התמלול המלא של the localhost:0008 | return of the workstation

then shared 39 00:02:00,836 --> 00:02:04,068 that at the ASU Agentic AI Summit. 40 00:02:04,307 --> 00:02:07,563 will.i.am made a guest appearance, which I'll share more on throughout the show. 41 00:02:07,831 --> 00:02:08,106 but 42 00:02:08,401 --> 00:02:09,782 Robert: Wow 43 00:02:09,923 --> 00:02:12,803 Neil: Chicago, and Phoenix in the last week. 44 00:02:12,911 --> 00:02:13,635 Frankcx: Welcome back 45 00:02:13,735 --> 00:02:14,627 Jacob: How do we keep up? 46 00:02:14,769 --> 00:02:15,952 Frankcx: Yeah, it's good to have you. 47 00:02:16,677 --> 00:02:17,168 Hey, Robert, 48 00:02:17,168 --> 00:02:18,703 Robert: I'm in the middle of… You can see though now 49 00:02:18,704 --> 00:02:22,079 my office is put back together, different pink color on the wall. 50 00:02:22,081 --> 00:02:23,137 We're in the middle of construction. 51 00:02:23,138 --> 00:02:26,243 You might hear, a patio stone cutter in the background, but, 52 00:02:26,279 --> 00:02:27,802 hopefully AI Will keep that out. 53 00:02:28,157 --> 00:02:30,960 And as soon as we're done here, I won't dox myself 'cause this will be-- 54 00:02:30,970 --> 00:02:34,361 I'll be home by the time this airs, but I'm on my way to Columbus, to go 55 00:02:34,365 --> 00:02:36,228 see our son play soccer this weekend. 56 00:02:36,229 --> 00:02:39,099 Maybe take in a Columbus Blue Jackets game while we're there, 57 00:02:39,720 --> 00:02:41,110 and, looking forward to it. 58 00:02:41,152 --> 00:02:43,096 Frankcx: the Ohio State University? 59 00:02:43,211 --> 00:02:44,927 Robert: it, happens to be there, yes. 60 00:02:45,257 --> 00:02:46,703 I won't be there, but yes 61 00:02:48,236 --> 00:02:49,116 Frankcx: He's… Yeah, my 62 00:02:49,376 --> 00:02:51,673 Jacob: Will you tell us the exact address you'll be at, Robert? 63 00:02:51,757 --> 00:02:56,355 Frankcx: he always wanted me to dot the I because he was a Ohio native, 64 00:02:56,355 --> 00:02:57,759 but I never got the chance to do 65 00:02:57,760 --> 00:02:57,979 that. 66 00:02:57,996 --> 00:02:58,939 Let my dad down, 67 00:02:59,098 --> 00:02:59,834 Robert: sorry 68 00:03:00,497 --> 00:03:03,720 Frankcx: Speaking of euphoniums and tubas, Jacob 69 00:03:05,750 --> 00:03:06,978 Jacob: How's it going, everyone? 70 00:03:07,245 --> 00:03:11,662 last week we didn't have the pod, but, I wasn't on because I was in San Francisco 71 00:03:11,695 --> 00:03:17,718 at a Best Buy in downtown San Francisco with, folks like co-founder of NVIDIA, 72 00:03:17,737 --> 00:03:23,185 folks like the founder of OpenClaw, Peter, folks like the founder of Comfy 73 00:03:23,185 --> 00:03:26,841 UI, which we showed on stage, Microsoft showed on stage a couple weeks ago. 74 00:03:27,358 --> 00:03:31,412 and about 400 people buying their first RTX Spark devices. 75 00:03:32,914 --> 00:03:34,723 Yes, which was awesome 76 00:03:34,863 --> 00:03:36,206 Frankcx: other OEMs too? 77 00:03:37,463 --> 00:03:41,350 Jacob: this is a chance on this

podcast to say that I don't believe any of the 78 00:03:41,363 --> 00:03:47,690 purchases were Surface Laptop Ultras because, that specific location did not 79 00:03:47,716 --> 00:03:50,537 assort Surface Laptop Ultras in person. 80 00:03:50,751 --> 00:03:51,747 Frankcx: transparency 81 00:03:51,780 --> 00:03:55,773 Jacob: sales online, tons of excitement about the platform, but, I was the 82 00:03:55,784 --> 00:03:59,270 lone representative for Microsoft there to, to help kinda connect 83 00:03:59,270 --> 00:04:00,923 the dots and, and be a part of it, 84 00:04:01,145 --> 00:04:03,565 Robert: keeping track, that was Jacob's third low-key 85 00:04:03,565 --> 00:04:05,333 brag already on this podcast. 86 00:04:05,334 --> 00:04:05,791 Just… 87 00:04:06,884 --> 00:04:11,052 Frankcx: Well, Neil worked in Neil I Am, or Neil I Am, Will.i.am. 88 00:04:11,752 --> 00:04:12,168 Neil: like that. 89 00:04:12,169 --> 00:04:12,454 My new 90 00:04:12,574 --> 00:04:14,056 Jacob: I feel I am 91 00:04:15,254 --> 00:04:16,260 Frankcx: You are a DJ. 92 00:04:16,346 --> 00:04:17,450 Is that your DJ name? 93 00:04:17,703 --> 00:04:19,234 Neil: no, I go with Neil Nebula. 94 00:04:19,288 --> 00:04:23,414 and I can share my, recent Redmond debut in the Microsoft tree houses, 95 00:04:23,527 --> 00:04:24,279 Frankcx: Ooh, 96 00:04:24,484 --> 00:04:26,209 Neil: recently connected as well, so 97 00:04:26,330 --> 00:04:28,614 Frankcx: I will cut in the video if you have some to share. 98 00:04:28,614 --> 00:04:31,853 I'll put it in the pod right here, so awesome. 99 00:04:32,267 --> 00:04:34,275 All right, well, Chauncey is out this week. 100 00:04:34,284 --> 00:04:37,799 I think he's down at a wedding or some other event down in Mexico. 101 00:04:37,870 --> 00:04:41,002 But we are honored to have Scott Slater here. 102 00:04:41,030 --> 00:04:44,194 Scott joins us from New York, in the house. 103 00:04:44,216 --> 00:04:48,423 Scott's old school, Microsoft Surface and Microsoft's, we call 104 00:04:48,423 --> 00:04:50,801 it an M2, like he's a sales leader. 105 00:04:51,095 --> 00:04:53,399 so Scott, I'll turn it over to you to kinda introduce 106 00:04:53,400 --> 00:04:55,641 yourself and how are you doing? 107 00:04:55,979 --> 00:04:58,874 Scott Slater: Yeah, and I'll say the honor's all mine to be with you 108 00:04:58,882 --> 00:05:01,337 fine gentlemen, and I wasn't sure where you were starting when you 109 00:05:01,338 --> 00:05:03,092 started sharing I thought it was old. 110 00:05:03,198 --> 00:05:03,410 Frankcx: old 111 00:05:03,667 --> 00:05:05,372 Scott Slater: yeah, pleasure to join you all. 112 00:05:05,372 --> 00:05:07,773 Looking forward to having a fun conversation. 113 00:05:08,112 --> 00:05:11,950 I'm one of the leaders in Microsoft's Windows specialist organization. 114 00:05:12,348 --> 00:05:17,203 no great stories this week, but I was in and out of New York City meeting 115 00:05:17,206 --> 00:05:20,897 with different

large customers, which is generally how I spend my 116 00:05:20,897 --> 00:05:24,280 weeks, meeting from customer to customer across different industries 117 00:05:25,974 --> 00:05:27,356 Frankcx: Scott, I'm gonna open it up. 118 00:05:27,358 --> 00:05:28,876 We're changing the format a little bit. 119 00:05:28,949 --> 00:05:33,116 I've got these top questions of the week that I'd love to be able to ask, 120 00:05:33,116 --> 00:05:37,052 and as kind of the host week, my first question is actually to Scott, because 121 00:05:37,052 --> 00:05:40,448 it sounds like making that trip into New York City, meeting with those 122 00:05:40,457 --> 00:05:45,790 valued customers at Microsoft, what are customers actually asking about 123 00:05:45,823 --> 00:05:49,278 in AI and specifically AI on device? 124 00:05:49,326 --> 00:05:52,972 I mean, sounds like you're on the Windows side now, so Windows is obviously 125 00:05:52,975 --> 00:05:56,481 more than just AI on device, but, coming from the Surface side, it'd be 126 00:05:56,482 --> 00:05:59,854 interesting to get your perspective as to how customers are thinking about it. 127 00:06:00,352 --> 00:06:02,005 Scott Slater: Yeah, it's a good question, Frank. 128 00:06:02,060 --> 00:06:06,397 I think, the conversation flips at some point with customers. 129 00:06:06,398 --> 00:06:09,186 I think some customers start talking about the hardware. 130 00:06:09,586 --> 00:06:12,925 They may ask about the chip, the silicone. 131 00:06:12,936 --> 00:06:15,064 They may ask what an NPU is. 132 00:06:15,076 --> 00:06:21,980 They may ask about memory and tops and GPUs and, and after getting a 133 00:06:21,981 --> 00:06:26,269 little bit into that conversation, it kinda changes to, you know, can 134 00:06:26,270 --> 00:06:31,387 I, can I use the devices I have today or do I need to get new devices? 135 00:06:31,606 --> 00:06:33,771 And that's kinda where the conversation flips. 136 00:06:34,086 --> 00:06:38,823 And then at some point the conversation goes from there to, where is my data 137 00:06:38,861 --> 00:06:44,797 and is it secure as I'm running these AI workloads on these new AI PCs? 138 00:06:45,165 --> 00:06:49,724 And then at some point it shifts from there to like, does everybody need 139 00:06:49,725 --> 00:06:53,588 this in my environment or do only some people need it in my environment? 140 00:06:54,044 --> 00:06:54,416 And so 141 00:06:54,417 --> 00:06:57,104 it's interesting how these conversations are going with customers. 