1 00:00:03,600 --> 00:00:05,839 Welcome to episode 403 2 00:00:05,839 --> 00:00:08,419 of the Microsoft Cloud IT Pro podcast 3 00:00:08,720 --> 00:00:11,779 recorded live 05/30/2025. 4 00:00:12,080 --> 00:00:14,375 This is a show about Microsoft three sixty 5 00:00:14,375 --> 00:00:16,614 five and Azure from the perspective of IT 6 00:00:16,614 --> 00:00:18,774 pros and end users, where we discuss a 7 00:00:18,774 --> 00:00:21,094 topic or recent news and how it relates 8 00:00:21,094 --> 00:00:24,074 to you. Microsoft build has come and gone. 9 00:00:24,214 --> 00:00:26,134 So today, we dive into some of the 10 00:00:26,134 --> 00:00:27,420 announcements from the conference. 11 00:00:27,899 --> 00:00:30,320 We'll catch you up on announcements around Microsoft 12 00:00:30,460 --> 00:00:32,640 three sixty five Copilot tuning, 13 00:00:33,020 --> 00:00:36,460 securing Copilot agents and enter ID, and a 14 00:00:36,460 --> 00:00:39,739 can't miss build session with Mark Russinovich and 15 00:00:39,739 --> 00:00:42,079 more. So let's dive into the show. 16 00:00:44,265 --> 00:00:46,104 It's been a week, Scott, for both of 17 00:00:46,104 --> 00:00:48,585 us. Time marches on. I I have not 18 00:00:48,585 --> 00:00:51,225 told you about my latest adventure. Your latest 19 00:00:51,225 --> 00:00:53,304 adventure? Not your latest toy, but your latest 20 00:00:53,304 --> 00:00:56,159 adventure? My latest adventure. Scott story time. Scott 21 00:00:56,159 --> 00:00:58,179 story time. So my latest adventure was 22 00:00:58,560 --> 00:01:01,760 my oldest graduated high school Congratulations. In the 23 00:01:01,760 --> 00:01:03,840 last couple weeks. Yeah. Thanks. Accomplishment. We will 24 00:01:03,840 --> 00:01:04,959 see. We need to get him out of 25 00:01:04,959 --> 00:01:06,640 the house still. As part of that, we 26 00:01:06,640 --> 00:01:08,560 threw him a graduation party. So we had 27 00:01:08,560 --> 00:01:10,665 people over to the house, and Yep. It's 28 00:01:10,665 --> 00:01:12,265 hot outside, but you can still hang out 29 00:01:12,265 --> 00:01:13,724 outside if you're in the shade. 30 00:01:14,025 --> 00:01:15,325 So we bought these 31 00:01:15,784 --> 00:01:18,265 pop up tents, like, 12 by 12 tents 32 00:01:18,265 --> 00:01:19,865 without sides. They just had a roof on 33 00:01:19,865 --> 00:01:21,865 them. Right. Like, the whole canopy thing, yeah, 34 00:01:21,865 --> 00:01:23,465 you see them at all the sporting event. 35 00:01:23,465 --> 00:01:25,700 Yeah. All that. Simple basic. We we we 36 00:01:25,700 --> 00:01:28,599 found some at a a bargain bin store. 37 00:01:28,979 --> 00:01:29,479 Like, 38 00:01:29,859 --> 00:01:31,939 there's a store in Jacksonville where this stuff 39 00:01:31,939 --> 00:01:33,700 just comes in on crates, and then they 40 00:01:33,700 --> 00:01:35,700 just mark it down to clear it all 41 00:01:35,700 --> 00:01:38,179 out. So it was whatever. They're they're normally 42 00:01:38,179 --> 00:01:39,700 a hundred $50, 43 00:01:39,700 --> 00:01:42,965 and so they were on sale for $60.70. 44 00:01:42,965 --> 00:01:44,165 I don't know. My wife went and grabbed 45 00:01:44,165 --> 00:01:45,365 a couple of them. I will say this 46 00:01:45,365 --> 00:01:47,125 is one thing I like about Florida because 47 00:01:47,125 --> 00:01:49,125 nobody has basements, nobody has places to store 48 00:01:49,125 --> 00:01:50,885 anything, so everybody gets rid of stuff if 49 00:01:50,885 --> 00:01:52,325 they don't need it. And I feel like 50 00:01:52,325 --> 00:01:53,844 because of that, you can find good deals 51 00:01:53,844 --> 00:01:55,525 and cheap stuff. Well, this was even brand 52 00:01:55,525 --> 00:01:57,420 new because this is, like, a wholesaler. 53 00:01:57,799 --> 00:02:00,840 So they're buying what's kinda like Amazon return 54 00:02:00,840 --> 00:02:02,939 bin kinda thing. They're buying a big 55 00:02:03,319 --> 00:02:05,079 40 foot shipping crate. They don't know what's 56 00:02:05,079 --> 00:02:06,439 in it, and then they're just bringing it 57 00:02:06,439 --> 00:02:07,960 into their store, and they're trying to clear 58 00:02:07,960 --> 00:02:09,625 it out and sell it. So so they 59 00:02:09,625 --> 00:02:11,944 must have had a pallet of these These 60 00:02:11,944 --> 00:02:14,264 canopies. Ready to go. Yeah. These canopies. So 61 00:02:14,264 --> 00:02:16,104 we bought a couple of them and stood 62 00:02:16,104 --> 00:02:17,705 them up for the day of the party, 63 00:02:17,705 --> 00:02:19,164 and it ended up raining. 64 00:02:19,625 --> 00:02:20,525 And so 65 00:02:20,824 --> 00:02:22,664 we ended up inside for a little bit, 66 00:02:22,664 --> 00:02:24,500 in, out, in, out. But I left them 67 00:02:24,500 --> 00:02:26,340 up after the party. So this was last 68 00:02:26,340 --> 00:02:28,500 week, and then we went through this week, 69 00:02:28,500 --> 00:02:29,000 and 70 00:02:29,780 --> 00:02:31,959 we've had a couple good rainstorms 71 00:02:32,259 --> 00:02:34,500 and squalls and things like that. So they're 72 00:02:34,500 --> 00:02:36,099 still up? You left them up? I did. 73 00:02:36,099 --> 00:02:37,754 I left them up. This wasn't, you know, 74 00:02:37,754 --> 00:02:40,314 hurricane stuff. I wasn't scared of something, like, 75 00:02:40,314 --> 00:02:42,715 blowing away and landing on a neighbor's house, 76 00:02:42,715 --> 00:02:45,754 things like that. So day one, they went 77 00:02:45,754 --> 00:02:48,155 up, party. It rained. It was it was 78 00:02:48,155 --> 00:02:50,074 fine. Like, all the water shed off the 79 00:02:50,074 --> 00:02:52,569 roof. Like, it was taut enough and and 80 00:02:52,569 --> 00:02:53,629 tight enough that 81 00:02:53,930 --> 00:02:54,990 all that stuff happened. 82 00:02:55,290 --> 00:02:58,090 At some point, it must have just loosened 83 00:02:58,090 --> 00:02:59,450 up or some of the some of the 84 00:02:59,450 --> 00:03:01,530 Velcro must have slipped or something that was 85 00:03:01,530 --> 00:03:03,150 holding that canopy onto 86 00:03:04,074 --> 00:03:05,834 onto the legs. So I went out there, 87 00:03:06,394 --> 00:03:07,854 one day after one of the rainstorms, 88 00:03:08,474 --> 00:03:10,814 and one of them just had this big 89 00:03:10,875 --> 00:03:12,794 bubble in a corner where it was all 90 00:03:12,794 --> 00:03:13,854 just leaning down. 91 00:03:14,155 --> 00:03:16,474 And we've got the lawn furniture and stuff 92 00:03:16,474 --> 00:03:18,669 underneath it, like, the patio furniture, and it 93 00:03:18,669 --> 00:03:20,909 was leaning down about to my head. So 94 00:03:20,909 --> 00:03:22,349 this is maybe a tent that's about eight 95 00:03:22,349 --> 00:03:23,409 feet high, and 96 00:03:23,710 --> 00:03:26,349 I'm I'm five seven. So it's coming down, 97 00:03:26,349 --> 00:03:28,349 like, two, two and a half feet and 98 00:03:28,349 --> 00:03:30,110 just about touching my head with this big 99 00:03:30,110 --> 00:03:31,710 bubble. And I'm like, oh, that's no good. 100 00:03:31,710 --> 00:03:33,284 So I'm in there and pushing it up 101 00:03:33,284 --> 00:03:34,885 and trying to get the water off it. 102 00:03:34,885 --> 00:03:36,085 And I take a look, and I'm like, 103 00:03:36,085 --> 00:03:38,325 oh, it came loose on on the side 104 00:03:38,325 --> 00:03:41,045 because the Velcro slipped. So I kinda, like, 105 00:03:41,045 --> 00:03:43,925 redid the Velcro. I went and pulled it 106 00:03:43,925 --> 00:03:45,685 all back down and tightened it back up. 107 00:03:45,685 --> 00:03:48,539 Fine. Great. Two more days go by. Rain. 108 00:03:48,539 --> 00:03:49,599 Rain. Okay. 109 00:03:50,139 --> 00:03:50,639 Yesterday, 110 00:03:50,939 --> 00:03:54,219 we had some massive kinda squalls come through, 111 00:03:54,219 --> 00:03:57,180 like sideways rain. I saw that. It was 112 00:03:57,180 --> 00:03:59,259 a little nutty. It it came down for 113 00:03:59,259 --> 00:03:59,759 basically 114 00:04:00,060 --> 00:04:01,840 two hours straight sideways. 115 00:04:02,594 --> 00:04:04,935 I was watching a little river form between 116 00:04:05,235 --> 00:04:07,875 my house and and my neighbor's house because 117 00:04:07,875 --> 00:04:09,634 there's kind of a natural divot there, and 118 00:04:09,634 --> 00:04:11,794 and that leads down to the preserve behind 119 00:04:11,794 --> 00:04:12,455 my house. 120 00:04:12,835 --> 00:04:15,074 And so I'm I'm looking. I'm looking. So 121 00:04:15,074 --> 00:04:16,730 I go outside, and I just I just 122 00:04:16,730 --> 00:04:18,089 open the door. I'm not I'm not going 123 00:04:18,089 --> 00:04:19,449 out in this weather, and I peek my 124 00:04:19,449 --> 00:04:21,769 head over, and there's not just one bubble 125 00:04:21,769 --> 00:04:24,009 coming down. Now there's two bubbles, and the 126 00:04:24,009 --> 00:04:26,250 whole thing is, like, leaning, leaning, leaning, leaning. 