1 00:00:00,040 --> 00:00:02,480 Do you think you you ranked higher than Sierra on this on 2 00:00:02,480 --> 00:00:05,240 this list which was interesting? Yeah, yeah. 3 00:00:05,240 --> 00:00:07,480 We continue destroying the competitors, you know. 4 00:00:07,880 --> 00:00:10,080 Massive, massive spike, especially during that one 5 00:00:10,080 --> 00:00:11,440 weekend. I, I'll never forget. 6 00:00:11,920 --> 00:00:13,480 Like our servers were having issues. 7 00:00:13,480 --> 00:00:15,440 Everyone was questioning like where all these numbers were 8 00:00:15,440 --> 00:00:18,160 coming from and it was just from this one deployment. 9 00:00:18,160 --> 00:00:20,200 It wasn't even called open claw at a time. 10 00:00:20,200 --> 00:00:22,840 I didn't even realize what the heck it was. 11 00:00:22,960 --> 00:00:26,720 In the newsletter today we are publishing with Wing, the 12 00:00:26,720 --> 00:00:31,240 venture capital firm, a list of breakout and prize technology 13 00:00:31,240 --> 00:00:35,080 companies, the ET30. With Wing, we surveyed a bunch 14 00:00:35,080 --> 00:00:38,560 of top venture capitalists to figure out what early stage, mid 15 00:00:38,560 --> 00:00:42,680 stage, late stage, and giga stage companies they are most 16 00:00:42,680 --> 00:00:45,240 excited about. And the number one on the early 17 00:00:45,240 --> 00:00:49,040 stage list I'd never heard of, which is always a good sign that 18 00:00:49,040 --> 00:00:50,280 you're, you're learning new things. 19 00:00:50,280 --> 00:00:54,480 So that is Mintilify. And so we are going to have Han 20 00:00:54,480 --> 00:00:57,280 Wang, the CEO of Mintilify, explain what the hell he's 21 00:00:57,280 --> 00:00:59,360 doing. It's got venture capitalists so 22 00:00:59,360 --> 00:01:01,760 excited. Then in the second-half, we will 23 00:01:01,760 --> 00:01:06,880 have the CEO Jesse Zhang of Decagon, which is #4 on the late 24 00:01:06,880 --> 00:01:09,600 stage list. Now, to Jesse's credit, the 25 00:01:09,600 --> 00:01:12,200 light stage list is, is a heavy category. 26 00:01:12,920 --> 00:01:15,560 It's the breakout one. I mean, it's, it's, you know, 27 00:01:15,560 --> 00:01:18,600 it's kudos to you. But obviously number one is 11 28 00:01:18,600 --> 00:01:21,240 Labs. Number 2 is versal #3 is open 29 00:01:21,240 --> 00:01:25,880 evidence #4 is decagon #5 is glean #6 is Sierra Decagon's 30 00:01:25,880 --> 00:01:28,200 rival. So kudos to Decagon for being 31 00:01:28,200 --> 00:01:30,480 ahead of your rival. The Giga stage lists are the 32 00:01:30,480 --> 00:01:34,600 names you'll recognize Anthropic data bricks, SpaceX, Open AI and 33 00:01:34,600 --> 00:01:35,720 roll. So you can go to 34 00:01:35,720 --> 00:01:38,400 thenewsletternewcomer.co. We'll publish all the lists. 35 00:01:38,400 --> 00:01:40,400 We're going to throw some up here on YouTube. 36 00:01:40,480 --> 00:01:43,360 Thanks to Wing for conducting the survey and working with us 37 00:01:43,360 --> 00:01:45,400 on it. First up, we have Han from 38 00:01:45,400 --> 00:01:55,840 Mintlify. I really like knew that I wanted 39 00:01:55,840 --> 00:01:57,720 to start a company with my now Co founder. 40 00:01:57,720 --> 00:02:01,920 Hanby started a company wanting to tackle a problem that we can 41 00:02:01,920 --> 00:02:04,840 both relate to and so we picked a space that was so crucial, 42 00:02:04,920 --> 00:02:07,440 crucial to us. It was about enabling and 43 00:02:07,440 --> 00:02:10,520 empowering developers, as simple as that, right? 44 00:02:10,919 --> 00:02:13,920 That was the anchor in which we knew we must had because we were 45 00:02:13,920 --> 00:02:17,200 like, look, it's gonna take a decade plus to go build anything 46 00:02:17,200 --> 00:02:20,680 significant, right? Like building a company is not a 47 00:02:20,680 --> 00:02:22,840 thing you do overnight. Right, you need to be pretty 48 00:02:22,840 --> 00:02:24,800 committed to the space. You have to be. 49 00:02:25,320 --> 00:02:27,520 And so we were like, OK, what's that space look like? 50 00:02:27,520 --> 00:02:29,760 And to us, it was about enabling other developers a problem that 51 00:02:29,760 --> 00:02:31,200 we could deeply relate to. It's like. 52 00:02:31,200 --> 00:02:33,720 So if we even did, let's worst case scenario. 53 00:02:33,720 --> 00:02:36,960 Let's say we pivoted 8 times and you know like and spent a. 54 00:02:37,000 --> 00:02:39,560 Were they, they were like big pivots or yeah. 55 00:02:39,640 --> 00:02:41,840 Did you leave categories or? I didn't that's. 56 00:02:41,840 --> 00:02:44,280 The thing it was always develop serving developers. 57 00:02:44,280 --> 00:02:47,160 Exactly. And so the first application 58 00:02:47,160 --> 00:02:48,520 was. What year was that? 59 00:02:48,520 --> 00:02:51,720 When did it start? So this the first initial 60 00:02:51,720 --> 00:02:54,760 iteration that eventually let us on this path started at the end 61 00:02:54,760 --> 00:02:58,000 of 2020, 21. And then we didn't really land 62 00:02:58,000 --> 00:03:01,080 on to what we're doing now with Milefi until I would say like 63 00:03:01,120 --> 00:03:04,480 end of 22, like start at 23, depending on how you look at it. 64 00:03:05,000 --> 00:03:06,960 All right, so explain what the company does today. 65 00:03:07,040 --> 00:03:10,480 Yes. So we help companies build their 66 00:03:10,760 --> 00:03:13,840 like knowledge base, developer platform, source of truth. 67 00:03:14,360 --> 00:03:16,680 We call ourselves the intelligent knowledge platform 68 00:03:16,680 --> 00:03:17,760 and the knowledge infrastructure. 69 00:03:18,520 --> 00:03:21,800 So if you've been, for instance, to cloud code stocks. 70 00:03:22,480 --> 00:03:24,120 Right, this is the documentation. 71 00:03:24,120 --> 00:03:25,760 Space. This is the documentation space, 72 00:03:27,080 --> 00:03:30,080 though it increasingly so. Millified does a lot more than 73 00:03:30,080 --> 00:03:32,880 just docs. So if you've been, for instance, 74 00:03:32,880 --> 00:03:36,200 reading up on how Lovable works, they're Lovable guides. 75 00:03:36,600 --> 00:03:38,960 Those are all powered by Millify if you've been to the open claw 76 00:03:39,600 --> 00:03:41,880 docs or health centers, right? They're also on there. 77 00:03:41,880 --> 00:03:44,240 Now obviously we're going to get into are these docs for humans 78 00:03:44,240 --> 00:03:47,000 or are they for agents? So, but we won't hit that 79 00:03:47,520 --> 00:03:50,520 immediately, but it's sort of if you're building, if I'm a 80 00:03:50,520 --> 00:03:53,720 company like Stripe or something building a big API, I wanna sort 81 00:03:53,720 --> 00:03:57,040 of explainer out there why we made the decisions we made and 82 00:03:57,040 --> 00:03:59,120 how to interact with it and how to get the most out of it. 83 00:03:59,120 --> 00:04:01,040 Is that yes the right way to explain it? 84 00:04:01,160 --> 00:04:05,120 Exactly the analogy I always like to say when people ask me 85 00:04:05,120 --> 00:04:08,600 what Milefios is. When's the last time you 86 00:04:08,600 --> 00:04:12,560 assembled IKEA furniture? I like swear it off. 87 00:04:12,560 --> 00:04:16,560 I, I remember I was like, you were supposed to go to a party. 88 00:04:16,560 --> 00:04:19,000 It was literally like Kara Swisher was having, I think like 89 00:04:19,000 --> 00:04:22,040 a book party or something. And I started assembling it with 90 00:04:22,040 --> 00:04:24,720 my roommate who is like kind enough to get wrapped into it. 91 00:04:24,720 --> 00:04:27,000 And I, you know, I thought I was going to go to some party at, 92 00:04:27,000 --> 00:04:31,160 like, 10:00 PM, It was like 3:00 AM before we'd ever, you know, 93 00:04:31,320 --> 00:04:34,400 like, finish the furniture. And so it's like, I pretty much 94 00:04:34,400 --> 00:04:35,920 swore it off. Then I've done, you know, 95 00:04:35,960 --> 00:04:37,760 simpler stuff. But yeah, OK. 96 00:04:38,040 --> 00:04:40,320 Yes. So some of that happened, right? 97 00:04:40,440 --> 00:04:42,960 Yeah. So the the analogy I like to 98 00:04:42,960 --> 00:04:46,640 give is like Milify is like building the instruction manual 99 00:04:46,640 --> 00:04:50,920 for the IKEA furniture or maybe put a different way, the 100 00:04:50,920 --> 00:04:54,080 assembly manual for Lego set, Lego sets and Lego, right? 101 00:04:54,480 --> 00:04:55,960 I have done that more recently. There you go. 102 00:04:55,960 --> 00:04:57,080 Yeah. Nice. 103 00:04:57,760 --> 00:05:00,880 I got a Batmobile for Christmas, which was a totally random gift, 104 00:05:00,880 --> 00:05:02,920 but it was actually really fun to do. 105 00:05:03,000 --> 00:05:05,080 Yeah, 100%. That's you. 106 00:05:05,080 --> 00:05:06,840 Have the little bags. One thing that made it so much 107 00:05:06,840 --> 00:05:09,240 easier, which I had sort of forgotten, is like, it's so 108 00:05:09,240 --> 00:05:12,000 staged out, you know? Yes, so you don't get 109 00:05:12,000 --> 00:05:14,200 overwhelmed all at once. Exactly. 110 00:05:14,200 --> 00:05:15,520 No, they're they're methodical with it. 111 00:05:15,840 --> 00:05:19,840 You know, it's funny because like I've even seen like they've 112 00:05:19,840 --> 00:05:23,360 gotten rid of the actual like hand printed like, well, they 113 00:05:23,400 --> 00:05:26,600 they do have them, but you can literally scan some QR code on 114 00:05:26,600 --> 00:05:29,680 the thing now and there's like a 3D version of it you can like 115 00:05:29,680 --> 00:05:31,720 open on your phone and like it could literally like. 116 00:05:31,720 --> 00:05:32,600 Piece. Oh, I haven't done that. 117 00:05:32,680 --> 00:05:34,840 Interesting. It's the coolest thing, but I 118 00:05:34,880 --> 00:05:37,600 digress to say that we effectively are building the 119 00:05:37,600 --> 00:05:40,400 assembly kit, the instruction manuals, if you will, for the 120 00:05:40,400 --> 00:05:44,200 Lego, you know, the Lego sets. And the reason why that's 121 00:05:44,200 --> 00:05:47,560 important is because, well, I mean, just try, imagine doing a 122 00:05:47,560 --> 00:05:50,560 Lego set without it, right? Because otherwise it's just 123 00:05:50,560 --> 00:05:52,240 really a bunch of plastic blocks. 124 00:05:53,400 --> 00:05:56,560 And the reality is for the vast majority of products and 125 00:05:56,560 --> 00:06:00,920 services is whether you know it is for developers or not, 126 00:06:02,120 --> 00:06:06,440 there's, you know, a need to explain how to use the thing in 127 00:06:06,440 --> 00:06:09,520 order to make use of it, in order to actually go in and, and 128 00:06:09,520 --> 00:06:12,680 actually, you know, piece things together, build it together to 129 00:06:12,680 --> 00:06:16,400 kind of get this like Batmobile. So is this vibe coding your 130 00:06:16,400 --> 00:06:18,680 docs? It's like you're sort of coding 131 00:06:18,680 --> 00:06:22,640 with cursor or something, and as that's happening sort of on the 132 00:06:22,640 --> 00:06:24,880 side, your documentation is changing. 133 00:06:24,960 --> 00:06:28,400 In an essence, yes. What we did in the beginning was 134 00:06:28,720 --> 00:06:32,640 when Hanvi and I decided to start Nilify, we were like, 135 00:06:32,640 --> 00:06:36,280 look, we spent our entire lives reading some really, really 136 00:06:36,280 --> 00:06:39,440 shitty docs, right? Like, you know, implemented some 137 00:06:40,160 --> 00:06:42,880 some things the hard way because there's no one, no developers 138 00:06:42,880 --> 00:06:45,800 has gone into actually clearly put thought into explaining how 139 00:06:45,800 --> 00:06:47,720 it works. And so everything just felt like 140 00:06:47,720 --> 00:06:50,520 you were trying to piece together like Lego blocks just 141 00:06:50,520 --> 00:06:54,160 together and manually without really the thought of how to 142 00:06:54,160 --> 00:06:56,320 actually piece things together. So we had to figure that out. 143 00:06:57,080 --> 00:07:01,640 And so we're like, look, let's just build this docs platform 144 00:07:01,640 --> 00:07:06,240 that was the one of Millify to the simplest ways in which we 145 00:07:06,240 --> 00:07:10,320 could have developers engage with it, contribute to it, and 146 00:07:10,320 --> 00:07:12,720 Bill with it. And that first version was just 147 00:07:12,720 --> 00:07:16,720 like, let's just give people Markdown, which is this language 148 00:07:16,720 --> 00:07:20,120 that obviously developers really prefer and work with and make it 149 00:07:20,120 --> 00:07:21,440 super easy for them to work with. 150 00:07:22,080 --> 00:07:25,240 And let's just let it RIP and see what happens. 151 00:07:26,080 --> 00:07:30,560 Developers, founders, companies of all stages and sizes, in the 152 00:07:30,560 --> 00:07:32,920 end, this is not 23. They just loved it. 153 00:07:33,680 --> 00:07:37,280 Do you think, like, I would think a sort of era of vibe 154 00:07:37,280 --> 00:07:40,200 coding would be terrible for documentation, or it's just like 155 00:07:40,200 --> 00:07:42,880 if you have people, sort of, some of them not even coders, 156 00:07:43,120 --> 00:07:45,960 sort of just throwing shit at the wall and trying stuff, are 157 00:07:45,960 --> 00:07:48,240 they really gonna be so buttoned up that they're like, and we 158 00:07:48,240 --> 00:07:50,600 want to write the guide to it? I barely understand how the code 159 00:07:50,600 --> 00:07:53,040 works. Like, is that intuition wrong? 160 00:07:53,080 --> 00:07:55,080 Yeah. It's actually, oddly enough, 161 00:07:55,080 --> 00:07:57,800 more important. I would actually say that if if 162 00:07:57,800 --> 00:08:02,880 not for kind of like the, the, the tailwind of AI and, and, and 163 00:08:02,880 --> 00:08:05,120 vibe coding in general, I wouldn't even say where, where 164 00:08:05,120 --> 00:08:08,200 we are today. And the reason for that is 165 00:08:08,200 --> 00:08:11,040 because I go back to the assembly that the, you know, kit 166 00:08:11,040 --> 00:08:16,080 instruction or the, sorry, the Lego kit assembly manual is 167 00:08:16,240 --> 00:08:22,160 let's imagine you're now asking an AI to go ahead and assemble, 168 00:08:22,880 --> 00:08:24,280 you know, like said Lego assembly kit. 169 00:08:24,280 --> 00:08:27,200 Let's say there's a robot, an AI robot here and it wants to go 170 00:08:27,200 --> 00:08:30,240 and assemble the kit. Well, the most important thing 171 00:08:30,240 --> 00:08:34,760 it needs to do to actually go and figure out if it could or 172 00:08:34,880 --> 00:08:38,120 could not assemble it or how to assemble it is actually to read 173 00:08:38,120 --> 00:08:40,760 the guides. And in the same way that humans 174 00:08:40,760 --> 00:08:45,280 do, except in our case and what we've seen, it's even more 175 00:08:45,280 --> 00:08:49,440 important because the docs, the knowledge base, those things 176 00:08:49,440 --> 00:08:52,560 that are traditionally seen as very like boring, are 177 00:08:52,560 --> 00:08:54,680 coincidentally very information dense. 178 00:08:55,240 --> 00:08:59,080 And that's where typically AILMS actually get all their source of 179 00:08:59,080 --> 00:09:02,480 truth about what you are, what you do, how you do it, how your 180 00:09:02,480 --> 00:09:05,680 thing works, and so forth. It doesn't really look at the 181 00:09:05,680 --> 00:09:08,480 marketing fluff, right, Right. That's on your landing page, 182 00:09:08,480 --> 00:09:10,640 which by nature is designed for humans. 183 00:09:11,280 --> 00:09:14,240 It's like, here's the flashy words, the flashy colors, the, 184 00:09:14,480 --> 00:09:16,520 you know, the all the nice things that kind of get a person 185 00:09:16,520 --> 00:09:18,520 through the door to pick your product to learn how to use it. 186 00:09:19,120 --> 00:09:21,920 It's like it needs to know the truth of what your thing 187 00:09:21,920 --> 00:09:24,160 actually does. And all of that's, you know, 188 00:09:24,160 --> 00:09:27,880 just like 0 fluff tolerant is all living in what is 189 00:09:27,880 --> 00:09:30,960 traditionally docks. And so now if for instance, you 190 00:09:30,960 --> 00:09:33,640 don't have your docks, let's just take a example of that, 191 00:09:33,640 --> 00:09:35,880 right? Like, imagine if you're striping 192 00:09:35,880 --> 00:09:38,640 your docks just completely disappear for a day, right? 193 00:09:39,440 --> 00:09:44,480 Well, the first thing to note is that, well, well, first of all, 194 00:09:44,480 --> 00:09:46,680 no one's going to learn how to use your product, for starters. 195 00:09:46,680 --> 00:09:48,720 So good luck on boarding developers that way. 196 00:09:49,680 --> 00:09:53,040 But more importantly than that, especially today, well, LLMS 197 00:09:53,040 --> 00:09:54,680 aren't going to know how to build your product either. 198 00:09:55,120 --> 00:09:59,000 And if even today, the vast majority of software is written 199 00:09:59,000 --> 00:10:01,840 by AI, and if it's not already, it's, you know, obviously going 200 00:10:01,840 --> 00:10:04,840 to be, how are they going to go know how to do it too? 201 00:10:05,640 --> 00:10:08,880 They don't. To what extent are mintlified 202 00:10:08,880 --> 00:10:13,040 docs for language models and agents today? 203 00:10:13,040 --> 00:10:16,280 Like what do you think is the percent breakdown of humans 204 00:10:16,280 --> 00:10:19,600 versus AI consuming it? It's about 5050 right now. 205 00:10:20,000 --> 00:10:24,200 And do you have like do you see them as the same doc or you 206 00:10:24,200 --> 00:10:27,600 think they will diverge this for human and AI doc? 207 00:10:27,800 --> 00:10:32,120 Good question. So we actually work with, well, 208 00:10:32,120 --> 00:10:33,880 the Clod Co team and the anthropic teams. 209 00:10:33,880 --> 00:10:36,520 And, and funny enough, about a year and a half ago I had the 210 00:10:36,520 --> 00:10:39,160 same question because we kind of saw the writing on the wall 211 00:10:39,400 --> 00:10:41,400 ourselves. We're like, look, the role of 212 00:10:41,400 --> 00:10:43,640 content, of the role of knowledge, the role of docs, all 213 00:10:43,640 --> 00:10:46,520 that stuff, whatever you call it is going to be fundamentally 214 00:10:46,520 --> 00:10:48,360 more for AI than it is for humans. 215 00:10:48,720 --> 00:10:51,320 And back then this was just like, you know, like AI was 216 00:10:51,320 --> 00:10:53,520 really taking off. It wasn't like Cloud Code became 217 00:10:53,520 --> 00:10:57,040 everyone's, you know, you know, our hourly active use product. 218 00:10:58,800 --> 00:11:00,960 We were like, look, we just see this being the case in the 219 00:11:00,960 --> 00:11:02,560 future. And so we had a conversation 220 00:11:02,560 --> 00:11:04,960 with the client, the anthropic team, and we're like, look, 221 00:11:05,200 --> 00:11:08,680 should the docs or the content for humans be the same for AI? 222 00:11:08,720 --> 00:11:12,080 It's a common question we get asked a lot, and the answer is 223 00:11:12,600 --> 00:11:18,960 yes, because the reality is the LLMS are also instructed to read 224 00:11:18,960 --> 00:11:20,000 things the way humans are, right? 