Transcript 0:00 [upbeat music] You're listening to ASOTU CON Sessions, brought to you by the More Than Cars Creative Shop. Let us help your business strategize, produce, and deploy modern creative content fast. Now, on to the episode. 0:11 We're live at the ASOTU CON Sessions podcast stage with, I think, the last podcast of the day, which means we can go for about two hours. Let's go. Let's go. That's, that's my type of pod. 0:23 So some people in the More Than Cars community, the ASOTU community don't know, uh, uh, you know, I don't know why we keep it so much a secret, but Ben and I are business partners in a, in a crazy venture we call Auto Genius. 0:35 Yep, right down there. There's a bunch of geniuses around here. Let's go. Shout if you're a genius. Yeah. We got genius- Yeah... there, genius there, genius there, back there. Back there. Let's go. All these geniuses. 0:44 Uh, man, I, I, I debated on this, what, what to title this podcast. Yeah. CLI or MCP, market product fit- CLI wins... product mar- Yeah, CLI wins, right? That was an easy question- Yeah... right? 1:00 AI, speed, there's so many things that we can talk about here. Um, I wanna start with this. You've been in the auto industry for quite some time, and one of the things that you, you point out are moments in time, right? 1:14 These, like, critical moments in time where major shifts happen. Yeah. 1:19 And over the, let's call it, like, since the late '90s from a technology perspective, there's been these moments in time that radically shift the way we operate, and we're kind of falling into one of those. 1:31 But I'd, I'd like you to do the walkthrough of, like, give us a couple moments in time and how we're seeing a similar moment in time right now, and what that means- Hmm... and what people should be thinking through. 1:41 We really have two hours? Yeah, yeah. Let's go. Oh, wow. Let's go. All right. We have, like, an hour. So- Maybe 40 minutes... ta- let's go back to, uh, m- people's main complaint of, uh, of the DMS, right? 1:53 Which is, like, it's this, like, box that, that connects to nothing. Um, you know, what's interesting is that auto, 2:00 because it ha- was financially, uh, healthy enough to afford the first servers, it's one of the areas that DMS could even exist so early on. 2:10 And so what, first, first market, uh, first, like, milestone I would point out is, man, how crazy is that in, like, the '80s, '90s, that we had this giant trillion-dollar industry that could, that could be the first adopter in such high-tech, such a high-tech thing as the DMS, right? 2:28 And then you fast-forward and you go, "Well, why are we all complaining about not being able to connect to it now?" 2:33 Well, right, like, just like, um, uh, if you look at India right now and their ability to, uh, to adopt, um, 5G as a wideband, right, on cellular networks way faster than us, it's also because they don't have to worry about creation of telephone poles. 2:50 Telephone poles, yeah. We have this, we have this subset of technology because we are the first adopters that we have to support. Okay, so s- let's go for the next point. 2:59 The next point to that, I would say, is probably somewhere around CRM or websites. Debatable. Probably let's go website. Website, dealer.com was founded in 1996. Yep. 3:12 I bet if we p- like, went around this entire audience and said, "Hey, do you have a website in 1996?" They would all say, "Hell no," right? What's weird about that is that you have two different versions. 3:26 You have, "I'm gonna be the first adopter in the world of stuff like a server on premise-" The next big one. And they're like, "Ah." "Ah, pump the brakes." [laughs] "Wait a second. 3:34 I don't know if that's g- uh, internet could be a fad," you know? Right. Kinda weird. Um, next one after that you would say is probably mobile. Yeah. Mobile, everybody remembers, like, uh- Oh '04, '05, right... 3:47 yeah, responsive, you know. Is it responsive? Yeah. Whatever. Or m dot something websites- Oh, wow... the secondary website. Yeah. 3:55 Um, okay, what happened there, uh, it was, like, every month for a while, you were seeing mobile traffic go from 12% to 13% to 14% to whatever. iPhone comes out, explodes. Um, 4:10 next one after that, probably digital retail, I would say. Yeah. Right? 2012-ish. Um, last one I can recall, maybe... 4:26 Okay, probably- CDP marketing automation probably- I think you have to-... starts to come in there... I, yeah, it's tough. Yeah, yeah. It's tough to say if CDP deserves a milestone. Sure. Yeah. 4:37 In other industries, I don't think you would say it does. In ours, because everything's so disconnected, I think, you know, unique to us, we have to give it. 4:44 It's the milestone of, like, starting to forgive the sin of the black box of the DMS. Yeah. Right? Yeah. Right. Right, right. And, and you gotta give it sort of its flowers in that, like, it came along. 4:54 Yeah, yeah, yeah, yeah. Exactly. Hey, good job. All right. Yeah. Now, okay, what's interesting about this is if you look at, um, the pattern there, you have on-prem, and then, uh, you have, um, in-household, right? 