127 00:04:26,250 --> 00:04:27,930 And it's coming down, and it's almost touching 128 00:04:27,930 --> 00:04:30,189 the patio furniture. Oh, man. Patio furniture's 129 00:04:30,569 --> 00:04:33,254 only, what, three feet maybe? Right. Considering, like, 130 00:04:33,254 --> 00:04:34,454 the top of the back of the chair, 131 00:04:34,454 --> 00:04:36,454 three, four feet. So it's coming, like, just 132 00:04:36,454 --> 00:04:38,055 down, down, down. I'm, like, I'm telling my 133 00:04:38,055 --> 00:04:39,894 wife, I said, this might be it. Like, 134 00:04:39,894 --> 00:04:42,074 it it it's it's it's probably gonna go. 135 00:04:42,294 --> 00:04:44,535 Absolutely did. So it got so heavy with 136 00:04:44,535 --> 00:04:45,194 the rain 137 00:04:45,610 --> 00:04:47,769 on with those two bubbles on the one 138 00:04:47,769 --> 00:04:49,550 side, it actually tore 139 00:04:49,930 --> 00:04:52,649 the the entire top. Like, it just sheared 140 00:04:52,649 --> 00:04:54,810 itself, like, right off. And then when it 141 00:04:54,810 --> 00:04:57,370 sheared itself with the weight, it also bent 142 00:04:57,370 --> 00:04:59,050 all all the metal and things because this 143 00:04:59,050 --> 00:05:01,425 is like a big accordion. So now this 144 00:05:01,425 --> 00:05:03,904 thing will not go back together. It's probably 145 00:05:03,904 --> 00:05:05,985 the best $70 I've spent for, like, a 146 00:05:05,985 --> 00:05:07,585 week of just being able to sit outside 147 00:05:07,585 --> 00:05:08,865 and things like that. I was kinda telling 148 00:05:08,865 --> 00:05:10,384 my wife. It's like, you know what? Worth 149 00:05:10,384 --> 00:05:11,665 it. Maybe I should just go back and 150 00:05:11,665 --> 00:05:13,285 see if we can buy a couple more. 151 00:05:13,425 --> 00:05:15,500 Like, we'll we'll just keep them on the 152 00:05:15,500 --> 00:05:17,519 side of the house in in stock and 153 00:05:17,660 --> 00:05:19,180 and and ready to go for the next 154 00:05:19,180 --> 00:05:21,100 one. So that that that's my next adventure 155 00:05:21,100 --> 00:05:23,180 is after I'm off this call, I'm gonna 156 00:05:23,180 --> 00:05:25,019 go grab a hacksaw, and I'm gonna go 157 00:05:25,019 --> 00:05:26,539 out in the backyard. And now I gotta 158 00:05:26,539 --> 00:05:29,745 cut all these metal spider legs off this 159 00:05:29,824 --> 00:05:32,464 12 by 12 monstrosity thing because there's no 160 00:05:32,464 --> 00:05:34,785 way it's going back together with how bent 161 00:05:34,785 --> 00:05:37,345 it is and everything else. And then because 162 00:05:37,345 --> 00:05:39,264 it sat there over my patio furniture so 163 00:05:39,264 --> 00:05:41,504 normally in my patio furniture, it sits outside 164 00:05:41,504 --> 00:05:43,504 and we leave the cushions on sometimes, and 165 00:05:43,504 --> 00:05:46,129 it doesn't get soaked soaked. It's got, like, 166 00:05:46,129 --> 00:05:47,889 Scotchgard or whatever on it. The water all 167 00:05:47,889 --> 00:05:49,889 runs off. Uh-huh. Because this sat on it 168 00:05:49,889 --> 00:05:52,290 for so long, it soaked through the material 169 00:05:52,290 --> 00:05:53,729 on the tent, and it soaked into all 170 00:05:53,729 --> 00:05:55,569 the patio cushions. So when I picked the 171 00:05:55,569 --> 00:05:57,169 patio cushions up and I just turned them 172 00:05:57,169 --> 00:05:59,605 from horizontal to vertical, they've just been leaking 173 00:05:59,605 --> 00:06:01,845 out for the past hour. Like, yep. Still 174 00:06:01,845 --> 00:06:03,365 watching them drip. I will not lie. I 175 00:06:03,365 --> 00:06:04,805 thought for sure one of those tents was 176 00:06:04,805 --> 00:06:06,884 gonna, like, take off and blow away. I 177 00:06:06,884 --> 00:06:08,725 thought that's what was gonna happen. Because when 178 00:06:08,725 --> 00:06:11,360 it starts raining sideways, I've seen I think 179 00:06:11,360 --> 00:06:12,879 it was actually a hurricane. I've seen a 180 00:06:12,879 --> 00:06:16,560 trampoline, like, blow from one house into the 181 00:06:16,560 --> 00:06:18,319 neighbor's yard in some of the wind that 182 00:06:18,319 --> 00:06:20,080 we've had. It had a great, like, vent 183 00:06:20,080 --> 00:06:22,560 on the top that Okay. Would crosscut, so 184 00:06:22,560 --> 00:06:24,480 it really wasn't gonna take off or go 185 00:06:24,480 --> 00:06:25,774 up and down. And we had it all 186 00:06:25,774 --> 00:06:27,935 tied down and everything. I I didn't use, 187 00:06:27,935 --> 00:06:30,495 like, all the hurricane ties, but it it 188 00:06:30,495 --> 00:06:32,995 was definitely, like, attached to the house and 189 00:06:33,055 --> 00:06:35,214 and where it needed to be. So fun 190 00:06:35,214 --> 00:06:37,790 fun little adventure for the week. Wow. Way 191 00:06:37,790 --> 00:06:39,389 more adventurous than my week. I don't think 192 00:06:39,389 --> 00:06:41,470 I have any fun adventure stories. So between 193 00:06:41,470 --> 00:06:43,470 that and and my new terminal toy, which 194 00:06:43,470 --> 00:06:45,310 we'll we'll talk about later. We'll we'll do 195 00:06:45,310 --> 00:06:47,329 a toy episode and and annoy folks 196 00:06:47,629 --> 00:06:50,670 sometime in the future. But, anyway, that's my 197 00:06:50,670 --> 00:06:52,524 news for the week. Why don't we get 198 00:06:52,524 --> 00:06:55,724 into the news from the last couple weeks 199 00:06:55,724 --> 00:06:57,425 because we've had events like build 200 00:06:57,805 --> 00:06:59,884 and all sorts of other stuff. Yeah. And 201 00:06:59,884 --> 00:07:03,345 some of those events, they're definitely developer heavy. 202 00:07:03,404 --> 00:07:06,384 Like, it's all different things. But there's 203 00:07:06,720 --> 00:07:08,560 always news, I feel like, that relates to 204 00:07:08,560 --> 00:07:10,960 IT pros because us as IT pros have 205 00:07:10,960 --> 00:07:13,040 to manage our developers and keep them in 206 00:07:13,040 --> 00:07:15,360 line and make sure they're behaving. And as 207 00:07:15,360 --> 00:07:18,615 a result, we get different things, and let's 208 00:07:18,694 --> 00:07:21,175 face it, there's Copilot stuff everywhere, so there's 209 00:07:21,175 --> 00:07:23,175 always Copilot news to talk about as well. 210 00:07:23,175 --> 00:07:24,535 And we haven't done much of a news 211 00:07:24,535 --> 00:07:27,014 episode, so some of these I think everything 212 00:07:27,014 --> 00:07:28,535 we kind of put together was the last 213 00:07:28,535 --> 00:07:29,814 couple weeks. Some of it may have been 214 00:07:29,814 --> 00:07:32,074 a little before build, all around the build 215 00:07:32,379 --> 00:07:34,779 time frame. We really should do these more 216 00:07:34,779 --> 00:07:36,300 often. Although, I don't feel like there's been 217 00:07:36,300 --> 00:07:38,220 a ton of news either, or I've just 218 00:07:38,220 --> 00:07:40,220 missed it. There's been quite a bit. I 219 00:07:40,220 --> 00:07:42,540 think it depends It's kinda what you're looking 220 00:07:42,540 --> 00:07:44,779 for. Where you sit and where you play 221 00:07:44,779 --> 00:07:46,860 around. Yeah. Yeah. There's been a bunch of 222 00:07:46,860 --> 00:07:48,615 new stuff come out across 223 00:07:49,235 --> 00:07:50,615 M 365. 224 00:07:51,074 --> 00:07:52,855 I was looking the other day 225 00:07:53,314 --> 00:07:54,615 at some of the 226 00:07:54,995 --> 00:07:57,654 latest announcements that have come out for 227 00:07:58,115 --> 00:08:00,514 Azure Kubernetes service and and some of the 228 00:08:00,514 --> 00:08:02,455 things going on in AKS land. 229 00:08:02,939 --> 00:08:05,360 Tons going on over there. 230 00:08:05,900 --> 00:08:09,199 The the trick is kinda sifting through the 231 00:08:09,580 --> 00:08:12,939 AI, AI, AI to what's what's the what's 232 00:08:12,939 --> 00:08:15,100 the real thing, or if AI was your 233 00:08:15,100 --> 00:08:18,055 thing, then, hey, maybe there was something real 234 00:08:18,055 --> 00:08:20,294 for you there too. Like, I definitely saw 235 00:08:20,294 --> 00:08:21,354 some stuff with Copilot 236 00:08:22,055 --> 00:08:24,454 that piqued my interest, and and maybe we 237 00:08:24,454 --> 00:08:25,754 could start there. So 238 00:08:26,375 --> 00:08:28,375 last week, we we talked to Do you 239 00:08:28,375 --> 00:08:30,430 wanna start with that one? Yeah. So so 240 00:08:30,430 --> 00:08:32,210 last week, we chatted with AEC 241 00:08:32,590 --> 00:08:34,669 about declarative agents, and that was kind of 242 00:08:34,669 --> 00:08:36,110 a follow-up to some stuff that we had 243 00:08:36,110 --> 00:08:39,870 talked about previously with Copilot Studio and doing 244 00:08:39,870 --> 00:08:40,529 kind of 245 00:08:40,990 --> 00:08:43,649 well, for you and I, no code declarative 246 00:08:43,710 --> 00:08:45,649 agents, just in Copilot Studio, 247 00:08:46,004 --> 00:08:48,485 next, next, next in context of an m 248 00:08:48,485 --> 00:08:49,285 $3.65 249 00:08:49,285 --> 00:08:50,745 subscription, things like that. 250 00:08:51,125 --> 00:08:52,024 They're introducing 251 00:08:52,804 --> 00:08:54,585 fine tuning for Copilot, 252 00:08:55,285 --> 00:08:57,684 which kinda takes this a whole another level 253 00:08:57,684 --> 00:08:59,929 because some of the at at least from 254 00:08:59,929 --> 00:09:01,610 my perspective and the things that I do 255 00:09:01,610 --> 00:09:03,370 in my day to day, because what I 256 00:09:03,370 --> 00:09:04,029 do today 257 00:09:04,490 --> 00:09:07,370 in something like Copilot Studio with declarative agents 258 00:09:07,370 --> 00:09:10,730 is I'm relying on RAG and and retrieval 259 00:09:10,730 --> 00:09:11,789 augmented generation. 260 00:09:12,375 --> 00:09:15,335 And this is really being able to take 261 00:09:15,335 --> 00:09:17,174 a model that already exists and the weights 262 00:09:17,174 --> 00:09:18,855 in that model that are already deployed out 263 00:09:18,855 --> 00:09:21,415 there and exist, and then fine tune your 264 00:09:21,415 --> 00:09:23,355 own model on top of it 265 00:09:23,654 --> 00:09:25,894 or fine tune and refine that model so 266 00:09:25,894 --> 00:09:27,274 that you can drive 267 00:09:27,809 --> 00:09:29,429 specific business processes. 268 00:09:30,850 --> 00:09:33,330 And the demo I saw around this really 269 00:09:33,330 --> 00:09:35,730 got my head turning. So I'm a product 270 00:09:35,730 --> 00:09:36,230 manager. 