142 00:06:57,110 --> 00:07:01,980 But I usually I see it starting with like understanding about today's new 143 00:07:01,989 --> 00:07:07,019 hardware and then it quickly shifts to who needs this and can I do this on 144 00:07:07,019 --> 00:07:08,914 what I have or do I need something new? 145 00:07:09,140 --> 00:07:12,947 Frankcx: I find that an interesting question because, as I was watching a lot 146 00:07:12,947 --> 00:07:18,479 of the, reviews of the RTX Spark, device, and whether that's a Surface Laptop Ultra 147 00:07:18,487 --> 00:07:22,504 or the other OEMs that are selling those devices, one of them that came to mind 148 00:07:22,504 --> 00:07:26,424 was Linus Tech Tips, a guy that's been in the internet for quite some time. 149 00:07:26,424 --> 00:07:29,617 He's got quite a bit of experience out there, sits up in Vancouver, 150 00:07:30,000 --> 00:07:31,456 watching over all of this. 151 00:07:31,461 --> 00:07:39,341 But his actual take on it was like, who on earth would spend $6,000 for an AI laptop 152 00:07:39,406 --> 00:07:44,060 when you can through a tool like Tailscale or another tool that allows you to 153 00:07:44,060 --> 00:07:47,542 tunnel into an AI device, into your home? 154 00:07:47,630 --> 00:07:51,289 And I'll, I'll say that firstfor- and foremost, that's honestly probably 155 00:07:51,290 --> 00:07:53,652 the way that I use AI the most today. 156 00:07:54,112 --> 00:07:58,063 I have a device that's literally right behind me, AI models. 157 00:07:58,091 --> 00:08:01,812 I'll talk a little bit about this, Ride Edge app that I've developed, 158 00:08:02,115 --> 00:08:03,713 but, that's the way I use it. 159 00:08:03,715 --> 00:08:09,371 I use my phone, in some ways to Tailscale, , a nice portal and a nice kind of tunnel 160 00:08:09,395 --> 00:08:13,382 application that cr- creates a secure loca- secure tunnel into the device. 161 00:08:13,803 --> 00:08:17,214 So I would wonder, like, why would people even buy a laptop? 162 00:08:17,220 --> 00:08:20,940 But Jacob, I know that you were at the Best Buy, and you were 163 00:08:20,943 --> 00:08:24,597 watching people walk out the door with a bunch of AI laptops. 164 00:08:24,614 --> 00:08:27,071 You know, what was the reason that they were doing that? 165 00:08:28,437 --> 00:08:31,688 Jacob: Yeah, I mean, first of all, I, I think, your perspective, Frank, 166 00:08:31,700 --> 00:08:34,336 is mostly because you're using the hardware you have available. 167 00:08:34,599 --> 00:08:38,271 Almost to, to, to Scott's point earlier, like customers are asking 168 00:08:38,271 --> 00:08:41,316 like, "Do I need to buy something new or can I use what I already have?" 169 00:08:41,570 --> 00:08:46,418 Like, if you've got dual 3090s, like you're already set up really well. 170 00:08:46,427 --> 00:08:49,792 Like that's gonna cost you… Like that whole setup is probably 10 171 00:08:49,792 --> 00:08:52,827 grand on the open market these days with the price of graphics cards. 172 00:08:53,308 --> 00:08:55,832 so I would say that yes, you're kind of an edge case in that you're 173 00:08:55,832 --> 00:08:59,907 already set up well, and you don't have one of these, from what I know, 174 00:08:59,998 --> 00:09:02,826 Laptop Ultras or RTX Spark devices yet 175 00:09:02,911 --> 00:09:03,589 Frankcx: one, but nobody 176 00:09:03,799 --> 00:09:05,023 Robert: Get in line 177 00:09:05,208 --> 00:09:05,382 Frankcx: as 178

00:09:05,508 --> 00:09:06,086 Jacob: Yeah 179 00:09:06,177 --> 00:09:08,988 Frankcx: influencer, now I'm not as big as, Neil I am 180 00:09:13,017 --> 00:09:17,034 Jacob: but what I will say is that customers are just looking for devices 181 00:09:17,034 --> 00:09:19,483 that match their needs in my perspective. 182 00:09:19,520 --> 00:09:22,430 And so the folks that walked out of that Best Buy, they were excited. 183 00:09:22,658 --> 00:09:25,969 it was a little bit more on the consumer side where folks are looking 184 00:09:25,969 --> 00:09:28,997 for their AI work, they're looking for their gaming work, they're 185 00:09:28,999 --> 00:09:30,680 looking for their development work. 186 00:09:30,969 --> 00:09:33,925 They're-- Like some of them were AI founders themselves. 187 00:09:33,925 --> 00:09:37,869 I got pitched on a number of different, startups that week on like, "Hey, do you 188 00:09:37,869 --> 00:09:41,400 have access to the Microsoft socials? Like you work in marketing. can you get 189 00:09:41,427 --> 00:09:44,498 me connected with the right person here or there?" I'm like: Okay, I'd love to 190 00:09:44,499 --> 00:09:46,143 hear your idea, but let's chat about it. 191 00:09:46,197 --> 00:09:50,834 But I think that users just want a device that they can turn on, that 192 00:09:50,835 --> 00:09:54,882 they can use, and by the way, a ton of great announcements this week and 193 00:09:54,882 --> 00:09:59,661 the last couple weeks from Microsoft and other partners on how it's easier 194 00:09:59,661 --> 00:10:03,889 than ever to just have local AI built into the tools that you're using. 195 00:10:03,965 --> 00:10:04,694 Robert: an RTX 196 00:10:04,797 --> 00:10:07,118 Jacob: like GitHub Copilot CLI, stuff like that. 197 00:10:07,167 --> 00:10:11,621 Robert: device, laptop as a server, or you can take it with you. 198 00:10:11,654 --> 00:10:15,323 Now, look it, full disclosure, I think we talked last podcast, do 199 00:10:15,324 --> 00:10:19,036 I spend six grand on a laptop or do I spend six grand on a patio? 200 00:10:19,832 --> 00:10:22,826 I think by, if you hear the noise You know, what my decision was. 201 00:10:23,540 --> 00:10:25,137 And no, it's not very mobile. 202 00:10:25,875 --> 00:10:29,414 but I… Look it, I love the idea of going to Neal's Point, going on a 203 00:10:29,415 --> 00:10:34,507 plane, going offline, and just grooving for four or five hours on a flight. 204 00:10:34,957 --> 00:10:37,341 That, dude, that to me is totally worth it 205 00:10:37,956 --> 00:10:39,151 Scott Slater: know, I, I'll 206 00:10:39,246 --> 00:10:41,510 Frankcx: Neil, I guess maybe the… Oh, go ahead, Scott 207 00:10:41,587 --> 00:10:44,499 Scott Slater: Yeah, I was just gonna say, I, I think I, I really agree 208 00:10:44,500 --> 00:10:46,270 with, with Jacob and Robert here. 209 00:10:46,339 --> 00:10:49,909 I don't think… If you're trying to save a few bucks on tokens, 210 00:10:49,965 --> 00:10:51,349 I don- I think Linus is right. 211 00:10:51,349 --> 00:10:53,191 You're, you're not buying the device, right? 212 00:10:53,557 --> 00:10:58,704 But, for, for builders, for creators, they're, they're buying freedom, right? 213 00:10:58,738 --> 00:11:04,268 Like, look, when you're running on a meter, your behavior changes, right? 214 00:11:04,381 --> 00:11:09,695 Like, if every try costs money, you're gonna play it safe a little 215 00:11:09,696 --> 00:11:14,411 more, but if you take that meter away, you're gonna try weird ideas. 216 00:11:14,411 --> 00:11:17,040 You're gonna run crazy tests overnight. 217 00:11:17,040 --> 00:11:20,483 You're gonna try 10 things versus one thing, right? 218 00:11:20,496 --> 00:11:24,621 Running on a meter, are more careful, right? 219 00:11:24,652 --> 00:11:30,588 And what you want is, is to take that away so you can drive the curiosity and, like, 220 00:11:31,000 --> 00:11:31,164 Robert: More 221 00:11:31,323 --> 00:11:32,229 Scott Slater: That, that's right. 222 00:11:32,260 --> 00:11:32,724 You know? 223 00:11:32,762 --> 00:11:36,850 And, and look, the people that are the most curious, that don't have 224 00:11:36,850 --> 00:11:40,819 that restriction and are the most creative, those are the ones who 225 00:11:40,821 --> 00:11:45,787 think of the next big things, I think that's the big thing here, to enable 226 00:11:46,001 --> 00:11:47,843 the freedom to get that outcome 227 00:11:48,790 --> 00:11:50,427 Neil: Totally agree with that, Scott. 228 00:11:50,874 --> 00:11:53,779 I was gonna say, there's been a big push from enterprises around 229 00:11:54,024 --> 00:11:57,805 token economics, and we're running this workload in the cloud. 230 00:11:57,814 --> 00:12:01,338 What facet of that workload can come to the edge, and what does that, 231 00:12:01,367 --> 00:12:04,485 correspond to in terms of an ROI model over the cost of this device? 232 00:12:04,863 --> 00:12:08,181 And additional value add is, what is the cost of lost experimentation? 233 00:12:08,210 --> 00:12:12,861 If I have a finite token budget and I have my role specifically in scope, I'm 234 00:12:12,862 --> 00:12:17,699 preserving those tokens to accelerate what I need to accomplish, and I'm not 235 00:12:17,708 --> 00:12:19,406 thinking about doing and achieving more. 236 00:12:19,407 --> 00:12:24,019 So this notion of unmetered intelligence, right, taking away that. 237 00:12:24,028 --> 00:12:29,123 finite token limit not only helps accelerate productivity, whether you're 238 00:12:29,125 --> 00:12:33,766 35,000 feet in the air, on-site with a client in an air-gapped environment, 239 00:12:34,029 --> 00:12:39,239 but it also unlocks, the opportunity to create tooling that wouldn't 240 00:12:39,239 --> 00:12:43,693 be possible when you're operating under just a fixed cap mindset 241 00:12:44,334 --> 00:12:45,382 Scott Slater: know, Neil, I, 242 00:12:46,412 --> 00:12:50,830 I had a developer say that exact thing last

