From x.ai to ChatGPT Ads: Scheduling, Bubbles, and Apple’s Mistake

Transcript generated automatically from the episode audio and lightly cleaned up (filler words removed, misheard names corrected where clear). Speaker labels were added afterward and may contain errors.

Rob May: Hi everyone and welcome to the AI in NYC show. I'm Rob May the co-founder and CEO at Neurometric

Ryan Eppley: and I'm Ryan Eppley, co-founder CEO of RootAccess.

Rob May: And our guest host today is Dennis Mortensen from Launch Brightly is your current company.

Dennis Mortensen: That is correct.

Rob May: Yeah. And really excited to have him here today because 20 year entrepreneur in New York City lots of companies, some successful, some not. So we're going to talk all about that, but why why don't you tell us the sort of key Dennis facts before we get started, things people should know about yourself.

Dennis Mortensen: So that's the I'll give you the short version and then you can get the email to my mom. She'll do the fourhour seminar which is spectacular. But the short version is that I actually which is the funny part of it did not want to be an entrepreneur because my dad, uncles, cousins are all entrepreneurs in their own right. So all I wanted was just to get my computer science degree and go work for IBM. I lived in Denmark, get my seven weeks of vacation and that was the plan. And I did game development to kind of earn a little bit of beer money in college and that company ended up going bankrupt and I was so pissed that the title which I've been working on together with a small team for a year and a half and the way you pay students as a side note is you pay them on delivery because they're just not reliable. So I ended up buying the assets of that company and got kind of sucked into entrepreneurship immediately right out of college and earned what I thought was a small fortune when I bought the software and sold it to some German distributor.

Wasn't really much but back then it felt like a lot and even on that little success I still thought that I was kind of destined I actually never talked to IBM but I was still destined to going to go work for IBM because they had an office in my town. So I took all of it, not even like, oh, let's split it up. No, let's take all of it and I put it into like an internet analytics company back in the mid 90s and I thought, you know what, I'll do this quick, lose it all, then go to my desk out at IBM. And then of course it was just good timing. We did analytics on web server logs kind of pre you putting a pixel on a site to get some statistics, right?

So we did that for four and a half years. Had a very good u exit in April. April 11th, 2000. So, I'm one of those kids for where you talk about the dot boom. Yeah. I got out right at the top. Yeah. Didn't know anything. It was just one of those amazing timing. You stumble into some sort of party where oh yeah, this is good. So, that was the beginning. But all my ventures I'm now in my sixth venture. Spent about five years on average on each have been data centric trying to figure out some either original data source or some way to extract new novel data then build something on top of it.

Rob May: Well, let's talk about I think one of your most interesting ventures and when I first came across you publicly but before we met in person but I was following you a little bit. You were doing x.ai. Yeah. Which was one of the first like real AI use cases where you would you know I so for people that don't know it you would email you would copy x.ai Amy or Andrew or somebody like that and it would just handle your scheduling the way a human would. And you guys raised like a ton of money for that. And I remember we were I was working on a customer support chatbot and we were trying to figure out like, wow, what do you guys know that you can make this work because I couldn't figure out how to make the AI work for that yet, but and your time was close. You were just barely prel. Tell us a little bit about that experience. What happened?

Dennis Mortensen: So there I think we all agree if you do any venture, you should try to at least not fall in love with the technology as in it's so easy if you're just a little bit of a geek. You know what? I can geek out on things not for weeks but for years on end with no commercial prospects. So I try very hard to figure out whether the very pain that I'm trying to solve for is true and honest. So this whole idea certainly going back in time and probably still today for where the three of us trying to meet up at a certain day at a certain time at a certain location in New York is not as easy as it sounds. Which is why if you ask people, what is your fantasy? The fantasy is always the same. You know what, Dennis? I dream of becoming SVP of Flimflam at Time Inc. So that I can get Tim, that human assistant, to do this [ __ ] for me. Yeah.