271 00:09:36,769 --> 00:09:38,710 I live in a land where 272 00:09:39,250 --> 00:09:40,789 I'm a I'm a remote employee. 273 00:09:41,294 --> 00:09:44,514 Communication and particularly written communication is super important. 274 00:09:44,815 --> 00:09:46,575 So we spend a lot of time on 275 00:09:46,575 --> 00:09:48,674 product requirements documents, PRDs, 276 00:09:49,054 --> 00:09:49,554 justifications, 277 00:09:50,174 --> 00:09:50,674 specifications, 278 00:09:51,375 --> 00:09:51,875 designs, 279 00:09:52,254 --> 00:09:53,934 and we have formal templates for all this 280 00:09:53,934 --> 00:09:56,095 stuff. One of my big problems with it 281 00:09:56,095 --> 00:09:56,595 is 282 00:09:57,190 --> 00:09:59,110 people who take the template and they massage 283 00:09:59,110 --> 00:10:01,029 it into their own thing or they don't 284 00:10:01,029 --> 00:10:02,950 know what a good one is. So they 285 00:10:02,950 --> 00:10:04,710 they kinda look at the template that we 286 00:10:04,710 --> 00:10:07,029 have today, and and they just get like 287 00:10:07,029 --> 00:10:09,110 a deer in headlights and and scared about 288 00:10:09,110 --> 00:10:09,610 it. 289 00:10:09,945 --> 00:10:12,605 So I like, with this, I could take 290 00:10:12,825 --> 00:10:13,725 all our existing 291 00:10:14,105 --> 00:10:16,585 PRDs that we've done for, like, at least 292 00:10:16,585 --> 00:10:18,264 in my time for the past five years 293 00:10:18,264 --> 00:10:19,085 on my team. 294 00:10:19,465 --> 00:10:20,445 I could train 295 00:10:20,904 --> 00:10:22,285 and fine tune 296 00:10:22,665 --> 00:10:24,924 the existing model on my PRDs, 297 00:10:25,580 --> 00:10:28,540 and I could actually have it tune on 298 00:10:28,540 --> 00:10:30,559 top of my templates as well. 299 00:10:31,019 --> 00:10:33,919 So then I could make it super turnkey 300 00:10:34,059 --> 00:10:36,320 to go in and turn these things out. 301 00:10:36,700 --> 00:10:39,524 And that's a very simple scenario for me. 302 00:10:39,524 --> 00:10:42,105 I saw a great demo on the Microsoft 303 00:10:42,165 --> 00:10:44,565 Mechanics YouTube channel, and I'll I'll put a 304 00:10:44,565 --> 00:10:46,004 link in the show notes, and I recommend 305 00:10:46,004 --> 00:10:47,764 folks go out and watch that one. That 306 00:10:47,764 --> 00:10:49,785 was generating legal documents, 307 00:10:50,165 --> 00:10:52,165 and it actually did a pretty good job. 308 00:10:52,165 --> 00:10:53,845 Like, I'm not a lawyer, all all that 309 00:10:53,845 --> 00:10:56,460 jazz. Yep. It's just to couch all that 310 00:10:56,460 --> 00:10:57,920 and and have that caveat. 311 00:10:58,379 --> 00:11:00,560 But that said, I I think that one's 312 00:11:00,620 --> 00:11:02,860 really cool and and really exciting, and it's 313 00:11:02,860 --> 00:11:05,259 something I'll have access to in my tenant 314 00:11:05,259 --> 00:11:07,580 and totally intend to go out and play 315 00:11:07,580 --> 00:11:09,194 around with. So I don't know if you 316 00:11:09,194 --> 00:11:11,115 had a chance to either play with this 317 00:11:11,115 --> 00:11:13,855 in your tenant or watch any of the 318 00:11:14,074 --> 00:11:16,315 watch any of the introduction videos or anything 319 00:11:16,315 --> 00:11:18,634 like that that's out there, but we would 320 00:11:18,634 --> 00:11:20,315 would love to get your thoughts on this 321 00:11:20,315 --> 00:11:22,720 one as well. I did watch the Microsoft 322 00:11:22,720 --> 00:11:24,639 Mechanics video while I was doing other stuff, 323 00:11:24,639 --> 00:11:26,720 so I was kinda half watching it. This 324 00:11:26,720 --> 00:11:28,559 is a really cool feature. I like how 325 00:11:28,559 --> 00:11:30,559 you can fine tune it that way. I 326 00:11:30,559 --> 00:11:32,500 would say the part that I was interested 327 00:11:32,559 --> 00:11:34,660 in, kinda coming at this from 328 00:11:35,040 --> 00:11:37,595 the security model is in my 329 00:11:37,975 --> 00:11:39,894 head too when you see this announcement. You're 330 00:11:39,894 --> 00:11:42,075 like, okay. You're now fine tuning 331 00:11:42,534 --> 00:11:43,195 an LLM. 332 00:11:43,815 --> 00:11:45,434 What does this mean for 333 00:11:46,054 --> 00:11:48,375 data security? What does this mean for how 334 00:11:48,375 --> 00:11:50,634 is my data now being used if I'm 335 00:11:50,695 --> 00:11:51,434 fine tuning 336 00:11:51,769 --> 00:11:53,950 an LLM? Is it going into the LLM? 337 00:11:54,009 --> 00:11:56,009 Is it training the LLM? And they talked 338 00:11:56,009 --> 00:11:58,090 about this in the Microsoft Mechanics video, and, 339 00:11:58,090 --> 00:12:00,250 ironically, I don't I probably don't have the 340 00:12:00,250 --> 00:12:03,129 business use cases for this yet, although I'm 341 00:12:03,129 --> 00:12:05,865 starting to play with it more. But it's 342 00:12:05,865 --> 00:12:08,205 more thinking about this from a client perspective 343 00:12:08,585 --> 00:12:11,304 of questions they're gonna ask about their data, 344 00:12:11,304 --> 00:12:13,225 their data security. Some of the stuff we 345 00:12:13,225 --> 00:12:14,665 even talked about a couple weeks ago when 346 00:12:14,665 --> 00:12:16,585 I was talking about how a lot of 347 00:12:16,585 --> 00:12:18,024 clients are coming to me now and saying 348 00:12:18,024 --> 00:12:19,945 we wanna prep for Copilot, we wanna make 349 00:12:19,945 --> 00:12:21,085 sure data is secure. 350 00:12:21,600 --> 00:12:23,679 Now what does fine tuning this mean? And 351 00:12:23,679 --> 00:12:25,279 they did a pretty good job, I would 352 00:12:25,279 --> 00:12:27,440 say, in the Microsoft Mechanics video from that 353 00:12:27,440 --> 00:12:30,179 perspective of, like, you still have your LLM, 354 00:12:30,480 --> 00:12:32,959 and it's not and I I could not 355 00:12:32,959 --> 00:12:34,879 explain this from a deep technical level on 356 00:12:34,879 --> 00:12:36,559 how it actually works, but they actually take 357 00:12:36,559 --> 00:12:38,855 your data. And the way they illustrated it 358 00:12:38,855 --> 00:12:41,115 is they, like, attach it to the LLM 359 00:12:41,335 --> 00:12:43,014 for the fine tuning. It's not like your 360 00:12:43,014 --> 00:12:44,855 data's gonna go into the LLM and they're 361 00:12:44,855 --> 00:12:48,154 retraining the core LLM. It's like something between 362 00:12:48,774 --> 00:12:51,990 RAG and actually having your own model, where 363 00:12:51,990 --> 00:12:54,069 it's a model with your data attached to 364 00:12:54,069 --> 00:12:56,149 it to do the fine tuning. And it 365 00:12:56,149 --> 00:12:57,909 sounded like from that video that it's not 366 00:12:57,909 --> 00:13:00,549 even always attached to the LLM. Like, it 367 00:13:00,549 --> 00:13:02,309 attaches to the LLM to do the fine 368 00:13:02,309 --> 00:13:04,409 tuning, and then it kinda goes and stands 369 00:13:05,154 --> 00:13:06,835 separate from it after you've done the fine 370 00:13:06,835 --> 00:13:08,995 tuning. It only reattaches when you do it, 371 00:13:08,995 --> 00:13:11,794 and it still maintains, like, your data and 372 00:13:11,794 --> 00:13:12,615 your own tenancy. 373 00:13:13,235 --> 00:13:15,475 They're not taking your data to train the 374 00:13:15,475 --> 00:13:16,535 core LLM. 375 00:13:17,154 --> 00:13:19,519 All your access lists and everything, that is 376 00:13:19,519 --> 00:13:20,500 all still maintained. 377 00:13:20,960 --> 00:13:23,039 So it was that that part to me 378 00:13:23,039 --> 00:13:25,840 was interesting to try to understand a little 379 00:13:25,840 --> 00:13:27,620 bit more from that perspective 380 00:13:28,480 --> 00:13:30,980 on how was it doing the fine tuning 381 00:13:31,194 --> 00:13:33,054 of the model without actually 382 00:13:33,434 --> 00:13:35,274 absorbing the data into the model? The way 383 00:13:35,274 --> 00:13:36,815 you can think about it is 384 00:13:37,674 --> 00:13:38,495 fine tuning, 385 00:13:38,954 --> 00:13:40,794 and and that's really what we're talking about 386 00:13:40,794 --> 00:13:42,495 here is is kinda 387 00:13:42,954 --> 00:13:45,674 taking an existing model that's already been trained 388 00:13:45,834 --> 00:13:49,079 Yep. Let's say chat GPT four o, and 389 00:13:49,079 --> 00:13:50,559 then you're gonna take that, and that that 390 00:13:50,679 --> 00:13:53,000 that's a stage in that model's life cycle, 391 00:13:53,000 --> 00:13:54,679 and you can use it as it is 392 00:13:54,679 --> 00:13:56,600 with the weights that exist and and and 393 00:13:56,600 --> 00:13:59,000 all of that stuff. Now what you can 394 00:13:59,000 --> 00:14:00,679 then do so you do all that, and 395 00:14:00,679 --> 00:14:03,455 and that's kinda like training slash pre training. 396 00:14:03,754 --> 00:14:06,095 And then you take that pre trained model 397 00:14:06,554 --> 00:14:09,535 and you fine tune it. And fine tuning 398 00:14:09,674 --> 00:14:13,769 is a refinement process. So you're taking this 399 00:14:14,090 --> 00:14:16,009 very it doesn't have to be very small, 400 00:14:16,009 --> 00:14:17,550 but you're taking a smaller 401 00:14:18,090 --> 00:14:19,070 task specific 402 00:14:19,529 --> 00:14:21,710 dataset. So, like, in my example, 403 00:14:22,170 --> 00:14:25,850 PRDs and and justifications. Templates and yeah. You're 404 00:14:25,850 --> 00:14:27,070 doing this to 405 00:14:27,929 --> 00:14:28,429 optimize 406 00:14:28,809 --> 00:14:29,309 performance 407 00:14:29,774 --> 00:14:30,274 and 408 00:14:30,735 --> 00:14:32,914 and and drive a business 409 00:14:33,294 --> 00:14:33,794 process. 