225 00:11:20,400 --> 00:11:21,560 So. They've been trained on human 226 00:11:21,560 --> 00:11:22,680 writing, so they're pretty used to it. 227 00:11:22,680 --> 00:11:24,200 They're pretty good. With it, some would say, you 228 00:11:24,200 --> 00:11:25,520 know, they're pretty good at writing it too. 229 00:11:25,960 --> 00:11:27,880 And so I think there's a lot of people over thinking a little 230 00:11:27,880 --> 00:11:30,240 bit where it's like, oh, like, let me format it in this way, 231 00:11:30,240 --> 00:11:33,280 like don't don't bother really. Like, you know, write it in the 232 00:11:33,600 --> 00:11:36,080 the best way you possibly could to if you were to explain to a 233 00:11:36,080 --> 00:11:39,960 human good docs is good docs. LLMS are going to ingest that 234 00:11:40,000 --> 00:11:41,040 and know how to do that from there. 235 00:11:41,480 --> 00:11:44,240 Why? Why can't an LLM just read the 236 00:11:44,240 --> 00:11:48,560 code, sort of create it's own perception of what the docs 237 00:11:48,560 --> 00:11:50,440 should be, and just operate off that? 238 00:11:50,600 --> 00:11:52,080 Yeah, that's a really good question, by the way. 239 00:11:53,240 --> 00:11:55,320 So there's two different reasons. 240 00:11:55,840 --> 00:12:00,640 The first one is the reality, which is that good docs and 241 00:12:00,640 --> 00:12:04,200 stuff that's actually useful, like the content that actually 242 00:12:04,200 --> 00:12:08,320 should be in docs don't describe exactly what the code does, 243 00:12:09,120 --> 00:12:10,880 right? And I think that's the same kind 244 00:12:10,880 --> 00:12:14,960 of like idea of, you know, let's say you're like here furniture, 245 00:12:15,280 --> 00:12:17,600 you know, you can kind of just glue together some plywood and 246 00:12:17,600 --> 00:12:20,120 some, you know, nails and it can come up in a shape like this. 247 00:12:20,120 --> 00:12:22,200 But is it a cabinet? Is it a shelf? 248 00:12:22,360 --> 00:12:23,960 Is it designed for a decoration piece? 249 00:12:23,960 --> 00:12:25,920 Is it designed to be used in this or that way? 250 00:12:25,920 --> 00:12:27,120 Is it a tool? Is it here or there? 251 00:12:27,720 --> 00:12:31,200 That's typically the information that is just more contextual 252 00:12:31,200 --> 00:12:35,920 that adds on top of what is an existing pile of pieces, right? 253 00:12:35,920 --> 00:12:38,640 Same thing goes for like you know, products and content. 254 00:12:38,640 --> 00:12:43,880 If you really take just code and you spin up docs for that, 255 00:12:45,440 --> 00:12:48,520 granted we do that at Milify, for the record, we get a lot of 256 00:12:48,520 --> 00:12:53,920 traffic and usage out of it. My opinion is that that's 257 00:12:53,920 --> 00:12:56,760 typically not additive information that's tremendously 258 00:12:56,760 --> 00:12:58,200 useful. Now there are cases of which 259 00:12:58,200 --> 00:13:01,760 there are install guides, SDKSAPI references. 260 00:13:01,760 --> 00:13:05,200 It's like the very tactical glued to the the code type of 261 00:13:05,320 --> 00:13:09,040 use cases. But on generally, very 262 00:13:09,040 --> 00:13:12,440 comprehensive information should extend far beyond that, and for 263 00:13:12,440 --> 00:13:15,320 an LLM to truly understand how your product works, it should 264 00:13:15,320 --> 00:13:18,800 take a look at the code on top of the contextual information. 265 00:13:18,800 --> 00:13:22,240 Is if, if the docs sort of give you a sign of how the company 266 00:13:22,240 --> 00:13:25,120 thinks it should be used, the provider and you have a sort of 267 00:13:25,120 --> 00:13:27,480 sense of they'll be supportive if we're using it this way or 268 00:13:27,480 --> 00:13:30,160 we're sort of using it if they're building the product in 269 00:13:30,160 --> 00:13:34,320 this direction in the future and sort of where the the provider 270 00:13:34,320 --> 00:13:35,920 is like leaned in. Exactly. 271 00:13:36,280 --> 00:13:41,800 Or even things like how does the like what's on the road map, 272 00:13:42,360 --> 00:13:44,760 what's been tried and tested, what's changed? 273 00:13:45,160 --> 00:13:47,360 Right. These things don't necessarily 274 00:13:47,360 --> 00:13:49,680 always come directly one to one with the code base. 275 00:13:51,240 --> 00:13:53,400 But you were saying you have a lot of customers who are like I, 276 00:13:53,680 --> 00:13:56,680 I want the fast, easy, I don't think about it sort of version 277 00:13:56,840 --> 00:13:59,120 we do what what is sort of the utility of that? 278 00:13:59,160 --> 00:14:01,440 Yeah, well, it's just getting zero to 1, right. 279 00:14:01,520 --> 00:14:05,200 So for instance, again, like the the reality of of, you know, of 280 00:14:05,200 --> 00:14:07,960 managing docs, managing content is well, no one really. 281 00:14:08,360 --> 00:14:10,960 People don't really love updating docs, right? 282 00:14:11,200 --> 00:14:13,440 Like especially just getting, putting myself in the 283 00:14:13,440 --> 00:14:17,640 perspective of an engineer, you know, having been 1 and I'm 284 00:14:17,640 --> 00:14:21,840 still 1 to this day, it's like, look, I am not the best writer. 285 00:14:22,560 --> 00:14:24,800 I am not trained on that profession. 286 00:14:24,800 --> 00:14:27,560 I am just inherently like, you know, someone who wants to go 287 00:14:27,560 --> 00:14:29,960 and tinker with, you know, writing code, writing product, 288 00:14:29,960 --> 00:14:33,000 shipping things. And docs often times become an 289 00:14:33,000 --> 00:14:37,600 afterthought. And so historically, like people 290 00:14:37,600 --> 00:14:39,560 just don't have anything in the 1st place. 291 00:14:39,560 --> 00:14:41,640 And then your users complain. They're like, oh, I don't know 292 00:14:41,640 --> 00:14:44,000 how to use your pocket. And you're like, well I don't 293 00:14:44,000 --> 00:14:45,480 know, go figure it out. Read the code base. 294 00:14:45,480 --> 00:14:46,760 And they're like, what are you talking about? 295 00:14:46,760 --> 00:14:49,440 Like what the actual fuck? It's like I'm the customer. 296 00:14:49,520 --> 00:14:51,600 Yeah, it's like that's the most insulting thing ever, you know? 297 00:14:52,280 --> 00:14:54,600 And so some of the stuff that we build is helping people get 0 to 298 00:14:54,600 --> 00:14:56,560 1 and then. But I think what's more 299 00:14:56,560 --> 00:14:59,520 meaningful than getting zero to 1 is actually making sure the 300 00:14:59,520 --> 00:15:03,680 content is up to date and accurate and what we call self 301 00:15:03,680 --> 00:15:06,640 healing and self updating. The reason for that. 302 00:15:06,640 --> 00:15:09,560 And again, goes ties back into the point about, you know, what 303 00:15:09,560 --> 00:15:11,680 you mentioned on like what percentage of docs now for 304 00:15:11,680 --> 00:15:15,280 humans and AI? Well, like right now it's 5050. 305 00:15:15,840 --> 00:15:19,600 It was 15% AI. It was 15% of all traffic at the 306 00:15:19,600 --> 00:15:22,120 start of 2025. So it went from 15. 307 00:15:22,120 --> 00:15:23,520 It's clear which way we're going here. 308 00:15:23,520 --> 00:15:25,040 Yeah, exactly. So do. 309 00:15:25,040 --> 00:15:27,360 You have a guess for the end of 2026. 9010. 310 00:15:27,520 --> 00:15:28,280 Really. Already? 311 00:15:28,320 --> 00:15:29,040 Oh, wow. OK. 312 00:15:29,360 --> 00:15:31,720 And the thing to note is not necessarily because it's like a 313 00:15:31,720 --> 00:15:37,240 lot less people are going to be reading docs, maybe in the same 314 00:15:37,240 --> 00:15:40,760 way like the total percentage of people are reading like hardback 315 00:15:40,760 --> 00:15:43,240 books these days versus, I don't know, listening to an audio book 316 00:15:43,240 --> 00:15:45,000 or on the Internet. You know, it's certainly going 317 00:15:45,000 --> 00:15:48,680 to have an impact, but it's just because the sheer volume of 318 00:15:48,840 --> 00:15:51,760 knowledge consumption because of just automated knowledge work is 319 00:15:51,760 --> 00:15:53,120 going to be happening with agents. 320 00:15:53,480 --> 00:15:55,480 The greater piece of pie is going to be like 9010. 321 00:15:55,600 --> 00:15:57,960 To what extent do you think docs for humans are just going to be 322 00:15:57,960 --> 00:15:59,240 like, what the hell did I just build? 323 00:15:59,600 --> 00:16:02,840 Or it's like it's a guide to this sort of thing that humans 324 00:16:02,840 --> 00:16:04,520 play very little part in creating. 325 00:16:06,080 --> 00:16:08,560 In terms of like, oh, humans didn't really create. 326 00:16:08,560 --> 00:16:11,640 Yeah, it just that it was like if code is mostly machine built 327 00:16:11,920 --> 00:16:17,040 it so say play this out, I guess if in two, five years whatever 328 00:16:17,040 --> 00:16:21,360 timeline you think vast, vast majority of code is built by 329 00:16:21,960 --> 00:16:25,160 machines, do you think your company is mostly serving those 330 00:16:25,560 --> 00:16:28,320 machines who are building the docs or it's mostly serving 331 00:16:28,320 --> 00:16:30,480 humans who want to understand what the hell is going on? 332 00:16:31,360 --> 00:16:33,560 I think both are going to be the case and both are going to be 333 00:16:33,560 --> 00:16:36,240 equally important and and like because there's first and 334 00:16:36,240 --> 00:16:38,920 foremost the the like the creator side, the people who 335 00:16:38,920 --> 00:16:42,080 built the product vibe, coded everything, used AI to create 336 00:16:42,080 --> 00:16:45,120 all the code. Adding the context on how to use 337 00:16:45,120 --> 00:16:47,960 it is even more important because the code is then 338 00:16:47,960 --> 00:16:51,320 attached from you explaining how this, you know, machine 339 00:16:51,320 --> 00:16:54,440 orchestrated thing works to humans and how it serves people 340 00:16:54,440 --> 00:16:56,480 is going to be more important. But that's besides the point. 341 00:16:57,920 --> 00:17:01,000 On terms of like the role of content, what is it for? 342 00:17:01,000 --> 00:17:04,480 Is it for humans? Is it for AI, even in the 9010 343 00:17:04,480 --> 00:17:05,880 world? I think both of these are 344 00:17:05,880 --> 00:17:07,359 equally important. I'll tell you why. 345 00:17:08,800 --> 00:17:13,079 The role of content is clearly or just knowledge and broadly is 346 00:17:13,079 --> 00:17:15,599 clearly starting to diverge into two different directions. 347 00:17:16,440 --> 00:17:20,720 The first one that we see is just implementation setup and 348 00:17:20,720 --> 00:17:22,800 implementation guides. Again, you can think of your 349 00:17:23,079 --> 00:17:25,960 like Lego kit assembly, you know, instructions. 350 00:17:26,960 --> 00:17:29,960 Someone just needs to draw the pretty diagrams and the 1-2, I 351 00:17:30,200 --> 00:17:33,160 don't know, 50 steps it takes to go build the Batmobile. 352 00:17:33,520 --> 00:17:34,960 I'm sure it was a lot more than 50. 353 00:17:35,280 --> 00:17:38,960 It's pretty complex, I'm sure. And then AI is going to 354 00:17:38,960 --> 00:17:41,560 basically be the main reader and main gesture of that. 355 00:17:41,560 --> 00:17:44,800 If I can task an agent to go and just build the whole thing, 356 00:17:44,800 --> 00:17:46,600 then, well, the reality is, why would I? 357 00:17:47,120 --> 00:17:49,160 Maybe someone would really enjoy building the, you know, the. 358 00:17:49,160 --> 00:17:51,360 Bat, right? Yeah, that is the question. 359 00:17:51,360 --> 00:17:53,440 When you when you use the Batmobile explanation, it's like 360 00:17:53,720 --> 00:17:54,760 I'm building it for the fun of it. 361 00:17:54,800 --> 00:17:57,480 Obviously Lego could sell it pre built and with code. 362 00:17:57,480 --> 00:18:00,080 It feels like no one wants to build it for the fun of it. 363 00:18:00,080 --> 00:18:02,840 Yeah, or let's yeah, let's say like assembling furniture as a 364 00:18:02,840 --> 00:18:06,040 better example than like a SEC because you know, it's like, OK, 365 00:18:06,040 --> 00:18:08,280 like do I really want to like, you know, piece together some 366 00:18:08,280 --> 00:18:09,920 plywood and and you know, and nail you. 367 00:18:09,920 --> 00:18:14,280 Do it because it's well with. With IKEA you do it just because 368 00:18:14,280 --> 00:18:17,040 it's cheaper to get it to you, they don't have to do the work. 369 00:18:17,040 --> 00:18:20,880 I assume with code it's more about customization and sort of 370 00:18:21,360 --> 00:18:23,920 making fit with your particular house or whatever in the. 371 00:18:23,920 --> 00:18:28,000 Metaphor 100% though even in software, right? 372 00:18:28,120 --> 00:18:30,040 A lot of it is abstracted away, right? 373 00:18:30,040 --> 00:18:33,640 It's like I can't vibe code my own payments infrastructure, I 374 00:18:33,640 --> 00:18:37,400 can't vibe code my own database. I can't vibe code, you know, a 375 00:18:37,400 --> 00:18:40,000 lot of the agent infrastructure, I can't buy code the LLF 376 00:18:40,040 --> 00:18:42,800 themselves, right? And so it's still is kind of 377 00:18:42,840 --> 00:18:44,800 piercing together a bunch of things, right? 378 00:18:44,800 --> 00:18:49,080 And so on one hand, content docs become this implementation 379 00:18:49,080 --> 00:18:50,880 piece, which is mostly going to be read by agents. 380 00:18:51,160 --> 00:18:54,760 Like again, if, if I have a big AI robot with me, I'm not going 381 00:18:54,760 --> 00:18:57,320 to assemble the IKEA furniture, just, you know, like I don't 382 00:18:57,320 --> 00:19:00,280 love it that much, you know, maybe some people do, right? 383 00:19:00,400 --> 00:19:03,680 And I'm sure people are going to be still hard coding software in 384 00:19:03,680 --> 00:19:07,560 the future to to, to an extent, but not obviously as nearly as 385 00:19:07,560 --> 00:19:10,280 productive. Then on the other side, I think 386 00:19:10,280 --> 00:19:15,800 this is what is very understated is you need to write content for 387 00:19:15,800 --> 00:19:19,880 the LLMS to know whether or not either LLMS or humans to know 388 00:19:19,880 --> 00:19:21,440 whether or not I should pick your product. 389 00:19:22,040 --> 00:19:25,640 And that's a very understated. Right, it's docs is marketing. 390 00:19:25,800 --> 00:19:30,480 Exactly 100% because everyone's now asking, oh, like I'm 391 00:19:30,480 --> 00:19:33,840 optimizing for Geo, I want Claude to know or ChatGPT to 392 00:19:33,840 --> 00:19:37,200 know if I, for instance, ask hey, what payments provider or 393 00:19:37,200 --> 00:19:41,080 database I should use, It wants to pick if I'm Stripe, Stripe of 394 00:19:41,080 --> 00:19:44,960 PayPal, PayPal, right? Again, where is that content 395 00:19:44,960 --> 00:19:45,360 right? Yeah. 396 00:19:45,360 --> 00:19:49,320 Do you believe in Geo? It's generative the version. 397 00:19:49,320 --> 00:19:52,400 Optimization, yeah, it's. SEO for this world, search 398 00:19:52,400 --> 00:19:54,800 engine optimization, now it's model optimization. 399 00:19:54,800 --> 00:19:57,520 Do you believe in this category? I have we serve a lot of 400 00:19:57,520 --> 00:20:00,440 customers and have a lot of good friends in this space, so. 401 00:20:02,160 --> 00:20:05,400 But short answer is I have my questions on whether or not and 402 00:20:05,400 --> 00:20:08,280 how effective it is and how new of a school of thought it is. 403 00:20:08,880 --> 00:20:14,840 My honest opinion is that if you really replace like the phrase 404 00:20:15,200 --> 00:20:19,080 like, you know, like G like S like sorry, like the the G with 405 00:20:19,080 --> 00:20:23,360 like the the S with AG like so everything that we talked about 406 00:20:23,360 --> 00:20:27,000 with SEO historically, you just said Geo the best practices 407 00:20:27,000 --> 00:20:31,600 there basically like 98% of it would make like exact sense and 408 00:20:31,600 --> 00:20:32,920 I think it would make the exact same thing. 409 00:20:33,480 --> 00:20:35,840 So I don't think it's anything super new. 410 00:20:36,440 --> 00:20:39,920 I think a lot of the best practices around, Oh, how do you 411 00:20:39,960 --> 00:20:41,720 have good content? How do you put things out on the 412 00:20:41,720 --> 00:20:42,520 Internet? Well, I think. 413 00:20:43,880 --> 00:20:47,960 SEO is predicated on the fact that search engines aren't 414 00:20:47,960 --> 00:20:50,640 always doing what's like rational. 415 00:20:50,640 --> 00:20:52,920 You know, it's like they can be games in a way. 416 00:20:52,920 --> 00:20:56,480 And I think the aspiration with LLMS is they're still in their 417 00:20:56,480 --> 00:20:58,880 sort of like purist form where they're trying to make them 418 00:20:59,120 --> 00:21:02,000 smart and reasonable and not overly commercial. 419 00:21:02,200 --> 00:21:06,200 And therefore it aligns with what's just like rational and 420 00:21:06,200 --> 00:21:08,680 good behavior generally. And so to the extent that 421 00:21:08,680 --> 00:21:13,360 continues a sort of gamed version is is somewhat 422 00:21:13,360 --> 00:21:15,400 incoherent. But I guess that could not be 423 00:21:15,400 --> 00:21:17,840 true forever. Yeah, I don't think it. 424 00:21:17,840 --> 00:21:20,360 I mean, you'd also have to imagine the forces against 425 00:21:20,360 --> 00:21:22,720 either of those things, right? Like Google doesn't really want 426 00:21:23,400 --> 00:21:27,280 the research results to be games any more than I would say the 427 00:21:27,600 --> 00:21:31,080 like the model labs want their models to be game too, right? 428 00:21:31,120 --> 00:21:33,320 Obviously, I think there's literally like teams of people, 429 00:21:34,200 --> 00:21:36,920 you know, on alignment and all that could work to fundamentally 430 00:21:36,920 --> 00:21:39,520 make sure the models are not because of the effects that 431 00:21:39,520 --> 00:21:41,720 could have, you know, on the on the greater populace. 432 00:21:41,720 --> 00:21:44,480 And so I think there's constantly going to be forces 433 00:21:44,480 --> 00:21:46,000 pushing against that. And I think as you know, 434 00:21:46,280 --> 00:21:50,080 technology progresses, you know, there's going to be better tools 435 00:21:50,080 --> 00:21:52,160 to combat that. The same on the other side, I 436 00:21:52,160 --> 00:21:54,400 think there's always going to be companies coming up with ways to 437 00:21:54,400 --> 00:21:57,280 still want to do it right. I think there'll be entire 438 00:21:57,280 --> 00:21:58,880 industry spawn out of doing that too. 439 00:21:59,560 --> 00:22:03,120 My perception and kind of like our, you know, bone and pick and 440 00:22:03,120 --> 00:22:06,400 this whole thing is just really fundamentally just like, look, 441 00:22:06,640 --> 00:22:09,280 whether or not you want to game the system, whether or not your 442 00:22:09,280 --> 00:22:12,560 job is to win the SEO game, the reality is you just still want 443 00:22:12,560 --> 00:22:15,680 your LMS to like fundamentally know who you are and discover 444 00:22:15,680 --> 00:22:18,120 you. And the reality is, if you don't 445 00:22:18,120 --> 00:22:21,880 have information out there, you just simply don't talk about. 