeah, so yeah, the old malt cask is the, is the group Hunter Liang that does the various editions and they bottle from all kinds of distilleries. So you can't just look at old malt Cask. You got to specifically go hunting for a long mourn edition.
A
Well, this is kind of fun. Look at all this stuff. Is this a subscription?
C
Absolutely. They're not because it's so hard to ship, but it's one of those things where you kind of have to pay attention. And your whole thing here is go to your boutique whiskey maker that might be on the Hunter Liang list, ask about certain editions, see if he can get a few.
A
How fun is this? Look at all these.
C
And that's the whole thing about these, these custom bottlers, right? Is that you get all these interesting things from it. And just like I said, they're rare, they're special, but they're a way to get very unique bottles of whiskey if you want to have something unusual. So that being said, I am in Denmark and today I was handed the bottle of Danish whiskey. No surprise. The good news is I'm staying for another week in Denma. They'll probably see you next week.
A
They see you coming, don't they, Richard?
C
They do see me coming. And you know, I'm here to help. The nice thing is I'm here long enough that I can open this bottle and share it with friends and not worry about leaving it behind. We'll be able to finish it, but next week you'll see me in a different location. I think I'll be over the scandic at that point and we'll.
A
Will you share this Danish whiskey with us next.
B
Next week?
C
Yeah. You know, I hate to not do a Danish whiskey. We've done one before, right? We have talked about Danish whiskey before, but this is a different one. So I'll leave it fun until I can finish the research on. It takes a little time.
A
Well, bless you. You kept it in under an hour and I appreciate that. And you think you're giving me time to do some more work. So thank you for that. We do Windows Weekly. Reason I say that if you're watching live is in about 10 minutes. We've got Robert Turcic, who was a former creative director at MTV and is a futurist, talking about AI on Intelligent Machines. And so that's going to be a lot of fun. Stick around if you want to watch that. If you're not watching live, well, you'll find it on the IM feed a little bit later. We do Windows Weekly every Wednesday before Intelligent Machines. We do it 11am Pacific, 2pm Eastern, 1800 UTC. You can even watch us live if you want. If you're in the club, of course. In the club, TWiT Discord. But also we stream to everyone on YouTube, Twitch, X.com, facebook, LinkedIn and Kik after the fact on demand versions of the show at TWiT TV WW. We have audio and video there. The video is also on YouTube on the windows Weekly dedicated channel. Good way to share clips of the show. And actually Richard's whiskey clips, you don't need to clip those out. We Kevin King goes through those. He's a little behind. Takes a while. But if you go to something weird from my closet.com you'll find that YouTube playlist with all the 100, more than 100 whiskey 150 and the story of whiskey and all sorts of great stuff.
C
That would be the first. The first eight episodes are me explaining the Scottish whiskey making process to my friends here. And it runs two and a half hours. Hours?
A
Well, it takes 20 years. So two and a half hours, that's nothing. Nothing.
C
Come quick. Yeah.
A
Paul thurat is@therot.com Become a Premium Member and you can actually get copies of his books from leanpub.com and that includes Field Guide to Windows 11, Windows Everywhere and the latest in D and Shitify Windows. Or you can buy them from leanpub.com Richard Campbell is@runasradio.com that's where you'll find his shows Run as Radio and.net Rocks or DNR, which does not mean do not resuscitate.
C
In this case for 2,000 episodes. Maybe you shouldn't resuscitate.
A
Yeah, no, it's living. It's alive. Do not pound on its chest. It's. It's fine.
C
Not quite dead.
A
It's just resting, getting better.
B
Thank you, Richard.
A
Thank you, Paul. Thanks to all our winners and dozers out there. We'll see you next Wednesday on Windows Weekly.
C
Bye Bye.
A
You can't reason with the sun. Trust us, we've tried.
C
This summer, it's time to put that angry ball of fire on mute.
A
Columbia's Omnishade technology is engineered to protect
B
you from the sun's harsh rays that
A
can burn and damage your skin. The sun is relentless. But so is our gear. Level up your summer@columbia.com to spend more time outside and less time slathering on aloe lotion. You're welcome, Columbia. Engineered for whatever.
B
Some follow the noise. Bloomberg follows the money. Whether it's the funds fueling AI or or crypto's trillion dollar swings, there's a
A
money side to every story.
B
Get the money side of the story.
A
Subscribe now@bloomberg.com
B
There's a new way to
A
Sweetgreen meat wraps handheld, hearty and made
B
for life on the move.
A
With bold, chef crafted flavors, fresh ingredients and over 40 grams of protein, they're built to satisfy without slowing you down.
B
Try wraps today in the app or@order sweetgreen.com available at all participating locations.
Date: June 3, 2026
Host: Leo Laporte
Co-hosts: Paul Thurrott, Richard Campbell
Main Theme: The rise of local AI, Microsoft Build and Computex announcements, "liminal" computing platforms, and the next evolution of Windows and AI-powered devices.
This episode dives deep into Microsoft’s latest news from Build 2026 and Computex, focusing on the transition toward local AI on Windows devices, the implications of new hardware (especially Nvidia’s RTX Spark and the Surface Laptop Ultra), and a look ahead at Microsoft’s new “liminal” platform vision led by Project Celera. The hosts also break down developer-focused enhancements, market trends, and how Microsoft positions itself for the post-cloud era of AI. The tone is a lively, occasionally irreverent but highly informed roundtable.
Nvidia and Microsoft's New AI Laptops
Nvidia RTX Spark: Announced as a GPU-heavy, AI-centric chip for Windows ARM devices.
Surface Laptop Ultra: Microsoft's new flagship AI laptop, expected second half of 2026.
Implications:
Market Insight:
Privacy & Cost:
Developer Reactions:
Windows Developer Experience:
CLIs and App Creation:
Project Celera & Liminal OS (The Forward-Looking Platform Vision)
Final Word:
"It's a really interesting time. We're entering an era where the AI agent is the platform, and what that means for software, devices, and even business models is still to be discovered. 'Liminal' computing may be a new buzzword, but its promise is genuinely transformative."
— Windows Weekly panel, June 3, 2026