410 00:14:34,334 --> 00:14:36,894 So, like, you you you're right. You're not 411 00:14:36,894 --> 00:14:39,875 loading this stuff into the original 412 00:14:40,414 --> 00:14:43,549 pretrained model. You're kind of creating your little 413 00:14:43,549 --> 00:14:46,110 bit of, like, an your own offshoot of 414 00:14:46,110 --> 00:14:47,709 a model, but you don't have to go 415 00:14:47,709 --> 00:14:49,629 back and redo all the training and redo 416 00:14:49,629 --> 00:14:51,870 all the weights for the original model. So 417 00:14:51,870 --> 00:14:54,350 you've kind of augmented it with this set 418 00:14:54,350 --> 00:14:57,409 of sidecar weights that can then be used 419 00:14:57,605 --> 00:14:58,425 as well to 420 00:14:58,965 --> 00:15:01,785 go ahead and make your responses better. 421 00:15:02,165 --> 00:15:04,245 Now the the the purpose of this and 422 00:15:04,245 --> 00:15:05,305 and the difference 423 00:15:05,764 --> 00:15:06,264 versus, 424 00:15:07,365 --> 00:15:08,825 say, something like rag 425 00:15:09,205 --> 00:15:09,705 is 426 00:15:10,460 --> 00:15:13,019 RAG has to go out and you you 427 00:15:13,019 --> 00:15:15,019 have to chunk all these documents, you have 428 00:15:15,019 --> 00:15:16,480 to put them in a vector database, 429 00:15:16,940 --> 00:15:19,340 and you have to compute and and and 430 00:15:19,340 --> 00:15:21,740 run all that stuff every time. This is 431 00:15:21,740 --> 00:15:23,600 just having that set of 432 00:15:23,975 --> 00:15:26,875 kinda fine tuned weights out there for you 433 00:15:27,175 --> 00:15:29,014 and ready to go. So it's it's it's 434 00:15:29,014 --> 00:15:30,535 a click stop if you think about it 435 00:15:30,535 --> 00:15:33,014 as a series of stages. So pre training 436 00:15:33,014 --> 00:15:34,934 is all about, let me get the model 437 00:15:34,934 --> 00:15:36,235 out there, and that's 438 00:15:36,759 --> 00:15:37,899 general representation 439 00:15:38,519 --> 00:15:40,919 of whatever it was trained on. And then 440 00:15:40,919 --> 00:15:43,339 fine tuning is let me take and 441 00:15:43,639 --> 00:15:45,820 take this model to the next level by 442 00:15:46,360 --> 00:15:48,440 really augmenting it in a way I shouldn't 443 00:15:48,440 --> 00:15:50,620 say augment because of the whole rag thing, 444 00:15:50,865 --> 00:15:51,365 by 445 00:15:51,825 --> 00:15:54,384 adapting it. Yeah. Fine tuning it. We're fine 446 00:15:54,384 --> 00:15:55,284 tuning it. Right? 447 00:15:55,825 --> 00:15:57,825 We're fine tuning it. So you're going but 448 00:15:57,825 --> 00:15:59,524 you're going from general representations 449 00:16:00,065 --> 00:16:04,004 to specific tasks or Right. Knowledge domains. 450 00:16:04,669 --> 00:16:06,829 Processes. Yeah. Processes. So you could do this 451 00:16:06,829 --> 00:16:08,850 multiple times. So when you hear about maybe 452 00:16:09,069 --> 00:16:10,850 a company that's working on 453 00:16:11,389 --> 00:16:14,190 AI to solve cancer, well, guess what? They 454 00:16:14,190 --> 00:16:15,709 did the same thing. Right? They they took 455 00:16:15,709 --> 00:16:18,084 that general model that was out there, and 456 00:16:18,084 --> 00:16:19,764 then they took a bunch of cancer research 457 00:16:19,764 --> 00:16:21,605 that had already been done previously, and then 458 00:16:21,605 --> 00:16:23,684 they fine tuned that model so that they 459 00:16:23,684 --> 00:16:26,404 could have a purpose built model that they 460 00:16:26,404 --> 00:16:28,105 could go and and work with 461 00:16:28,485 --> 00:16:28,985 to 462 00:16:29,605 --> 00:16:31,284 attack cancer, attack your 463 00:16:31,879 --> 00:16:34,459 in your case, maybe your business processes 464 00:16:35,159 --> 00:16:38,120 around statements of work and how you put 465 00:16:38,120 --> 00:16:40,120 those together for clients. I should try one 466 00:16:40,120 --> 00:16:41,820 of those. You've done that a whole bunch 467 00:16:41,879 --> 00:16:44,779 throughout the years with things like document libraries 468 00:16:44,919 --> 00:16:47,924 and document sets, and this is I I 469 00:16:47,924 --> 00:16:49,125 I think you should give it a shot, 470 00:16:49,125 --> 00:16:50,725 like, if you're licensed for it because it's 471 00:16:50,725 --> 00:16:52,084 a little bit of a next level thing 472 00:16:52,084 --> 00:16:53,924 for you. I should absolutely go try that. 473 00:16:53,924 --> 00:16:56,324 I I didn't statements of work didn't click 474 00:16:56,324 --> 00:16:57,684 in my head until we were just talking 475 00:16:57,684 --> 00:16:59,944 about it. I'm like, yeah, master service agreements. 476 00:17:00,470 --> 00:17:02,070 And maybe it's because I don't put a 477 00:17:02,070 --> 00:17:04,070 ton in there. Statements of work, I put 478 00:17:04,070 --> 00:17:05,589 a little bit more in there. Master service 479 00:17:05,589 --> 00:17:06,890 agreements are a lot of addresses 480 00:17:07,349 --> 00:17:09,690 and that. But I have had clients ask 481 00:17:09,910 --> 00:17:11,750 can't even remember which client it was. Where 482 00:17:11,750 --> 00:17:13,430 they did, they're like, well, if I have 483 00:17:13,430 --> 00:17:15,984 a template of a document and I want 484 00:17:15,984 --> 00:17:18,884 to use AI to fill out the template, 485 00:17:19,105 --> 00:17:20,964 can you do it? And I'm like, well, 486 00:17:21,025 --> 00:17:22,085 not really. 487 00:17:22,384 --> 00:17:23,445 The way it was 488 00:17:24,065 --> 00:17:26,224 two months ago even or before they talked 489 00:17:26,224 --> 00:17:28,619 about this fine tuning, it's like, doesn't really 490 00:17:28,619 --> 00:17:30,779 do a great job at that. And this 491 00:17:30,779 --> 00:17:31,519 is where 492 00:17:32,059 --> 00:17:34,539 now I think as clients get this, again, 493 00:17:34,539 --> 00:17:36,700 licensed for it, willing to build it out, 494 00:17:36,700 --> 00:17:38,859 it's a lot more gets you a lot 495 00:17:38,859 --> 00:17:40,539 closer to being able to have templates that 496 00:17:40,539 --> 00:17:42,539 you can auto fill out. Not auto fill 497 00:17:42,539 --> 00:17:44,855 out, but you still as a user 498 00:17:45,234 --> 00:17:47,734 can yeah. Yeah. So so you can generate 499 00:17:47,795 --> 00:17:48,295 those 500 00:17:48,755 --> 00:17:51,575 and you as a user providing inputs. So 501 00:17:51,875 --> 00:17:53,815 the the way I think about it and 502 00:17:54,115 --> 00:17:56,115 the way I see a lot of folks 503 00:17:56,115 --> 00:17:57,174 use generative 504 00:17:57,555 --> 00:17:58,295 AI today 505 00:17:58,769 --> 00:18:01,590 is they kind of look at it and 506 00:18:02,289 --> 00:18:03,670 and I've definitely seen, 507 00:18:04,130 --> 00:18:06,289 particularly, like, my PRD example, I've seen some 508 00:18:06,289 --> 00:18:08,210 PMs that go out there, and they just 509 00:18:08,210 --> 00:18:10,070 say, hey. Write me the document. 510 00:18:10,450 --> 00:18:12,869 And this isn't, hey. Write me the document. 511 00:18:13,285 --> 00:18:16,325 This is write me help me write the 512 00:18:16,325 --> 00:18:19,545 document so that it conforms to this process, 513 00:18:19,605 --> 00:18:20,744 this knowledge domain, 514 00:18:21,444 --> 00:18:24,085 any anything like that, you know, that exists 515 00:18:24,085 --> 00:18:26,740 out there. So, yeah, it's it's a tool 516 00:18:26,740 --> 00:18:27,559 in the toolbox 517 00:18:28,019 --> 00:18:30,660 versus a do my job for me kinda 518 00:18:30,660 --> 00:18:33,160 thing. So if you take an agent, 519 00:18:33,700 --> 00:18:35,460 and I don't know that they talked about 520 00:18:35,460 --> 00:18:37,380 this, and I don't know if you've seen 521 00:18:37,380 --> 00:18:39,505 it, Could you potentially take like, let's say 522 00:18:39,505 --> 00:18:42,164 you go do the co the fine tuning, 523 00:18:42,305 --> 00:18:43,984 and, really, you're creating an agent. So that's 524 00:18:43,984 --> 00:18:45,825 the other aspect about this. This isn't fine 525 00:18:45,825 --> 00:18:47,125 tuning like your general 526 00:18:47,585 --> 00:18:50,669 Microsoft three sixty five Copilot responses. You're creating 527 00:18:50,669 --> 00:18:52,849 an agent that's fine tuned 528 00:18:53,230 --> 00:18:53,809 to do 529 00:18:54,109 --> 00:18:55,409 a specific task. 530 00:18:55,710 --> 00:18:56,769 Can you combine 531 00:18:57,149 --> 00:18:59,630 agents with actions yet where you could do 532 00:18:59,630 --> 00:19:01,970 a fine tuned agent going back to the 533 00:19:02,029 --> 00:19:04,335 proposals or statements of work where you're then 534 00:19:04,335 --> 00:19:05,474 using an action 535 00:19:05,775 --> 00:19:08,275 to prompt or to get certain information 536 00:19:08,734 --> 00:19:11,134 to feed into a fine tuned agent to 537 00:19:11,134 --> 00:19:14,255 generate these particular outputs? I haven't tried to 538 00:19:14,255 --> 00:19:16,255 chain them together like that yet. I I 539 00:19:16,255 --> 00:19:17,934 I think that's where all this stuff goes. 540 00:19:17,934 --> 00:19:19,154 So if you think about 541 00:19:19,819 --> 00:19:22,700 MCP servers and the whole model context protocol 542 00:19:22,700 --> 00:19:23,200 thing, 543 00:19:23,500 --> 00:19:26,779 that's where these agentic things are going. And 544 00:19:26,779 --> 00:19:28,400 you see a little bit of this today 545 00:19:28,460 --> 00:19:28,960 in 546 00:19:29,420 --> 00:19:31,339 in maybe in the declarative agents that we 547 00:19:31,339 --> 00:19:33,965 already talked about where there's a connector for 548 00:19:33,965 --> 00:19:36,305 SAP, there's a connector for Workday, 549 00:19:36,924 --> 00:19:39,105 things like that. Like, you have this agentic 550 00:19:39,164 --> 00:19:40,924 system on the side that