Now, that's probably going to take some time. I'll give it a few decades, but at some point, I'll kind of graduate and cross that inflection point on the corporate ladder for I'm given a human. But that's the fantasy. So we always thought that if we could somehow replicate that fantasy of you not having to pay 60k, pay me 10 bucks instead and kind of do that particular task that might be a good idea and it certainly was a good idea. Everybody wanted it. Now delivering on that promise was a little harder. So we took a chance and many times again just as a side note and you guys would agree right as as entrepreneurs we take typically one of two kind of bets. Is either a market bet or a science bet. Yeah, this was not a market bet. Kind of like the self-driving car is not a market bet. We all wanted it took a little longer to kind of engineer than we had imagined. This was not a market bet.

This was one for where can we at this very moment 2015 when we got started carve out a corner of the language universe and solve for that. Not a kind of chatbot who can speak kind of in general things but just a an agent that could converse on this one particular topic and we got it to work as you said in the end but damn it took a lot of money and a lot of time to get there. We did the whole thing as that's too simple of a kind of title to it but it was primarily based on supervised learning. So we had just in scope a little over a hundred people alone just doing labeling on in the end about what short of 50 intents like new meeting, cancel meeting, running late, add participant, extend duration, so on so forth doing predictions on all of those and three entities. So temporal location and people and we handlabeled about 32 million labels.

Ryan Eppley: Wow.

Dennis Mortensen: Over a 4-year period, but we did cross into that kind of other side for where the very use of the product generated new labels for where we could then train in that. So all the labels ended up going on June 1st noon EST which was midnight Manila which was a very kind of you know ecstatic moment right where we kind of got to the other side the agent do exactly as we wanted to do.

Rob May: Well, it's also a hard it's an ambiguous domain. I actually know funny story. I know a guy in Boston who did his PhD work on NLP and what does soon mean?

Dennis Mortensen: Yeah. Because that's a Yeah. So, so just to give you some context here and again not to turn this into something overly kind of geeky, but if everybody would just be somewhat sane and not lie like just two concepts, then this could probably be solved in a whole lot less time. But the fact is that people are not sane and they lie all the time. Not intentional lies, but they'll say things like, at 1:00 a.m. You email me saying, "Dennis, I got a great idea. Can we meet up first thing tomorrow morning?" But you really mean today morning, it's just that you haven't gone to bed yet. So the idea of saying today morning kind of doesn't feel right. Right.

Okay. At 1:00 am you and we can do an if then statement do a little bit of ML perhaps but at what point at 5:00 am and you say tomorrow that's probably what in 27 hours or something like that or or when is it and then of course there's the kind of next week [ __ ] Americans and this whole idea of the week beginning on Sunday or having the idea of the week beginning on Sunday like next week if I say that Sunday 6 p.m. That's in seven days. Like we just had very different ideas. It was very hard to solve for. So to give you some idea of the kind of long tail. So we tested against 22,000 edge cases whenever we did an update to our models to see if it would still kind of reliable predict what it did yesterday.

Ryan Eppley: And as my engineering friends and I always joke, how many systems break because of daylight savings? So many things always crash because of that.

Dennis Mortensen: So here here's another story, right? So there's two types. They're not, right? But there's certainly two buckets of errors. So we did all of the things that machines do well. Daylight saving, the fact that India is on a half hour time zone, so on so forth. All those kind of silly things that we kind of forget. Oh yes, London, New York is only four hours over those three weeks over the summer and so on so forth. The machine just killed it in that because that's the easy part. No, it was the errors that humans just couldn't understand. Such as we spent probably eight months working on some idea of common sense that allowed elasticity in your scheduling hours. So you might have scheduling hours that go from 9 to 5 on Wednesdays, but we haven't seen each other for years. We used to be college buddies. I'm in from Singapore. And that whole kind of backdrop will have any human existence say, "Oh, I'm probably allowed to stretch your hours." Oh, it sounds like a happy hour or dinner or drinks after. Like many kind of good arguments.

Doesn't mean that all of a sudden I'm at an event until 11 p.m. But if you can come in at 5:30, not even you don't even have to ping me, of course. Right. So, we spent eight months on that. But when we did it, people ended up kind of getting a little bit annoyed as hey, so how many tens of millions of dollars? Because our funding was public. Do you need to do an if then statement? You muppet. That's what I got on Twitter all the time as in 5:30 is later than five.