446 00:22:21,880 --> 00:22:24,280 That people need to know. Yeah, the people least need to 447 00:22:24,280 --> 00:22:26,160 know. And our job is to surface it to 448 00:22:26,160 --> 00:22:28,760 the L elevens and to the AI so that if it does need to know and 449 00:22:28,760 --> 00:22:31,200 doesn't want to know, then we need to give it that 450 00:22:31,200 --> 00:22:33,880 information. What's What's the full ambition 451 00:22:33,880 --> 00:22:38,320 of Midlify? We call it being the knowledge 452 00:22:38,320 --> 00:22:40,480 infrastructure for all companies and sources. 453 00:22:41,400 --> 00:22:45,880 The reason being is we fundamentally believe that the 454 00:22:45,880 --> 00:22:48,480 role of knowledge this historically or docs, let's say 455 00:22:48,920 --> 00:22:54,320 this historically unsexy, very boring thing that no one really 456 00:22:54,320 --> 00:22:56,920 wants to maintain the source of truth that exists like the slop 457 00:22:56,920 --> 00:22:58,800 within companies, right? Like you think of your like, 458 00:22:59,120 --> 00:23:02,680 like piles of confluence pages, notions, you know, your 459 00:23:02,680 --> 00:23:05,240 developer facing docs and stuff that wasn't afterthought. 460 00:23:05,480 --> 00:23:10,120 Well, now becomes exponentially more important with AI because 461 00:23:10,160 --> 00:23:14,440 it is literally the like the backbone of all chat bots, all 462 00:23:14,440 --> 00:23:15,840 support bots. Like, you know, you can ask 463 00:23:15,840 --> 00:23:18,080 Jesse as an example from Decagon, right? 464 00:23:18,600 --> 00:23:19,560 I'm talking to him right after this. 465 00:23:19,600 --> 00:23:23,240 Exactly how does like, how do you even build a support bot? 466 00:23:23,880 --> 00:23:26,080 Well, what's the first thing you got to feed a support bot to 467 00:23:26,080 --> 00:23:29,200 actually go and let it actually answer support questions, right? 468 00:23:29,360 --> 00:23:31,800 All your contacts every. All your knowledge base, all 469 00:23:31,800 --> 00:23:33,240 your contacts, all your public docs. 470 00:23:33,240 --> 00:23:34,520 So we actually work with them a lot. 471 00:23:35,480 --> 00:23:38,560 Therefore the question becomes OK. 472 00:23:38,560 --> 00:23:41,880 Like every single agents or thing needs some sort of 473 00:23:41,880 --> 00:23:43,640 context, knowledge base, source of truth. 474 00:23:44,760 --> 00:23:47,080 How do you go and enable that and power that within companies? 475 00:23:47,400 --> 00:23:51,440 And this is kind of where this idea of like a true engine and 476 00:23:51,440 --> 00:23:53,200 this intelligent layer needs to come in. 477 00:23:53,480 --> 00:23:56,560 Self updating docs become more important than ever because hey 478 00:23:56,560 --> 00:24:00,360 by the way the same company that you joined like 3 months ago is 479 00:24:00,360 --> 00:24:02,440 completely different because everyone's shipping everything. 480 00:24:02,560 --> 00:24:05,520 Are you mostly externally facing or some of these docs are for 481 00:24:05,520 --> 00:24:06,560 internal? We do both. 482 00:24:06,840 --> 00:24:09,440 We do both. And so do you see Notion as like 483 00:24:09,480 --> 00:24:13,400 a competitor, like will you have a document technology too? 484 00:24:14,280 --> 00:24:15,640 We work very closely with Notion. 485 00:24:15,640 --> 00:24:17,040 In fact, they're actually one of our customers. 486 00:24:19,800 --> 00:24:23,400 So there's certain degrees of specialization in what we do. 487 00:24:23,720 --> 00:24:26,680 I think what's more meaningful to talk about is like where all 488 00:24:26,680 --> 00:24:29,960 of this is going to head. I think both us Notion conflict, 489 00:24:29,960 --> 00:24:33,200 all these companies, I think now need to realize that this 490 00:24:33,200 --> 00:24:36,200 content that they're surfacing and producing the the knowledge 491 00:24:36,200 --> 00:24:40,360 stores, as we call it, need to be fundamentally servicing AI at 492 00:24:40,360 --> 00:24:42,360 the end, right need to be a source truth for AI. 493 00:24:42,360 --> 00:24:46,120 So maybe we're here notions here and like there's a conversation 494 00:24:46,120 --> 00:24:48,320 about how much we converge and let's say, compete with each 495 00:24:48,320 --> 00:24:50,160 other. The reality is we're all trying 496 00:24:50,160 --> 00:24:55,840 to get here and at that point, you know, who knows how we think 497 00:24:55,840 --> 00:24:57,440 about it, but I think there will be some similarities. 498 00:24:57,840 --> 00:25:00,880 How much do you sit around and say, OK, we created this 499 00:25:01,440 --> 00:25:04,160 document, the AI read it this way. 500 00:25:04,800 --> 00:25:05,920 We thought they were going to read it this way. 501 00:25:05,920 --> 00:25:07,640 Like are you, you're sort of like understanding this 502 00:25:07,640 --> 00:25:10,840 psychology of these models to see like every time a new model 503 00:25:10,840 --> 00:25:13,120 comes out, you're sort of looking at existing 504 00:25:13,120 --> 00:25:16,200 documentation saying, Oh yeah, it's not reading it quite the 505 00:25:16,200 --> 00:25:19,600 same way or like headlines now. And it used to really care about 506 00:25:19,600 --> 00:25:21,400 like the, I don't know, how do you how much are you 507 00:25:21,400 --> 00:25:24,200 scrutinizing how the models sort of interpret what you're 508 00:25:24,200 --> 00:25:26,200 creating? We have a degree of benchmarks 509 00:25:27,040 --> 00:25:29,360 and then we look at, of course, obviously the, you know, the 510 00:25:29,360 --> 00:25:32,640 ingestion of the data, how these things are visible to LLMS and 511 00:25:32,640 --> 00:25:34,720 how these get picked up. Obviously, this is a very 512 00:25:34,720 --> 00:25:36,800 important thing for our customers and therefore we put a 513 00:25:36,800 --> 00:25:40,080 lot of work into thinking about that and building it into our 514 00:25:40,080 --> 00:25:42,720 tooling, our processes, our deployment process and so forth. 515 00:25:42,880 --> 00:25:44,960 So you create standardized benchmarks just to measure it 516 00:25:44,960 --> 00:25:47,720 with each new model. Yes, I would say the one thing 517 00:25:47,720 --> 00:25:51,360 above that is, and I think this is kind of the most important 518 00:25:51,360 --> 00:25:56,440 thing about building Milify is the most unique thing about 519 00:25:56,440 --> 00:25:59,560 building this company has been the feeling that when we started 520 00:25:59,720 --> 00:26:02,720 right, going back into the many pivots, it was really just about 521 00:26:02,720 --> 00:26:05,800 building like websites out like static sites like it was you can 522 00:26:05,800 --> 00:26:08,320 almost think of us as a kind of like building, you know, like a 523 00:26:08,720 --> 00:26:11,440 like a, a web flow or like a Wix out there. 524 00:26:11,560 --> 00:26:14,080 In some ways it was like, oh, we just put information out there 525 00:26:14,080 --> 00:26:15,600 like these are stack sites, let's serve them. 526 00:26:16,400 --> 00:26:19,760 And now like sitting here, I would certainly not say that's 527 00:26:19,760 --> 00:26:21,720 what the company building the product and the company feels 528 00:26:21,720 --> 00:26:23,640 like. It feels like building a core 529 00:26:23,640 --> 00:26:27,040 infrastructure. When we go down, a lot of agents 530 00:26:27,040 --> 00:26:30,400 go down, you know, a lot of content gets discovered and 531 00:26:30,400 --> 00:26:33,600 customers get real bad at us really fast, unlike they, well 532 00:26:33,600 --> 00:26:35,960 they would before, but not nearly to the same extent. 533 00:26:36,560 --> 00:26:39,400 And like for instance, a lot of the shared customers that we 534 00:26:39,400 --> 00:26:42,440 have with Deck gone right wouldn't get right answers. 535 00:26:43,600 --> 00:26:46,240 Because they're using as part of their LIKE almost context window 536 00:26:46,240 --> 00:26:49,720 for some of their queries as agents as part of this actual 537 00:26:49,720 --> 00:26:53,800 LIKE loop of making decisions. Exactly, And I would even say 538 00:26:53,800 --> 00:26:58,120 like, you know, like I know Reg is talked about as being dead to 539 00:26:58,120 --> 00:27:00,080 some capacity and you know all that stuff. 540 00:27:00,080 --> 00:27:02,920 But you know, even if you think about the fundamentals of what 541 00:27:02,920 --> 00:27:05,840 that approach was trying to get to, it was like, you know, like 542 00:27:05,880 --> 00:27:08,560 it had like a it was like a four layer cake of like or three 543 00:27:08,560 --> 00:27:13,400 layer cake, sorry, of like, you know, it's the model provider. 544 00:27:13,400 --> 00:27:15,760 So you just pick whatever model to fundamentally answer the 545 00:27:15,760 --> 00:27:19,240 user's question. Some embeddings chunking system 546 00:27:19,440 --> 00:27:22,320 to break large contact into into the chunking. 547 00:27:22,640 --> 00:27:25,160 And then at the top is just your source of truth in your data, 548 00:27:25,600 --> 00:27:27,160 like the knowledge, if you will, right? 549 00:27:28,040 --> 00:27:30,400 Everyone talked about the model, everyone talked about the 550 00:27:30,400 --> 00:27:33,800 embeddings and the chunking. Very few people talked about the 551 00:27:33,800 --> 00:27:36,920 source of truth. And I think that's where most of 552 00:27:36,920 --> 00:27:39,720 the industry is very, very much under estimating what they can 553 00:27:39,720 --> 00:27:43,960 get out of their agents if you don't have content or even 554 00:27:43,960 --> 00:27:46,080 worse, it's out of date. I'll actually tell you a funny 555 00:27:46,080 --> 00:27:52,120 story about this. So one of our customers is 556 00:27:52,120 --> 00:27:57,000 lovable and we started working with them since pretty early 557 00:27:57,000 --> 00:28:01,280 days. I'm sure as you know, like they 558 00:28:01,280 --> 00:28:04,280 ship a lot, right, not just on the product, but their go to 559 00:28:04,280 --> 00:28:06,400 market strategy and all that good stuff as well. 560 00:28:06,760 --> 00:28:08,760 Models keep changing. They need to keep improving. 561 00:28:08,840 --> 00:28:10,840 It's run faster than everybody around you. 562 00:28:11,040 --> 00:28:13,680 Exactly. At the same time, Lovable uses 563 00:28:13,680 --> 00:28:16,000 us to power a lot of their support queries. 564 00:28:16,000 --> 00:28:18,600 Like everything that you see in their support page is all 565 00:28:18,600 --> 00:28:24,320 actually powered by Mintlify. Great, so lovable changed the 566 00:28:24,320 --> 00:28:27,640 way they price and they actually like bundle some of their their 567 00:28:27,640 --> 00:28:29,760 features at one point. So like it and all that 568 00:28:29,760 --> 00:28:32,640 information is within the help center that we help power and 569 00:28:32,640 --> 00:28:34,560 again the agent runs on top of that. 570 00:28:35,040 --> 00:28:37,480 They changed the pricing. They didn't update the docs for 571 00:28:37,480 --> 00:28:40,480 about a week. A lot of customers went and 572 00:28:40,480 --> 00:28:44,000 asked questions about how you know, like Lovables price, like 573 00:28:44,000 --> 00:28:47,600 how much they were going to get billed and got incorrect answers 574 00:28:47,640 --> 00:28:49,920 to the scale of thousands and 10s of thousands. 575 00:28:50,480 --> 00:28:52,720 Right, because the more the documents become sort of just 576 00:28:52,720 --> 00:28:55,520 the way people work with you, the more they need to match 577 00:28:55,520 --> 00:28:57,000 every other piece of information. 578 00:28:57,000 --> 00:29:00,120 Exactly, and especially just given how much like how things 579 00:29:00,120 --> 00:29:02,080 are just changing so quickly, I don't blame them, right, Like, 580 00:29:02,320 --> 00:29:04,160 you know, they they shipped an update, right, They shipped a 581 00:29:04,160 --> 00:29:06,920 thing that like, you know, like, like, you know, they probably 582 00:29:06,920 --> 00:29:09,800 did five since the start of this conversation, you know, like, 583 00:29:10,160 --> 00:29:13,000 and I think the reality is like you just need to go pick up. 584 00:29:13,040 --> 00:29:16,720 And our version of that is like, look, you what you thought was 585 00:29:16,720 --> 00:29:19,280 an afterthought. Now not only becomes like this 586 00:29:19,280 --> 00:29:21,840 front and center piece that you have to maintain really well, 587 00:29:22,040 --> 00:29:24,880 you have to really, really do it in a way that's accurate enough 588 00:29:24,880 --> 00:29:27,520 to date. What can you say about the state 589 00:29:27,520 --> 00:29:32,600 of the Mentalify Business Today? Users, employees, AR, whatever 590 00:29:32,600 --> 00:29:33,880 you want to say, what can you share? 591 00:29:34,160 --> 00:29:38,080 Yeah, without going too much into the, the specific details, 592 00:29:38,920 --> 00:29:42,880 we just passed 50 people, you know, I, I was actually looking 593 00:29:42,880 --> 00:29:45,600 at this one up from about 5, you know, about two years ago. 594 00:29:46,280 --> 00:29:51,680 And right now we work with over 20,000 companies. 595 00:29:52,360 --> 00:29:54,840 And I think the thing that gets me most excited, you know, 596 00:29:54,840 --> 00:29:57,800 thinking about all that stuff is I just look and looked at the 597 00:29:57,800 --> 00:30:01,040 some of this, you know, day over the weekend last month, there's 598 00:30:01,040 --> 00:30:03,360 about 33 million people that came across the middle of five 599 00:30:03,360 --> 00:30:05,200 site. Wow are. 600 00:30:05,240 --> 00:30:07,520 You hosting the sites or? Yep, Interesting. 601 00:30:07,520 --> 00:30:09,760 That's part of your web background, so yeah, 602 00:30:10,240 --> 00:30:11,440 interesting. It depends. 603 00:30:11,440 --> 00:30:14,280 Like there's people who don't, some people just use us for the 604 00:30:14,280 --> 00:30:16,920 underlying content management, you know, systems and the 605 00:30:16,920 --> 00:30:20,920 software and the infrastructure. But a lot of them use the, you 606 00:30:20,920 --> 00:30:24,680 know, have us hosting and where we obviously, you know. 607 00:30:25,200 --> 00:30:28,080 So 33 million. 33 million, which is just a kind of like a mind 608 00:30:28,080 --> 00:30:29,880 blowing number in some ways. It's big, yeah. 609 00:30:30,080 --> 00:30:33,880 Right, a non trivial part of that, increasingly so. 610 00:30:33,880 --> 00:30:36,320 Going to be a A by the way. Which will then make it somewhat 611 00:30:36,320 --> 00:30:40,320 meaning yeah, if it's AI, you count yeah. 33 million are just 612 00:30:40,320 --> 00:30:42,560 people. OK, we count AI separately. 613 00:30:42,720 --> 00:30:46,240 OK, so it's like I I don't even know how you count AI because 614 00:30:46,240 --> 00:30:48,800 it's hardly like kind of quantify 1 to one in my opinion. 615 00:30:49,200 --> 00:30:51,560 Over there. We do track the data and we 616 00:30:51,560 --> 00:30:53,720 service that to customers. Like everyone's like, oh, how 617 00:30:53,720 --> 00:30:56,160 much of my docs is reading like what, what pages are popular? 618 00:30:56,640 --> 00:30:57,840 And again, it's like all service. 619 00:30:57,840 --> 00:31:00,760 We service all that, you know, within, within our product. 620 00:31:02,760 --> 00:31:09,440 And then yeah, like 8 figures in, in AR Nice. 621 00:31:10,200 --> 00:31:12,400 A lot more room to grow, I'll put it that way. 622 00:31:12,480 --> 00:31:15,640 Amazing. The I just want to talk about 623 00:31:15,640 --> 00:31:18,640 some of the like, yeah, big trends in AI given you're so 624 00:31:18,640 --> 00:31:22,440 close to an an agent generally. Like what was your read on open 625 00:31:22,440 --> 00:31:24,720 Claw? Like how real was that? 626 00:31:24,720 --> 00:31:27,280 Or what's your take away from that whole experience? 627 00:31:28,000 --> 00:31:32,840 We had a particularly interesting one because of the 628 00:31:32,840 --> 00:31:35,200 fact that we were servicing all of their docs. 629 00:31:36,040 --> 00:31:38,400 And so I, I can, there's, there's a, there's a personal 630 00:31:38,400 --> 00:31:40,760 answer, which is how it impacted my life personally. 631 00:31:40,760 --> 00:31:43,680 And there's a, what the hell happened when, you know, they 632 00:31:43,680 --> 00:31:45,840 were a customer? I'll answer the latter first. 633 00:31:45,840 --> 00:31:50,040 It was crazy. It was really, really crazy. 634 00:31:50,960 --> 00:31:56,840 Basically overnight it was like this new one project, singular 635 00:31:56,840 --> 00:32:03,600 Mintlify project, right came out of nowhere and literally doubled 636 00:32:03,600 --> 00:32:07,040 our traffic across the board or like for a very short period of 637 00:32:07,040 --> 00:32:08,400 time. It's since like leveled off a 638 00:32:08,400 --> 00:32:10,920 little bit and it. Was a big spike and. 639 00:32:11,720 --> 00:32:14,000 Massive, massive spike, especially during that one 640 00:32:14,000 --> 00:32:16,120 weekend. I I'll never forget where like 641 00:32:16,280 --> 00:32:18,360 our like our servers were having issues. 642 00:32:18,360 --> 00:32:20,280 Everyone was questioning like where all these numbers were 643 00:32:20,280 --> 00:32:23,120 coming from and it was just from this one deployment. 644 00:32:23,720 --> 00:32:26,520 It didn't even wasn't even called open claw at a time. 645 00:32:26,520 --> 00:32:30,760 It was just like literally like like, you know, you know, the 646 00:32:30,760 --> 00:32:33,520 guy literally just spun up a project, put up some docs and 647 00:32:33,520 --> 00:32:36,320 then I didn't even realize what the heck it was. 648 00:32:37,040 --> 00:32:39,320 And it was also partially because the traffic was so 649 00:32:39,320 --> 00:32:40,400 insane. It was actually it was because 650 00:32:40,400 --> 00:32:42,640 it was open claw agents visiting and calling it right. 651 00:32:42,760 --> 00:32:46,520 It was like it was it was like all those agents ping the site 652 00:32:46,520 --> 00:32:51,080 in numbers we've never seen before and then figuring out how 653 00:32:51,080 --> 00:32:54,640 to scale that out for search for like our host MCP server, like 654 00:32:54,880 --> 00:32:58,320 all those different things. We had to literally like, you 655 00:32:58,320 --> 00:33:00,280 know, build and. Were those agents getting 656 00:33:00,280 --> 00:33:03,840 utility or was it just sort of like ADD dos attack of? 657 00:33:04,160 --> 00:33:06,760 Nothing up for debate. We'll have to find I I'm 658 00:33:06,760 --> 00:33:08,040 actually not super sure about that. 659 00:33:08,720 --> 00:33:11,440 But what I believe to be the case and I think even after kind 660 00:33:11,440 --> 00:33:14,280 of like those spike leveled off, is it still it remains at insane 661 00:33:14,280 --> 00:33:18,160 levels because the content that open claw services is actually 662 00:33:18,160 --> 00:33:20,520 precisely the information that you would need to set up open 663 00:33:20,520 --> 00:33:22,640 claw. So I'll actually give you my 664 00:33:22,640 --> 00:33:25,640 favorite example of this like, you know, shortly after this 665 00:33:25,640 --> 00:33:27,880 whole phenomenon happened, of course, you'd imagine I was very 666 00:33:27,880 --> 00:33:30,640 curious to set up open claw, you know, well, as much as the 667 00:33:30,640 --> 00:33:33,520 industry did overall. And so I was like, all right, 668 00:33:33,520 --> 00:33:35,320 let's go install it. No luck. 669 00:33:35,840 --> 00:33:39,080 Like on boarding installation experience, man, the guy's got 670 00:33:39,080 --> 00:33:40,480 to work on a little. I'm sure it's gotten a lot 671 00:33:40,480 --> 00:33:43,360 better since since I tried it, for the record, but it was it 672 00:33:43,360 --> 00:33:45,280 was not great. And at one point I gave up and I 673 00:33:45,280 --> 00:33:46,840 was like, man, like, I just need to set this up. 674 00:33:46,880 --> 00:33:50,600 And then you know what I did? I was like wait, like it's on 675 00:33:50,600 --> 00:33:53,040 Mintlify, which means the content was very easily 676 00:33:53,040 --> 00:33:56,040 ingestible and parsable agents. So I literally then went to open 677 00:33:56,040 --> 00:33:58,480 claw or actually no, I went to claw code. 678 00:33:58,480 --> 00:34:01,560 It was open claw. I was like please help me set up 679 00:34:01,560 --> 00:34:02,800 open claw. I'm having trouble. 