it can reach 551 00:19:40,924 --> 00:19:43,164 out to and sidecar that knowledge in from 552 00:19:43,164 --> 00:19:44,065 that other system 553 00:19:44,445 --> 00:19:46,045 and bring it in. I always think it's 554 00:19:46,045 --> 00:19:48,809 super funny, at least in today's world where 555 00:19:48,809 --> 00:19:49,549 we're still 556 00:19:50,009 --> 00:19:52,190 very early days in this stuff, when 557 00:19:52,730 --> 00:19:55,690 you you see the funny two agents talking 558 00:19:55,690 --> 00:19:56,970 to each other kinda thing. I don't I 559 00:19:56,970 --> 00:19:59,049 don't know if you've seen the video. It 560 00:19:59,049 --> 00:20:00,490 came out a couple months ago, but it 561 00:20:00,490 --> 00:20:03,015 was like a robo calling agent called another 562 00:20:03,015 --> 00:20:05,174 robo caller. And then they both figured out 563 00:20:05,174 --> 00:20:07,015 that they were robo callers, so they just 564 00:20:07,015 --> 00:20:10,294 basically started talking to themselves in it sounded 565 00:20:10,294 --> 00:20:12,615 like an old, like, bot modem going off. 566 00:20:12,615 --> 00:20:14,430 Like I think I did see that. Yes. 567 00:20:14,590 --> 00:20:16,509 Back and forth because they knew there was 568 00:20:16,509 --> 00:20:18,910 no human there and, hey. Like like, let's 569 00:20:18,910 --> 00:20:20,990 go down a different path kinda thing. So 570 00:20:20,990 --> 00:20:22,590 they're still a little wild and off the 571 00:20:22,590 --> 00:20:24,750 rails. But, ultimately, I think that's where these 572 00:20:24,750 --> 00:20:27,330 things do go. So if I think about 573 00:20:27,585 --> 00:20:28,085 an 574 00:20:28,704 --> 00:20:31,505 agentic chain to build a PRD, well, great. 575 00:20:31,505 --> 00:20:33,444 So I've got the fine tuned 576 00:20:33,744 --> 00:20:35,845 process where I've come in and said, hey. 577 00:20:35,904 --> 00:20:38,384 Help me write this within these guard re 578 00:20:38,625 --> 00:20:40,544 these guardrails, or help me generate this thing 579 00:20:40,544 --> 00:20:42,804 within these guardrails and and these boundaries. 580 00:20:43,440 --> 00:20:46,019 But part of my PRDs would be customer 581 00:20:46,079 --> 00:20:46,579 stories. 582 00:20:47,039 --> 00:20:49,599 Well, maybe I wanna go build a little 583 00:20:49,599 --> 00:20:52,639 agent that does nothing other than reason over 584 00:20:52,639 --> 00:20:53,779 my customer feedback. 585 00:20:54,079 --> 00:20:56,159 And then there's another agent that I might 586 00:20:56,159 --> 00:20:58,259 wanna go build that reasons over 587 00:20:58,865 --> 00:21:01,265 the financial aspects of my business and and 588 00:21:01,265 --> 00:21:03,265 the cogs of it and and how that 589 00:21:03,265 --> 00:21:05,904 comes together. So all these little things have 590 00:21:05,904 --> 00:21:06,565 their own 591 00:21:07,025 --> 00:21:08,705 click stops along the way, and, eventually, I 592 00:21:08,705 --> 00:21:10,305 think, yeah, that that's where you end up 593 00:21:10,305 --> 00:21:12,880 is potentially a bunch of agents just chatting 594 00:21:12,880 --> 00:21:14,720 back and forth with each other, and one 595 00:21:14,720 --> 00:21:16,319 throws this over here, and then it throws 596 00:21:16,319 --> 00:21:17,679 it back this, then it throws it into 597 00:21:17,679 --> 00:21:19,859 a new one, and out the other end 598 00:21:20,000 --> 00:21:22,559 comes something else. And I I think that's 599 00:21:22,559 --> 00:21:25,200 the thing that we we all haven't, like, 600 00:21:25,200 --> 00:21:27,759 really figured out as Right. I saw the 601 00:21:27,839 --> 00:21:29,744 I think it was the CEO of Claude 602 00:21:29,884 --> 00:21:31,964 this week, last week. He was doing an 603 00:21:31,964 --> 00:21:34,384 interview, and he was talking about basically, like, 604 00:21:34,444 --> 00:21:35,644 50% 605 00:21:35,644 --> 00:21:38,045 of it was some crazy number, 20%, fifty 606 00:21:38,045 --> 00:21:40,525 %. It was really high of kinda low 607 00:21:40,525 --> 00:21:43,200 end knowledge worker jobs going away because of 608 00:21:43,440 --> 00:21:46,799 AI and and Gen AI. So nobody knows 609 00:21:46,799 --> 00:21:48,799 how this stuff's gonna go and where it 610 00:21:48,799 --> 00:21:50,799 comes out. I'm trying not to be, like, 611 00:21:50,799 --> 00:21:52,799 the doom and gloom person about it. Like, 612 00:21:52,799 --> 00:21:54,899 I very much do look to this stuff 613 00:21:55,519 --> 00:21:57,299 to augment my job today 614 00:21:57,674 --> 00:22:00,315 and help me be faster? Like, I don't 615 00:22:00,315 --> 00:22:02,475 know. IT Pro land, all we've ever been 616 00:22:02,475 --> 00:22:04,235 doing is trying to automate ourselves out of 617 00:22:04,235 --> 00:22:06,715 jobs. This is just another step in that 618 00:22:06,715 --> 00:22:08,475 journey is the way I kinda look at 619 00:22:08,475 --> 00:22:10,654 it. And then just go watch Mission Impossible. 620 00:22:10,875 --> 00:22:12,875 Have you seen the latest Mission Impossible yet? 621 00:22:12,875 --> 00:22:14,450 I haven't. Don't spoil it for me. I 622 00:22:14,450 --> 00:22:16,289 will not spoil it for you, but all 623 00:22:16,289 --> 00:22:18,230 I could think about was AI and Copilot 624 00:22:18,289 --> 00:22:19,670 during Mission Impossible. 625 00:22:25,090 --> 00:22:27,250 Do you feel overwhelmed by trying to manage 626 00:22:27,250 --> 00:22:29,605 your Office three sixty five environment? Are you 627 00:22:29,605 --> 00:22:32,904 facing unexpected issues that disrupt your company's productivity? 628 00:22:33,125 --> 00:22:35,044 Intelligink is here to help. Much like you 629 00:22:35,044 --> 00:22:36,964 take your car to the mechanic that has 630 00:22:36,964 --> 00:22:39,125 specialized knowledge on how to best keep your 631 00:22:39,125 --> 00:22:42,105 car running, Intelligink helps you with your Microsoft 632 00:22:42,164 --> 00:22:45,710 cloud environment because that's their expertise. Intelligent keeps 633 00:22:45,710 --> 00:22:47,649 up with the latest updates in the Microsoft 634 00:22:47,710 --> 00:22:50,029 cloud to help keep your business running smoothly 635 00:22:50,029 --> 00:22:51,789 and ahead of the curve. Whether you are 636 00:22:51,789 --> 00:22:54,190 a small organization with just a few users 637 00:22:54,190 --> 00:22:57,009 up to an organization of several thousand employees, 638 00:22:57,414 --> 00:22:59,335 They want to partner with you to implement 639 00:22:59,335 --> 00:23:00,075 and administer 640 00:23:00,375 --> 00:23:02,154 your Microsoft cloud technology. 641 00:23:02,775 --> 00:23:06,315 Visit them at inteliginc.com/podcast. 642 00:23:06,535 --> 00:23:13,339 That's intelligink.com/podcast 643 00:23:13,640 --> 00:23:15,799 for more information or to schedule a thirty 644 00:23:15,799 --> 00:23:17,900 minute call to get started with them today. 645 00:23:18,119 --> 00:23:21,480 Remember, Intelligink focuses on the Microsoft cloud so 646 00:23:21,480 --> 00:23:23,259 you can focus on your business. 647 00:23:25,480 --> 00:23:27,825 Alright. So Copilot Tuning, folks can go out 648 00:23:27,825 --> 00:23:29,505 and sign up for that. I believe you 649 00:23:29,505 --> 00:23:30,865 have to sign up for a preview for 650 00:23:30,865 --> 00:23:33,105 that one or or kinda put yourself Yeah. 651 00:23:33,105 --> 00:23:36,144 This one's still it's early access. About the 652 00:23:36,144 --> 00:23:37,924 Copilot Tuning early access, 653 00:23:38,819 --> 00:23:41,539 customers with more than so this is a 654 00:23:41,539 --> 00:23:43,940 high barrier to entry, so I probably cannot 655 00:23:43,940 --> 00:23:45,460 do this yet either. But it does say 656 00:23:45,460 --> 00:23:47,059 early access, you have to have more than 657 00:23:47,059 --> 00:23:47,960 5,000 658 00:23:48,579 --> 00:23:51,399 Microsoft three 60 five Copilot seats to participate, 659 00:23:51,805 --> 00:23:53,345 EAP candidates undergo 660 00:23:53,805 --> 00:23:55,964 screening, make sure they have scenarios that align 661 00:23:55,964 --> 00:23:58,365 with it. Like, Microsoft is at that point, 662 00:23:58,365 --> 00:24:00,525 they're sounds like they really wanna learn from 663 00:24:00,525 --> 00:24:02,365 this before they just throw it out there 664 00:24:02,365 --> 00:24:04,445 to GA for everyone. This is the same 665 00:24:04,445 --> 00:24:06,829 thing you saw with Copilot before. So 666 00:24:07,210 --> 00:24:09,450 this shouldn't be a model that surprises folks 667 00:24:09,450 --> 00:24:11,069 at this point. They have a couple 668 00:24:11,450 --> 00:24:13,690 people that are already using it. Ernst and 669 00:24:13,690 --> 00:24:14,909 Young, McCarthy, 670 00:24:15,369 --> 00:24:17,049 I don't even know how to pronounce that 671 00:24:17,049 --> 00:24:18,029 word, Tyrolt, 672 00:24:18,970 --> 00:24:19,869 Land O'Lakes. 673 00:24:20,224 --> 00:24:22,545 They have some big companies using it. So 674 00:24:22,545 --> 00:24:24,785 if you're a large customer and wanna go 675 00:24:24,785 --> 00:24:27,025 play with us, early access program, if you're 676 00:24:27,025 --> 00:24:29,424 not, be patient. You'll get it eventually. Don't 677 00:24:29,424 --> 00:24:30,945 let that stop you from going out and 678 00:24:30,945 --> 00:24:33,025 watching some videos. Like, the mechanics video is, 679 00:24:33,025 --> 00:24:35,159 like, ten, twelve minutes. Totally worth your time, 680 00:24:35,159 --> 00:24:36,919 I think. This looks like a build video. 