5:08 So, like, you actually have this expansion and contration, and contraction moment that is happening between, uh, where the data lives and what's designed around it. 5:18 So, you know, like, uh, if you, if you rethink of mobile as, um, cloud, then you kind of, like, have this, like, all right, uh, you know, back and forth motion that happens between, uh, a DMS on-prem, website is desktop. 5:36 Then you go back to mobile, and then what's really all the data's sitting on a server somewhere, on premise it's somewhere. And then you go back and forth, back and forth, back and forth. 5:44 And now we have AI, obviously is, is probably the next big milestone. And it's des- it's not destroying, but it's reconfiguring both of those. Both of them, yeah. Simultaneously. 5:59 It's actually streaming them between each other and then-Making a whole new thing on the side. Right. Yeah. Like, I have my AIs that are running on-prem. I have my local AIs. 6:09 I have the ones that I have in the cloud, my frontier models. Uh, I have access to them both on mobile and both on, uh, desktop. You know, it's like, it's really, it's, it's collapsing both pieces. 6:21 Um- I, I would say, like, uh, I, I'm, I'm gonna kinda give a new phrase, like the commodity of compression. Yeah. Age of compression, bro. Right? Yeah. But it's not just the, it's the commodity of it, right? 6:33 What do you mean by commodity? So like, a- as things become more accessible, they become commodities as, uh, uh, uh, as opposed to being... Like, they are just readily available, right? 6:45 Like, you go from nobody can access data, DMS, telecom companies, it doesn't get touched. Yeah. What even is Meta doing, you know, or, or, or Facebook doing to 6:58 someone has access to the world's most intelligent super compute on their phone. Bro, how insane is it? I talked to a dealer, must, uh, probably like five months ago. 7:13 Actually, shout out, it was on David Long's All Things. All Things. Yeah. Used cars. Yeah. Yeah. It was a Friday clubhouse move. And, uh, we got into a little bit of an argument 'cause I was like, "Look, 7:23 it's $30 a month to give your employee the largest trained knowledge base-" In the history..."ever in the history of human." [laughs] And they're like, "Ah." "Pump the brakes. A dollar a day, you say, 7:41 for the largest..." I'm like, that's crazy. I mean, it, it, the other piece I think that it's, like, tangential- Yeah... but it's the amount of value that $30 a month can create- Yes... challenges companies like ours. 7:54 Like, I have to deliver so much value now as a service, a software service, because my compe- competition of, of value creation is a $30 product that can make other products. Yeah. It's insane. 8:10 Meanwhile, you got guys like Nathan, uh, Grechlik, Nathaniel Grechlik. He hates when you say Nathan. Nathaniel. Who's- I'm glad I know that... 8:18 who's coding, Auto Genius guy right there, who's coding some of the coolest, like, in-house CRMs. Like- Lit- full CRM for a collision body shop. Crazy. One guy using an LLM. It's insane. It's wild. Yeah. Yeah. 8:36 So when you think about that and you look around companies that are selling products that have to overdeliver on service, Auto Genius, right, we have to think about this on a daily basis- Yep... 8:48 is what are we compressing to, and what's the moat around technology? Is there a moat on technology, and how long will that moat still exist? Yeah. I don't know. 8:59 I think, uh- [laughs] You're like, "Hopefully long enough." 9:02 [laughs] Well, that's, that's a good question because, you know, if everything collapses, right, everything compresses, commoditize, then, you know, the argument that all there is left is taste is a really fair one. 9:15 And yet we live sort of in industry that people pretend to sort of have brand and pretend to have taste. Uh, but if we're honest with ourselves, like, no one's really taking branding that seriously, right? Yeah. 9:33 Very few examples. Um, because, uh, the, the ir- the, the big irony is, um, the commoditization happens because of metrics, right? 9:48 If something is measurable, it can be more quickly commoditized. If AI has seen it before, it's easier for, for it to produce a clone, so it's easier to make repeated copy/paste products. 10:02 [smacks lips] Um, so what's the difference between something, uh, you know, one thing versus the other? Well, that probably falls more in line with, like, how it makes you feel. Yeah. The whole vibe here, 10:17 the whole vibe at this conference- The whole thing. Right? You're go- you got, you got it. You know. You're feeling the difference. 10:24 Um, but imagine if you had, like, some investor that came up to you and was like, "Yeah, yeah, yeah, but bottle this up. 10:30 Bottle the feeling of electricity that's going through the room right now and give me an ROI on that." Yeah. Right. So tough. 