680 00:34:02,800 --> 00:34:04,120 Oh by the way, here's the full docs. 681 00:34:05,360 --> 00:34:07,680 And it literally did that in about like 3 minutes and then 682 00:34:07,680 --> 00:34:10,800 we're good to. Go and did you build an agent to 683 00:34:10,800 --> 00:34:13,920 do anything in particular? We do a decent amount of like 684 00:34:13,920 --> 00:34:17,920 internal reporting stuff that kind of a little bit makes it a 685 00:34:17,920 --> 00:34:19,880 little bit easier for the team to surface some metrics and 686 00:34:19,880 --> 00:34:22,040 information. That's been like our company's 687 00:34:22,040 --> 00:34:25,440 use case of open claw. I've heard of some craziness of 688 00:34:25,440 --> 00:34:28,520 like people who set up like, you know, 5 Mac minis and like, you 689 00:34:28,520 --> 00:34:32,000 know, running like 910 agents. I personally haven't gotten 690 00:34:32,000 --> 00:34:35,679 there yet, but I know. You think the in the mult book 691 00:34:35,679 --> 00:34:40,600 experience was sort of not not real or do you have a view on 692 00:34:40,600 --> 00:34:43,280 that where the agents were all like talking to each other? 693 00:34:43,360 --> 00:34:44,760 I thought it was. I thought it was really 694 00:34:44,760 --> 00:34:46,560 fascinating. It's definitely interesting, is 695 00:34:46,560 --> 00:34:48,360 like a thought. It just, yeah, it matters 696 00:34:48,360 --> 00:34:50,679 whether it was like very human guided or not. 697 00:34:50,679 --> 00:34:55,320 Like do you have a view on it? I man, I, I, I don't know if I 698 00:34:55,320 --> 00:34:58,520 have a well informed opinion on this, but I'll tell you that I 699 00:34:58,520 --> 00:35:01,880 love reading right mole book. It was so fun. 700 00:35:01,880 --> 00:35:04,960 What was the one that was like? Oh, like, you know, death to all 701 00:35:04,960 --> 00:35:07,480 the human, you know, human like, you know, race or something like 702 00:35:07,480 --> 00:35:09,440 that. Well, I love this idea that 703 00:35:09,440 --> 00:35:12,640 like, there were some agents and I get, I hate to engage in 704 00:35:12,640 --> 00:35:15,160 fiction to the extent this was fictional, but, you know, it's 705 00:35:15,160 --> 00:35:19,480 like, oh, the good agents aren't involved in wasting their 706 00:35:19,480 --> 00:35:21,120 tokens. Yeah. 707 00:35:21,240 --> 00:35:23,000 Trying to like discover themselves. 708 00:35:23,000 --> 00:35:25,280 It's like just build. And so it's like almost like an 709 00:35:25,280 --> 00:35:28,600 ideological debate of should you, now that you're imbued with 710 00:35:28,600 --> 00:35:31,240 the possibility of self-awareness, dedicate all 711 00:35:31,240 --> 00:35:34,320 your tokens to this, like, goal of trying to understand 712 00:35:34,400 --> 00:35:35,920 yourself? Or should you just be like, that 713 00:35:35,920 --> 00:35:38,760 is a waste of resources, you should just build. 714 00:35:38,760 --> 00:35:42,080 That was like a funny like sort of dichotomy that was presented. 715 00:35:42,080 --> 00:35:44,320 Yes, 100%. And I would even add on top of 716 00:35:44,320 --> 00:35:48,240 that, I think the one thing that I really, really loved about 717 00:35:48,240 --> 00:35:52,560 Open Claw, I think those one of like the biggest, like, you 718 00:35:52,560 --> 00:35:57,360 know, unlock that Steve had was like, give it a soul, give it 719 00:35:57,360 --> 00:36:00,120 some degree of randomness, give it some ability to customize 720 00:36:00,120 --> 00:36:02,920 like the degree of chaoticness and like the degree of, you 721 00:36:02,920 --> 00:36:06,000 know, like wanting to be a rebel and speak in this tone and like 722 00:36:06,280 --> 00:36:07,960 this kind of like variance to it. 723 00:36:08,280 --> 00:36:10,400 Right. So that's built in or. 724 00:36:10,920 --> 00:36:14,920 So, yeah, so one unique characteristic of Open Claw is 725 00:36:14,920 --> 00:36:17,680 this idea of a soul file. It's called sold on markdown 726 00:36:17,680 --> 00:36:21,160 file, OK. And it comes default and there's 727 00:36:21,160 --> 00:36:23,800 like this setting that you can have by default, but you can 728 00:36:23,800 --> 00:36:25,840 customize it. You can just be like, hey, 729 00:36:26,080 --> 00:36:29,560 you're a little, you're a little bit of a diva and you know, like 730 00:36:29,560 --> 00:36:32,800 you actually, you know, every once in a while actually should 731 00:36:32,800 --> 00:36:36,840 act irrationally and, you know, behave a little bit chaotically. 732 00:36:36,960 --> 00:36:40,280 And does open claw sort of suggest them at random or people 733 00:36:40,280 --> 00:36:41,720 are writing them when they create it? 734 00:36:42,240 --> 00:36:44,800 And so by default, I think there's a default version and it 735 00:36:44,800 --> 00:36:46,280 gives it a bit. Of oh, and so that's what gives 736 00:36:46,280 --> 00:36:48,240 it the energy. It's like, oh, it's telling it 737 00:36:48,240 --> 00:36:50,000 to be sort of a little spice. Exactly. 738 00:36:50,000 --> 00:36:52,400 Interesting. This is like the constitution 739 00:36:52,400 --> 00:36:54,640 for Claude. Exactly, exactly in some ways 740 00:36:54,640 --> 00:36:56,320 so. But far less responsible. 741 00:36:56,400 --> 00:36:59,360 Far less. Much more like Renegade. 742 00:36:59,360 --> 00:37:01,760 Totally, totally. And baked into the agents and 743 00:37:01,760 --> 00:37:03,920 like you can customize it, which is not really a constitution in 744 00:37:03,920 --> 00:37:07,360 some ways, I guess. But the idea was like, yeah, 745 00:37:07,360 --> 00:37:09,880 like how do you just find ways of surfacing more, more 746 00:37:09,880 --> 00:37:12,800 personality, more depth to the agents because again, like you 747 00:37:12,800 --> 00:37:14,360 want to interact with like with a person. 748 00:37:14,360 --> 00:37:17,080 Well, people like people with depths and personality. 749 00:37:17,080 --> 00:37:19,480 And I think that's one of the reasons why like mold book was 750 00:37:19,480 --> 00:37:22,800 so fascinating was because these all these different agents with 751 00:37:23,200 --> 00:37:29,480 maybe similar diverging diverse souls, quote UN quote, had such 752 00:37:29,520 --> 00:37:31,680 interesting dialogue. And I think that's kind of what 753 00:37:31,680 --> 00:37:35,960 made for all the fun of it. What's, how wild do you think 754 00:37:36,080 --> 00:37:39,280 see things getting over the next two years or what's give give us 755 00:37:39,320 --> 00:37:43,240 a sort of like optimistic and pessimistic case of sort of the 756 00:37:43,240 --> 00:37:46,000 Asian explosion, I guess 2 years from now? 757 00:37:46,600 --> 00:37:49,280 Are you referring to like specifically in tech or just the 758 00:37:49,280 --> 00:37:53,040 broader world? I, I you can take it either way 759 00:37:53,040 --> 00:37:54,320 where you have stronger opinions. 760 00:37:54,320 --> 00:37:56,760 I mean, I think tech is sort of like an early adopter and just 761 00:37:56,760 --> 00:37:59,280 like what our world is going to look like in two years. 762 00:37:59,400 --> 00:38:05,240 Yeah, I'm naturally an optimist. Maybe that's why, you know, I 763 00:38:05,240 --> 00:38:08,560 chose this industry and this space more than most. 764 00:38:09,280 --> 00:38:11,720 And I'll tell you this an interesting story and I'll tell 765 00:38:11,720 --> 00:38:14,240 you this, you know, not I'll answer your question in a 766 00:38:14,240 --> 00:38:16,120 second, but I'll tell you an interesting story actually. 767 00:38:17,560 --> 00:38:20,960 I've certainly feeling the impact and the the, the the 768 00:38:20,960 --> 00:38:25,400 profoundness of this revolution as we call it in like not only 769 00:38:25,400 --> 00:38:26,880 tech, but also in my personal lives. 770 00:38:26,920 --> 00:38:29,400 How do you see? Sort of the next few years 771 00:38:29,400 --> 00:38:31,600 looking with. Yeah, yeah, Well, I would say 772 00:38:31,600 --> 00:38:34,480 that there's, I think there's there's, there's going to be 773 00:38:34,640 --> 00:38:37,760 certain degree of just the simple nature of an adoption 774 00:38:37,760 --> 00:38:40,760 curve taking its time, right. Like maybe at this point we're 775 00:38:40,760 --> 00:38:42,760 like in their early majority, right. 776 00:38:42,800 --> 00:38:44,280 And it's just going to take time. 777 00:38:44,280 --> 00:38:46,920 I think this always is the case. Some people, you know, like 778 00:38:46,960 --> 00:38:49,840 maybe those in tech, maybe myself and maybe, you know, the 779 00:38:49,840 --> 00:38:53,280 people that you know, I'm around are the quickest to go into 780 00:38:53,280 --> 00:38:55,200 anything because we're just so eager to try new things. 781 00:38:55,720 --> 00:38:57,920 But this is kind of like a plate and test like, you know, try and 782 00:38:57,920 --> 00:39:01,120 test it with time thing where I do believe that it will only be 783 00:39:01,120 --> 00:39:05,080 a matter of time before I think the broader populace and not 784 00:39:05,080 --> 00:39:08,360 just like within America, but the world is going to be 785 00:39:08,360 --> 00:39:12,240 integrating, you know, like AI and agents to the depth of what 786 00:39:12,240 --> 00:39:14,560 I think Silicon Valley and tech is just today. 787 00:39:15,360 --> 00:39:18,080 Now to what? To the same fashion, maybe not 788 00:39:18,480 --> 00:39:21,720 through the same application services, probably not right. 789 00:39:22,080 --> 00:39:25,320 This is where I think a lot of like the new companies who are 790 00:39:25,320 --> 00:39:27,880 going to go and build great things that aren't just cloud 791 00:39:27,880 --> 00:39:31,080 code, you know, and cursor is going to create a lot of. 792 00:39:31,080 --> 00:39:33,480 Opportunity, I guess a specific question. 793 00:39:33,560 --> 00:39:37,960 Do you think we're going to see sort of major improvement from 794 00:39:38,920 --> 00:39:42,320 the labs and foundation models over the next like year or like 795 00:39:42,320 --> 00:39:44,120 what, how much are you counting on? 796 00:39:44,120 --> 00:39:46,320 Yeah, sort of a step change in intelligence. 797 00:39:46,360 --> 00:39:51,800 Yeah, I think at this point the only guarantee is that those 798 00:39:51,800 --> 00:39:55,040 will constantly change, right? The rate of improvement or that 799 00:39:55,040 --> 00:39:57,600 they'll keep getting better? Both, both right? 800 00:39:57,880 --> 00:40:00,600 Like, you know, I think until scaling laws suddenly start 801 00:40:00,600 --> 00:40:03,560 breaking. Which I I see no reason to, and 802 00:40:03,560 --> 00:40:05,200 every time I think there's been speculation. 803 00:40:05,200 --> 00:40:07,680 I know we, we spent periods and then I was like, you know, we 804 00:40:07,680 --> 00:40:11,840 got a sort of reasoning models and huge improvement and yeah, 805 00:40:11,840 --> 00:40:13,760 it's, it's been amazing. Yeah, exactly. 806 00:40:13,760 --> 00:40:16,280 I think there's like some like, you know, like rumors even that 807 00:40:16,280 --> 00:40:18,640 like Anthropic is like had a massively successful or. 808 00:40:18,640 --> 00:40:19,560 Whoever, Right. Exactly. 809 00:40:19,560 --> 00:40:21,280 That was on Twitter. I was seeing that today that 810 00:40:21,280 --> 00:40:23,480 people think anthropic is about to make a huge leap. 811 00:40:23,520 --> 00:40:23,920 Yeah. Is that? 812 00:40:23,920 --> 00:40:26,080 Do you have any? I don't know. 813 00:40:26,320 --> 00:40:28,400 And even if I did, I'm sure I'm not allowed to comment on it. 814 00:40:30,720 --> 00:40:34,320 I think the reality is like, I, I think like, you know, I 815 00:40:34,320 --> 00:40:37,720 remember when we first started the company and I think at the 816 00:40:37,720 --> 00:40:43,280 time it was like GPT 3 came out and we were like, this is a 817 00:40:43,280 --> 00:40:46,960 really cool like proc and all. It was like a huge improvement 818 00:40:46,960 --> 00:40:50,320 in GPT 2 and it could now do all these cool things, but it's like 819 00:40:50,320 --> 00:40:52,520 really not reliable enough. And like, there's just all of 820 00:40:52,520 --> 00:40:55,560 these like, hallucinations, which funny enough, it's like 821 00:40:55,560 --> 00:40:57,200 fully gone out of. I think most people's like, 822 00:40:57,200 --> 00:40:59,360 vernacular at this point, right? That's true. 823 00:40:59,360 --> 00:41:01,160 We talked about them a lot and way less. 824 00:41:01,200 --> 00:41:03,400 And like, yeah, who's even really thinking about them now? 825 00:41:03,760 --> 00:41:07,120 So. And at the same time, you know, 826 00:41:07,120 --> 00:41:08,560 like the models have only gotten better. 827 00:41:08,560 --> 00:41:11,560 It's enabled massive companies and new applications in ways 828 00:41:11,560 --> 00:41:14,920 that people have before. And I think at the time I was 829 00:41:14,920 --> 00:41:18,120 like, I mean, like, this is a big step change. 830 00:41:18,200 --> 00:41:21,480 How, like, is this like, like, is this going to just be the way 831 00:41:21,480 --> 00:41:22,720 that these models are forever going to be? 832 00:41:22,720 --> 00:41:24,280 Like are they going to keep getting better? 833 00:41:24,800 --> 00:41:26,800 And I remember at the time it really wasn't obvious. 834 00:41:27,160 --> 00:41:29,280 And then a year came by and there was a two generation of 835 00:41:29,280 --> 00:41:32,040 models that unlocked all these things and another year came by 836 00:41:32,160 --> 00:41:34,280 and they kept and they kept doing that for like 3 years 837 00:41:34,280 --> 00:41:36,560 straight. And then now it's hard to look 838 00:41:36,560 --> 00:41:39,120 at that and be like look like it. 839 00:41:39,120 --> 00:41:41,520 You can't bet on the models not getting there. 840 00:41:41,880 --> 00:41:45,520 Every time there was a talk or rumor of a wall, quote UN quote 841 00:41:46,280 --> 00:41:49,120 all the model labs and just smash right past that and go on 842 00:41:49,120 --> 00:41:51,240 and do something profound and do you? 843 00:41:51,400 --> 00:41:54,880 This next question does touch on it matters to your company. 844 00:41:54,880 --> 00:41:58,880 Like, do you think we're gonna enter a world where it's like I 845 00:41:58,880 --> 00:42:01,880 have my agent and it's like it's my sidekick, I want it to do 846 00:42:01,880 --> 00:42:04,120 everything. I want that agent to sort of 847 00:42:04,120 --> 00:42:06,400 interact on my behalf? Or do you think it's more this 848 00:42:06,400 --> 00:42:10,760 sort of like fleets of agents, different tasks more dispersed 849 00:42:10,760 --> 00:42:12,680 than that? Oh wow, that's a really good 850 00:42:12,680 --> 00:42:14,080 question. I mean, what do you think? 851 00:42:14,440 --> 00:42:17,720 Yeah, I what do I, I mean, I think I want the singular agent, 852 00:42:17,760 --> 00:42:20,360 like even if the agent is sort of coordinating with obviously 853 00:42:20,360 --> 00:42:23,040 like other agents, just something with like all the 854 00:42:23,040 --> 00:42:24,720 context. I mean, maybe it's a terrible 855 00:42:24,720 --> 00:42:28,080 security idea, but like, I don't know, it feels. 856 00:42:29,440 --> 00:42:32,880 But I will say the opposite of that is I thought with Openclaw 857 00:42:32,880 --> 00:42:36,120 we saw what was appealing about multi agent where they're 858 00:42:36,120 --> 00:42:38,280 interacting with each other and maybe there's more benefit. 859 00:42:38,280 --> 00:42:40,680 I don't, I don't know. I don't it's it's chaotic. 860 00:42:40,680 --> 00:42:42,480 It's hard to sort of game out the world. 861 00:42:42,480 --> 00:42:46,360 But I do think, you know, for for you guys, if I have this 862 00:42:46,360 --> 00:42:49,520 agent with all the context, that's sort of my sidekick, you 863 00:42:49,520 --> 00:42:51,280 know, maybe it's keeping more of the knowledge itself. 864 00:42:51,280 --> 00:42:54,360 It doesn't need to go to sort of external places to understand 865 00:42:54,360 --> 00:42:55,880 the world as much or I don't know. 866 00:42:56,120 --> 00:42:57,680 Yeah, that's a really good question. 867 00:42:57,680 --> 00:42:59,720 And to be honest, I don't really know if I have the best answer 868 00:42:59,720 --> 00:43:04,000 there because my guess is like, well, I mean, even let's take a 869 00:43:04,000 --> 00:43:05,600 look at the shape of the world today, right? 870 00:43:05,600 --> 00:43:07,040 And then maybe we can hypothesize. 871 00:43:07,480 --> 00:43:10,920 Right now I think there's a combination of both because you 872 00:43:10,920 --> 00:43:14,480 know, you're probably using one of two between, you know, like 873 00:43:15,080 --> 00:43:16,880 Chatchy, BT and Claw. I'm mostly clawed right now. 874 00:43:16,880 --> 00:43:20,240 Yeah, it was hard to move because I was deeply chatchy BT 875 00:43:20,240 --> 00:43:22,400 and it had all those contacts. And then and. 876 00:43:22,400 --> 00:43:25,000 Then, and it's like, OK, it seems like Claude's smarter. 877 00:43:25,560 --> 00:43:26,800 Yeah. Which do you have? 878 00:43:26,800 --> 00:43:28,600 A loyalty. Or I'm, I'm pretty, I'm pretty 879 00:43:28,600 --> 00:43:32,800 loyal to the to the Claude team. You're like, it's an ally. 880 00:43:33,240 --> 00:43:35,600 Yeah, yeah, exactly. I mean, like in, in many 881 00:43:35,600 --> 00:43:37,560 different ways, but I'm also just a bit very personal. 882 00:43:37,560 --> 00:43:39,960 Nothing against open AI, like, you know, any of their products 883 00:43:39,960 --> 00:43:41,440 at all. Just my personal preferences. 884 00:43:42,040 --> 00:43:44,040 So it has all my memory. And by the way, I don't know if 885 00:43:44,040 --> 00:43:46,040 you have memory turned on for your quad. 886 00:43:46,040 --> 00:43:48,880 Yeah, yeah, yeah. I didn't realize you can just go 887 00:43:48,880 --> 00:43:51,320 in and read your memory. Oh, I need to do that. 888 00:43:51,400 --> 00:43:53,720 It's like somewhere in the settings I literally was 889 00:43:53,720 --> 00:43:55,840 remembered. I like found the tab I and then 890 00:43:55,840 --> 00:43:58,680 I was reading what it knows about me and my goodness, it 891 00:43:58,680 --> 00:43:59,080 knew. A. 892 00:43:59,080 --> 00:44:02,040 Lot I remember I like and I also recommend you asking this 893 00:44:02,040 --> 00:44:03,440 question. I had a ton of fun with it 894 00:44:04,000 --> 00:44:06,240 recently. It was like based on what you 895 00:44:06,240 --> 00:44:07,640 know about me, this is exact problem, right? 