681 00:24:36,919 --> 00:24:38,839 I haven't clicked on this one. This one's 682 00:24:38,839 --> 00:24:41,000 probably, like, an hour long. Oh, no. This 683 00:24:41,000 --> 00:24:43,799 one's ten minutes. There's an introduction video too 684 00:24:43,799 --> 00:24:46,359 on the article, so definitely go out and 685 00:24:46,359 --> 00:24:48,139 learn from it. Start building agents. 686 00:24:48,445 --> 00:24:50,365 That's I've been playing more and more with 687 00:24:50,365 --> 00:24:52,765 building some agents as well, which that leads 688 00:24:52,765 --> 00:24:54,625 to another news article. This one's 689 00:24:54,924 --> 00:24:57,184 one that caught my eye is 690 00:24:57,884 --> 00:24:59,025 you now have 691 00:24:59,884 --> 00:25:00,384 Microsoft 692 00:25:00,845 --> 00:25:02,144 Entra Agent 693 00:25:02,630 --> 00:25:05,750 ID, Secure and Manager AI agents. So this 694 00:25:05,750 --> 00:25:07,930 one was, I think, right before 695 00:25:08,710 --> 00:25:12,150 Build, and I saw Merrill posted something about 696 00:25:12,150 --> 00:25:13,910 this. I saw that. I saw a few 697 00:25:13,910 --> 00:25:16,934 blog articles about it. Alex Simmons has a 698 00:25:16,934 --> 00:25:19,494 announcement on the Entra blog about it. But 699 00:25:19,494 --> 00:25:22,054 this was interesting to me because, again, we've 700 00:25:22,054 --> 00:25:24,294 been talking about we're building these agents. We're 701 00:25:24,294 --> 00:25:25,835 giving them access to 702 00:25:26,375 --> 00:25:28,474 SharePoint sites, to different data. 703 00:25:29,519 --> 00:25:31,519 And this is starting to look at, so 704 00:25:31,519 --> 00:25:33,299 how do we actually go in 705 00:25:33,920 --> 00:25:37,119 and start securing those agents, securing what those 706 00:25:37,119 --> 00:25:38,259 agents can do? 707 00:25:38,559 --> 00:25:40,019 And now if 708 00:25:40,320 --> 00:25:42,000 you're an admin and you're a tenant, you 709 00:25:42,000 --> 00:25:44,204 can go out to Entra, You can actually 710 00:25:44,204 --> 00:25:44,944 go out 711 00:25:45,325 --> 00:25:45,984 to Entra 712 00:25:46,285 --> 00:25:47,984 and go to your enterprise applications. 713 00:25:48,444 --> 00:25:49,984 And if you look at application 714 00:25:50,365 --> 00:25:50,865 types, 715 00:25:51,325 --> 00:25:53,025 you now have an agent 716 00:25:53,404 --> 00:25:53,904 ID 717 00:25:54,684 --> 00:25:55,984 type in your 718 00:25:56,285 --> 00:25:57,424 enterprise applications. 719 00:25:57,820 --> 00:25:59,840 And this is this is cool 720 00:26:00,220 --> 00:26:01,359 just from a pure 721 00:26:01,820 --> 00:26:04,700 reporting perspective right now, and that it will 722 00:26:04,700 --> 00:26:07,840 give you a list of every single agent 723 00:26:08,059 --> 00:26:10,320 that has been created in your environment 724 00:26:10,940 --> 00:26:12,160 by any users, 725 00:26:12,644 --> 00:26:15,384 whether it's an agent created in Copilot Studio 726 00:26:15,765 --> 00:26:19,125 or Azure AI Foundry that are registered a 727 00:26:19,125 --> 00:26:21,785 part of your Microsoft three sixty five tenant. 728 00:26:21,845 --> 00:26:24,085 This is one that there's some videos out 729 00:26:24,085 --> 00:26:25,945 there. I went and did it. My tenant 730 00:26:26,299 --> 00:26:27,679 saw all the agents there. 731 00:26:27,980 --> 00:26:29,599 Great for reporting perspective 732 00:26:29,980 --> 00:26:31,900 for right now. I would say from, like, 733 00:26:31,900 --> 00:26:33,200 a security perspective, 734 00:26:33,740 --> 00:26:36,140 there's not as much you can do, but 735 00:26:36,140 --> 00:26:37,819 there's a lot of stuff coming. So they 736 00:26:37,819 --> 00:26:39,494 have a whole part of this that's what 737 00:26:39,575 --> 00:26:41,815 what's next with these EntraID agents now that 738 00:26:41,815 --> 00:26:44,134 they're in EntraID, now that they're showing up 739 00:26:44,134 --> 00:26:45,515 as enterprise applications. 740 00:26:45,974 --> 00:26:47,654 It means that you can start going in 741 00:26:47,654 --> 00:26:50,954 and looking at, like, eventually what Graph API 742 00:26:51,015 --> 00:26:52,714 access they have or 743 00:26:53,049 --> 00:26:55,769 even creating conditional access policies for your agent. 744 00:26:55,769 --> 00:26:57,930 So now someone can interact with this agent. 745 00:26:57,930 --> 00:27:00,029 Being able to go in and do 746 00:27:00,490 --> 00:27:04,910 granular conditional access and all that detailed permissioning 747 00:27:05,130 --> 00:27:06,509 that comes from that perspective 748 00:27:06,890 --> 00:27:07,390 because, 749 00:27:08,115 --> 00:27:09,154 I mean, I mean, we're getting to the 750 00:27:09,154 --> 00:27:11,075 point with these things, like, going back a 751 00:27:11,075 --> 00:27:12,515 little bit to what we talked about earlier 752 00:27:12,515 --> 00:27:13,015 with 753 00:27:13,394 --> 00:27:16,914 agentic workflows and these things effectively talking to 754 00:27:16,914 --> 00:27:19,634 each other. Like, you almost are treating them 755 00:27:19,634 --> 00:27:22,820 as individuals within your business. So it makes 756 00:27:22,820 --> 00:27:24,119 sense from that perspective 757 00:27:24,500 --> 00:27:26,900 to secure them in in the same way 758 00:27:26,900 --> 00:27:29,140 that you would your users and other things 759 00:27:29,140 --> 00:27:30,580 that are out there. So this will give 760 00:27:30,580 --> 00:27:32,359 you additional operational controls 761 00:27:33,140 --> 00:27:36,500 and, frankly, like, comfort food around locking some 762 00:27:36,500 --> 00:27:38,835 of this stuff down when it comes to 763 00:27:38,835 --> 00:27:40,835 what do they access, when do they access 764 00:27:40,835 --> 00:27:42,695 it, how do they access it. 765 00:27:43,075 --> 00:27:44,994 The the the I think the granular graph 766 00:27:44,994 --> 00:27:46,994 stuff will be great. Like, hey, making sure 767 00:27:46,994 --> 00:27:48,214 this thing only has 768 00:27:48,755 --> 00:27:51,690 a read permission versus a read all, or 769 00:27:51,769 --> 00:27:53,609 does it does it have the ability to 770 00:27:53,609 --> 00:27:54,109 actually 771 00:27:55,049 --> 00:27:57,529 generate content in that environment? Like, oh, no, 772 00:27:57,529 --> 00:27:59,450 I wanna I wanna block rights for this 773 00:27:59,450 --> 00:28:01,210 one because I don't actually want it to 774 00:28:01,210 --> 00:28:03,210 go in and I don't know. Somebody goes 775 00:28:03,210 --> 00:28:06,089 and builds the agent to rename titles in 776 00:28:06,089 --> 00:28:06,750 the GAL. 777 00:28:07,525 --> 00:28:09,444 You you you might not want that that 778 00:28:09,444 --> 00:28:10,105 that thing 779 00:28:10,404 --> 00:28:12,565 running that way. Yep. It's also good just 780 00:28:12,565 --> 00:28:13,224 from a 781 00:28:13,605 --> 00:28:16,325 visibility perspective for these things. Like, it's another 782 00:28:16,325 --> 00:28:19,045 way to kind of track adoption in your 783 00:28:19,045 --> 00:28:21,305 organization, maybe go talk to some users, 784 00:28:22,309 --> 00:28:23,690 see where they're getting 785 00:28:24,070 --> 00:28:26,309 ROI out of out of these kinds of 786 00:28:26,309 --> 00:28:26,809 things, 787 00:28:27,430 --> 00:28:29,990 and better manage the life cycle of them. 788 00:28:29,990 --> 00:28:32,170 I still see a lot of places where 789 00:28:32,390 --> 00:28:34,329 you go in and you look at their 790 00:28:35,054 --> 00:28:37,875 their intro tenant, and it's just the graveyard 791 00:28:38,014 --> 00:28:40,174 of lost stuff that that has not been 792 00:28:40,174 --> 00:28:42,335 cleaned up over the years and hasn't been 793 00:28:42,335 --> 00:28:44,575 attributed to a user. Hey. What's this old 794 00:28:44,575 --> 00:28:47,154 enterprise app ID doing out here that has, 795 00:28:47,214 --> 00:28:50,579 like, basically, god permissions in your environment? Well, 796 00:28:50,579 --> 00:28:52,019 like, what did you do, and why was 797 00:28:52,019 --> 00:28:53,859 this a deity? Oh, that was created way 798 00:28:53,859 --> 00:28:55,940 back when five years ago when we spun 799 00:28:55,940 --> 00:28:58,019 up our tenant. Great. Let's clean it up 800 00:28:58,019 --> 00:29:00,404 today. Right. And I think some of those 801 00:29:00,404 --> 00:29:02,105 two going back to life cycle, 802 00:29:02,484 --> 00:29:05,045 above and beyond this announcement, permissions change too 803 00:29:05,045 --> 00:29:07,465 with Graph API. Right? Like, there's old applications 804 00:29:07,605 --> 00:29:10,325 out there that have full control over your 805 00:29:10,325 --> 00:29:11,625 entire SharePoint environment 806 00:29:11,924 --> 00:29:12,424 because 807 00:29:12,965 --> 00:29:13,465 granular 808 00:29:14,419 --> 00:29:17,940 site level API access in the graph wasn't 809 00:29:17,940 --> 00:29:20,440 there when these apps were originally created. So 810 00:29:20,579 --> 00:29:21,720 as things continue 811 00:29:22,500 --> 00:29:25,799 to evolve, permissions continue to evolve over time, 812 00:29:26,099 --> 00:29:27,150 being able to, 813 00:29:28,565 --> 00:29:30,484 see that life cycle, see what they're doing, 814 00:29:30,484 --> 00:29:32,325 see what they have access to, some of 815 00:29:32,325 --> 00:29:34,325 the auditing and monitoring you'll be able to 816 00:29:34,325 --> 00:29:35,464 do around 817 00:29:36,005 --> 00:29:38,805 logs and visibilities into what these agents are 818 00:29:38,805 --> 00:29:40,184 doing, what they're accessing. 