week, and they wanted 243 00:12:50,841 --> 00:12:52,893 the right to fail cheaply, 244 00:12:53,193 --> 00:12:54,845 Neil: Mm. 245 00:12:54,993 --> 00:12:59,469 Scott Slater: wanted the right to not ask permission, and I think that's part 246 00:12:59,469 --> 00:13:03,157 of why people want these devices where they don't have to ask permission, 247 00:13:03,159 --> 00:13:04,869 and the cloud doesn't give them that. 248 00:13:05,183 --> 00:13:07,909 You know, in these big enterprise companies, they need to ask 249 00:13:07,924 --> 00:13:10,165 permission to have more tokens 250 00:13:12,325 --> 00:13:15,644 Robert: if we wanna tie this back to the original question to Scott about 251 00:13:15,645 --> 00:13:18,749 what customers are saying and then the token economics piece, I learned this 252 00:13:18,749 --> 00:13:23,821 week through talking with customers the amount of overspend that is happening in 253 00:13:23,823 --> 00:13:32,825 the cloud, either due to token usage or token cost, raw token cost, dwarfs the 254 00:13:32,834 --> 00:13:38,524 increased price is for RAM or storage or GPUs or anything in the hardware. 255 00:13:38,590 --> 00:13:40,598 It absolutely dwarfs it. 256 00:13:40,618 --> 00:13:45,917 So for some of us that are in the Windows and PC game, we're like, "Oh 257 00:13:45,917 --> 00:13:48,898 my God, the world's on fire." To these guys, it's like, "Eh, it's a drop in 258 00:13:48,899 --> 00:13:51,213 the bucket" for, for the potential 259 00:13:51,266 --> 00:13:51,506 Jacob: Угу 260 00:13:51,582 --> 00:13:51,972 Robert: could save. 261 00:13:52,630 --> 00:13:53,839 It was a huge eye-opener 262 00:13:54,975 --> 00:13:58,323 Frankcx: this is a question back to Scott too, and I'd love other people's input 263 00:13:58,326 --> 00:14:01,334 because it feels like everybody's doing a little bit of vibe coding out there. 264 00:14:01,799 --> 00:14:06,805 But Scott, as you look at, new sales reps that you're hiring on and, bringing 265 00:14:06,805 --> 00:14:10,901 into the fold, obviously different reps have different skills of relationship 266 00:14:10,919 --> 00:14:12,908 compared to technology background. 267 00:14:13,265 --> 00:14:17,454 But with the vibe coding story that's out there these days, what happens when every 268 00:14:17,455 --> 00:14:22,679 seller can build software of their own and be able to tell a story to their customers 269 00:14:22,688 --> 00:14:24,840 to showcase the value of these devices? 270 00:14:24,842 --> 00:14:26,431 I don't know if you've started to see that. 271 00:14:26,697 --> 00:14:30,112 Obviously, Robert works on your team, so that might be one of your 272 00:14:30,344 --> 00:14:34,238 example reps that, you have out there that can showcase, like, I can 273 00:14:34,240 --> 00:14:39,347 customize something to really sell the customer on the story of what AI has. 274 00:14:39,681 --> 00:14:41,429 But I'm wondering, do you have a take on that? 275 00:14:41,867 --> 00:14:44,561 Scott Slater: Yeah, I've thought about that, and I think what worries 276 00:14:44,576 --> 00:14:50,045 me a little bit is that, look, every person, every seller can build today. 277 00:14:50,382 --> 00:14:55,525 They can build apps, dashboards, trackers, automation, right? 278 00:14:55,861 --> 00:14:58,108 Building isn't the hard part anymore. 279 00:14:58,166 --> 00:14:59,726 That's the easy part now. 280 00:15:00,025 --> 00:15:06,191 But keeping something running and, that's a hard thing to do, or knowing 281 00:15:06,379 --> 00:15:11,271 which problem you should actually build something for, I think that's the 282 00:15:11,272 --> 00:15:13,312 harder part today, in my opinion, right? 283 00:15:13,533 --> 00:15:17,526 The building part is easier and, and I think what we're gonna see soon 284 00:15:17,526 --> 00:15:21,923 is, like, thousands of these little apps and dashboards and trackers 285 00:15:22,179 --> 00:15:25,144 that one person owns or nobody owns. 286 00:15:25,178 --> 00:15:30,301 And it means is IT's not owning this, and I think that's a problem that's gonna 287 00:15:30,301 --> 00:15:33,198 really raise up soon in all organizations 288 00:15:33,198 --> 00:15:38,101 As all these things get built, they're not necessarily solving for the right things 289 00:15:39,348 --> 00:15:39,994 Neil: I have an ig- 290 00:15:40,217 --> 00:15:43,332 Frankcx: for the rest of the team here, like being able to customize something 291 00:15:43,333 --> 00:15:48,015 that kind of specifically aims at the customer's voice, if you've seen 292 00:15:48,023 --> 00:15:50,061 value in that type of storytelling? 293 00:15:51,400 --> 00:15:52,287 Neil: 100%. 294 00:15:52,305 --> 00:15:59,361 I walk into every customer partner meeting with a custom demo pack, and to Scott's 295 00:15:59,365 --> 00:16:04,602 point, art of possible, doing my homework, meeting with the reps that already know 296 00:16:04,602 --> 00:16:09,960 that customer, doing my research online, so that I can come prepared with a custom 297 00:16:09,979 --> 00:16:14,305 artifact, functional local AI solution in every customer meeting has been my 298 00:16:14,306 --> 00:16:16,278 mindset for the last eight, nine months. 299 00:16:16,611 --> 00:16:19,634 Keep in mind, I have an electrical computer engineering background. 300 00:16:19,839 --> 00:16:22,282 I hadn't touched code in 15 years, right? 301 00:16:22,282 --> 00:16:26,956 And so the ability to now, again, create these prototypes in my favorite language, 302 00:16:26,970 --> 00:16:30,623 English, meet customers where they are on their journey, extremely impactful. 303 00:16:30,916 --> 00:16:32,881 I do wanna highlight the inverse of that, though. 304 00:16:32,940 --> 00:16:37,102 In a world where you can build anything and creation is no longer 305 00:16:37,103 --> 00:16:40,588 the bottleneck, choosing which problems to solve, super important. 306 00:16:40,890 --> 00:16:45,332 I met with a

customer, and they had, essentially a rogue experiment gone wrong. 307 00:16:45,608 --> 00:16:48,830 Someone came up with the idea of a brand-new ERP platform. 308 00:16:49,118 --> 00:16:54,977 They consumed six figures in tokens, only to find out that they had already licensed 309 00:16:54,978 --> 00:16:57,327 a perfectly functional ERP platform. 310 00:16:57,630 --> 00:17:02,091 So that's over 100 grand down the drain on top of that time investment. 311 00:17:02,489 --> 00:17:05,560 So again, when in a world where you can build anything, make sure 312 00:17:05,561 --> 00:17:06,662 you're building the right thing. 313 00:17:06,934 --> 00:17:10,478 Avoid that, disaster scenario of 100 grand in tokens down the 314 00:17:10,478 --> 00:17:12,526 drain to build a duplicate system 315 00:17:13,329 --> 00:17:19,720 Robert: And the, the other side to that is say you have 1,000 things people 316 00:17:19,720 --> 00:17:22,896 built, but they're built by 1,000 different people, and then what becomes 317 00:17:22,920 --> 00:17:27,101 the standard for… I mean, everybody's building something that they think 318 00:17:27,101 --> 00:17:31,740 is awesome, but now we have like 100 different versions of something, and 319 00:17:31,740 --> 00:17:35,406 nobody knows which is the, you know, which is the blessed one and, and whatnot. 320 00:17:35,407 --> 00:17:38,777 I had to build a customer one-pager today, for one of our 321 00:17:38,777 --> 00:17:40,767 OEMs and couldn't find anything. 