10:36 You'd be like, "Well, but, you know, go to another conference and be bored as hell, and you'll know what the ROI is," right? Yeah. [laughs] Like... 10:43 But, sorry, that's, like, that's the reality of today- That's the difference. Yeah, yeah, yeah... 10:47 is you either get excited by something, it creates feeling and captures your attention, or you're, um, you know, you're playing the game of Costco Kirkland. Solid product, not knocking Kirkland T-shirts. 10:59 I own them, but, you know, whatever. Dime a dozen. Dime a dozen. Right. Yeah. Yeah. All right, so pivot a little bit. Why, 11:08 w- why is Auto Genius as a technology company hyper-focused on speed, the edge, cloud compute- Hmm... s- security in the edge? Why is that, 11:22 why i- i- in relationship with all that, because if you're thinking about all that, why are those the things that, that locking that down, especially in retail auto, is the imfirst, important front for, for frontier- Yeah... 11:33 to not stray from before it's right? 11:36 So, okay, if you agree with what I've said so far, right, which is, um, metrics are basically, uh, driving people into average behavior, and average behavior is easier to copy and paste, uh, and the only way to differentiate yourself is sort of through feeling, interestingly, um, speed e- exists on both sides of that equation. 11:58 It's a unique, um, thing where it is measurable, but it is also felt. Hmm.Yeah Right? You can feel faster. You can feel- Oh, for sure. Yeah. But tell me- And you can feel angry slower. Totally. 12:14 But can you feel the quality of a lead before you call it? Not really. No. Yeah. Right? So i- you know, if you had the spectrum of, like, quantitative and qualitative, it exists on both sides of that spectrum. 12:26 Uh, meaning that you have a product or, um, a, an, a opportunity, um, to solve something that both enhances your brand but also enhances all of the metrics, which is rare. 12:39 There's not too many other of those, uh, things that exist that, that you can, you can win on both sides of the equation. 12:44 And, and, and someone that was a great logo designer would argue that that would increase the metrics, but the direct line to attribution on those metrics is much tougher to paint- And also, not if- Just like you said at the beginning- Not if you don't see it... 12:55 of this event, right? Not if it's loading. Right. [laughs] You know what I mean? Not if it's a blank white screen. Blank white screen. I don't get, you know? Yeah. Yeah. 13:02 Um, so w- what are you thinking about building on top of that? Uh, you know, we, we have a lot of conversations on, like, but that's only the starting point. 13:10 That's only the, the, the micro of the opportunity of once you, once you get the metrics and the feeling of speed, you have, you have a next opportunity. Is that... Are, are we thinking AI- Yes... 13:22 and, and the opportunity there, or, or is it the human user, or do both get the benefit of the next, the next piece? Okay. Exactly. So, so now you have the spectrum: quantitative, qualitative. 13:32 Say that humans exa- exist more on the qualitative, right? Our relationships with people are not about metrics. They're about feeling, right? So put humans over on this side. 13:44 Qua- uh, on quantitative, where I'm trying to metric everything and bounce rates and blah, blah, blah. Okay, cool. Let's, let's put that on the other. AI is a, is an, uh, entity that exists on that spectrum. 13:56 Humans, entity that exists on that spectrum. Um, both want speed. Both actually require completely different versions of measurement, too. So I think the next frontier is, uh, once we solve speed, 14:11 y- you have to then move it into personalization. There's no reason that you should be getting 45,000 visitors to a website and not producing 45,000 different experiences, right? Hmm. 14:22 Right now we've settled with, you know, one website experience for 45,000 people, and we just kinda say, "Deal with it." Eventually, it's gonna have to turn into, okay, you get a different... 14:33 You get, you know, Oprah, you get a different website. You get a different... Now- Well, in retail auto, that makes a lot of sense, too, because in the showroom- Yeah... 14:40 a lot of times we have, "How do you wanna buy this car?" Right? Totally. Yeah, but online we're just like, "Everybody's the same." Everybody's the same. One of those, one of those people is a robot. Yep. 14:51 That, that's the other consideration. So of those 45,000 visitors, less than 1% of them is ChatGPT, an agent, or whatever today. 14:59 But just like we were talking about before, where you saw mobile going from 11% to 12% every month over- It's gonna... Yeah... it's absolutely rapidly growing. 15:07 The conversion rates on those agents is way higher than a lead left through traditional means. It's weird now to also call digital traditional. [laughs] I think that might be a milestone, too. That just messes me up. 15:21 Yeah, yeah, yeah. 