Then we explained to them and that actually carved out this whole new category of what we called perceived errors and the whole bucket of what we call perceived errors was much much harder to kind of solve for for where that's actually the predictions are running at very high accuracy. The design is doing exactly what we wanted to do. But that ended up having us kind of re-engineer the agent to not be humanized because as soon as we humanized it, there was a set of expectations that was built into this conversational UI that people couldn't kind of escape. So when we started to make it machine-like, then they started to kind of better accept that kind of bucket of perceived errors that wasn't errors. So a perceived error for us for us was one for where the user thought this didn't go according to plan and I'll now complain even though the very system did exactly what it wanted to do and on explanation you also agree. Yeah. I mean, sorry, this is turning into a little bit of

Ryan Eppley: No, it's very funny. It's very funny cuz, you know, I just had this actually happen. My my friend from England was over and she said, "Meet me at half like 6:30 or like half six." Half six. In my mind, you know, that's halfway to 6:00. Yeah. Right. Like that's 5:30. That's like, you know, half. It's the same for Dan, right? Nah, not 6:30 in Britain. I was like, whoa, whoa, whoa. Like that's halfway to six o'clock. Not 6:30. So I you know can only imagine how many small things like that just go over our heads dayto day and then scheduling just sounds like a nightmare to do that.

Dennis Mortensen: But I'll I'll take the whole thing back back to the top and we can certainly go into into details here just because we spent half a decade which is a funny side note for where there's no dinner party where I get to stand in a corner speaking to somebody who will say Dennis can we talk about meeting scheduling for the next two or three hours like there's no room for me to talk about this PhD which I've assembled over half a decade.

So it's just me. So any opportunity for people to ask who hold my hand. But let me roll it back to the top. If I were to kind of go back and kind of state what had the most impact of all the stuff we did and we had a spectacular data science team and an unfair amount of labelers and really good backdrop to do all the sexy stuff. But the things that had the most impact was product design as in taking decisions for what should the product look like and do and little things like so we had an overall success metric which we called meeting success. So from request to actual insert there's some sort of natural ceiling around 91ish% for where many meetings upon request will end up not happening because you get sick. It got solved over Slack. The flight was cancelled and so on so forth. So it's a very natural ceiling. So we were kind of kind of running towards that ceiling.

But the kind of step functions towards that were and I'll give you the biggest one. When we renamed the agent from Amy and Andrew to scheduler. It was one of those like any startup you run a thousand experiments, right? And not only did I [ __ ] hate that experiment because I was hey Amy and Andrew. Yeah. They're my kids, right? And not only that, I had people in place like Anna out of Harvard who worked on folklore for ages to build a character. So when you spoke to the agent on Monday, it felt like the same agent on Friday. Like an unfair amount of detail were put in to how the agent kind of communicated.

When we renamed that to scheduler, multiple single percentage points increase in meeting success. Like another thing that that worked really well in our favor that I also hated. Like the most successful experiments we ran were mostly experiments that I hated. But I like the idea of the emails being mostly asy. It's a little cheat. So we'll still use kind of bold, italic, a few bullets and so on so forth. But the concept of you could have written that right as in ah yeah you just typed that out right now as a human would do. As soon as we inserted HTML as in no need for somebody to write back saying next Wednesday at is okay and lane is indeed optional. If I can predict with a high probability that that is what you want just give them a button as in don't be allergic to good UI paradigms that already exist but I was so allergic for the longest period of time any yes I worked for that on that for yeah half a decade.

Rob May: Wow. So tell us a little bit about you've been in the New York ecosystem for a long time and you know I moved here from Boston but the reason that I'm so I moved to Boston from Louisville, Kentucky in 2010 because in that was my first startup and I was looking at the Boston versus the New York ecosystem and like Boston was like two or three times the size of New York at that point from a tech perspective and I would say like 2013 2014 New York sort of caught up and now has just like really surpassed Boston. So, like what what was it like when you got here and like what do you think have been some of the drivers of the those changes over the years?