322 00:17:40,860 --> 00:17:43,274 Took a long while to do it, and the first thought that came to my mind is, 323 00:17:43,277 --> 00:17:47,485 "Oh, I should share this." And I'm like, "No." I know what that's gonna create 324 00:17:47,485 --> 00:17:51,794 is the problem of this proliferation, and then the governance and ownership of 325 00:17:52,255 --> 00:17:53,989 what you should or shouldn't be sharing 326 00:17:55,191 --> 00:17:58,432 Frankcx: I'd love to turn this to Jacob, because Jacob, you saying that 327 00:17:58,432 --> 00:18:01,703 you stood in a Best Buy with a bunch of influential people, but also watched 328 00:18:01,704 --> 00:18:03,420 a lot of devices go out the door. 329 00:18:03,870 --> 00:18:07,102 I'm reminded of, when I was at Build one time and somebody came up and 330 00:18:07,103 --> 00:18:12,880 literally bought a $4,000 Surface laptop, and now I see that there 331 00:18:12,880 --> 00:18:14,491 is a return of the workstation. 332 00:18:14,845 --> 00:18:17,039 the workstation, I would say, once died. 333 00:18:17,093 --> 00:18:20,482 we didn't see a value because people were pushing so much up to Azure. 334 00:18:20,781 --> 00:18:25,978 Laura, coworker of, Robert's and Neil's in the past, or still Neil, she pointed 335 00:18:25,978 --> 00:18:30,481 out, like, I spent a decade pushing people to start using Azure in the 336 00:18:30,481 --> 00:18:34,997 cloud, and not that Azure-- I think the Azure numbers are still secure, but 337 00:18:34,997 --> 00:18:38,237 we're starting to see people start to think about a workstation again, like 338 00:18:38,237 --> 00:18:43,699 the power of a super powerful device, a device that can make money, for you. 339 00:18:43,707 --> 00:18:47,322 So I'm wondering, Jacob, when people were dishing out that much money… 340 00:18:47,629 --> 00:18:50,730 Well, of course, everybody in San Francisco's got unlimited cash flow. 341 00:18:52,408 --> 00:18:53,328 Jacob: Not necessarily 342 00:18:53,428 --> 00:18:55,138 Frankcx: I'm like, what do you think about that? 343 00:18:55,221 --> 00:18:57,012 Like, the return of the workstation? 344 00:18:57,988 --> 00:19:03,928 Jacob: the first person that bought an RTX Spark device in San Francisco, he was 345 00:19:03,929 --> 00:19:08,835 on camera and shouted out to everyone, like, essentially, so excited to use this 346 00:19:08,852 --> 00:19:16,656 to build without limits and to be able to not have to worry about my token costs and 347 00:19:16,664 --> 00:19:19,782 also do some awesome gaming and stuff." 348 00:19:19,848 --> 00:19:22,440 So he was going to build his business around it. 349 00:19:22,448 --> 00:19:24,173 What his business was, I have no clue, 350 00:19:24,269 --> 00:19:24,361 Robert: to 351 00:19:24,589 --> 00:19:26,609 Jacob: but he was gonna build his business around this investment. 352 00:19:26,821 --> 00:19:27,372 Robert: How long did it 353 00:19:27,387 --> 00:19:27,865 Jacob: What's that? 354 00:19:28,118 --> 00:19:28,512 Robert: that? 355 00:19:30,529 --> 00:19:30,791 thought he 356 00:19:30,894 --> 00:19:32,858 Jacob: you can look at the PCMag video. 357 00:19:32,876 --> 00:19:34,726 Like, they've got an Instagram reel up. 358 00:19:34,738 --> 00:19:37,741 They've probably got TikToks up where it highlights what he said 359 00:19:37,771 --> 00:19:38,926 when he was walking out of that. 360 00:19:39,248 --> 00:19:42,129 But it was… Maybe I butchered it, but it was something to that effect 361 00:19:42,608 --> 00:19:45,074 They're like, "Oh, we wanna clap," but, "Oh, he keeps on going. 362 00:19:45,085 --> 00:19:48,805 Oh, he keeps on going." It was like a quasi interview question, and everyone was 363 00:19:48,805 --> 00:19:50,296 cheering and clapping and it was exciting. 364 00:19:50,460 --> 00:19:54,178 That was the type of sentiment that it was kind of a hybrid of what Scott and 365 00:19:54,178 --> 00:19:57,745 Robert and y- like what we were already talking about that like kind of opens up 366 00:19:57,745 --> 00:20:01,578 your ability to create, your ability to, to not feel like you're compromised by the 367 00:20:01,579 --> 00:20:07,104 amount of, tokens you have or even by like the sovereignty of your data, like where 368 00:20:07,104 --> 00:20:12,766 it's going, like what, what, you know, what, who has access to your sensitive. 369 00:20:12,774 --> 00:20:15,756 For a lot of these founders that I talk to, like who has access 370

00:20:15,759 --> 00:20:17,523 to their sensitive information? 371 00:20:18,152 --> 00:20:22,567 and they're not just gonna use a standard, you know, consumer, AI model 372 00:20:22,976 --> 00:20:26,345 in the cloud for building their own business that may be competing with 373 00:20:26,381 --> 00:20:29,116 that cloud solution that already exists. 374 00:20:29,271 --> 00:20:29,647 Frankcx: Jacob, 375 00:20:29,793 --> 00:20:31,090 Jacob: And so that's something interesting. 376 00:20:31,097 --> 00:20:36,089 There are obviously enterprise versions of these that have license agreements, 377 00:20:36,117 --> 00:20:41,119 but it's still an interesting thing like to think about that privacy 378 00:20:41,189 --> 00:20:44,111 Frankcx: Well, I know that Jacob, you like to make fun of me because I make reference 379 00:20:44,111 --> 00:20:48,244 to some of the older technologies that I've worked with, but it would be 380 00:20:48,246 --> 00:20:50,520 that, you know, we watch that pendulum 381 00:20:50,660 --> 00:20:51,239 Jacob: Got it, thanks 382 00:20:51,540 --> 00:20:55,293 Frankcx: where we talk about going from like a mainframe to a PC, and 383 00:20:55,293 --> 00:20:59,006 I would argue like Lotus 1-2-3 was probably like what caused a lot of 384 00:20:59,015 --> 00:21:03,768 people to start buying a PC what is that application that allowed them to do that? 385 00:21:03,786 --> 00:21:06,907 I always name drop that I worked for this company, Silicon Graphics, that 386 00:21:06,908 --> 00:21:11,251 got put literally out of business by Nvidia at the time by selling a $300 387 00:21:11,264 --> 00:21:15,937 gaming card, that business caused people to go from like a large device 388 00:21:15,938 --> 00:21:20,774 to a small device again because of this $300 gaming card and be able to do 389 00:21:20,785 --> 00:21:22,774 things like Photoshop on your device. 390 00:21:23,236 --> 00:21:27,172 But I would say like, what's that thing that's really moving the needle or moving 391 00:21:27,173 --> 00:21:32,432 the dial, you know, that causes people to buy a workstation, like at $6,000? 392 00:21:32,448 --> 00:21:33,966 Is it open weight models? 393 00:21:33,969 --> 00:21:35,756 Is it agentic coding? 394 00:21:36,131 --> 00:21:37,912 Like, you know, what are you seeing out there? 395 00:21:37,912 --> 00:21:41,064 And I guess for Neil and Robert and Scott, as well, like what are 396 00:21:41,066 --> 00:21:44,039 you guys seeing that is kind of pushing people to have interest in 397 00:21:44,040 --> 00:21:45,724 these workstation class devices? 398 00:21:46,418 --> 00:21:47,130 Robert: you go first 399 00:21:47,964 --> 00:21:49,208 Scott Slater: yeah, you have the honor. 400 00:21:49,299 --> 00:21:53,219 So you go, then I'll go, 'cause I think I'll have a different perspective 401 00:21:53,646 --> 00:21:58,605 Robert: I think the open weight models is the one that's getting 402 00:21:58,605 --> 00:22:04,723 people hooked and interested,

as the first kind of step to this. 403 00:22:05,056 --> 00:22:08,041 I think it goes back to, I don't know who said it, but about 404 00:22:08,052 --> 00:22:12,061 the, like, the unlimited, like, limitless creativity, right? 405 00:22:12,085 --> 00:22:17,805 I have the ability to do this without a connection, without oversight, without 406 00:22:18,076 --> 00:22:20,282 token spend, all of these kind of things. 407 00:22:20,582 --> 00:22:26,100 I would love nothing more than to have a AI workstation, just to 408 00:22:26,231 --> 00:22:28,311 run everything I wanna do off of. 409 00:22:28,311 --> 00:22:35,172 There's probably 80% of the things I don't do right now is because I don't have AI 410 00:22:35,186 --> 00:22:37,199 workstation sitting right here beside me. 411 00:22:37,486 --> 00:22:41,038 there's a whole bunch of stuff I wanna try out and do, but I'm not gonna burn 412 00:22:41,075 --> 00:22:42,896 tokens to go, to go figure that out. 413 00:22:42,901 --> 00:22:45,458 I'll wait till, the chains come off, so to speak 414 00:22:47,224 --> 00:22:48,179 Scott Slater: I'll add to that. 415 00:22:48,192 --> 00:22:53,186 what I'm excited about is what's going to happen, right? 416 00:22:53,240 --> 00:22:57,063 Like, so I know we're talking about on-device AI and, you 417 00:22:57,064 --> 00:22:58,238 know, what's exciting there. 