819 00:29:40,960 --> 00:29:42,960 You've even seen that some with Copilot and 820 00:29:42,960 --> 00:29:44,340 some of the new stuff around, 821 00:29:45,039 --> 00:29:48,420 sensitivity labels and being able to do DLP 822 00:29:48,960 --> 00:29:50,820 and audit user interactions 823 00:29:51,200 --> 00:29:52,420 with Copilot, 824 00:29:52,720 --> 00:29:55,600 now being able to audit agent interactions with 825 00:29:55,600 --> 00:29:57,855 stuff. Enterprises, this is I feel like it's 826 00:29:57,855 --> 00:29:59,455 something you have to have when you start 827 00:29:59,455 --> 00:30:02,495 getting into AI. If there's legal questions that 828 00:30:02,495 --> 00:30:04,414 come up, questions that come up about how 829 00:30:04,414 --> 00:30:06,575 something was generated, being able to dive into 830 00:30:06,575 --> 00:30:08,990 these logs and see this does become super 831 00:30:08,990 --> 00:30:10,750 important. And this is one that's out there 832 00:30:10,750 --> 00:30:12,750 today. Like, folks can go play with this 833 00:30:12,750 --> 00:30:14,829 one. Yes. You can see them. Like, they 834 00:30:14,829 --> 00:30:16,269 have a whole list. Again, I could see 835 00:30:16,269 --> 00:30:17,869 it in mine. It's labeled as preview, so 836 00:30:17,869 --> 00:30:20,130 all the preview stuff, it's in preview. 837 00:30:20,444 --> 00:30:22,044 I'd say don't use it in production, but 838 00:30:22,044 --> 00:30:23,644 it's just a preview feature that shows up 839 00:30:23,644 --> 00:30:24,304 in production. 840 00:30:24,684 --> 00:30:28,204 So don't put production workloads with it, but 841 00:30:28,204 --> 00:30:30,284 then a lot of this other stuff that 842 00:30:30,284 --> 00:30:32,684 we talked about, the conditional access, the auditing, 843 00:30:32,684 --> 00:30:33,744 life cycle management, 844 00:30:34,400 --> 00:30:37,440 in this blog article from May 19, it 845 00:30:37,440 --> 00:30:37,940 says, 846 00:30:38,319 --> 00:30:40,400 this is all in the coming months. A 847 00:30:40,400 --> 00:30:41,700 lot of these capabilities 848 00:30:42,159 --> 00:30:42,659 around 849 00:30:42,960 --> 00:30:44,500 zero trust security posture, 850 00:30:45,119 --> 00:30:47,700 all of that is going to be rolling 851 00:30:47,759 --> 00:30:48,740 out to these 852 00:30:49,144 --> 00:30:52,984 agent IDs in Entra. So that's one I'm 853 00:30:52,984 --> 00:30:54,825 looking forward to playing with, keeping an eye 854 00:30:54,825 --> 00:30:56,345 on. What other one you wanna talk about? 855 00:30:56,345 --> 00:30:58,825 We maybe have time. One more announcement. Why 856 00:30:58,825 --> 00:31:01,704 don't we do the Russinovich session? So he 857 00:31:01,704 --> 00:31:02,684 he did a 858 00:31:03,144 --> 00:31:06,950 app build. Mark Russinovich did a session on 859 00:31:07,650 --> 00:31:09,269 inside Azure innovations. 860 00:31:09,730 --> 00:31:10,769 I don't I don't know if you had 861 00:31:10,769 --> 00:31:12,930 a chance to go and watch that one. 862 00:31:12,930 --> 00:31:15,250 I failed on my homework to watch that. 863 00:31:15,250 --> 00:31:15,750 So 864 00:31:16,049 --> 00:31:17,775 I am curious for you to fill me 865 00:31:17,775 --> 00:31:19,055 in on that one. I'll put a link 866 00:31:19,055 --> 00:31:21,215 in the show notes for everybody, but that's, 867 00:31:21,615 --> 00:31:22,674 b r k 868 00:31:23,775 --> 00:31:25,394 one ninety five. 869 00:31:25,855 --> 00:31:28,434 Slides are out there, recordings out there, 870 00:31:28,894 --> 00:31:30,494 all all that good kind of stuff, so 871 00:31:30,494 --> 00:31:32,015 folks can go and take a look at 872 00:31:32,015 --> 00:31:34,710 it. But, effectively, like, this talks about 873 00:31:35,410 --> 00:31:36,150 the internals 874 00:31:36,690 --> 00:31:38,950 of Microsoft and Azure infrastructure. 875 00:31:39,650 --> 00:31:43,670 So what's the latest in boost and DPUs? 876 00:31:44,289 --> 00:31:46,789 What's going on with FPGAs? 877 00:31:47,845 --> 00:31:49,305 How are these things being 878 00:31:49,684 --> 00:31:51,465 leveraged within the 879 00:31:52,005 --> 00:31:53,384 within the Azure fabric. 880 00:31:53,845 --> 00:31:55,465 One of the cool things 881 00:31:56,005 --> 00:31:57,365 that was in there, or at least I 882 00:31:57,365 --> 00:31:59,305 think it was cool more because, like, 883 00:32:00,164 --> 00:32:01,705 some folks on my team were 884 00:32:02,200 --> 00:32:03,580 were were involved in it, 885 00:32:04,600 --> 00:32:07,240 is there's a demo in there where he 886 00:32:07,240 --> 00:32:11,000 talks about blob storage and scaled accounts. And 887 00:32:11,000 --> 00:32:12,600 so one of the things that we've been 888 00:32:12,600 --> 00:32:15,160 doing in storage over the years is trying 889 00:32:15,160 --> 00:32:17,180 to enable these large 890 00:32:17,644 --> 00:32:18,784 AI training companies 891 00:32:19,085 --> 00:32:21,484 who are doing these large pre training runs, 892 00:32:21,484 --> 00:32:24,065 so OpenAI, so the world, things like that. 893 00:32:24,284 --> 00:32:24,784 And 894 00:32:25,325 --> 00:32:28,224 as these customers are bringing in petabytes 895 00:32:28,684 --> 00:32:30,924 and petabytes and petabytes of data, it's not 896 00:32:30,924 --> 00:32:33,819 like single digit, It's it's double to triple 897 00:32:33,819 --> 00:32:34,319 digit 898 00:32:34,619 --> 00:32:36,779 to exabyte scale because you got, like, the 899 00:32:36,779 --> 00:32:39,099 world's knowledge in there. Like, downloading the New 900 00:32:39,099 --> 00:32:41,200 York Times archive is a pretty big thing. 901 00:32:41,339 --> 00:32:43,359 Do that across a bunch of 902 00:32:43,660 --> 00:32:46,160 different sessions that exist out there, and 903 00:32:46,545 --> 00:32:48,725 it it gets quite a bit bigger. So 904 00:32:49,025 --> 00:32:50,085 we did a demo, 905 00:32:50,545 --> 00:32:52,545 and we've been on this journey with scaled 906 00:32:52,545 --> 00:32:55,184 accounts to ever increase the amount of throughput 907 00:32:55,184 --> 00:32:57,684 that we can give them. So ingress, egress, 908 00:32:57,744 --> 00:32:59,265 the amount of IOPS we can give them 909 00:32:59,265 --> 00:33:01,490 in the form of TPS, all that. So 910 00:33:01,490 --> 00:33:03,109 we did a demo where 911 00:33:03,570 --> 00:33:05,890 we were able to, on a single storage 912 00:33:05,890 --> 00:33:06,390 account, 913 00:33:06,769 --> 00:33:07,269 provision 914 00:33:07,570 --> 00:33:08,390 a workload 915 00:33:09,090 --> 00:33:10,150 that ran 916 00:33:10,450 --> 00:33:13,509 at 15 plus terabits a second of ingress, 917 00:33:14,204 --> 00:33:15,005 25 918 00:33:15,005 --> 00:33:18,684 plus terabits a second of egress, and this 919 00:33:18,684 --> 00:33:19,505 is all random 920 00:33:19,964 --> 00:33:21,505 random IO through 921 00:33:21,964 --> 00:33:23,744 through an HPC benchmarking 922 00:33:24,125 --> 00:33:25,105 suite called 923 00:33:25,724 --> 00:33:28,605 IIOR. Okay. We were able to just run 924 00:33:28,605 --> 00:33:29,424 this for hours. 925 00:33:29,910 --> 00:33:31,930 We didn't do this in a special place, 926 00:33:32,150 --> 00:33:33,850 like, just public fleet. 927 00:33:34,470 --> 00:33:36,250 This this is stuff our our customers 928 00:33:37,109 --> 00:33:38,809 can go in and do today. 929 00:33:39,190 --> 00:33:41,350 So so I thought that was a super 930 00:33:41,350 --> 00:33:43,105 cool one. And like I said, important to 931 00:33:43,105 --> 00:33:45,025 me because while my team doesn't work on, 932 00:33:45,025 --> 00:33:46,384 like, the perf and scale stuff, one of 933 00:33:46,384 --> 00:33:48,384 the key components we work on is AI 934 00:33:48,384 --> 00:33:49,365 and ML clients. 935 00:33:49,904 --> 00:33:51,924 So I own our my team, 936 00:33:52,304 --> 00:33:54,565 we we do our SDKs, client tools. 937 00:33:54,865 --> 00:33:57,765 BlobFuse is one of those. So BlobFuse was 938 00:33:57,940 --> 00:33:59,619 heavily used in that demo and ready to 939 00:33:59,619 --> 00:34:01,059 go. So we've done a bunch of work 940 00:34:01,059 --> 00:34:02,500 there just to be able to, like, spin 941 00:34:02,500 --> 00:34:03,160 up workload 942 00:34:03,539 --> 00:34:05,220 in BlobFuse and have it be able to 943 00:34:05,220 --> 00:34:06,519 take, like, full advantage 944 00:34:06,820 --> 00:34:07,320 of 945 00:34:07,700 --> 00:34:10,179 a VM's NIC and soak it all the 946 00:34:10,179 --> 00:34:12,605 way through. So I I would recommend everybody 947 00:34:12,605 --> 00:34:14,945 go watch that session. Like, if you're interested 948 00:34:15,405 --> 00:34:15,885 in the 949 00:34:16,684 --> 00:34:19,405 either the internals of Azure or even a 950 00:34:19,405 --> 00:34:21,405 little bit more about the internals of of 951 00:34:21,405 --> 00:34:24,224 the way some of these workloads compose. So 952 00:34:24,559 --> 00:34:26,400 there was some stuff in there about Linux 953 00:34:26,400 --> 00:34:29,280 Guard and kind of how code integrity is 954 00:34:29,280 --> 00:34:29,780 handled 955 00:34:30,159 --> 00:34:33,059 within Linux, some cool demos for that one. 956 00:34:33,840 --> 00:34:37,300 There was a bunch of stuff around 957 00:34:38,320 --> 00:34:39,219 virtual machines 958 00:34:39,534 --> 00:34:41,954 and kinda how we do host OS upgrades, 959 00:34:42,335 --> 00:34:44,195 hot patch, driver swaps, 960 00:34:44,894 --> 00:34:45,715 full reboots, 961 00:34:46,255 --> 00:34:48,835 and and how those things go within the 962 00:34:49,054 --> 00:34:51,375 within the infrastructure. There was some insights into 963 00:34:51,375 --> 00:34:52,434 how we do RDMA 964 00:34:53,289 --> 00:34:56,009 on the networking side, like, all sorts of 965 00:34:56,009 --> 00:34:58,809 cool stuff. And because it's a Recenavit session, 966 00:34:58,809 --> 00:35:00,889 it's also demo driven, so you actually get 967 00:35:00,889 --> 00:35:02,650 to, like, see it in action. It's not 968 00:35:02,650 --> 00:35:04,650 just a bunch of slideware. Yeah. I'm flipping 969 00:35:04,650 --> 00:35:05,150 through 970 00:35:05,609 --> 00:35:07,130 the slides here, and it looks like there 971 00:35:07,130 --> 00:35:09,394 were, like, 10 or 12 demos in there. 