Dennis Mortensen: Good question. I can certainly remember when I arrived. So, I arrived on us having sold our prior prior venture to Yahoo and one of the deal terms were actually that I moved to New York. So, it was not even one of those things on your bucket list or feels like a good destination to kind of go work. This was part of the terms put in place for me to cash out to get that summer money in my pocket that we had negotiated and the funny thing is I pitched very hard for the west coast Silicon Valley all that good juice but nine hours and we did that venture out of Budapest. So I spent four years in Budapest and we bootstrapped it. Which you could do back then in Eastern Europe. It's obviously nice when you sell to Yahoo.

Don't have to send out any checks to anybody else, just in your pocket. So they put me here and I thought, yeah, I'll just hang around for a little while, then move out west. But New York, it's just very hard once you arrive and you kind of embed yourself into the system to escape it. And I've had multiple kind of moments where there was kind of an opportunity, a juncture for where Dennis, there's an exit coming 100 meters. You can take it if you want to. And I tried but I just I just can't. So when I arrived I certainly had this of moment of despair is a little too dramatic, right? But you build up a network of people wherever you're at, right? So back home Europe you have some sort of network people you can call or rely on good engineers a scientist or kind of possible customers and this and that kind of industry.

So you arrive kind of empty here but I remember the first like week one New York tech meetup and I did it every month probably for the first three four years. So the very first one that was MongoDB who presented not even an early version just they put a little bit together they presented it and I thought hey I first of all I took my CS degree on multiple versions of traditional relational DBs right what is this document database these queries kind of in half JavaScript come on this is laughable Not serious. This is not serious at all. You guys are not serious people. Yes, if I've had that quote back then. Of course now we built a whole new venture on top of a DynamoDB. Yeah.

Which is kind of just MongoDB in a different flavor. But I saw that. But the one thing that I certainly think both back then today is this kind of willingness which I don't think is very different to kind of the west coast of wanting to help the next guy as in there was just not a single player here which I couldn't ring or ping or get a meeting with and that was just super empowering as in ah you want to talk money you want to talk tech, you want to talk talent, you want to talk anything. And it was just very hard to see where else can I get that within 15 minutes on the subway. It's not even like, oh yes, perhaps one day I'll drive to Palo Alto.

No, no, no. Got time at 1. Yeah, I'll pop by at 1 today. That was super empowering and I still think that counts today in a different way to many other destinations which is that it's very easy both because the whole thing is so compressed but also because of I think the kind of culture attached to the fact that yeah we can do things in the city today at 1:00 if you want to and I do want to and I have auler for that. Yeah. And yeah, I have a schedule for that. Or did now what I think I've changed is perhaps if I were to pick one parameter perhaps the glitter attached to entrepreneurship where out west if we're comparing those two kind of places perhaps even Boston which I have less exposure to but certainly out west there used to be a lot of glitter attached to entrepreneurship or running a startup as in if you can't play the guitar, you're still a rockstar if you can code, right?

From the get-go. A little less so here. You're either a celebrity or you're not. And if you code, you're not. But I think over time this whole idea of you're working on some venture or some idea in any kind of vertical which of course is just very hard to compete with for any other place for you can be big in finance or fashion or any number of other things because it's in the city. So I do think that have changed so that you can be in your lift down on Wall Street where I live and tell the guy, "Oh yeah, me and my buddies got started again. Just 10 guys in a basement working on a new venture and it doesn't look like you at you like you're an idiot." Like, "Oh yeah, sexy. I've been wanting to kind of quit Morgan Stanley for years." Yeah, I know. But you're never going to do it. But hey, you could. It feels like I'm I'm the sexier of these two people. Or perhaps this is just what I tell myself. I'm living I'm living a lie, I think.

Rob May: Yeah. I don't know. It's a good and a bad thing I think that entrepreneurship has kind of encouraged and so part of it is it's such a hard thing and it's so stressful and like I don't know why like like you like I keep going back to it like particularly after like I mean at least one company that failed spectacularly and everything and like but like you're just you're drawn back to like the fight and everything and I think part of it is that they build up this mythos and everything else is to get people to try it because otherwise if you were if you were entirely sane you would never do it. If you just looked at a spreadsheet of kind of kind of things you could do with your life.