418 00:22:58,238 --> 00:23:04,243 I think what the exciting part is the stuff that users don't even notice, right? 419 00:23:04,555 --> 00:23:08,054 And I think that's just gonna continue to evolve, right? 420 00:23:08,055 --> 00:23:14,011 So we know, like, now there's AI models that make video calls better, right? 421 00:23:14,037 --> 00:23:19,969 Or that can do live captions, or you can search files faster, right? 422 00:23:20,110 --> 00:23:23,873 Or stuff that can work when you're 30,000 feet on a plane with no 423 00:23:23,881 --> 00:23:28,141 connectivity, and like, that's the stuff that gets me real excited. 424 00:23:28,142 --> 00:23:31,802 Like, don't get me wrong, I'm still excited about local models and, you 425 00:23:31,803 --> 00:23:35,781 know, I'm prepared to talk about local models today, and I know we announced 426 00:23:35,783 --> 00:23:37,786 a whole bunch, about a week ago, right? 427 00:23:37,786 --> 00:23:40,845 Like MAI, which stands for Microsoft AI, by the way. 428 00:23:40,896 --> 00:23:41,379 Robert: Yeah 429 00:23:41,632 --> 00:23:43,840 Scott Slater: A super creative name that we came up with, right? 430 00:23:44,086 --> 00:23:45,550 I'm really excited about that. 431 00:23:45,565 --> 00:23:51,646 Like, the stuff that nobody knows and it's just there that the new hardware enables. 432 00:23:51,997 --> 00:23:56,573 And, new hardware plus, the models and the apps that are tapping into those models. 433 00:23:56,613 --> 00:24:01,967 I'll try not to talk a lot about Microsoft products, but, the hybrid intelligence 434 00:24:01,980 --> 00:24:04,768 stuff really I'm excited about, right? 435 00:24:04,775 --> 00:24:10,295 'Cause the idea of that, that was announced, is that a user wouldn't 436 00:24:10,304 --> 00:24:16,420 know, Like, the hybrid intelligence would decide if it's leveraging a 437 00:24:16,420 --> 00:24:21,691 cloud model or a local model based on some efficiency metric, right? 438 00:24:21,709 --> 00:24:24,044 Like performance, latency, cost. 439 00:24:24,417 --> 00:24:25,721 Like, that's really… 440 00:24:26,050 --> 00:24:26,354 Robert: Yeah. 441 00:24:26,407 --> 00:24:26,884 Scott Slater: Privacy. 442 00:24:26,894 --> 00:24:28,313 Like, that's really interesting. 443 00:24:28,328 --> 00:24:32,241 Like, if you don't need to know as a user and you never notice it, but 444 00:24:32,242 --> 00:24:34,390 you have the right hardware, right? 445 00:24:34,634 --> 00:24:38,583 Frankcx: Yeah, Neil, I wonder, like, when you're talking to customers, that 446 00:24:38,584 --> 00:24:42,128 story that, Scott is, talking about it is really starting to resonate 447 00:24:42,167 --> 00:24:45,680 with a lot of our commercial or enterprise-grade customers out there? 448 00:24:46,636 --> 00:24:47,588 Neil: 100%. 449 00:24:47,613 --> 00:24:53,464 we've been promoting local hybrid intelligence, for years, and now that it's 450 00:24:53,484 --> 00:24:59,506 coming essentially baked into the tools customers already own, that barrier to 451 00:24:59,507 --> 00:25:02,286 adopt hybrid intelligence has disappeared. 452 00:25:02,562 --> 00:25:06,042 Before, you had to go out and either buy or build a routing 453 00:25:06,045 --> 00:25:09,571 mechanism, which frontier customers are doing, but now we're meeting 454 00:25:09,572 --> 00:25:11,005 them right where they are, right? 455 00:25:11,020 --> 00:25:15,103 that auto-routing capability in GitHub Copilot with HydroFusion, 456 00:25:15,384 --> 00:25:16,493 absolute game changer. 457 00:25:16,785 --> 00:25:20,401 I'll highlight on the topic of unlimited creativity, I learned 458 00:25:20,401 --> 00:25:23,750 during my time at ASU, will.i.am is actually a professor there. 459 00:25:24,053 --> 00:25:28,926 He has 200 students, in the creative department, and NVIDIA gave him a grant. 460 00:25:28,937 --> 00:25:33,762 All 200 of his students have a Spark device, to tinker with, build 461 00:25:33,763 --> 00:25:34,781 with throughout their semester. 462 00:25:34,960 --> 00:25:37,347 Frankcx: Can I sit in on that course? 463 00:25:37,708 --> 00:25:39,876 Neil: going back to school myself to attend this one. 464 00:25:40,136 --> 00:25:41,498 and so think about that, right? 465 00:25:41,538 --> 00:25:45,176 Unlimited creativity for will.i.am students throughout the semester, 466 00:25:45,392 --> 00:25:49,455 essentially creating a digital twin of their creative, persona, whether 467 00:25:49,455 --> 00:25:51,490 they're, a musician, artist or more. 468 00:25:51,716 --> 00:25:54,701 we've really just hit the tip of the iceberg in terms of what's

possible 469 00:25:54,707 --> 00:25:59,534 when we bring unmetered intelligence, and hybrid intelligence together, 470 00:26:00,215 --> 00:26:01,916 unlocking a new era of computing 471 00:26:02,539 --> 00:26:03,233 Scott Slater: So Neil 472 00:26:03,734 --> 00:26:05,990 Jacob: I mean, you guys have been doing this for a while. 473 00:26:05,992 --> 00:26:10,868 Like, I remember Robert, Neil, you showed me a demo two years ago of, essentially 474 00:26:10,868 --> 00:26:14,908 deciding where that workload should go, you know, whether it's cloud or it's not. 475 00:26:14,998 --> 00:26:17,446 and I remember being blown away by that, going like, "Okay." 476 00:26:17,446 --> 00:26:20,470 And that took some effort to do. 477 00:26:20,501 --> 00:26:22,973 Like, we would talk about, oh, it's just some lines of code, 478 00:26:23,013 --> 00:26:24,209 but, it took some effort to do. 479 00:26:24,209 --> 00:26:26,647 You'd have to have the right accounts, the right things, and, 480 00:26:26,647 --> 00:26:28,066 having that built in is awesome. 481 00:26:28,113 --> 00:26:32,488 And I know Robert, shameless plug for the Surface IT Pro blog, but we've got, 482 00:26:32,548 --> 00:26:34,984 a blog post that just landed this week. 483 00:26:35,258 --> 00:26:37,579 it was actually authored by one of our very own, so we've got a 484 00:26:37,579 --> 00:26:38,864 thought leader here on this call. 485 00:26:38,876 --> 00:26:40,402 Robert, you wanna talk at all about that? 486 00:26:41,448 --> 00:26:42,989 Robert: how awkward. 487 00:26:43,251 --> 00:26:44,145 Frankcx: How awkward. 488 00:26:45,714 --> 00:26:48,925 Robert: listen, to your point, Jacob, full disclosure, I was in Surface. 489 00:26:48,951 --> 00:26:49,793 Now I'm in Windows. 490 00:26:49,798 --> 00:26:53,515 Before I left, we wrote a hybrid intelligence blog, waited till the 491 00:26:53,515 --> 00:26:56,474 announcement came, so the people knew what we were talking about 492 00:26:56,474 --> 00:27:00,166 with hybrid intelligence, which essentially now replaces unmetered 493 00:27:00,169 --> 00:27:02,355 intelligence, in the nomenclature. 494 00:27:02,358 --> 00:27:04,446 But it's all about, what we've been talking about. 495 00:27:04,476 --> 00:27:08,549 Lookit, it's all about what our blog or what our podcast is about, and 496 00:27:08,550 --> 00:27:10,137 just talks about it in those terms. 497 00:27:10,370 --> 00:27:13,055 Frankcx: kinda hoping you'd title it Super Intelligence, 498 00:27:13,275 --> 00:27:13,730 Robert: Oh. 499 00:27:14,534 --> 00:27:16,751 I did, but it did not make it past legal 500 00:27:19,320 --> 00:27:22,274 There is another one coming out, by the way, on, on device 501 00:27:22,278 --> 00:27:24,029 economics as well that, I think 502 00:27:24,456 --> 00:27:25,107 Frankcx: Definitely check 503 00:27:25,152 --> 00:27:26,480 Robert: our audience will enjoy 504 00:27:26,594 --> 00:27:28,021 Frankcx: blog that Robert put together. 505 00:27:28,042 --> 00:27:29,379 Super insightful. 506 00:27:30,412 --> 00:27:32,399 I want to turn the wheels a little bit. 507 00:27:32,449 --> 00:27:36,637 we talk about this idea of the return of the workstation, but I'm also 508 00:27:36,639 --> 00:27:40,300 reminded of, what Satya has put out there in the past in some of his blog 509 00:27:40,301 --> 00:27:43,177 posts around keeping the intelligence. 510 00:27:43,191 --> 00:27:48,450 You know, you shouldn't have to share your own intelligence with a cloud model. 511 00:27:48,474 --> 00:27:51,777 And I think, Scott, you made mention of these new models from Microsoft, 512 00:27:51,777 --> 00:27:58,249 like Microsoft AI, and what Mustafa out there is doing with his big giant 513 00:27:58,259 --> 00:28:00,246 team down in the Silicon Valley area. 514 00:28:00,532 --> 00:28:01,882 beautiful office, by the way. 515 00:28:01,883 --> 00:28:04,401 I got to visit once or twice, and, pretty amazing. 