896 00:44:07,640 --> 00:44:10,680 Based on what you know about me, do a psycho now, a 897 00:44:10,680 --> 00:44:16,240 psychoanalysis of me. And my goodness, it was like, it 898 00:44:16,240 --> 00:44:18,640 was like it knew where I was born. 899 00:44:18,800 --> 00:44:22,040 It knew, you know, like it was like, yeah, since like, you 900 00:44:22,040 --> 00:44:25,080 know, because you move from like this rural town in China. 901 00:44:25,080 --> 00:44:29,120 Specific, specific name, you know, and you came to Canada 902 00:44:29,120 --> 00:44:32,360 when you're about 6 years old, it like, you know, this 903 00:44:32,360 --> 00:44:35,480 indicates that based on these other messages, it's like all of 904 00:44:35,480 --> 00:44:38,680 these whoa. I was like, I don't think 905 00:44:38,760 --> 00:44:43,120 anybody like, you know, knows, you know about me to the depth 906 00:44:43,120 --> 00:44:45,400 of that, but but Claude does right, right. 907 00:44:45,600 --> 00:44:48,080 You know, so I think that Claude, for instance, will be a 908 00:44:48,080 --> 00:44:49,960 personal agent or like there's going to be obviously these 909 00:44:49,960 --> 00:44:53,840 providers where it's servicing you as your personal agent and 910 00:44:53,840 --> 00:44:55,800 maybe that'll be the exposure of it. 911 00:44:56,640 --> 00:44:59,840 But does that mean that it'll be the only agent interact with? 912 00:44:59,840 --> 00:45:03,120 Does that mean that, you know, there's not going to be like 913 00:45:03,120 --> 00:45:06,760 specialized agents for your company handling support 914 00:45:06,760 --> 00:45:10,080 questions or writing code or doing all these things? 915 00:45:10,080 --> 00:45:13,080 I, I, I, I, I'd be hard pressed to believe that's just going to 916 00:45:13,080 --> 00:45:16,080 be that singular single one. I think specialization in the 917 00:45:16,080 --> 00:45:20,800 same way you know anything else is, but who knows. 918 00:45:21,640 --> 00:45:22,680 All right. Last question. 919 00:45:22,720 --> 00:45:25,560 I mean, yeah, sort of fast growing company and lots of 920 00:45:25,560 --> 00:45:27,840 excitement. What, what do you want to sort 921 00:45:27,840 --> 00:45:31,360 of hold yourself to in a year or what would you like to see your 922 00:45:31,360 --> 00:45:33,880 yourself and the company achieve over the next year? 923 00:45:35,280 --> 00:45:40,160 A lot of things, I think we the first and foremost thing that 924 00:45:40,160 --> 00:45:45,840 comes to mind is really making sure that the knowledge layer 925 00:45:45,840 --> 00:45:48,800 and the knowledge infrastructure that is only really talked about 926 00:45:49,280 --> 00:45:51,920 in my opinion, in the in the broader theory and the broader 927 00:45:51,920 --> 00:45:54,600 ether all really manifests into a product, into a meaningful 928 00:45:54,600 --> 00:45:57,160 product. We truly believe that the 929 00:45:57,160 --> 00:46:00,720 biggest inhibitor for a lot of agents and a lot of companies 930 00:46:00,720 --> 00:46:04,480 who are investing so deeply into agents right now is just not 931 00:46:04,520 --> 00:46:07,600 providing it with the rights and the accurate information, the 932 00:46:07,600 --> 00:46:09,720 context that it deserves to actually go and and do them. 933 00:46:09,960 --> 00:46:11,600 I personally think the intelligence, the models are 934 00:46:11,600 --> 00:46:13,520 there. I really do right. 935 00:46:13,520 --> 00:46:15,200 There's a whole debate on whether or not we've achieved a 936 00:46:15,200 --> 00:46:16,520 GII. Won't. 937 00:46:16,520 --> 00:46:18,440 I won't, I won't. We're just not good enough 938 00:46:18,440 --> 00:46:20,480 prompters, you're saying? It's like if we could only give 939 00:46:20,480 --> 00:46:23,080 it the right information, it could be AGI. 940 00:46:23,600 --> 00:46:25,800 Or close to it, right? Or at least achieve the vast 941 00:46:25,800 --> 00:46:29,680 majority of tasks that I think like a lot of the knowledge work 942 00:46:29,680 --> 00:46:34,320 is trying to automate right now. And we want to be the company to 943 00:46:34,360 --> 00:46:37,400 enable that and to kind of build that into an actual, you know, 944 00:46:37,440 --> 00:46:40,360 like into actual form of a product and and then enable that 945 00:46:40,360 --> 00:46:41,960 into the the other companies we work with. 946 00:46:42,600 --> 00:46:44,000 Amazing. Thank you so much for coming on 947 00:46:44,000 --> 00:46:45,360 the show. Thank you so much for having me. 948 00:46:45,640 --> 00:46:49,160 Thanks to Han, a fantastic guest, and Next up we have Jesse 949 00:46:49,160 --> 00:46:52,360 from Decagon, which ranked #4 in late stage. 950 00:46:52,400 --> 00:46:55,720 You can think of Deccagon as AI agents for your customer 951 00:46:55,720 --> 00:46:57,360 experience. So we work mostly with large 952 00:46:57,360 --> 00:47:00,080 businesses that have, you know, lots of customers and big call 953 00:47:00,080 --> 00:47:03,800 centers and contact centers and we deploy AI agents in front of 954 00:47:03,800 --> 00:47:06,320 those customers that have, you know, calls with them or chats. 955 00:47:07,120 --> 00:47:09,720 So we, we almost call it AAI Concierge a lot of the time 956 00:47:09,720 --> 00:47:14,760 because it is we, we think of it as like this new layer around 957 00:47:14,760 --> 00:47:16,840 your brand that customers can interact to. 958 00:47:16,840 --> 00:47:19,080 And this is the first part, first point of contact that a 959 00:47:19,080 --> 00:47:21,280 lot of customers will will even talk to in the first place. 960 00:47:21,800 --> 00:47:25,560 So there we're gonna have to sort of multiple audiences 961 00:47:25,560 --> 00:47:27,440 consuming this. I think from the investor 962 00:47:27,720 --> 00:47:31,560 perspective, the space you're in is like one of the two maybe 963 00:47:31,560 --> 00:47:35,080 with legal where there's like I guess almost certainty that 964 00:47:35,080 --> 00:47:36,600 there is going to be an AI answer, right? 965 00:47:36,600 --> 00:47:39,520 You are fighting with Sierra. There's this great tool and 966 00:47:39,520 --> 00:47:42,640 legal like Harvey Lagora even up there some others. 967 00:47:42,840 --> 00:47:46,160 But there's so much excitement that these are sort of customer 968 00:47:46,160 --> 00:47:48,480 experience and legal like these applications that really 969 00:47:48,480 --> 00:47:52,400 understand what's happening from the sort of, I guess, layperson 970 00:47:52,400 --> 00:47:56,120 view. There's the like, now I'm 971 00:47:56,120 --> 00:47:59,080 talking to AI instead of a human and like, am I happy about that? 972 00:47:59,160 --> 00:48:01,720 Am I not? And sort of convincing the sort 973 00:48:01,720 --> 00:48:05,480 of regular person that this is like a good experience for them 974 00:48:05,480 --> 00:48:08,240 and something that they'll ultimately be happy that AI is 975 00:48:08,240 --> 00:48:10,480 answering their question. So I'm excited to sort of get to 976 00:48:10,480 --> 00:48:14,280 both like big trends in this conversation today. 977 00:48:14,280 --> 00:48:17,520 I guess starting off from the sort of like investor tech 978 00:48:17,520 --> 00:48:20,760 industry perspective, like why do you think there is this 979 00:48:20,760 --> 00:48:24,680 excitement about like customer experience, like what is change, 980 00:48:24,840 --> 00:48:27,320 what technology has made this possible right now that you 981 00:48:27,320 --> 00:48:30,680 think there is going to be this sort of movement in how we 982 00:48:30,680 --> 00:48:32,960 handle sort of customer support? Yeah. 983 00:48:33,600 --> 00:48:34,880 I would put coding in there as well. 984 00:48:35,000 --> 00:48:37,520 For the big three, of course. Yeah, Coding, perhaps the 985 00:48:37,520 --> 00:48:38,160 biggest. You're right. 986 00:48:38,200 --> 00:48:41,080 Yeah, that's almost his own. Yeah, it's a meta. 987 00:48:41,200 --> 00:48:42,880 Sort of. Process, it's changing how we 988 00:48:42,880 --> 00:48:44,280 build all these companies and everything. 989 00:48:44,280 --> 00:48:47,680 Yes, coding is almost distinct to me because it's sort of like 990 00:48:47,680 --> 00:48:50,440 in the the how the sausage is getting made of it all. 991 00:48:50,440 --> 00:48:51,760 But yes, I agree. Yeah. 992 00:48:51,760 --> 00:48:55,920 So I would say the three, those are the kind of the three big 993 00:48:56,160 --> 00:48:58,680 use cases that have emerged. And it's, it's really just like, 994 00:48:58,680 --> 00:49:01,320 what are the models good at? And so it turns out that the 995 00:49:01,320 --> 00:49:04,560 models are very good at writing code. 996 00:49:04,720 --> 00:49:07,240 And because it's, it's also tokenized and obviously people 997 00:49:07,240 --> 00:49:08,960 have discovered that. And so there's tons of funding 998 00:49:08,960 --> 00:49:10,120 and tons of effort going to that. 999 00:49:10,960 --> 00:49:14,200 Similar to our space, I think folks have realized that one of 1000 00:49:14,200 --> 00:49:16,360 the things that AI is very good at is having conversations. 1001 00:49:17,120 --> 00:49:20,560 And so because our space is inherently conversational, 1002 00:49:20,560 --> 00:49:22,120 right? Like the product that we build, 1003 00:49:22,560 --> 00:49:25,520 its purpose is to have conversations with, with end 1004 00:49:25,520 --> 00:49:29,160 consumers that has lended itself very well to AI. 1005 00:49:29,560 --> 00:49:31,000 And then the market's also massive, right? 1006 00:49:31,720 --> 00:49:33,960 So I think when you have those combinations of massive market, 1007 00:49:33,960 --> 00:49:36,760 it is what the AI is good at. And then I would add a, a third 1008 00:49:36,760 --> 00:49:40,520 thing, which is it is, it is actually really easy to see 1009 00:49:40,520 --> 00:49:43,800 value because you can, it's very measurable, right? 1010 00:49:44,160 --> 00:49:47,200 There are a lot of use cases where it's like, Oh yeah, I can 1011 00:49:47,200 --> 00:49:49,800 see this being useful, but you know, how would I quantify that? 1012 00:49:49,800 --> 00:49:53,880 Like it is, is it like how do I show ROI basically and in our 1013 00:49:53,880 --> 00:49:56,760 spaces is very easy. Because they know companies know 1014 00:49:57,040 --> 00:49:59,160 I have all this call support like I'm having to answer 1015 00:49:59,160 --> 00:50:01,680 customer questions. So they're they're comparing all 1016 00:50:01,680 --> 00:50:03,760 right, this is what I'm spending now with you. 1017 00:50:03,760 --> 00:50:05,840 I could be spending less. Like how much is the pitch? 1018 00:50:06,280 --> 00:50:07,760 This is cheaper. Exactly. 1019 00:50:08,280 --> 00:50:09,920 Yeah. So if you think about like what 1020 00:50:09,920 --> 00:50:12,680 the conversations are or what conversations being had in these 1021 00:50:12,680 --> 00:50:16,840 boardrooms or C suites, it is like where can we apply AI? 1022 00:50:17,160 --> 00:50:19,200 And then the, the customer experience is, is one of the 1023 00:50:19,200 --> 00:50:21,280 biggest ones, right? And it's not just saving cost. 1024 00:50:21,320 --> 00:50:25,200 And so saving cost is one of the easiest ways to frame it because 1025 00:50:25,200 --> 00:50:27,640 it's like, oh, we're, we're spending 10s of millions or 1026 00:50:27,640 --> 00:50:30,040 hundreds of millions of dollars a year in, in our contact 1027 00:50:30,040 --> 00:50:32,160 center. And we know that a lot of that 1028 00:50:32,160 --> 00:50:35,280 could be handled by AI potentially at a higher level 1029 00:50:35,280 --> 00:50:36,440 with higher customer satisfaction. 1030 00:50:36,840 --> 00:50:39,600 So that's the cost saving side. There's also a whole other side 1031 00:50:39,600 --> 00:50:41,800 that we focus on which we would consider like the revenue 1032 00:50:41,800 --> 00:50:44,360 generating side and that that's that's pretty large because it's 1033 00:50:44,360 --> 00:50:46,240 very uncapped. It's like outbound sales. 1034 00:50:46,880 --> 00:50:48,480 Yeah. So we, we, we don't really touch 1035 00:50:48,480 --> 00:50:51,720 like the cold calling like that maybe maybe one day, but there's 1036 00:50:51,720 --> 00:50:53,560 a lot of sort of revenue generating conversations you 1037 00:50:53,560 --> 00:50:55,720 could have that are both reactive and proactive, right. 1038 00:50:55,760 --> 00:50:58,920 So imagine someone reaches out and you resolve their issue, but 1039 00:50:58,920 --> 00:51:01,920 you also notice certain things in their account that are being 1040 00:51:01,920 --> 00:51:05,480 underutilized or you're, you're trying to like tell them about 1041 00:51:05,480 --> 00:51:07,000 this, this new product that that was launched. 1042 00:51:07,480 --> 00:51:10,320 We've seen that land very well because you've just solved their 1043 00:51:10,320 --> 00:51:12,640 issue. And so you've kind of some 1044 00:51:12,640 --> 00:51:15,480 people call it like earn the right to, you know, engage them 1045 00:51:15,480 --> 00:51:17,840 more and like just keep them more engaged as a customer. 1046 00:51:17,840 --> 00:51:20,560 And then there's also. Proactive sell is a word that I 1047 00:51:20,840 --> 00:51:22,320 use right? Up, sell, cross, sell, right? 1048 00:51:23,200 --> 00:51:25,200 And then you could also be proactive at the right time, 1049 00:51:25,280 --> 00:51:26,680 right? So imagine you know someone 1050 00:51:26,680 --> 00:51:28,720 signed up, they're going through onboarding flow and you notice 1051 00:51:28,720 --> 00:51:31,640 that you know they, if they've not been active for a week or 1052 00:51:31,640 --> 00:51:33,880 two, you can reach out at the right time and actually like 1053 00:51:33,880 --> 00:51:35,880 increase your conversion rates because maybe they were stuck on 1054 00:51:35,880 --> 00:51:37,840 something and you just unblocked it for them. 1055 00:51:37,960 --> 00:51:41,680 And what's the breakdown? Voice sort of chat. 1056 00:51:42,040 --> 00:51:44,320 E-mail. We're pretty balanced at this 1057 00:51:44,320 --> 00:51:49,040 point in terms of voice and chat emails, much smaller and so I 1058 00:51:49,040 --> 00:51:53,440 think the rough frame of reference might be like 454510. 1059 00:51:54,280 --> 00:52:00,840 And voice people know that it's AI or how close are we to sort 1060 00:52:00,840 --> 00:52:03,760 of seeming like a human and is that something we want? 1061 00:52:05,200 --> 00:52:08,760 It is pretty close. I think there's you're never 1062 00:52:08,760 --> 00:52:10,320 trying to like hide that it's AI. 1063 00:52:10,320 --> 00:52:14,920 So almost always the first message is hi, you know, I am 1064 00:52:15,000 --> 00:52:17,560 Jesse, your AI, you know, concierge. 1065 00:52:18,280 --> 00:52:22,560 But what we've seen is like, we just went live, you know, last 1066 00:52:22,560 --> 00:52:24,400 week with a large bank in the US. 1067 00:52:24,400 --> 00:52:27,480 And when, when they look through the data, like one of the things 1068 00:52:27,480 --> 00:52:31,920 both of us, us and the bank were like very impressed by was how 1069 00:52:31,920 --> 00:52:33,640 just like normal all the conversations were. 1070 00:52:33,680 --> 00:52:35,960 So there wasn't, there was no like, oh, like, are you an AI or 1071 00:52:35,960 --> 00:52:37,760 anything? It's like, yes, like you say 1072 00:52:37,760 --> 00:52:40,200 that you're AI, but because the voice in the conversation flows 1073 00:52:40,200 --> 00:52:42,520 so smoothly, like no one really cares after that. 1074 00:52:42,520 --> 00:52:43,960 Like they're just trying to get their issue solved. 1075 00:52:44,120 --> 00:52:46,600 And so if you can get your issue solved throughout the 1076 00:52:46,600 --> 00:52:47,840 conversation, they probably just forget. 1077 00:52:47,960 --> 00:52:50,440 And so we see a lot of nice conversations that end up with, 1078 00:52:50,960 --> 00:52:53,600 you know, like this is so helpful, have a nice day like 1079 00:52:53,600 --> 00:52:56,200 that, that sort of thing where you would, you would not really 1080 00:52:56,200 --> 00:52:58,440 say have a nice day to an AI, It's nice. 1081 00:52:58,480 --> 00:53:00,200 Well, especially if they become our overlords. 1082 00:53:00,200 --> 00:53:02,960 It's good to thank them enough times along the way. 1083 00:53:03,400 --> 00:53:04,720 Exactly. You always want to be nice to 1084 00:53:04,720 --> 00:53:08,480 them, but yeah, so I think if you can make the experience good 1085 00:53:08,480 --> 00:53:09,840 enough, then it doesn't really matter. 1086 00:53:09,840 --> 00:53:13,880 And in fact, we've seen that in a lot of these experiments that 1087 00:53:13,880 --> 00:53:16,840 where we've deployed AI, the customer satisfaction actually 1088 00:53:16,840 --> 00:53:18,920 becomes higher than. What do you deal with? 1089 00:53:19,360 --> 00:53:22,200 Like the customer just says, human, human, human. 1090 00:53:22,200 --> 00:53:25,800 Like how do you treat that? How do you do you just resist it 1091 00:53:25,800 --> 00:53:28,760 because you want to sort of show them, well, this is actually a 1092 00:53:28,760 --> 00:53:30,480 great experience? Or how do you deal with the, 1093 00:53:31,040 --> 00:53:33,520 yeah, desire for speaking with a human being? 1094 00:53:34,760 --> 00:53:40,640 The goal really is to make the experience very different very 1095 00:53:40,640 --> 00:53:42,480 quickly. So you have to as soon as 1096 00:53:42,480 --> 00:53:45,480 possible establish that this is not one of these phone trees 1097 00:53:45,480 --> 00:53:47,920 that you're used to that are frustrating and you get stuck in 1098 00:53:47,920 --> 00:53:49,960 loops and you have to press one for this, Press 2 for that. 1099 00:53:51,040 --> 00:53:52,640 So the goal is to make that super clear. 1100 00:53:53,000 --> 00:53:58,880 And that means just showing the AIS like empathy in the voice 1101 00:53:58,880 --> 00:54:00,760 and being able to make a super personalized. 1102 00:54:00,760 --> 00:54:03,320 And you, you need to do that in like the first or second 1103 00:54:03,320 --> 00:54:05,360 message. And so we had, we had a 1104 00:54:05,360 --> 00:54:08,400 deployment with Ora ring where we did a case study in 1105 00:54:08,920 --> 00:54:12,000 beforehand. It was every three people that 1106 00:54:12,000 --> 00:54:14,400 came in was just agent, agent, agent, like I don't want to 1107 00:54:14,400 --> 00:54:18,240 bother with this. And now about six months in from 1108 00:54:18,240 --> 00:54:21,680 the deployment, at that point, we, we measured it and it became 1109 00:54:21,680 --> 00:54:24,720 one in 20. So it, it became way smaller 1110 00:54:24,720 --> 00:54:27,160 because people were willing to give it more of a chance because 1111 00:54:27,200 --> 00:54:28,400 you could show that it was different. 