Dennis Mortensen: Yeah, this is the shittiest column. Yeah, I don't I'm not doing that one. I haven't seen the title yet, but I'm certainly not doing it. Yeah, I've come to conclude since so I've done all of them back to back. No satical like back to back. No, we'll spend six months in Barbados. Just I'll quit Monday. Somewhere else and tonight 8 pm we get started on a new thing. Wow. Or really we've done the overlap, right? That's always what we kind of do. We do all the research and all of that. So done them back to back, but I think you have to do it and that's why I think there might be two or two types that's unfair, but you have to do it for the love of the sport. Yeah.

As in if you're in it for the trophies and the medals like in sport, you're not going to get any of them. But if any for the love of the sport where I actually just like to play soccer then you play soccer seven days a week. It's not a job. It's just say me just kind of fooling around. And who was it? Was it Paul Graham who said it perhaps? But this whole idea of competing against somebody who does it. He didn't call it for the love of the sport but does it because they think it's fun. It's just very hard to win against people where I'm just I don't even care if I win. I just care if I play. Yeah. Because that means your kind of really shitty days, it's just another kind of fun day at the gym for me.

Ryan Eppley: That's cool. I'll never surrender.

Rob May: Did you Did you guys ever hear Michael Jordan had this clause in his contract called for the love of the game contract, love of the game clause. And it allowed him he could play any like pickup game. He could play basketball anytime, anywhere. And if he was injured, he got treated like he was playing in an NBA game in terms of how his contract played out. And because and he insisted on it because he was like, "But I love basketball so much, I'm not going to pass up a chance to play."

Ryan Eppley: Awesome. Awesome. Yeah, I'm going to look that up.

Rob May: Yeah, it was a really really really cool thing. Well, let's talk about a couple things going on in the AI world today. I the one of the things I'm most interested in how you guys think about it is so ChatGPT is going to start offering ads and this happens for so many consumer tech companies, right? It has over time. Even people that say they won't do it or whatever, they always end up monetizing via ads. And now all the TV companies, you know, Netflix and Amazon Prime and all that [ __ ] you got to look at ads. And all roads and consumer lead to ads. Yeah. Yeah. So, like, what do we think about? Are we worried about that? Is this what was expected? Are we Do we think they'll back off? We think they'll be successful.

Dennis Mortensen: Who was it? I just bought the book a month back, so I thought I'd just scan it quickly. That journalist who invented the word incitification where

Rob May: oh is that Cory Doctorow

Dennis Mortensen: I think you might be right yes it's on my kind of bookshelf along with the 20 other things I want to read next week but it seems inevitable that we will get to a point for where there's people who are unwilling or unable to pay 20 bucks a month or whatever kind of addition they might be beyond I'm not so sure I'm allergic to it but I do this time is that there was some sort of procedure put in place for where you can pay to not get ads for where on Google there was never an option. There was never a $10 subscription to Google for where I don't want AdWords.

What's funny though on Google was that I think probably even today there was some sort of inflection point where the ads might have been better than the organic results as in well there's been it's been hyper optimized to the degree for where the organic results can't really be trusted anymore in certain kind of terms and phrases and that might not happen in this space so I'm not allergic to it.

The one thing that of the many things we could talk about that irks me a little bit is once the agent arrives in spaces for where it's not obvious what is biased or have been bought and what is not as in sure you can label things in a visual UI what do we do in some sort of audio UI or what do we do in some setting for where oh I really just want the action you're doing the prediction you're taking the action but the action was taken with what in mind and sure anybody will say yeah but there'll be all sorts of kind of preferences and settings and so on and so forth sure but most people won't touch any of that. So that's probably the path where that's going to be interesting where you can all of a sudden start to kind of direct hundreds if not millions of people in certain buy them to go in certain directions that feels a little kind of that irks me a little bit.

Ryan Eppley: Yeah, I think people are overreacting with this because it's been Merburn for a while. Almost everything that's free has some type of ad in involved and even things that are paid. You know, I think the idea of, oh, we can't have LLMs feed you up an ad. As long as they're documented what they are, we walk by a billboard, we watch commercials all the time. I think it's just a part of a natural product development that you include ads at some point if it's a consumerf facing app. And I think many times ads are not bad. It's not like I buy Wired Magazine and I throw up on every third page because there's an ad. That's pretty cool. Or I'm in the Uber and there's like, you know, the picture of the beer on like the screen in front of me. Some of these ads are cool, right?