516 00:28:04,824 --> 00:28:08,471 But, by the way, they offer free food down there, which they don't give 517 00:28:08,510 --> 00:28:12,301 on the Microsoft campus, so I guess that's a great in- incentive as to 518 00:28:12,451 --> 00:28:15,304 Scott Slater: I think they grow grapes on the rooftop also 519 00:28:16,884 --> 00:28:17,850 Frankcx: Of course they do. 520 00:28:18,506 --> 00:28:21,806 but one of the things that really stands out to me is that new model 521 00:28:21,806 --> 00:28:25,525 is now available in Azure as a model that you could actually load 522 00:28:25,525 --> 00:28:30,599 yourself on your expensive Azure infrastructure, but it's all yours, 523 00:28:30,663 --> 00:28:32,506 meaning, like, it's not trained. 524 00:28:32,518 --> 00:28:35,185 It doesn't learn from, the things that you share from it. 525 00:28:35,198 --> 00:28:37,344 It's private into that space. 526 00:28:37,344 --> 00:28:41,057 And I wonder, like, that seems to be another reason to me, 527 00:28:41,069 --> 00:28:44,642 like keeping the workstation and keeping the data, into that space. 528 00:28:44,659 --> 00:28:48,025 And I'm wondering even Jacob or Neil or Robert or even Scott, like, 529 00:28:48,025 --> 00:28:52,271 as you're talking to customers, this idea of IP protection. 530 00:28:52,281 --> 00:28:57,134 Like, I can't share that information, with anybody outside of my own space. 531 00:28:57,152 --> 00:29:01,350 And I also talk about this idea of this second wave of AI, like these 532 00:29:01,360 --> 00:29:04,894 customers that just couldn't take part in the first wave because they 533 00:29:04,894 --> 00:29:08,916 didn't want to share with Sam Altman or Dario Amodei, you know, all of their 534 00:29:08,916 --> 00:29:10,876 internal, you know, trade secrets. 535 00:29:11,152 --> 00:29:15,645 But now with these new devices and, you know, Microsoft artificial intelligence 536 00:29:15,691 --> 00:29:20,662 models or having workstations that the data stays on that device, if we're 537

00:29:20,662 --> 00:29:24,160 starting to see some customers start to think about that into that space. 538 00:29:25,785 --> 00:29:28,570 Scott Slater: Yeah, I'll… Well, I, I could start giving you just 539 00:29:28,571 --> 00:29:30,074 a quick little thought on it. 540 00:29:30,182 --> 00:29:33,755 I think it's important for us to talk with customers about these new 541 00:29:33,764 --> 00:29:40,132 local models that exist and how large models, some of them are being shrunk, 542 00:29:40,513 --> 00:29:46,687 to provide just as good capability locally than on the cloud, right? 543 00:29:47,022 --> 00:29:51,572 And, a lot of these models you'll see have this word flash at the 544 00:29:51,580 --> 00:29:53,110 end of the name of the model. 545 00:29:53,366 --> 00:29:54,135 That's mean 546 00:29:54,152 --> 00:29:57,694 that means that it's fast, it's lightweight, it's built for 547 00:29:57,696 --> 00:30:01,616 speed and cost, and It's not intended to be the biggest model. 548 00:30:02,031 --> 00:30:05,610 And, I like to, for customers that don't understand this, 549 00:30:05,630 --> 00:30:07,181 to give them some analogies. 550 00:30:07,185 --> 00:30:11,150 So one that everybody understands is, like, an image, right? 551 00:30:11,254 --> 00:30:17,279 a full-resolution photo versus, like, a compressed JPEG, right? 552 00:30:17,308 --> 00:30:19,221 And I think most people understand that. 553 00:30:19,221 --> 00:30:23,451 And for most people, you look at that photo, you won't know the difference. 554 00:30:23,476 --> 00:30:28,637 But if you zoom in on the compressed one, it may be a little more blurry, right? 555 00:30:28,639 --> 00:30:31,511 But for most people, the compressed photo is good enough. 556 00:30:31,910 --> 00:30:38,185 Or in music, it's like a master studio file versus, a compressed 557 00:30:38,191 --> 00:30:40,258 file for streaming, right? 558 00:30:40,494 --> 00:30:43,283 For most people, you're not gonna notice the difference when you're 559 00:30:43,296 --> 00:30:46,510 driving in your car and you're listening to it in your car. 560 00:30:46,511 --> 00:30:50,484 But, maybe an audiophile audio engineer, they'll know the 561 00:30:50,485 --> 00:30:52,323 difference, but most people wouldn't. 562 00:30:52,721 --> 00:30:57,235 And I think that's, a nice way to compare, a cloud model that may have been 563 00:30:57,236 --> 00:31:00,169 shrunk to now be a local supported model 564 00:31:03,377 --> 00:31:05,900 Jacob: Yeah, I would just point out that there's concern sometimes from 565 00:31:05,900 --> 00:31:10,116 folks going, "Well, doesn't it lose some precision when you compress it?" And while 566 00:31:10,116 --> 00:31:15,251 yes, in theory it does, when you quantum model down, it does lose some precision. 567 00:31:15,261 --> 00:31:19,209 But nowadays, even with frontier models, all of the work that 568 00:31:19,209 --> 00:31:21,374 folks are doing should be checked. 569 00:31:21,767 --> 00:31:24,491

And honestly, it's often checked by another model. 570 00:31:24,725 --> 00:31:26,817 Like that's built into most workflows nowadays. 571 00:31:27,093 --> 00:31:30,590 And so you have one model that's shrunk down that's doing some great work, 572 00:31:30,590 --> 00:31:34,004 heavy lifting, and all of that work should be checked by another model. 573 00:31:34,169 --> 00:31:37,104 It might be another version of that same model or a different one 574 00:31:37,106 --> 00:31:38,734 that's tuned for different things. 575 00:31:39,043 --> 00:31:42,273 And so together, those two smaller models make an output that's 576 00:31:42,274 --> 00:31:45,356 better than even one larger model could do, which is kind of cool. 577 00:31:45,529 --> 00:31:46,086 Scott Slater: models? 578 00:31:47,373 --> 00:31:48,704 Okay. 579 00:31:51,707 --> 00:31:55,827 Frankcx: I think that's pretty common these days in writing applications. 580 00:31:55,941 --> 00:31:59,958 I'm writing an application right now for a partner company of ours that 581 00:31:59,959 --> 00:32:03,752 does meeting room applications, and I'm writing it with Codex, but I'm 582 00:32:03,760 --> 00:32:08,011 using Claude to do all of the check, so that they balance each other. 583 00:32:08,280 --> 00:32:12,920 And then even this last week, I noticed that, Google released Argon, and so 584 00:32:12,920 --> 00:32:16,902 I've been using Argon to be almost a third-party reviewer because there 585 00:32:16,902 --> 00:32:19,096 are blind spots that these AIs create. 586 00:32:19,109 --> 00:32:22,989 And so I asked Argon to go through And do a full scrub of the code base. 587 00:32:23,283 --> 00:32:25,737 And sure enough, it came up with a couple, blind spots that 588 00:32:25,737 --> 00:32:27,536 Claude and Codex, didn't see. 589 00:32:27,644 --> 00:32:31,665 so it's amazing to watch, how you can drive application development 590 00:32:31,877 --> 00:32:36,030 by trusting three agentic coding agents to drive that, so 591 00:32:36,728 --> 00:32:39,931 Jacob: And by the way, I've got three, different models running 592 00:32:39,932 --> 00:32:43,465 on one device next to me, an RTX Spark Surface Laptop Ultra. 593 00:32:43,865 --> 00:32:47,629 they're all about f- 20 to 40 billion parameter models all running at 594 00:32:47,630 --> 00:32:49,610 the same time in the same workflow. 595 00:32:49,636 --> 00:32:52,023 So th- like that, that is totally possible and 596 00:32:52,153 --> 00:32:52,728 Robert: by the way. 597 00:32:52,767 --> 00:32:53,686 Frankcx: Four, yeah, 598 00:32:53,836 --> 00:32:54,826 Robert: low-key brags 599 00:32:54,947 --> 00:32:55,264 Frankcx: yeah 600 00:32:55,540 --> 00:32:57,660 Jacob: Hey, we're here to talk about local AI, you 601 00:32:57,747 --> 00:32:58,659 Robert: ran my first, 602 00:32:58,750 --> 00:32:59,766 Jacob: as well, talk about what we're doing 603 00:33:00,331 --> 00:33:02,327 Robert: model, a 32 billion 604 00:33:02,369 --> 00:33:02,379 Frankcx: Oh 605 00:33:02,628 --> 00:33:04,167 Robert: MoE two-bit model. 606 00:33:04,552 --> 00:33:05,642 and I was Very impressed. 607 00:33:05,903 --> 00:33:10,082 Actually, Jacob and I did a rogue podcast, I think earlier this week, and 608 00:33:10,082 --> 00:33:12,243 we talked about models checking models 609 00:33:13,768 --> 00:33:14,992 Jacob: we did, yes. 610 00:33:15,165 --> 00:33:16,948 Scott Slater: of the new Transformers model? 