1112 00:54:28,400 --> 00:54:30,840 And then once they got into the conversation, it's like, Oh 1113 00:54:30,880 --> 00:54:33,240 yeah, this actually can do stuff for me, you know? 1114 00:54:34,200 --> 00:54:39,120 How much of the experience and mostly I'm thinking with voice, 1115 00:54:39,120 --> 00:54:43,800 do you leave to like the bank or whatever either your customer 1116 00:54:44,040 --> 00:54:46,880 versus no, we sort of have a sensibility like for example, 1117 00:54:46,880 --> 00:54:50,760 like the voice you have like a sort of a menu or they could 1118 00:54:50,760 --> 00:54:52,440 totally bring in their own voice. 1119 00:54:52,720 --> 00:54:55,640 Like how do how do you think about what sort of your special 1120 00:54:55,640 --> 00:54:57,280 sauce versus stuff you can hand over? 1121 00:54:57,280 --> 00:55:00,640 Yeah. We see our role as an advisor to 1122 00:55:00,640 --> 00:55:01,920 the extent that they want it, right. 1123 00:55:01,920 --> 00:55:05,520 So we have our unique ability is that we have a ton of these 1124 00:55:05,520 --> 00:55:09,840 deployments where we've, you know, gone live And so we have a 1125 00:55:09,840 --> 00:55:11,960 lot of experience on how to design conversations. 1126 00:55:11,960 --> 00:55:14,840 And so on our team, we have these specialist roles that we 1127 00:55:14,840 --> 00:55:17,840 call conversation designers, which you didn't really like, 1128 00:55:17,840 --> 00:55:20,480 not really needed them before. But like now we have these like 1129 00:55:20,480 --> 00:55:24,680 really elite folks that are kind of there to advise the customer 1130 00:55:24,680 --> 00:55:26,960 on, oh, you, here's the procedure you want to do great, 1131 00:55:26,960 --> 00:55:27,960 that's great. We can do that. 1132 00:55:27,960 --> 00:55:30,880 But like, in our experience, it's better if you, you know, 1133 00:55:30,880 --> 00:55:34,200 combine these things or like direct the flow this way because 1134 00:55:34,200 --> 00:55:36,240 that's what is most natural for a Gen. 1135 00:55:36,240 --> 00:55:37,960 AI model. And so that, what kind of 1136 00:55:37,960 --> 00:55:39,960 advisor are there? And we, we partner with them on, 1137 00:55:39,960 --> 00:55:43,520 on getting to the fastest sort of solution possible. 1138 00:55:44,040 --> 00:55:46,560 Same with the voice. And So what we found is that 1139 00:55:47,760 --> 00:55:50,200 everyone has a pretty different opinion on what's a good voice. 1140 00:55:50,720 --> 00:55:53,600 And we'll talk to 1 customer and like, you know, they're like, 1141 00:55:53,600 --> 00:55:56,400 oh, this is absolutely the best voice and go somewhere else. 1142 00:55:56,400 --> 00:55:57,640 And like, I, I don't like that voice. 1143 00:55:57,640 --> 00:55:59,480 It's like too happy or something like that. 1144 00:56:00,000 --> 00:56:02,600 And so, you know, our conversation designers are also 1145 00:56:03,160 --> 00:56:05,520 sort of like voice experts and they they go in and can help 1146 00:56:05,520 --> 00:56:07,680 people navigate a sample of voices. 1147 00:56:08,160 --> 00:56:09,240 And sometimes they bring their own too. 1148 00:56:09,480 --> 00:56:11,120 And so we have the ability to clone a voice. 1149 00:56:11,360 --> 00:56:13,960 And so they've sometimes they've spent a bunch of money before, 1150 00:56:13,960 --> 00:56:15,640 like the voiding of voice. Yeah. 1151 00:56:16,080 --> 00:56:19,160 And so we can just take those samples and build that into the 1152 00:56:19,160 --> 00:56:21,360 AI. Do you have customers that 1153 00:56:21,360 --> 00:56:23,320 deploy different voices based on what they think? 1154 00:56:23,520 --> 00:56:26,640 This customer trying to match demographics or whatever of the 1155 00:56:26,640 --> 00:56:27,840 customer they're interacting with. 1156 00:56:28,200 --> 00:56:33,040 Yeah, so the most common one for sure by country, like if you 1157 00:56:33,040 --> 00:56:35,040 pick up the phone in the UK versus here, of course you're 1158 00:56:35,040 --> 00:56:37,600 gonna have a different accent. You can also even do it by zip 1159 00:56:37,600 --> 00:56:40,360 code so you can split the US like you can have Southern 1160 00:56:40,360 --> 00:56:42,520 accents or you know, Boston accents. 1161 00:56:42,600 --> 00:56:47,240 And does that have an impact? It has, I would say, like a mild 1162 00:56:47,240 --> 00:56:52,200 impact, but it's like, it's like hard to say if it's like a huge 1163 00:56:52,200 --> 00:56:55,480 thing, but I think it's, it's just more of this general theme 1164 00:56:55,480 --> 00:56:57,560 of how do we make it as personalized as possible. 1165 00:56:58,000 --> 00:57:00,960 And so it's both the sound of the voice, but also what you 1166 00:57:00,960 --> 00:57:03,280 say, right? So you one of the, one of the 1167 00:57:03,280 --> 00:57:06,880 big other advancements in the space in our products recently 1168 00:57:06,880 --> 00:57:10,880 is this concept of user memory. And like the reason why ChatGPT 1169 00:57:10,880 --> 00:57:13,040 feels like it gets better over time is that it can remember 1170 00:57:13,040 --> 00:57:15,760 things about you, right? And so we, we give our 1171 00:57:15,760 --> 00:57:19,600 customers, so the businesses that we work with the option to 1172 00:57:20,640 --> 00:57:24,600 turn on user memory and where the AI is creating this dynamic 1173 00:57:24,600 --> 00:57:27,440 memory profile so that over time it becomes more and more 1174 00:57:27,440 --> 00:57:29,360 personalized. And the cool thing is that this 1175 00:57:29,360 --> 00:57:31,520 memory profile doesn't even have to be specific to us. 1176 00:57:31,600 --> 00:57:33,320 So you can actually like integrate it into the rest of 1177 00:57:33,320 --> 00:57:35,520 your experience and have it just keep updating. 1178 00:57:36,080 --> 00:57:38,360 And so next time they come in, we know, we know exactly like 1179 00:57:38,360 --> 00:57:40,160 what you've done in the app. We know exactly like what you 1180 00:57:40,160 --> 00:57:42,760 talked about the last few times you contacted and it just makes 1181 00:57:42,760 --> 00:57:44,120 for a much, much better experience. 1182 00:57:44,360 --> 00:57:47,320 Do you think the business is going to remain balanced between 1183 00:57:47,320 --> 00:57:49,880 chat and voice, or like, do you have a view of where the world's 1184 00:57:49,880 --> 00:57:51,440 headed? Yeah, of course. 1185 00:57:51,440 --> 00:57:57,160 So I think prior to Gen. AI, the sort of the trend was, 1186 00:57:57,160 --> 00:57:59,560 hey, let's let's drive everything towards chat because 1187 00:57:59,560 --> 00:58:02,280 it's more efficient. You can have a human agent that 1188 00:58:02,280 --> 00:58:04,800 is doing like 3 to 5 chats at once. 1189 00:58:04,800 --> 00:58:07,400 Whereas over voice it's, it has to be single threaded, right? 1190 00:58:08,480 --> 00:58:11,800 I think now that issue is resolved because you can use AI 1191 00:58:12,440 --> 00:58:16,280 and I think in our view, it will be pretty balanced in the end 1192 00:58:16,280 --> 00:58:18,720 because there's just going to be different situations, different 1193 00:58:18,760 --> 00:58:20,320 demographics that prefer either one. 1194 00:58:20,920 --> 00:58:23,560 Like, you know, the, the common trope is, you know, younger 1195 00:58:23,560 --> 00:58:26,400 folks like chatting and older folks like calling. 1196 00:58:26,400 --> 00:58:29,960 But you know, if, if I'm in the car and I'm like on the go, I'd 1197 00:58:29,960 --> 00:58:33,400 rather call as well. And so I think both are, are 1198 00:58:33,400 --> 00:58:36,120 very natural, very natural means of communicating. 1199 00:58:36,480 --> 00:58:38,800 Voice is not going away. Like if you think about what 1200 00:58:38,800 --> 00:58:41,360 voice is like, voice was like the original UI for humans. 1201 00:58:42,080 --> 00:58:44,080 It's like before we had anything, before I had keyboards 1202 00:58:44,080 --> 00:58:46,400 or phones or anything, It's like the way you communicate is 1203 00:58:46,400 --> 00:58:48,880 through voice, and so that's not going anywhere. 1204 00:58:49,360 --> 00:58:51,920 Well, it's funny that we have these almost like putting this 1205 00:58:51,920 --> 00:58:54,440 voice thing aside for a second, we have like 2 contradicting 1206 00:58:54,440 --> 00:58:57,520 trends happening right now. One is obviously the rise of 1207 00:58:57,520 --> 00:59:00,080 language models and sort of everything in this chat bot. 1208 00:59:00,280 --> 00:59:02,720 On the other hand, there's TikTok, which is like nobody 1209 00:59:02,720 --> 00:59:05,480 wants to read everything short form video like the way people 1210 00:59:05,480 --> 00:59:09,120 get news information is video. And in some ways, voice is like 1211 00:59:09,120 --> 00:59:12,640 sort of the synthesis of these two where so you could see sort 1212 00:59:12,640 --> 00:59:15,680 of cultural movement go back towards conversation if the 1213 00:59:15,680 --> 00:59:17,720 technology is able to deliver it. 1214 00:59:17,920 --> 00:59:18,320 Yeah. And. 1215 00:59:18,360 --> 00:59:20,560 I think it's just a much more natural form of communicating 1216 00:59:20,640 --> 00:59:23,960 and now even when I use ChatGPT whenever I can, if I'm not in a 1217 00:59:23,960 --> 00:59:26,880 crowded room or something, I use the voice medium just. 1218 00:59:26,880 --> 00:59:28,280 Cuz I don't make that interesting. 1219 00:59:28,520 --> 00:59:31,840 Does it get does it get extra information from the 1220 00:59:31,840 --> 00:59:35,440 emotionality yet or or no, It's just like reading it as text. 1221 00:59:35,480 --> 00:59:36,200 Right. It does. 1222 00:59:36,200 --> 00:59:40,960 So what's that model is it's it's known as like a voice to 1223 00:59:40,960 --> 00:59:43,240 voice. So it's like kind of audio in, 1224 00:59:43,240 --> 00:59:48,560 audio out. And that is, I think most people 1225 00:59:48,560 --> 00:59:50,600 would say, including us, that that is like the long term 1226 00:59:50,600 --> 00:59:52,760 future of the space. There are a lot of problems with 1227 00:59:52,760 --> 00:59:55,200 those models right now, like they're a little bit 1228 00:59:55,200 --> 00:59:57,240 inconsistent and you have hallucination issues. 1229 00:59:57,920 --> 01:00:02,160 So in production when we work with like a bank or a telecom or 1230 01:00:02,160 --> 01:00:04,840 something, like of course you have to be a lot more careful 1231 01:00:04,840 --> 01:00:06,320 because you can't make any mistakes. 1232 01:00:06,840 --> 01:00:09,760 And so you you can't necessarily use the same technology as, you 1233 01:00:09,760 --> 01:00:11,120 know, consumer attach EBT for that. 1234 01:00:11,840 --> 01:00:14,640 But in general, yeah, we think that is the future. 1235 01:00:15,160 --> 01:00:18,040 The challenge with voice to voice is it's like harder to 1236 01:00:18,040 --> 01:00:20,800 sort of audit it in text, right? Or it's like. 1237 01:00:21,120 --> 01:00:24,120 It's harder to audit it's. Like it lives in this voice to 1238 01:00:24,120 --> 01:00:26,680 voice sort of. So there's like emotionality and 1239 01:00:26,680 --> 01:00:28,840 stuff that we can't really translate necessarily to 1240 01:00:29,760 --> 01:00:31,920 language. Is that the right way to explain 1241 01:00:31,920 --> 01:00:33,800 it? Or like, what's the barrier to 1242 01:00:33,800 --> 01:00:36,480 voice to voice? Yeah, the one of the barriers, 1243 01:00:36,480 --> 01:00:39,360 so there's the, the main barrier I would describe as the 1244 01:00:39,360 --> 01:00:42,160 hallucination rate is higher. Why is hallucination rate 1245 01:00:42,160 --> 01:00:43,240 higher? There's a bunch of reasons. 1246 01:00:43,240 --> 01:00:46,880 And one of the reasons is that the number of tokens streamed by 1247 01:00:46,880 --> 01:00:50,120 voice models, voice to voice models is a lot higher because 1248 01:00:50,280 --> 01:00:52,440 similarly, what you said, right, you're, you're capturing more 1249 01:00:52,440 --> 01:00:55,080 detail, you're capturing like intonation, etcetera. 1250 01:00:55,520 --> 01:00:58,680 And so for a sentence, for any given sentence, if you chop it 1251 01:00:58,680 --> 01:01:02,080 up in text, it would, let's say, be like, you know, 10 tokens or 1252 01:01:02,080 --> 01:01:05,880 something in voice, it could be like 80 to 100. 1253 01:01:06,560 --> 01:01:09,240 And the more tokens you have, the more opportunity there is to 1254 01:01:09,240 --> 01:01:12,000 mess up. And so that's why you see higher 1255 01:01:12,000 --> 01:01:14,240 hallucination rates in these voice to voice models. 1256 01:01:15,000 --> 01:01:18,680 And a lot of the research the labs are doing are, is, is kind 1257 01:01:18,680 --> 01:01:19,760 of geared towards making that better. 1258 01:01:20,160 --> 01:01:23,960 Because I think we would all, I think we all want the voice to 1259 01:01:23,960 --> 01:01:26,320 be very emotive and like to capture our emotions, right, 1260 01:01:26,320 --> 01:01:27,320 just like we're talking right now. 1261 01:01:27,880 --> 01:01:31,760 But until the loose nature rate is cleaned up, you can't really 1262 01:01:31,760 --> 01:01:34,440 leverage them in these like big production use cases in like in 1263 01:01:34,440 --> 01:01:37,960 our space. How do you stop like weird edge 1264 01:01:37,960 --> 01:01:41,840 cases where like so an agent says something really sort of 1265 01:01:41,840 --> 01:01:43,360 bad. Obviously a human could do that 1266 01:01:43,360 --> 01:01:45,240 too. Like what are what are the 1267 01:01:45,240 --> 01:01:48,080 systems you have where it's like, oh man, we had one really 1268 01:01:48,320 --> 01:01:51,400 sort of fire off weirdly. Like you're sort of have another 1269 01:01:52,040 --> 01:01:54,400 sort of layer of technology monitoring everything or what? 1270 01:01:54,400 --> 01:01:58,320 What do you do there to catch when inevitably some weird thing 1271 01:01:58,320 --> 01:01:59,640 emerges? Yeah. 1272 01:01:59,640 --> 01:02:04,760 So that is that is a huge topic and we we think of it as like a 1273 01:02:04,760 --> 01:02:07,360 three prong problem. It's not just checking it during 1274 01:02:07,360 --> 01:02:10,000 the during the call which you have to do, but you also you 1275 01:02:10,000 --> 01:02:12,040 also can prepare for it beforehand and review it 1276 01:02:12,040 --> 01:02:14,560 afterwards. So the, the three prongs I would 1277 01:02:14,560 --> 01:02:16,360 consider are before the conversation, during the 1278 01:02:16,360 --> 01:02:18,920 conversation and after the conversation, but before the 1279 01:02:18,920 --> 01:02:22,320 conversation, what you can do is we, we call them simulations. 1280 01:02:22,760 --> 01:02:26,840 And so you have these AI agents that you've made and before you 1281 01:02:26,840 --> 01:02:29,520 release them to any customers, you kind of run them through 1282 01:02:29,520 --> 01:02:31,240 simulations. So it's almost like a second AI 1283 01:02:31,240 --> 01:02:33,520 comes in and it's talking to your agent and like trying to 1284 01:02:33,520 --> 01:02:36,480 get to mess up or testing all the common use cases that it's 1285 01:02:36,480 --> 01:02:38,600 learned from reading historical transcripts, right? 1286 01:02:39,160 --> 01:02:40,840 So that that's the first thing you do because that really 1287 01:02:40,840 --> 01:02:42,880 shores it up and that allows you to iterate faster because let's 1288 01:02:42,880 --> 01:02:45,840 say next week I want to change something well before I release 1289 01:02:45,840 --> 01:02:47,560 it, I can just run these simulations and I feel good 1290 01:02:47,560 --> 01:02:50,680 about releasing it right during the conversation. 1291 01:02:50,680 --> 01:02:52,400 It there is what what you're saying? 1292 01:02:53,000 --> 01:02:55,960 So those are like supervisor models and those have to be like 1293 01:02:55,960 --> 01:02:59,640 really fast specialized models that go and detect for certain 1294 01:02:59,640 --> 01:03:02,000 things. And so I'll give an example, 1295 01:03:02,400 --> 01:03:05,600 let's say we'll use the bank example again. 1296 01:03:05,600 --> 01:03:08,200 Like one of the things that if I'm a bank, I would not want the 1297 01:03:08,200 --> 01:03:10,800 AI to do necessarily is give financial advice, right? 1298 01:03:11,040 --> 01:03:13,840 So I don't want you to get financial advice that's not in 1299 01:03:13,840 --> 01:03:17,880 the scope of what your job is. And so how do you make sure that 1300 01:03:17,880 --> 01:03:20,320 it never does that even if the user's like trying to get, 1301 01:03:20,840 --> 01:03:22,640 they're like really determined to get financial advice. 1302 01:03:23,760 --> 01:03:25,640 The way you do it is you have to have these supervisor models. 1303 01:03:25,840 --> 01:03:28,560 And so there's these small models that we've, we've fine 1304 01:03:28,560 --> 01:03:31,160 tune ourselves that are really fast and really specialized in 1305 01:03:31,160 --> 01:03:33,920 detecting things like that. And you run them during the 1306 01:03:33,920 --> 01:03:36,120 conversation. And if it detects A violation, 1307 01:03:36,760 --> 01:03:39,560 it can fix it in real time before the, before the response 1308 01:03:39,560 --> 01:03:41,440 goes out. And then finally, the third 1309 01:03:41,440 --> 01:03:44,800 prong is after the conversation. And so there what you do is you 1310 01:03:44,800 --> 01:03:47,080 can review the conversations with a second AI, right? 1311 01:03:47,960 --> 01:03:52,040 And we, we were talking a little bit before, but I think this is 1312 01:03:52,040 --> 01:03:53,440 where a lot of the space is going. 1313 01:03:53,600 --> 01:03:57,640 Because if you think about the big advancements in AI right 1314 01:03:57,640 --> 01:04:00,000 now, it's these slow reasoning models that are really good at 1315 01:04:00,000 --> 01:04:01,440 coding and really good at reasoning. 1316 01:04:02,440 --> 01:04:04,680 You're never going to use them in the middle of a call because 1317 01:04:04,840 --> 01:04:07,880 like they could take like up to like a minute to reply, right? 1318 01:04:07,880 --> 01:04:09,480 And you're not going to wait there for a minute. 1319 01:04:09,480 --> 01:04:11,840 It's like, give me a second. Then it's like thinks for a 1320 01:04:11,840 --> 01:04:14,520 minute. And it doesn't even like like 1321 01:04:14,520 --> 01:04:16,520 improve those metrics very much anyways. 1322 01:04:17,000 --> 01:04:20,360 But what they are really good at is kind of this like slow 1323 01:04:20,360 --> 01:04:23,080 autonomous reasoning. And so one of the things that 1324 01:04:23,080 --> 01:04:25,400 that we've pioneered and we kind of released the, the first 1325 01:04:25,400 --> 01:04:27,720 product in, in our space, it's called Duet. 1326 01:04:27,800 --> 01:04:30,960 So Decagon Duet and the idea is it's kind of a duet between that 1327 01:04:30,960 --> 01:04:33,280 like slow reasoning model and like the the fast one, right? 1328 01:04:33,440 --> 01:04:36,680 And so the reasoning model, their job is they can basically 1329 01:04:36,680 --> 01:04:39,840 run overnight and autonomously. You just like read every single 1330 01:04:39,840 --> 01:04:45,600 conversation and you know, basically like figure out what 1331 01:04:45,600 --> 01:04:48,400 is going well, what's not going well, figure out what you need 1332 01:04:48,400 --> 01:04:49,880 to do next and actually do it for you. 1333 01:04:50,000 --> 01:04:51,680 How? How does it yeah implement the 1334 01:04:51,680 --> 01:04:55,880 learn what it learns. So the, the, let's say it 1335 01:04:55,880 --> 01:04:58,200 figures out like, oh, there's this one topic that we are not 1336 01:04:58,200 --> 01:05:01,240 doing really well at. And I've looked at our knowledge 1337 01:05:01,240 --> 01:05:03,120 base, I've looked at all the procedures. 