972 00:35:09,394 --> 00:35:11,315 And it is like, I always try to 973 00:35:11,315 --> 00:35:12,614 go watch Russinovich's. 974 00:35:13,235 --> 00:35:15,315 It has been a time constraint more than 975 00:35:15,315 --> 00:35:17,635 lack of desire to watch his yet. But, 976 00:35:17,635 --> 00:35:19,474 yeah, all this stuff, there's some stuff around 977 00:35:19,474 --> 00:35:20,590 that. A confidential 978 00:35:21,289 --> 00:35:21,789 compute, 979 00:35:22,250 --> 00:35:24,349 a bunch of stuff around Azure confidential 980 00:35:24,890 --> 00:35:26,190 GPU VMs. 981 00:35:26,809 --> 00:35:28,030 Yeah. His are always, 982 00:35:28,409 --> 00:35:29,789 like you said, super 983 00:35:30,090 --> 00:35:32,625 demo driven, but I would also say super 984 00:35:32,625 --> 00:35:33,125 nerdy. 985 00:35:33,905 --> 00:35:34,565 Like, for 986 00:35:35,505 --> 00:35:37,664 super nerdy or super technical. Maybe I should 987 00:35:37,664 --> 00:35:40,545 be politically correct. Super technical, like, the amount 988 00:35:40,545 --> 00:35:43,025 of detail that are in his sessions and 989 00:35:43,025 --> 00:35:44,625 the amount of I don't know if he 990 00:35:44,625 --> 00:35:46,390 just sits there and practices all these to 991 00:35:46,390 --> 00:35:48,630 get all these technical details down or if 992 00:35:48,630 --> 00:35:50,789 he just has this much technical knowledge floating 993 00:35:50,789 --> 00:35:53,510 around us in his head. Either way, I 994 00:35:53,510 --> 00:35:55,769 am always super impressed with the 995 00:35:56,230 --> 00:35:59,714 level of technical depth that he's able to 996 00:35:59,714 --> 00:36:02,135 go to in these discussions and these presentations. 997 00:36:02,594 --> 00:36:05,155 So I always find them fascinating. They're great 998 00:36:05,155 --> 00:36:06,054 to go watch. 999 00:36:06,355 --> 00:36:07,255 Mark is 1000 00:36:07,795 --> 00:36:09,414 super deep in the platform 1001 00:36:10,194 --> 00:36:12,434 and and does understand a bunch of this 1002 00:36:12,434 --> 00:36:14,739 stuff. Like, he's not out there just 1003 00:36:15,119 --> 00:36:17,359 spouting it off based on a script that 1004 00:36:17,359 --> 00:36:19,300 I think somebody else wrote for him. So, 1005 00:36:20,000 --> 00:36:21,539 yeah, I I encourage everybody 1006 00:36:22,159 --> 00:36:23,760 to to to go and watch that one. 1007 00:36:23,760 --> 00:36:25,994 It's it's it's worth an hour if you're 1008 00:36:25,994 --> 00:36:28,154 into Azure in general, and you and you're 1009 00:36:28,154 --> 00:36:30,315 kinda looking for what's the latest in hardware 1010 00:36:30,315 --> 00:36:34,255 innovations there across really all the core services, 1011 00:36:34,474 --> 00:36:37,514 so compute, networking, storage. Yep. And if you're 1012 00:36:37,514 --> 00:36:39,179 only interested in one or two of those, 1013 00:36:39,260 --> 00:36:40,859 skip around the session. You can watch it 1014 00:36:40,859 --> 00:36:43,339 on YouTube, and it's got chapters and everything 1015 00:36:43,339 --> 00:36:44,699 in there for you. What is this? The 1016 00:36:44,699 --> 00:36:47,199 world's first analog optical computer 1017 00:36:47,500 --> 00:36:48,239 for accelerating 1018 00:36:48,539 --> 00:36:49,039 AI 1019 00:36:49,420 --> 00:36:49,920 inference. 1020 00:36:50,539 --> 00:36:53,179 Potentially a hundred times more energy efficient than 1021 00:36:53,179 --> 00:36:53,679 GPUs. 1022 00:36:54,094 --> 00:36:56,114 If that is at all intriguing to you 1023 00:36:56,414 --> 00:36:58,994 about light doing massive parallel computations 1024 00:36:59,375 --> 00:37:01,855 for AI, go watch the end of this 1025 00:37:01,855 --> 00:37:03,454 session. Right? Go watch the end of it. 1026 00:37:03,454 --> 00:37:05,554 So power consumption is important. Sustainability 1027 00:37:05,934 --> 00:37:08,434 is important. Yep. I saw and and 1028 00:37:08,980 --> 00:37:11,059 don't quote me exactly again on the percentage, 1029 00:37:11,059 --> 00:37:13,880 but it's something like three to 4% 1030 00:37:14,099 --> 00:37:16,260 of the world's power now is going to 1031 00:37:16,260 --> 00:37:17,000 data centers, 1032 00:37:17,300 --> 00:37:19,719 and it's been driven on this general uptick 1033 00:37:19,940 --> 00:37:20,440 around 1034 00:37:21,059 --> 00:37:23,140 more and more CPU and particularly more and 1035 00:37:23,140 --> 00:37:25,985 more GPU usage because these things are so 1036 00:37:25,985 --> 00:37:27,445 power hungry. So 1037 00:37:27,744 --> 00:37:29,825 I I think it is important that all 1038 00:37:29,825 --> 00:37:31,844 these companies that are working towards this stuff 1039 00:37:31,985 --> 00:37:34,005 continue to strive towards sustainability 1040 00:37:34,625 --> 00:37:35,525 and innovations 1041 00:37:36,144 --> 00:37:36,644 in 1042 00:37:37,664 --> 00:37:40,769 not only, like, capping power consumption, like, at 1043 00:37:40,769 --> 00:37:42,470 some point, it can't keep growing forever, 1044 00:37:42,849 --> 00:37:45,250 but also bringing it back down. So things 1045 00:37:45,250 --> 00:37:46,309 like like analog 1046 00:37:46,849 --> 00:37:50,150 optical computer AOC, not not the other AOC, 1047 00:37:50,369 --> 00:37:51,030 this AOC 1048 00:37:51,465 --> 00:37:53,545 fall into that category of let's make the 1049 00:37:53,545 --> 00:37:55,465 world a better place. Very cool. Well, those 1050 00:37:55,465 --> 00:37:56,845 are some fun announcements, 1051 00:37:57,385 --> 00:37:59,305 interesting news. We spent, like, the whole time 1052 00:37:59,305 --> 00:38:01,305 in, like, three news articles. You should see 1053 00:38:01,305 --> 00:38:02,585 the number of blog posts I have to 1054 00:38:02,585 --> 00:38:04,125 go through. I think we went through 1055 00:38:04,739 --> 00:38:06,739 a a couple hundred of them that are 1056 00:38:06,739 --> 00:38:08,820 out there. So the other thing that I'll 1057 00:38:08,820 --> 00:38:10,679 put in the show notes for folks, 1058 00:38:11,059 --> 00:38:12,440 the the book of news 1059 00:38:13,380 --> 00:38:15,380 is out and available as well for that, 1060 00:38:15,380 --> 00:38:17,944 and and that's publicly available. So if you're 1061 00:38:17,944 --> 00:38:19,804 interested in maybe the more 1062 00:38:20,264 --> 00:38:23,625 broad roundup of, hey, what were the bigger 1063 00:38:23,625 --> 00:38:24,125 announcements 1064 00:38:24,505 --> 00:38:27,224 at Microsoft Build, things like that, I would 1065 00:38:27,224 --> 00:38:30,125 encourage folks to go out and 1066 00:38:30,585 --> 00:38:32,400 take a look at the book of news. 1067 00:38:32,799 --> 00:38:33,299 Maybe 1068 00:38:33,599 --> 00:38:36,319 throw it into your favorite LLM and have 1069 00:38:36,319 --> 00:38:38,480 it summarize it for you. Absolutely. And if 1070 00:38:38,480 --> 00:38:39,380 you have any questions, 1071 00:38:39,679 --> 00:38:41,619 topics you want us to talk about, 1072 00:38:42,000 --> 00:38:43,519 news you want us to cover in future 1073 00:38:43,519 --> 00:38:45,775 episodes as well, let us know. I would 1074 00:38:45,775 --> 00:38:48,014 say either if you're part of the membership, 1075 00:38:48,014 --> 00:38:49,954 let us know in Discord. If you're not, 1076 00:38:50,014 --> 00:38:50,514 LinkedIn, 1077 00:38:51,375 --> 00:38:54,514 I'm quickly leaning towards as being the social 1078 00:38:54,815 --> 00:38:57,954 media social media probably social media, social network 1079 00:38:58,094 --> 00:39:00,239 that I am spending the most time on. 1080 00:39:00,239 --> 00:39:01,219 All the other ones 1081 00:39:01,840 --> 00:39:04,099 Yeah. I feel very hit or miss. LinkedIn 1082 00:39:04,159 --> 00:39:06,079 is probably the best place to get both 1083 00:39:06,079 --> 00:39:08,099 Scott and I and get our attention, 1084 00:39:08,880 --> 00:39:11,280 if you have topics or questions about the 1085 00:39:11,280 --> 00:39:13,214 show. You can tag the podcast there. You 1086 00:39:13,214 --> 00:39:14,734 can also go to the website and just 1087 00:39:14,734 --> 00:39:15,795 use our contact 1088 00:39:16,335 --> 00:39:16,835 contact 1089 00:39:17,454 --> 00:39:18,594 contact us form. 1090 00:39:18,894 --> 00:39:21,315 It's alright. It's Friday. It's time to 1091 00:39:21,614 --> 00:39:23,614 call it a day at 11:00 in the 1092 00:39:23,614 --> 00:39:26,099 morning and go relax for the rest of 1093 00:39:26,099 --> 00:39:28,099 the day or go to meetings. I'll probably 1094 00:39:28,099 --> 00:39:30,500 do the former. But as always, thanks, Ben. 1095 00:39:30,500 --> 00:39:32,980 Appreciate the conversation. Thank you, and enjoy your 1096 00:39:32,980 --> 00:39:34,679 weekend. We'll talk to you next time. Yep. 1097 00:39:36,500 --> 00:39:38,739 If you enjoyed the podcast, go leave us 1098 00:39:38,739 --> 00:39:40,934 a five star rating in iTunes. It helps 1099 00:39:40,934 --> 00:39:42,695 to get the word out so more IT 1100 00:39:42,695 --> 00:39:44,855 pros can learn about Office three sixty five 1101 00:39:44,855 --> 00:39:45,515 and Azure. 1102 00:39:46,055 --> 00:39:47,735 If you have any questions you want us 1103 00:39:47,735 --> 00:39:49,894 to address on the show, or feedback about 1104 00:39:49,894 --> 00:39:52,215 the show, feel free to reach out via 1105 00:39:52,215 --> 00:39:54,429 our website, Twitter, or Facebook. 1106 00:39:54,829 --> 00:39:56,588 Thanks again for listening, and have a great 1107 00:39:56,588 --> 00:39:57,088 day.