611 00:33:18,448 --> 00:33:19,210 Robert: Ooh, do tell. 612 00:33:19,235 --> 00:33:19,451 Do 613 00:33:19,698 --> 00:33:23,311 Scott Slater: So that's my way of referring if you heard of NVIDIA's 614 00:33:23,335 --> 00:33:27,267 Nemotron model, it sounds like Megatron, so I'm just calling 615 00:33:27,268 --> 00:33:28,731 it the new Transformers model. 616 00:33:28,832 --> 00:33:30,799 I think it will be a fun way to talk about it 617 00:33:32,699 --> 00:33:32,821 Frankcx: It's 618 00:33:33,078 --> 00:33:36,121 Jacob: Yeah, I've been using Nemotron or Transformer, whatever 619 00:33:36,122 --> 00:33:38,651 you wanna call it, 3.5 Bolt. 620 00:33:38,987 --> 00:33:46,534 It's a 70 billion parameter MoE model, which fits into 28 gigs of VRAM, which 621 00:33:46,534 --> 00:33:50,687 is awesome because, you know, we've been talking about the $6,000 device. 622 00:33:50,711 --> 00:33:54,134 Like, you don't need the $6,000 version to run that model. 623 00:33:54,144 --> 00:33:58,715 You can actually run it on a $4,000 version, and still get great results. 624 00:33:58,725 --> 00:33:59,024 So 625 00:34:00,008 --> 00:34:00,456 Frankcx: All right. 626 00:34:00,507 --> 00:34:00,978 Jacob: it's been cool. 627 00:34:01,448 --> 00:34:03,202 Frankcx: Let's finish up the podcast this week. 628 00:34:03,204 --> 00:34:07,034 I know, Jacob, you gotta get going pretty soon, but I'd love to go around the horn 629 00:34:07,035 --> 00:34:08,665 and talk about what we're working on. 630 00:34:09,006 --> 00:34:12,307 many of you, you know, on the podcast, you're aware that I do a little bit of, 631 00:34:12,308 --> 00:34:15,786 Friday night and Saturday night Uber driving so I can figure out what the cool 632 00:34:15,786 --> 00:34:18,006 kids are doing in my retirement age here. 633 00:34:18,394 --> 00:34:22,109 but one of the things that I love to be able to do is that device that's literally 634 00:34:22,110 --> 00:34:27,067 sitting behind me, this workstation, it collects twenty-three API feeds. 635 00:34:27,094 --> 00:34:30,753 It looks at FlightAware to understand when planes are landing at SeaTac. 636 00:34:31,048 --> 00:34:34,596 It looks at when is my next bus, which is an interesting API that can 637 00:34:34,597 --> 00:34:38,427 tell you when the next bus or train is supposed to pull into a station. 638 00:34:38,854 --> 00:34:42,289 so I use it to understand, like, what are trains coming from Seattle 639 00:34:42,525 --> 00:34:44,283 and, like, did a concert let out? 640 00:34:44,283

--> 00:34:45,956 Did a Mariners game let out? 641 00:34:45,960 --> 00:34:47,438 Did a Seahawks game let out? 642 00:34:47,448 --> 00:34:50,417 And I can actually pick up people, you know, in that space. 643 00:34:50,635 --> 00:34:54,422 But it also looks at, there's an API that understands what is traffic. 644 00:34:54,422 --> 00:34:58,464 Like, you know how when you look at, like you're going to a restaurant, and Google 645 00:34:58,465 --> 00:35:02,846 can tell you that this is higher traffic period or it's a lower traffic period? 646 00:35:03,119 --> 00:35:07,217 Well, I can understand by subscribing to it what are all the traffic, spaces that 647 00:35:07,217 --> 00:35:11,033 I can understand where are all the people and what restaurants are they going to. 648 00:35:11,034 --> 00:35:16,339 And so it all runs on, my phone, and my phone can access my device that's 649 00:35:16,340 --> 00:35:20,783 behind me, but my device will look across all that data, and it uses a 650 00:35:20,789 --> 00:35:25,765 QUEN model to scrub all that data that understands where should I go next in 651 00:35:25,766 --> 00:35:29,547 order to see the highest propensity of where riders could show up. 652 00:35:29,898 --> 00:35:33,066 And so I've been using it, and it's actually starting to pay off a little bit. 653 00:35:33,110 --> 00:35:33,678 I don't know. 654 00:35:33,692 --> 00:35:37,056 It's hard to say whether it's, like, better or worse, but it's kinda cool 655 00:35:37,057 --> 00:35:40,725 to have that intelligence to be able to say, "Where should I be and where 656 00:35:40,726 --> 00:35:44,292 should I position myself to find the highest propensity of riders?" 657 00:35:44,306 --> 00:35:45,316 So I'm 658 00:35:45,508 --> 00:35:46,931 Jacob: You're gonna publish it and sell it, 659 00:35:47,037 --> 00:35:47,101 Frankcx: I 660 00:35:47,190 --> 00:35:47,693 Jacob: Frank? 661 00:35:47,990 --> 00:35:50,008 Frankcx: maybe it'll be an app that'll be out there. 662 00:35:50,016 --> 00:35:52,822 It would have to run on a device, whether that's in Azure or 663 00:35:52,822 --> 00:35:54,450 whether that's on my local device. 664 00:35:54,768 --> 00:35:59,466 But you could go out and buy a RTX Spark or a DGX Spark device and set up 665 00:35:59,473 --> 00:36:03,482 Tailscale into your own environment and be able to run it that way, where you're 666 00:36:03,490 --> 00:36:08,013 hosting it yourself rather than sharing it as a shared app across many people. 667 00:36:08,411 --> 00:36:12,153 But what I like to point out is, are there commercial instances where 668 00:36:12,419 --> 00:36:16,710 collecting a ton of API data and then asking for an instant in time 669 00:36:16,711 --> 00:36:20,537 in order to make a decision with an AI model, is that something that, 670 00:36:20,550 --> 00:36:24,717 Scott or Neil or Robert or even Jacob would be like, do you see customers 671 00:36:25,021 --> 00:36:27,083 that could utilize cases like that? 672 00:36:27,114 --> 00:36:28,854 I'm not trying to push it as my service. 673 00:36:28,870 --> 00:36:33,356 I'm more trying to say like, you know, instead of buying an RTX laptop, 674 00:36:33,394 --> 00:36:36,366 buying an RTX-based PC in the future. 675 00:36:36,659 --> 00:36:39,981 You know, do we see use cases like that of being able to have an always-on 676 00:36:39,982 --> 00:36:43,974 device that is collecting and monitoring, but you're using a more lower scale 677 00:36:43,975 --> 00:36:48,369 device in order to access that data, even though that's all local AI? 678 00:36:48,531 --> 00:36:52,498 So curious input, but also, hey, what's running on your device? 679 00:36:53,122 --> 00:36:55,570 Scott Slater: Frank, before we go there, you just got me thinking. 680 00:36:55,581 --> 00:36:56,736 You might be onto something. 681 00:36:56,745 --> 00:37:00,550 Were you talking about RTX Spark devices in cars? 682 00:37:00,887 --> 00:37:02,720 This may be a whole new market. 683 00:37:02,773 --> 00:37:08,905 Like, I think 16-year-old Scott wanted a car stereo system with two 10-inch 684 00:37:08,909 --> 00:37:13,687 subwoofers in my trunk in my car, but maybe now this is a new market where 685 00:37:13,688 --> 00:37:16,898 you want a RTX Spark device in the car. 686 00:37:17,177 --> 00:37:17,688 I don't know 687 00:37:17,952 --> 00:37:20,980 Frankcx: it makes you like a Waymo car, it has to rely on its own 688 00:37:20,988 --> 00:37:23,255 local AI in order to drive, right? 689 00:37:23,279 --> 00:37:23,731 It can't 690 00:37:23,842 --> 00:37:27,045 Jacob: All cars nowadays have GPUs in them. 691 00:37:27,418 --> 00:37:31,857 Not all of-- Like, I mean, Tesla famously a decade ago was like, "Hey, you know, 692 00:37:31,857 --> 00:37:35,961 you can plug in your Xbox to this because it's got a, it, it's got a, a 693 00:37:35,964 --> 00:37:40,656 GPU that's an Xbox level GPU at that time." But like all cars are taking this 694 00:37:40,670 --> 00:37:44,052 information in from all these inputs, especially if they have any self-driving 695 00:37:44,619 --> 00:37:47,044 or, or assisted driving capabilities. 696 00:37:47,099 --> 00:37:47,276 Neil: I'm, 697 00:37:47,480 --> 00:37:47,887 Frankcx: Ja 698 00:37:47,926 --> 00:37:48,697 Neil: some work with 699 00:37:48,977 --> 00:37:49,468 Jacob: It's me, thank you 700 00:37:49,552 --> 00:37:54,428 Neil: services firms, and this notion of does everyone need a $6,000 device? 701 00:37:55,070 --> 00:37:59,702 Probably not, but exploring the usage of RTX, DGX Spark, 702 00:37:59,971 --> 00:38:02,112 and scheduling specific jobs. 703 00:38:02,112 --> 00:38:04,010 So imagine like tax preparation. 704 00:38:04,488 --> 00:38:07,168 Johnny runs all his clients on Monday morning. 705 00:38:07,391 --> 00:38:11,550 Tuesday-- Suzy gets access to run jobs for her clients Tuesday night. 706 00:38:11,750 --> 00:38:14,771 I've been doing some testing with RTX Spark. 707 00:38:14,775 -->