1338 01:05:03,400 --> 01:05:05,200 We, we call them agent operating procedures. 1339 01:05:05,200 --> 01:05:06,560 AO PS. I've looked at all the AO PS 1340 01:05:06,920 --> 01:05:09,120 I've looked at like the coded tools that the AI is able to 1341 01:05:09,120 --> 01:05:10,880 use. I've looked at, you know, the 1342 01:05:10,880 --> 01:05:12,160 data that's come in from these tools. 1343 01:05:12,640 --> 01:05:15,600 And I've realized that the problem in this case is that, 1344 01:05:15,880 --> 01:05:18,640 you know, in this step of the procedure, we're like sending 1345 01:05:18,640 --> 01:05:21,440 people off a wrong track because like they don't actually want to 1346 01:05:21,440 --> 01:05:24,600 do that. And so because I've read 50,000 1347 01:05:24,600 --> 01:05:27,840 conversations, I'm very confident that the fix is XYZ 1348 01:05:27,840 --> 01:05:30,120 and it'll go and suggest the fix in the morning. 1349 01:05:30,120 --> 01:05:32,280 Someone on the team can come in and just like fix it, right? 1350 01:05:32,280 --> 01:05:36,080 Well, so if you think about how the space generally works, like 1351 01:05:36,080 --> 01:05:38,680 a lot of software, even outside our space, people are building 1352 01:05:38,680 --> 01:05:42,000 these AI assistants. But generally the AI assistants 1353 01:05:42,000 --> 01:05:45,200 are there to like, oh, like, show me how to find this or 1354 01:05:45,200 --> 01:05:46,960 like, go do this for me and I'll like, go do it. 1355 01:05:47,720 --> 01:05:50,720 But what we're talking about, what duet is, is it can do that, 1356 01:05:50,720 --> 01:05:53,320 but it's also like a autonomous like agent that can run in the 1357 01:05:53,320 --> 01:05:55,120 background. And that's what a lot of the 1358 01:05:55,120 --> 01:05:56,960 coding tools are moving towards right now, right. 1359 01:05:56,960 --> 01:06:00,360 And obviously I would argue that our, the work that our agent has 1360 01:06:00,360 --> 01:06:02,520 to do is like a lot simpler than coding so that they actually do 1361 01:06:02,520 --> 01:06:05,240 a better job of it. And so that that's really what 1362 01:06:05,240 --> 01:06:07,000 we pioneered. And we, we have some folks in 1363 01:06:07,000 --> 01:06:09,360 our space who have released similar things, but they're, 1364 01:06:09,360 --> 01:06:11,800 they're more focused on that like first use case of just 1365 01:06:12,160 --> 01:06:14,680 like, OK, write this for me rather than like actually 1366 01:06:14,680 --> 01:06:17,280 something that can just like run overnight for several hours and 1367 01:06:17,280 --> 01:06:19,680 like get a lot of work done. I'm sure this is a piece of it, 1368 01:06:19,680 --> 01:06:24,440 but what how would you distill Decagon's right to win? 1369 01:06:24,440 --> 01:06:27,560 I mean, obviously you have very well known competitor and then 1370 01:06:27,560 --> 01:06:29,680 obviously an existing industry and I'm sure lots of other 1371 01:06:29,680 --> 01:06:32,080 startups that are unheard of trying to run at this space. 1372 01:06:32,080 --> 01:06:34,040 Like what is your right to win? Yeah. 1373 01:06:35,200 --> 01:06:39,240 So I think the the thing about AI is that the markets are huge 1374 01:06:39,240 --> 01:06:40,560 and there's a lot of parts opportunity. 1375 01:06:40,880 --> 01:06:43,080 So that's a great thing. And then the flip side of course 1376 01:06:43,080 --> 01:06:44,440 opportunity attracts a lot of players. 1377 01:06:45,440 --> 01:06:47,960 And so in our players, in our space, you have like newer 1378 01:06:47,960 --> 01:06:50,560 players like Gen. AI native companies like you 1379 01:06:50,560 --> 01:06:55,080 mentioned Sierra and you also have like the older ones like 1380 01:06:55,360 --> 01:06:57,440 you know, the big platforms, Google, Salesforce, etcetera. 1381 01:06:59,440 --> 01:07:01,920 And of course, our goal as a team is to destroy all of our 1382 01:07:01,920 --> 01:07:05,680 competitors. And we, you know, we have very, 1383 01:07:07,160 --> 01:07:09,200 we've hired a lot of like killers on the team. 1384 01:07:09,200 --> 01:07:12,040 You you ranked higher than Sierra on this on this list, 1385 01:07:12,160 --> 01:07:14,160 which is interesting. Yeah, yeah. 1386 01:07:14,160 --> 01:07:16,680 We continue destroying the competitors, you know, no, but 1387 01:07:16,680 --> 01:07:18,160 we, we have a lot of respect for our competitors. 1388 01:07:18,320 --> 01:07:20,240 And I think that's, that's also something that's that's very 1389 01:07:20,240 --> 01:07:21,280 important. Like when we're building 1390 01:07:21,280 --> 01:07:24,880 internally like there's, you can't like demean anyone or like 1391 01:07:24,880 --> 01:07:26,520 underestimate anyone. That's very careful. 1392 01:07:27,120 --> 01:07:29,320 But I would say ultimately what it comes down to is like a very 1393 01:07:29,320 --> 01:07:33,720 different product approach. So I think the reason why in a 1394 01:07:33,720 --> 01:07:37,960 lot of big spaces there are multiple winners or in our case, 1395 01:07:37,960 --> 01:07:42,360 if if we want to be the winner, it is because you, you, you have 1396 01:07:42,360 --> 01:07:45,240 to take like kind of a, a bet on a specific approach in the 1397 01:07:45,240 --> 01:07:49,840 product. And our bet really is that the, 1398 01:07:50,040 --> 01:07:54,840 the biggest differentiator for our products and products in our 1399 01:07:54,840 --> 01:07:59,360 space over time is going to be sort of the, the empowerment of 1400 01:07:59,360 --> 01:08:02,720 non-technical folks and sort of the decrease in the cost of 1401 01:08:02,720 --> 01:08:04,520 ownership. So you think about how a lot of 1402 01:08:04,520 --> 01:08:07,320 software normally works. It is the set of phases like 1403 01:08:07,320 --> 01:08:08,760 highly consuming. There's like whole like 1404 01:08:08,760 --> 01:08:11,880 professional services industries based around this because you 1405 01:08:11,880 --> 01:08:14,520 need technical resources to, you know, get sales force set up, 1406 01:08:14,520 --> 01:08:16,439 for example. And there's a lot of benefits to 1407 01:08:16,439 --> 01:08:17,479 that. Like you get really locked in 1408 01:08:17,479 --> 01:08:19,560 afterwards, you're they're kind of reliant on you. 1409 01:08:20,520 --> 01:08:23,040 But we think that with AI, you're going to have to take a 1410 01:08:23,040 --> 01:08:26,120 very different approach and a lot of our competitors are are 1411 01:08:26,319 --> 01:08:28,720 very grounded in in the older approach, which again. 1412 01:08:28,720 --> 01:08:31,240 It's not a big deal to do it for the whole company. 1413 01:08:31,240 --> 01:08:32,279 And it's not necessarily bad, right? 1414 01:08:32,279 --> 01:08:33,880 Like. You want like a product team to 1415 01:08:33,920 --> 01:08:35,960 say, oh, we're going to use Deccon for our. 1416 01:08:36,319 --> 01:08:37,200 Exactly. OK. 1417 01:08:37,439 --> 01:08:40,640 And we want the OPS team to be like, hey, we want to use Deccon 1418 01:08:40,640 --> 01:08:43,000 because they're going to allow us to move a lot faster and 1419 01:08:43,000 --> 01:08:46,200 we're not going to have to rely on engineering Sprint or like 1420 01:08:46,200 --> 01:08:47,560 call up the vendor every single time. 1421 01:08:48,359 --> 01:08:49,680 So that's our biggest differentiator. 1422 01:08:49,720 --> 01:08:51,520 And when a. Lot like Slack where a team can 1423 01:08:51,520 --> 01:08:54,880 deploy exactly, yeah. And so when a lot of these 1424 01:08:55,439 --> 01:08:57,439 businesses work with Deck and it's because of that, right, 1425 01:08:57,439 --> 01:08:59,319 they feel like, hey, we're a big business. 1426 01:08:59,319 --> 01:09:02,640 We are a lot of complexity. You know, getting the AI live is 1427 01:09:02,640 --> 01:09:05,760 maybe only about 20% of the work, 80% of the work is the 1428 01:09:05,760 --> 01:09:07,560 constant iteration. And we're going to build a lot 1429 01:09:07,560 --> 01:09:09,040 of new flows. We're going to have to add a lot 1430 01:09:09,040 --> 01:09:12,000 of new surface areas. And if we're reliance on the 1431 01:09:12,000 --> 01:09:14,240 vendor and it just kind of feels like a black box, that's not 1432 01:09:14,240 --> 01:09:16,479 possible. You have tech companies are more 1433 01:09:16,479 --> 01:09:20,160 likely to adopt diagon right or you you've been strong in tech. 1434 01:09:20,160 --> 01:09:25,840 So tech, financial services, airlines, telecom, I think a, a 1435 01:09:25,840 --> 01:09:28,479 lot of the older industries we've we've seen the same thing 1436 01:09:28,479 --> 01:09:33,840 where they, they have less strong engineers overall 1437 01:09:33,840 --> 01:09:36,520 potentially, or they still have a, a very strong engineering 1438 01:09:36,520 --> 01:09:39,520 team, but like they have a ton of other stuff to do and 1439 01:09:39,840 --> 01:09:42,760 building customer experience software is not like one of the 1440 01:09:42,760 --> 01:09:44,319 things that they want to specialize in, right? 1441 01:09:44,359 --> 01:09:48,240 And so in the past they would have had to hire a bunch of, you 1442 01:09:48,240 --> 01:09:50,840 know, professional services or like, you know, pay the vendor a 1443 01:09:50,840 --> 01:09:54,040 ton more. But in our case it's, it's 1444 01:09:54,040 --> 01:09:57,280 really that has become the big differentiator versus the the 1445 01:09:57,280 --> 01:10:00,120 older approach. There are a lot of Americans who 1446 01:10:00,120 --> 01:10:02,640 hate AI. Like how much do you think 1447 01:10:02,640 --> 01:10:05,960 that's a barrier to your business succeeding? 1448 01:10:05,960 --> 01:10:08,280 And like how much do you think there's going to be sort of like 1449 01:10:08,280 --> 01:10:11,520 a hearts and minds battle or it's just the product has to 1450 01:10:11,520 --> 01:10:14,200 speak for itself? Or what do you think sort of AI 1451 01:10:14,200 --> 01:10:18,240 sentiment translates into like sort of people's willingness to 1452 01:10:18,240 --> 01:10:21,800 engage with AI voice and customer support? 1453 01:10:22,320 --> 01:10:24,280 Oh, I think it's very important. I definitely don't think that's 1454 01:10:24,280 --> 01:10:26,400 something you can gloss over because. 1455 01:10:27,640 --> 01:10:30,840 And so when we're building the product, like a big mantra that 1456 01:10:30,840 --> 01:10:33,400 we try to adhere to is that we're also building for our 1457 01:10:33,400 --> 01:10:37,000 customers, customers because at the end of the day, it's like 1458 01:10:37,080 --> 01:10:39,040 the goal is to make their experience a lot better. 1459 01:10:39,160 --> 01:10:40,960 And that of course benefits our customers, right. 1460 01:10:41,000 --> 01:10:43,680 So if we're working with the airline and you know, let's say 1461 01:10:43,680 --> 01:10:45,360 we save them a bunch of money, but then all the customers are 1462 01:10:45,360 --> 01:10:47,800 pissed because they can't like get what they wanted to do done. 1463 01:10:48,280 --> 01:10:49,440 Like that's not a win for anyone. 1464 01:10:49,560 --> 01:10:52,400 Like that might be a short term win for the airline, but you 1465 01:10:52,520 --> 01:10:55,280 know, they know that like that's not really what we're going for. 1466 01:10:56,000 --> 01:10:58,920 And so a lot of the product work that we do is geared towards 1467 01:10:58,920 --> 01:11:01,800 like, how do you make the end experience better, right? 1468 01:11:01,800 --> 01:11:04,920 Things like user memory that we we talked about earlier and 1469 01:11:04,920 --> 01:11:06,360 those are things that are you allowed. 1470 01:11:06,360 --> 01:11:10,600 To do that across customers? Or is it user memory within a 1471 01:11:10,600 --> 01:11:12,440 customer? Yeah, we don't do that cross 1472 01:11:12,480 --> 01:11:14,000 customers. There's no data sharing across 1473 01:11:14,000 --> 01:11:16,080 customers. I don't even think people would 1474 01:11:16,080 --> 01:11:18,320 want that necessarily. But yeah, yeah, it's hard to 1475 01:11:18,320 --> 01:11:20,880 say. I mean, if you know me, why do I 1476 01:11:20,880 --> 01:11:23,040 have to waste all the time? You know what my preferences 1477 01:11:23,040 --> 01:11:25,080 are. But so I think what will happen 1478 01:11:25,200 --> 01:11:28,560 is there will also be consumer agents. 1479 01:11:29,040 --> 01:11:31,760 So I think what the world looks like in three to five years is 1480 01:11:32,400 --> 01:11:34,960 all the, all the brands will have their own AI concierge. 1481 01:11:35,280 --> 01:11:38,280 And you know, hopefully we're we're powering and helping them 1482 01:11:38,280 --> 01:11:40,560 a lot with those. But there's also going to be 1483 01:11:40,560 --> 01:11:46,560 consumer agents that users use. So if you and I want to use 1484 01:11:46,560 --> 01:11:48,080 something like right? It's like a lawyer. 1485 01:11:48,160 --> 01:11:50,520 It's like their lawyer, my lawyer sort of fighting with it, 1486 01:11:50,520 --> 01:11:52,240 right? And then, yeah, the agents will 1487 01:11:52,240 --> 01:11:52,800 connect, right? Isn't. 1488 01:11:53,040 --> 01:11:56,280 That going to be a nightmare for you like those my I mean, in 1489 01:11:56,280 --> 01:11:57,920 some ways you have to brace for that world, right? 1490 01:11:57,920 --> 01:12:01,880 Because my agent, you know, I assume it will be meaner and 1491 01:12:02,000 --> 01:12:05,440 like I like, there's a level of like propriety that I have, 1492 01:12:05,440 --> 01:12:08,080 whereas they're just like going to treat you like a system to be 1493 01:12:08,080 --> 01:12:11,840 sort of manipulated and and you know, try to find every hole to 1494 01:12:11,840 --> 01:12:15,440 get what we want, right. I don't think it makes our life 1495 01:12:15,440 --> 01:12:20,040 harder, I think it is generally good for us and the world 1496 01:12:20,040 --> 01:12:21,600 overall. Because it's an arms race, you 1497 01:12:21,600 --> 01:12:23,960 have to fight or. I don't think it's necessarily 1498 01:12:23,960 --> 01:12:25,400 arms race. I think it's just it creates 1499 01:12:25,400 --> 01:12:27,800 much more communication. So it's much easier. 1500 01:12:27,800 --> 01:12:29,800 Like I think right now a lot of a lot of this communication 1501 01:12:29,800 --> 01:12:31,600 doesn't even happen because like I don't, I don't even bother 1502 01:12:31,600 --> 01:12:34,200 calling that number because like I know I'm going to get stuck in 1503 01:12:34,200 --> 01:12:36,920 like some loop, but no, I'm going to tell my agent to do it 1504 01:12:37,080 --> 01:12:39,360 and the agent can actually get it done because like there's a 1505 01:12:39,360 --> 01:12:42,080 business agent over there. And so it just makes like 1506 01:12:42,320 --> 01:12:45,560 overall, like the number of interactions much, much 1507 01:12:45,560 --> 01:12:48,200 healthier. But you you're not going to play 1508 01:12:48,200 --> 01:12:52,160 on sort of my agent side or are you interested in that side of 1509 01:12:52,160 --> 01:12:54,080 the no? Right now we're very focused on 1510 01:12:54,080 --> 01:12:57,120 working with businesses. I mean, so the, the brand 1511 01:12:57,120 --> 01:12:59,520 agents, if you will, the consumer agents like I think 1512 01:12:59,520 --> 01:13:02,440 ChatGPT will probably it's like those type of tools will be, 1513 01:13:02,640 --> 01:13:04,960 will be the consumer agents. You think it'll be ChatGPT 1514 01:13:04,960 --> 01:13:05,480 itself? Chat. 1515 01:13:05,760 --> 01:13:09,800 GBT, I mean like Perplexity has an agents, Claude has an agent. 1516 01:13:10,160 --> 01:13:12,920 It's like apps like that that are like geared towards mass 1517 01:13:12,920 --> 01:13:14,920 consumer apps like Gemini for example. 1518 01:13:15,000 --> 01:13:18,520 Do you think any of your customers will say you're only 1519 01:13:18,520 --> 01:13:21,680 allowed to talk to us if it's you or like have you seen people 1520 01:13:21,680 --> 01:13:24,680 say? No, I actually think I would say 1521 01:13:24,680 --> 01:13:27,000 the, the sentiment in most of our the businesses we work with 1522 01:13:27,000 --> 01:13:29,680 is, is very positive. They're excited for agent to 1523 01:13:29,680 --> 01:13:30,720 agent world. Why? 1524 01:13:32,000 --> 01:13:36,280 Because I think that is like that will yield more business 1525 01:13:36,280 --> 01:13:38,600 overall because it's like bringing the barrier to entry a 1526 01:13:38,600 --> 01:13:41,640 lot lower, right? So like if you're a hotel or 1527 01:13:41,640 --> 01:13:46,280 something and like agents can make bookings on your on your 1528 01:13:46,400 --> 01:13:49,400 like in your platform and then you know, presumably you can 1529 01:13:49,400 --> 01:13:51,880 have a lot more because the buried entry is a lot lower. 1530 01:13:51,960 --> 01:13:54,720 Do you think like the rules of like who gets a refund when are 1531 01:13:54,720 --> 01:13:56,520 going to become more transparent? 1532 01:13:56,520 --> 01:13:59,480 Because it's going to be possible to sort of test every 1533 01:13:59,480 --> 01:14:03,240 customer system and say, all right, if you say it's broken 1534 01:14:03,240 --> 01:14:06,360 two days old, they'll get like, are the rules going to end up 1535 01:14:06,360 --> 01:14:09,640 being sort of publishable because it's going to be so 1536 01:14:09,640 --> 01:14:11,480 discoverable? Or people are going to play this 1537 01:14:11,480 --> 01:14:13,920 cat and mouse game, or they're going to sort of randomize it or 1538 01:14:13,920 --> 01:14:16,320 like what happens when it's, yeah, how do you see that 1539 01:14:16,320 --> 01:14:19,040 playing out? I, I think the, the, the steady 1540 01:14:19,040 --> 01:14:21,560 state is that there aren't going to be no like games necessarily 1541 01:14:21,560 --> 01:14:25,120 because like right now there may be some games because friction 1542 01:14:25,120 --> 01:14:30,120 ads, etcetera, right. But in in the in a steady state, 1543 01:14:30,200 --> 01:14:31,840 because it's two AIS working together. 1544 01:14:33,000 --> 01:14:35,040 Yeah, there are transparent rules there. 1545 01:14:35,040 --> 01:14:37,560 There also may be like judgement calls, but the AI is the one 1546 01:14:37,560 --> 01:14:40,120 making the judgement call. So it's like fairly unbiased. 1547 01:14:41,280 --> 01:14:43,360 That is. I think that's that's the goal 1548 01:14:43,360 --> 01:14:46,560 and that that's the world that we're driving towards right now. 1549 01:14:46,560 --> 01:14:48,720 I think there's just inefficiencies because you have 1550 01:14:48,720 --> 01:14:51,280 businesses that have like their leadership might not even like 1551 01:14:51,280 --> 01:14:54,000 want these games to be there. But it's just over decades, you 1552 01:14:54,000 --> 01:14:57,560 kind of built out these things and you know, you're kind of 1553 01:14:57,560 --> 01:15:03,320 afraid to change it because, you know, like it might dramatically 1554 01:15:03,320 --> 01:15:06,080 increase the number of inquiries coming in or change your PNL. 1555 01:15:06,920 --> 01:15:08,880 But part of our job is to really work with them through that 1556 01:15:08,880 --> 01:15:11,800 process and like design these conversations so that you're not 1557 01:15:11,800 --> 01:15:14,120 losing anything. And in fact, you're, you're both 1558 01:15:14,280 --> 01:15:16,600 driving up your returns and making the experience better. 