'Cause, like, what's traditional? I don't know. But, like, also, like, you know- Okay, fine... traditional traffic. Yeah. Um, uh, so yeah. You- but now you also have to measure it completely different. 15:32 So we were obviously part of, uh, ASC Council. Yep. Loved what, uh, Brian Pash was trying to do with that. Um, but that measuring stick is a human measuring stick. Yeah. 15:43 Meaning if a ChatGPT agent goes to my website and digests the entire page in 200 milliseconds and leaves, that's a good thing. Yeah. That's positive. If you use a human measuring stick, that would be called bounce rate. 15:59 Yeah. Right? We would categorize it as a bad thing. No scroll, no touches, no nothing like that. Yep. 16:04 So, so sort of working backwards from personalization, you go solve speed first, then bifurcate analytics to judge, um, robot, good robot versus bad robot traffic, good human traffic. 16:17 So I need two different, um, ways of measuring, two different measuring sticks for them. Then you can start saying optimization per entity that is visiting website. Wild. All right. 16:29 Last week you posted on LinkedIn that you spent how many billion tokens? I put a billion. It was a billion tokens a week. Okay. 16:37 Technically, I have three of those accounts, so it was closer to 3 billion, but I was too lazy to, like, post three screenshots. And I was like, "Ah, I have to log into three different accounts." Yeah. "Nah, never mind. 16:47 I'll get my point across with a billion." How... Uh, uh, some people are like, "Yo, I just started asking GPT, like, enhanced Google Search questions." 16:57 How are you spending a billion tokens, and, and what should people thi- be thinking about that changes the game for development or personal enhancement? So one- And, and relationally, like, 17:11 a, a simple search like that is probably gonna cost you, like, a couple thousand tokens maybe if it's, like, deep research, right? A, yeah, a billion tokens probably costs, on a frontier model, 17:25 probably, like, 10,000 bucks- Yeah... I'd say. Um, it depends on, like, is it output or is it input and whatever, but call it 10 or $15,000. Uh, so 17:42 I think, I think the new way of, of development is less about single player, which is normally what we have is, um, Agile, right? 17:53 Which you have a bunch of, um, developers working together, solve- getting, getting issued tickets, solving problems, blah, blah, blah. Okay. 18:00 The, the future is actually going to be swarm and orchestration, where, you know, what I'm doing is basically producing 2 to 300 Jira tickets or Linear tickets-And then creating 10 to 20, um, sub-agents, and then ge- assigning the, the 2 to 300 tickets out to those sub-agents with prerequisites, and then just saying, "Go," right before I go to bed. 18:25 And then I wake up, and then see what, see what comes- See what happened... out of the oven. Yeah, yeah. You know, along that journey, um, 18:34 you- there's some unpretty things that are really, really helpful like, uh, um, a concept called auto research, where if you create an evaluation harness, effectively you take things like how fast can this thing load, and I want it to load under, you know, one second or something. 18:54 Um, and you just basically brute force. You, you, you have a... That's your test, one-second load. You have 20 agents basically go, "Oh, what if I change the code base like this? Oh, that was 1.2. 19:07 What if I change it like this? Oh, that was 1.1. What if I change it like this? Oh,.9. Yay." Move on to the next one. 19:13 So along that swarm mentality, you also are kri- kind of doing this thing, uh, which is kind of categorized as auto research to sort of like, um- Attack a- A- yeah... attack a thing, right? 19:24 Attack a thing and then- And no- And know what successes... now we, now we got Claude and GPT coming out with, or Claude and Codex coming out with- Slash Goal... Slash Goal. Oh, it's over. Slash, Slash Goal is so silly. 19:33 Slash Goal is just changing the game. Yeah, yeah. Uh- I, I, I've done it twice and I'm like, "Just blew my mind." [laughs] So good. Just came out. 19:43 Yeah, and I did it twice so far since they announced it yesterday- Yeah... at time of recording. So. Wh- what, um- [laughs] What was the other one? Uh, Dangerously- Dangerously Accept. Yeah, yeah. 19:54 Yeah, Dangerously Accept definitely. I never did that one because that's different than Goal. It's YOLO. Goal is like just win it. It's YOLO. [laughs] YOLO is Dangerously Accept. Yeah. Uh, well, YOLO. 20:01 Ben, it's always great hanging out with you. Obviously we, we get to talk a lot, but not in front of mics a lot, so. You said two hours. I know. I gotta like, you know, do stuff. You know, event things. Oh, yeah, yeah. 20:12 Sorry. Thanks for joining us- Got all the other commitments... on SodaConcessions with Tasin, dude. Let's go. [upbeat music] Thanks for listening to SodaConcessions, brought to you by the More Than Cars Creative Shop. 20:19 If you wanna learn more about how we can help your business strategize, produce, and deploy modern creative content fast, visit morethancarscreative.com.