00:38:19,255 I threw 64 jobs at it at the same time, and was able to reach a 708 00:38:19,297 --> 00:38:23,529 peak aggregate throughput of over 1,000 tokens per second, right? 709 00:38:23,534 --> 00:38:28,541 So pretty incredible as you think about this notion of a token factory. 710 00:38:28,857 --> 00:38:34,229 We've got token factories on desks, and for folks that may be using this 711 00:38:34,243 --> 00:38:39,145 once a week, once a month, that edge server appliance can meet those demands. 712 00:38:39,152 --> 00:38:41,102 So very bullish on this 713 00:38:42,496 --> 00:38:46,233 Frankcx: I know for the novices out there, like a token is like three quarters of a 714 00:38:46,459 --> 00:38:47,839 Robert: Yeah 715 00:38:48,577 --> 00:38:51,621 Frankcx: almost a thousand words per second. 716 00:38:52,366 --> 00:38:55,160 that's the type of power that these RTX Spark devices will bring. 717 00:38:55,268 --> 00:38:56,374 So pretty amazing. 718 00:38:56,550 --> 00:38:56,892 Scott Slater: And, 719 00:38:56,970 --> 00:38:59,226 Robert: changed our hybrid intelligence story now. 720 00:38:59,226 --> 00:39:01,126 It's not just device and cloud. 721 00:39:01,139 --> 00:39:05,882 Now there's an edge component, and customers are really rethinking 722 00:39:05,887 --> 00:39:08,169 how they go about this and where they're making their investments 723 00:39:09,285 --> 00:39:12,337 Scott Slater: I think those DGX stations are super interesting. 724 00:39:12,349 --> 00:39:15,696 Like, you know, where can those fit for companies? 725 00:39:15,696 --> 00:39:19,869 And I was having a conversation with a customer last week about it, and, the 726 00:39:19,869 --> 00:39:25,676 conversation got to maybe they could be needed besides for, an AI team or 727 00:39:25,678 --> 00:39:30,774 a developer team or a research lab or something like that, but where companies 728 00:39:30,774 --> 00:39:33,053 are constrained with capacity, right? 729 00:39:33,053 --> 00:39:37,600 Like, if they're building their own AI infrastructure and they have capacity 730 00:39:37,607 --> 00:39:42,447 constraints, or if they can't rent capacity from third parties, you could 731 00:39:42,457 --> 00:39:46,694 drop one of these in those scenarios and not slow down your business. 732 00:39:46,740 --> 00:39:50,480 I think it's gonna be really interesting for companies to have them 733 00:39:50,481 --> 00:39:54,344 available for those scenarios, ' cause we know capacity's not unlimited 734 00:39:55,517 --> 00:39:58,885 Frankcx: Okay, you're just getting me super fired up here because 735 00:39:59,543 --> 00:40:03,444 I'm gonna push it on back to you, Scott, of imagine a device, Scott, 736 00:40:03,449 --> 00:40:06,371 that your sellers could sell that costs a hundred thousand dollars. 737 00:40:06,835 --> 00:40:09,805 How many sellers do not wanna sell a device like that? 738 00:40:10,598 --> 00:40:13,971 the flip side would be is how many Surface laptops do you have to sell 739 00:40:13,997 --> 00:40:18,032 to make that type of number where you could be positioning a DGX 740 00:40:18,040 --> 00:40:19,942 station device at a hundred grand? 741 00:40:20,238 --> 00:40:26,304 But this device can run in a private environment, a frontier model that 742 00:40:26,304 --> 00:40:28,164 is just as capable of anything. 743 00:40:28,174 --> 00:40:32,329 Like, you could run that MIA or, you know, MAI 744 00:40:32,405 --> 00:40:32,981 Jacob: And may I? 745 00:40:33,270 --> 00:40:36,866 Frankcx: on that device, and I would say that that is probably 746 00:40:36,900 --> 00:40:39,297 Pavan and, S-Satya's, dream. 747 00:40:39,676 --> 00:40:44,044 But Scott, I'm gonna vote for you to build an incubation team within Windows that 748 00:40:44,044 --> 00:40:47,744 goes out and sells a hundred thousand dollar device 'cause I don't think there's 749 00:40:47,744 --> 00:40:51,453 a seller in the world that wouldn't love to get an order for a few of those 750 00:40:51,736 --> 00:40:52,766 Robert: Hey, sign me up. 751 00:40:52,766 --> 00:40:54,703 I'll, I'll sell the hell out of those. 752 00:40:54,972 --> 00:40:55,557 Scott Slater: There we go 753 00:40:55,615 --> 00:41:00,011 Robert: said before, the ROI for that, 100 grand one time opposed 754 00:41:00,011 --> 00:41:02,620 to every month for a branch? 755 00:41:02,749 --> 00:41:03,495 Frankcx: can make money. 756 00:41:03,533 --> 00:41:03,947 I mean, this 757 00:41:04,324 --> 00:41:05,078 Robert: Absolutely 758 00:41:05,342 --> 00:41:06,727 Frankcx: that just serves up data. 759 00:41:06,753 --> 00:41:07,950 It actually makes 760 00:41:08,259 --> 00:41:08,586 Robert: Yeah. 761 00:41:08,795 --> 00:41:09,367 Frankcx: in that way. 762 00:41:09,377 --> 00:41:09,553 It 763 00:41:09,614 --> 00:41:10,025 Robert: it's… 764 00:41:10,294 --> 00:41:11,229 Frankcx: factory of its 765 00:41:11,374 --> 00:41:12,112 Robert: There you go 766 00:41:12,561 --> 00:41:13,361 Scott Slater: Teñido Frank 767 00:41:13,805 --> 00:41:16,177 Jacob: I heard a rumor there'll be one in San Francisco next 768 00:41:16,177 --> 00:41:18,216 week at GitHub's Universe event. 769 00:41:18,274 --> 00:41:20,249 It's just a rumor, unconfirmed. 770 00:41:20,941 --> 00:41:21,983 San Francisco, if you're there 771 00:41:22,631 --> 00:41:28,049 Scott Slater: from a client seller, that $100,000 may seem like a lot. 772 00:41:28,472 --> 00:41:33,558 From an infrastructure seller, that $100,000 is not a lot, right? 773 00:41:33,638 --> 00:41:36,748 So anybody that sold infrastructure and data centers, 774 00:41:37,034 --> 00:41:37,271 Frankcx: Sure 775 00:41:37,719 --> 00:41:39,200 Scott Slater: that's a very reasonable amount 776 00:41:41,090 --> 00:41:41,961 Frankcx: All right. 777 00:41:42,280 --> 00:41:45,105 Well, I think, that kind of brings us as a good point. 778 00:41:45,105 --> 00:41:48,551 I think we're all excited for this DGX Station device, what it 779 00:41:48,553 --> 00:41:50,916 can offer in a Windows ecosystem. 780 00:41:50,947 --> 00:41:55,649 I think that Surface personally tees it up with Surface Laptop Ultra and, 781 00:41:55,649 --> 00:41:57,271 the dev boxes that are out there. 782 00:41:57,272 --> 00:42:02,372 But I think as customers look at how they can climb into these spaces and be 783 00:42:02,372 --> 00:42:06,303 able to move into infrastructure like that, and even colo facilities that 784 00:42:06,305 --> 00:42:11,175 are housing, NVIDIA and CUDA-based, architecture, I think this is a good 785 00:42:11,176 --> 00:42:12,937 time, to be able to position that. 786 00:42:13,738 --> 00:42:17,247 any closing words before we close out this week's pod? 787 00:42:18,830 --> 00:42:20,033 Robert: Great to have Scott on. 788 00:42:20,034 --> 00:42:20,631 Well done. 789 00:42:20,641 --> 00:42:21,026 Thank you, 790 00:42:21,076 --> 00:42:21,835 Neil: Thank you, Scott. 791 00:42:21,893 --> 00:42:23,029 Bringing in the heavy hitters. 792 00:42:23,479 --> 00:42:24,800 Scott Slater: thank you for having me. 793 00:42:24,828 --> 00:42:26,256 And, just one final question. 794 00:42:26,256 --> 00:42:29,701 Is Neil IM gonna take us out with some outro music? 795 00:42:31,259 --> 00:42:31,800 Frankcx: We will. 796 00:42:31,814 --> 00:42:36,837 We, Neil, I c- I will, send me some DJ, you know, exit music, and we'll put 797 00:42:36,837 --> 00:42:38,800 that on the, this week's pod as our exit 798 00:42:39,160 --> 00:42:39,914 Neil: Sounds great 799 00:42:39,963 --> 00:42:43,862 Robert: know that at one point at a Surface Airlift, Neil actually 800 00:42:43,979 --> 00:42:46,104 was my live DJ during a session? 801 00:42:46,969 --> 00:42:47,698 Neil: Absolutely. 802 00:42:47,704 --> 00:42:49,278 Helped launch my DJ career. 803 00:42:49,296 --> 00:42:51,780 Always, always remember and appreciate that moment. 804 00:42:51,841 --> 00:42:55,555 and if I recall, I actually, provided walk-up music for you as well, Frank. 805 00:42:55,600 --> 00:42:57,990 I think it was Fool's Gold was the track, if I recall. 806 00:42:58,030 --> 00:42:58,301 Frankcx: Yeah, 807 00:42:58,658 --> 00:42:59,018 Neil: Yep 808 00:42:59,273 --> 00:43:02,590 Frankcx: and I think maybe that's the music unless we get sued, that maybe we 809 00:43:02,788 --> 00:43:03,473 Neil: There we go. 810 00:43:05,600 --> 00:43:06,326 Jacob: Aaron Carter? 811 00:43:07,709 --> 00:43:08,241 Frankcx: All right. 812 00:43:08,277 --> 00:43:11,239 Well, thanks everyone, and, we'll see you again on the 813 00:43:11,239 --> 00:43:13,238 next edition of The Local Host 814 00:43:14,135 --> 00:43:14,623 Neil: Thank you 815 00:43:14,885 --> 00:43:15,475 Jacob: Thanks everyone 816 00:43:18,432 --> 00:43:19,948 Frankcx: There's no place like home.