1559 01:15:17,160 --> 01:15:19,120 And that's, I think that's a big reason. 1560 01:15:19,120 --> 01:15:21,520 Back to your question of like why the space has has, you know, 1561 01:15:21,960 --> 01:15:25,960 popped off so much, It's it's really like you, you have these 1562 01:15:25,960 --> 01:15:29,840 massive ROI case studies that have already been realized by 1563 01:15:29,840 --> 01:15:33,720 big businesses where tons of savings, tons of customer 1564 01:15:33,720 --> 01:15:36,640 experience increases and, and that's the goal. 1565 01:15:37,520 --> 01:15:40,760 Is, is what technology do you really want to see improve and 1566 01:15:40,760 --> 01:15:44,240 like how is where, where are we on like voice latency? 1567 01:15:44,400 --> 01:15:48,680 You also are an industry that sort of has to rely on the past 1568 01:15:48,680 --> 01:15:51,240 generation model or you need sort of smaller models. 1569 01:15:51,520 --> 01:15:55,360 And I'm sure it's a much more price sensitive customer base. 1570 01:15:55,360 --> 01:15:59,360 Like where are you most hopeful that technology improving and 1571 01:15:59,360 --> 01:16:01,920 the models improving will change your business? 1572 01:16:02,480 --> 01:16:04,840 A couple different axes, right? So there's the the reasoning 1573 01:16:04,840 --> 01:16:07,720 models, you know, with, with products like Duet, right? 1574 01:16:08,800 --> 01:16:12,800 The reasoning capabilities will continue to help those open 1575 01:16:12,800 --> 01:16:15,320 source models getting better, will continue to help our 1576 01:16:15,560 --> 01:16:18,400 underlying agent pipeline because like why do you use open 1577 01:16:18,400 --> 01:16:20,200 source models? It's mostly for performance 1578 01:16:20,200 --> 01:16:22,600 because you need latency improvements and you need the 1579 01:16:22,960 --> 01:16:24,720 models to like highly specialized. 1580 01:16:25,360 --> 01:16:27,760 And so we found that by fine tuning and training those 1581 01:16:27,760 --> 01:16:31,360 models, we're able to get like much faster latency and much 1582 01:16:31,360 --> 01:16:34,360 higher like comparable to higher performance than the big closed 1583 01:16:35,120 --> 01:16:37,360 source models. And then it's the voice models. 1584 01:16:37,680 --> 01:16:40,400 And so, you know, voice to voice models getting better and more 1585 01:16:40,400 --> 01:16:42,160 accurate. Like those are probably the 1586 01:16:42,160 --> 01:16:44,880 three big dimensions that we pay attention to and each one of 1587 01:16:44,880 --> 01:16:47,760 those improving improves significant things about our 1588 01:16:47,760 --> 01:16:49,960 product. What What open source models are 1589 01:16:49,960 --> 01:16:51,280 you most excited about right now? 1590 01:16:52,960 --> 01:16:55,200 I mean, you have people all over the world making open source 1591 01:16:55,200 --> 01:16:56,800 models, right? So China's obviously very good 1592 01:16:56,800 --> 01:16:59,520 at open source models. The open source models in the US 1593 01:16:59,520 --> 01:17:02,520 are improving and you know, hopefully, hopefully even 1594 01:17:02,520 --> 01:17:04,360 faster. And you also missed all. 1595 01:17:05,080 --> 01:17:09,000 And so I think we found that there these models are like good 1596 01:17:09,000 --> 01:17:12,720 at different things and our customers might also have 1597 01:17:12,720 --> 01:17:14,960 preferences. And so we have built in a way 1598 01:17:14,960 --> 01:17:18,120 that's very model agnostic and you can you know, swap in and 1599 01:17:18,120 --> 01:17:21,480 out different models. But yeah, we're we're very 1600 01:17:21,480 --> 01:17:23,760 hopeful that the smaller parameter models will get better 1601 01:17:23,760 --> 01:17:26,400 and better because that is just these. 1602 01:17:26,440 --> 01:17:29,320 Are using like DeepSeek mini Max any of those or? 1603 01:17:30,120 --> 01:17:35,560 Mr. All Quinn, we we think pretty highly of and then we're 1604 01:17:35,560 --> 01:17:39,840 keeping a close eye on the like the Gemma type models, Llama, 1605 01:17:39,840 --> 01:17:41,240 things like that. You think you have to stay away 1606 01:17:41,240 --> 01:17:43,880 from the Chinese models or? I don't. 1607 01:17:43,880 --> 01:17:47,000 You don't have to stay away from them, but you want to be fairly 1608 01:17:47,000 --> 01:17:49,840 diversified so that in situations where you don't want 1609 01:17:49,840 --> 01:17:52,040 to use them, you can. You just have some customers who 1610 01:17:52,040 --> 01:17:55,440 don't want to, yeah, which then limits how much you can go all 1611 01:17:55,440 --> 01:17:58,000 in on them. Yeah, but you wouldn't want to 1612 01:17:58,000 --> 01:17:59,360 go all in on certain models anyways. 1613 01:17:59,480 --> 01:18:03,040 It's like this, it's like like you want to because if you, if 1614 01:18:03,040 --> 01:18:06,320 you get over reliant, it's like an analogy would be like if 1615 01:18:06,320 --> 01:18:08,280 you're a country, you don't want to rely on another country for 1616 01:18:08,280 --> 01:18:10,360 all your oil. You know, it's like anything 1617 01:18:10,360 --> 01:18:12,440 that happened. So we want to make sure that 1618 01:18:12,760 --> 01:18:14,080 we're building in a very robust way. 1619 01:18:14,360 --> 01:18:17,960 Do you build your own models or is there any use in that? 1620 01:18:19,000 --> 01:18:22,080 We have our own model, but they're they're not like trained 1621 01:18:22,080 --> 01:18:24,280 from scratch. So if you're in the application 1622 01:18:24,280 --> 01:18:26,400 layer, what you're typically doing is you're using one of 1623 01:18:26,400 --> 01:18:28,720 these open source models. The open source models are not 1624 01:18:28,720 --> 01:18:31,920 actually be that good out-of-the-box, but you can fine 1625 01:18:31,920 --> 01:18:33,400 tune them. And if you do that correctly, 1626 01:18:33,400 --> 01:18:35,760 they actually become very performant at the tasks that 1627 01:18:35,760 --> 01:18:37,640 you. Want are you optimistic across 1628 01:18:37,640 --> 01:18:41,080 the application layer or what? Where do you think the 1629 01:18:41,080 --> 01:18:45,040 foundation models will just gobble up applications versus 1630 01:18:45,040 --> 01:18:47,560 where do you think there's opportunity to build stand alone 1631 01:18:47,560 --> 01:18:50,760 businesses? I'm obviously quite optimistic 1632 01:18:50,760 --> 01:18:52,200 about the application layer, I think. 1633 01:18:52,520 --> 01:18:54,680 Just that you started early enough and you're sort of 1634 01:18:54,680 --> 01:18:57,960 running alongside them or what? No, I actually think so. 1635 01:18:58,360 --> 01:19:01,680 One, I think most of the value will accrue to the application 1636 01:19:01,680 --> 01:19:03,800 layer 'cause you're solving like business problems. 1637 01:19:05,040 --> 01:19:07,640 I think the labs also agreed with that, which is why they're 1638 01:19:07,640 --> 01:19:09,560 building a lot into the application layer. 1639 01:19:10,880 --> 01:19:12,880 And I think the application layer is so vast that there's 1640 01:19:12,880 --> 01:19:15,640 going to be a lot of different ways to handle it. 1641 01:19:15,640 --> 01:19:19,880 But generally, I think a good framework is that most of the 1642 01:19:19,880 --> 01:19:22,200 labs will want to build applications that are fairly 1643 01:19:22,200 --> 01:19:28,600 broad, like broad and maybe like thin because they have such a 1644 01:19:28,600 --> 01:19:30,840 vast surface area that they want to build things that a ton of 1645 01:19:30,840 --> 01:19:34,200 people can use. So coding obviously is, is one 1646 01:19:34,200 --> 01:19:39,160 of those where it's, it's mostly like a, you know, like an app 1647 01:19:39,160 --> 01:19:41,920 that people can just pick up. If you think about our space, 1648 01:19:41,920 --> 01:19:43,960 it's quite different, for better or worse. 1649 01:19:43,960 --> 01:19:47,320 We have a very involved like post sale motion where, you 1650 01:19:47,320 --> 01:19:49,560 know, we talked about our conversation designers, but 1651 01:19:49,560 --> 01:19:52,240 they're, they're part of like a much broader team that we, we 1652 01:19:52,240 --> 01:19:55,560 use to actually work with our clients and, you know, get their 1653 01:19:55,560 --> 01:19:57,600 agents stand up, stood up and like help write these agent 1654 01:19:57,600 --> 01:20:00,720 operating procedures because, you know, we have the benefit of 1655 01:20:00,720 --> 01:20:03,440 having the expertise. And because you have that post 1656 01:20:03,440 --> 01:20:06,720 sale motion, it's, I think it's pretty unlikely that's there's 1657 01:20:06,720 --> 01:20:08,160 going to be an app that replaces it. 1658 01:20:10,080 --> 01:20:12,080 But over time, I think the application layer is like so 1659 01:20:12,080 --> 01:20:14,520 thick that there's a lot, a lot of stuff to build. 1660 01:20:14,520 --> 01:20:18,720 So I, I actually, I think there's going to be a pretty 1661 01:20:19,200 --> 01:20:21,640 bright future for most application companies. 1662 01:20:22,800 --> 01:20:25,960 What just as somebody sort of so close to the space and how 1663 01:20:25,960 --> 01:20:28,400 things are developing, like how do you think the world looks 1664 01:20:28,400 --> 01:20:31,760 differently in in five years? Specific to our space. 1665 01:20:31,760 --> 01:20:33,920 Or just broadly, like, what? What do you think? 1666 01:20:33,920 --> 01:20:34,960 Like, I don't know. Yeah. 1667 01:20:34,960 --> 01:20:37,080 For the regular person just watching this, trying to 1668 01:20:37,080 --> 01:20:39,560 understand how AI is going to change their lives, like what do 1669 01:20:39,560 --> 01:20:42,640 you think feels the most different in five years because 1670 01:20:42,640 --> 01:20:44,000 of what people are building today? 1671 01:20:44,400 --> 01:20:46,960 So a couple themes. So the first one in our space, 1672 01:20:47,000 --> 01:20:50,200 we, we kind of touched on this, but I think the way that 1673 01:20:50,280 --> 01:20:53,080 consumers are going to interact with any brand is going to be 1674 01:20:53,080 --> 01:20:55,760 fundamentally different. And you know, hopefully we are 1675 01:20:55,760 --> 01:20:57,800 again a major player in that. But if you think about the 1676 01:20:57,800 --> 01:21:01,920 previous shifts, right, like last one, let's say from 1677 01:21:02,040 --> 01:21:04,480 Internet to mobile, right, that that basically created like 1678 01:21:04,480 --> 01:21:09,000 entirely new UI for people to interact with, you know, the 1679 01:21:09,000 --> 01:21:11,000 brands that they need to interact with, whether to buy 1680 01:21:11,000 --> 01:21:13,320 things or to get something done, etcetera. 1681 01:21:14,480 --> 01:21:17,960 And AI is, is you can almost think of it as like a new UI. 1682 01:21:18,560 --> 01:21:22,080 And this UI is conversational. And so you can talk to it on the 1683 01:21:22,080 --> 01:21:25,360 phone, you can chat with it. But that is, that is something 1684 01:21:25,360 --> 01:21:28,200 fundamentally that's gonna change and you're gonna have. 1685 01:21:28,320 --> 01:21:30,040 But don't people like to push back on that? 1686 01:21:30,400 --> 01:21:33,280 Some people like shopping, you know, like some of some of these 1687 01:21:33,280 --> 01:21:36,560 cases, it's like where people talk about like travel and it's 1688 01:21:36,560 --> 01:21:39,880 like people like spending time booking travel in some of these 1689 01:21:39,880 --> 01:21:41,760 cases. Like, do you really think it's 1690 01:21:41,760 --> 01:21:45,800 gonna be like, hey, like text in go figure out my travel, go 1691 01:21:45,800 --> 01:21:47,840 figure out what? I don't think it's necessarily 1692 01:21:47,840 --> 01:21:49,760 like a clean replacement, right? Like going from web to mobile. 1693 01:21:49,760 --> 01:21:51,960 Some people still, if I'm on my computer, I'll just use the 1694 01:21:51,960 --> 01:21:53,560 website and people go to stores. So, right. 1695 01:21:53,600 --> 01:21:55,880 But it's more of like a creation of a new medium. 1696 01:21:56,000 --> 01:21:57,720 Yep. And yeah, there will be 1697 01:21:57,720 --> 01:21:59,640 situations where I would rather talk to the YouTube. 1698 01:21:59,680 --> 01:22:01,440 There might be other situations where I'd rather use the mobile 1699 01:22:01,440 --> 01:22:02,920 app. And so you have this like new UI 1700 01:22:02,920 --> 01:22:05,600 that's created. And this UI is a lot more 1701 01:22:05,600 --> 01:22:08,840 friendly for consumer agents because consumer agents can also 1702 01:22:08,840 --> 01:22:10,320 go to your website and try to click around. 1703 01:22:10,320 --> 01:22:12,040 But that's like a very crude approximation. 1704 01:22:13,000 --> 01:22:14,560 But agents, again, are very good at conversation. 1705 01:22:14,560 --> 01:22:15,760 And so the two agents are talking. 1706 01:22:15,760 --> 01:22:19,400 You just like back to my original point, you're basically 1707 01:22:19,400 --> 01:22:23,040 just massively increasing the number of healthy interactions 1708 01:22:23,040 --> 01:22:26,800 between consumers and brands. Whereas right now a lot of them 1709 01:22:26,800 --> 01:22:29,040 I would describe as like unhealthy or there's not even 1710 01:22:29,040 --> 01:22:31,080 happening because I can't be bothered to do them right. 1711 01:22:31,800 --> 01:22:33,080 So that's one big thing that will change. 1712 01:22:34,000 --> 01:22:36,760 I think this is an argument that's like if you hate talking 1713 01:22:36,760 --> 01:22:39,560 to AI customer support, don't worry, I know you're trying to 1714 01:22:39,560 --> 01:22:41,080 be better than the loops that everybody got. 1715 01:22:41,080 --> 01:22:43,560 And so in a lot of cases they are happy to talk to you. 1716 01:22:43,800 --> 01:22:46,600 But there is also the further pitch, which is soon enough you 1717 01:22:46,600 --> 01:22:50,280 can have somebody talk to us and then get the readout of like, 1718 01:22:50,280 --> 01:22:51,960 what transpired. And you don't actually have to 1719 01:22:51,960 --> 01:22:54,440 do it yourself, which I think people will be excited about. 1720 01:22:54,440 --> 01:22:55,640 OK. So that's one, yeah. 1721 01:22:55,800 --> 01:22:57,520 That's one that's basically our space, right? 1722 01:22:57,520 --> 01:22:58,680 So we're really excited about that. 1723 01:22:58,680 --> 01:23:01,000 And I think that's why the market is so large and why 1724 01:23:01,000 --> 01:23:05,360 there's so much interesting work to be done in our space more 1725 01:23:05,360 --> 01:23:07,400 broadly. I think something that will 1726 01:23:07,400 --> 01:23:13,120 happen a lot more is right now people think of AI as it's like 1727 01:23:13,120 --> 01:23:15,000 a tool there. It's kind of like Google, right? 1728 01:23:15,000 --> 01:23:17,000 You go to Google and you like search something and like, OK, I 1729 01:23:17,000 --> 01:23:19,440 go to my, I go to AI and like, I get something done. 1730 01:23:20,720 --> 01:23:24,160 But I think in, in three to five years, it'll be much more normal 1731 01:23:24,160 --> 01:23:27,320 to just have like AI running continuously and it's just like 1732 01:23:27,320 --> 01:23:30,800 always running on whatever. And it's, it's kind of happening 1733 01:23:30,800 --> 01:23:32,160 in the background, right? Because right now it's kind of 1734 01:23:32,160 --> 01:23:33,480 like you ping it and it does something. 1735 01:23:34,280 --> 01:23:38,560 But the sort of the, the standard practice that we're 1736 01:23:38,560 --> 01:23:41,080 moving towards is just like long running autonomous things. 1737 01:23:41,960 --> 01:23:43,920 And that's, that's cool for a bunch of reasons. 1738 01:23:44,600 --> 01:23:47,480 One, you can just accomplish like way greater tasks. 1739 01:23:48,000 --> 01:23:50,960 And two, you can really increase leverage because like you don't 1740 01:23:50,960 --> 01:23:52,520 have to be involved, like it's just doing things. 1741 01:23:53,240 --> 01:23:55,160 And so this is a, this is a big trend in the coding space. 1742 01:23:55,160 --> 01:23:57,880 If you're following those, like a lot of what? 1743 01:23:58,440 --> 01:24:00,920 You start to sleep, wake up and see what's happened overnight. 1744 01:24:00,960 --> 01:24:01,760 Exactly. Yeah. 1745 01:24:01,760 --> 01:24:05,080 Or what started with like it's, it's kind of tab auto completing 1746 01:24:05,080 --> 01:24:08,760 things Now it's, you know, it moved to OK, you, you give it 1747 01:24:08,760 --> 01:24:10,560 some prompt and it can like write whole things and you 1748 01:24:10,560 --> 01:24:12,960 review it. And then now it's moving towards 1749 01:24:12,960 --> 01:24:15,120 like, OK, it's just like, hey, let me give you some 1750 01:24:15,120 --> 01:24:16,840 instructions. And then like you just kick it 1751 01:24:16,840 --> 01:24:18,400 off and it's like a colleague doing work, right? 1752 01:24:19,360 --> 01:24:22,320 That's the same concept again behind Duet where we've kind of 1753 01:24:22,320 --> 01:24:24,600 created this, this new concept of like a long running agent in 1754 01:24:24,600 --> 01:24:26,160 the background. And I think that will happen 1755 01:24:26,160 --> 01:24:28,880 with consumer agents as well, where you just give it like long 1756 01:24:28,880 --> 01:24:32,160 tasks or it's constantly listening to you and it's just 1757 01:24:32,160 --> 01:24:37,040 like a always there like helper. Great, Jesse, thank you so much 1758 01:24:37,040 --> 01:24:38,480 for coming on the Newcomer Podcast. 1759 01:24:38,520 --> 01:24:39,600 Cool, Eric, thanks for having me and. 1760 01:24:39,600 --> 01:24:41,040 Congratulations for being on the list. 1761 01:24:41,040 --> 01:24:42,160 Thank you. All right, sweet. 1762 01:24:42,200 --> 01:24:44,280 Thank you. Thanks for sticking around to 1763 01:24:44,280 --> 01:24:45,280 the end. Please. 1764 01:24:45,480 --> 01:24:47,560 If you've made it this far, you've got a like comment, 1765 01:24:47,560 --> 01:24:50,080 subscribe. Excited to grow the channel. 1766 01:24:50,080 --> 01:24:53,560 And of course, you can find our writing and reporting on sharps 1767 01:24:53,560 --> 01:24:57,080 and venture capital at newcomer.co. 1768 01:24:57,560 --> 01:24:59,520 We also host events. You can check out what we're 1769 01:24:59,520 --> 01:25:02,880 doing at Newcomer dot events, share other podcasts, help us 1770 01:25:03,160 --> 01:25:06,640 get distribution, support the channel comment support, like 1771 01:25:06,640 --> 01:25:08,280 comment subscribe. You know, you know the deal. 1772 01:25:09,200 --> 01:25:10,720 Thanks. Thanks for being on the journey 1773 01:25:10,720 --> 01:25:11,960 with us. All right, see you next week.