Dennis Mortensen: As I actually don't mind. What I mind, of course, is when I don't know why you told me something. That's the part that irks me where ah I asked for a top three list of something or hey I need kind of a module for my document database I have something in mind and then you give me an answer. Yeah, but is that based on my query or question or based on the fact that somebody has some agenda and they are likely to sell me something and you would rather I be exposed to that than something else and it becomes not sure. Yeah, in the beginning it'll all be kind of these kind of square boxes with a little kind of ad on top of it. But there comes a point where Yeah, but I just asked my agent and he got voiced out or I asked Claw to do something and you just pick it. Oh yeah, but who who who actually picked it?

Rob May: Well, that would be like like if I was going to make a counterpoint to Ryan's sort of approval of this and have offer some disapproval, I would say my my concern would be sort of at the intersection of that and this idea of something that you mentioned earlier about x.ai, AI, which was like you have these like category of perceived errors or you have these spaces where people don't always know what they want or don't know what the what the truth is or whatever. And I would tie them together by saying like like there have been a bunch of companies I've looked at as an angel investor. They come along and they're like, you know what we want to do? We want to we want to use AI to like weed out fake news and like tell people what's true. And I'm like, but the humans can't agree. Like how do you even build a labeled data set like that?

Because two people on different sides of the spectrum would say that label that differently. And so there's there's areas like that that are either just ambiguous or people are misinformed. You know and so if you know if you ask these so part of the problem is these things are so broad and they do so much. If you ask it a question about something controversial like vaccines or down the food pyramid or like whatever like what's it going to say and then how are you going to tie ads to that and what might those ads be and might like it all like how much of this is nefarious or how much like I don't know what you guys think but like x.ai AI has changed or X.com, whatever Twitter's called now. Twitter, let's call it Twitter.

Dennis Mortensen: I call it Twitter. It's spite.

Rob May: It's very It's very different since like the flow like everything's different since Elon Musk bought it. Like I've even noticed like if you say very pro MAGA kinds of things. You'll get a lot more views just naturally than if you don't. Like I can see that in my analytics. Or if you just talk about something else like it's weird.

Dennis Mortensen: I think it's one of the that's not true, but it's one of the first times where like you I've actually thought about kind of what is going on on the other side. Oh, so sorry. Just outside of my own kind of thought bubble where most of the time I've kind of selected myself into a thought bubble and I kind of just live in that universe. I'm kind of what slightly center left or something like that. I'm Danish, right? So you can paint your own picture and then all of a sudden given oh you build up a little bit of a following that's kind of where you used to get your news right or kind of geeky new things and then all of a sudden it turns into something else where it's not like I would have moved to some other channel out of choice. It was just I was here already but then it moved. Yeah, that's been very kind of eye opening.

Ryan Eppley: So going back to the article then it's really about are ads going to manipulate our experience with LLMs. Is that what we're really saying? Because I don't think any of us disagree that it's wrong to get lobbed up an ad that feels relevant to the query or like the conversation you're having. It's really is it trying to like pull me away from what I was trying to achieve by using the tool.

Dennis Mortensen: I'll bring us back to the top. What I do like is if this is a tool in some business setting where I don't need an ad. We started out this whole journey where there is an option where you can buy it as a tool. Then there might be people who will use it as in lack of any better kind of moniker as a search engine where hey I'm just a consumer. I'm looking for kind of an way to do this particular cake. Great. Go use your your agent. There might be some flour you can go buy that comes along next to it. I don't think there's any any harm in that. And good on you. King George sells flour. You can just add that to basket right away. All all thumbs up.

Rob May: Well, let's move on to talking about the AI bubble, which I feel like we talk about every week. But something interesting happened this week which is or last week I guess but like TSMC's chief executive was asked on earnings call what he thought about the burb and he basically said like look I'm very nervous about it because we're going to invest like $50 billion in capex next year and everything else and he said so I went to my customers and to their customers and I talked to everybody I looked at their books. He's like I'm convinced that at least right now the demand is real. So, what what do you got? Did it did it change your either of you change your perception when you heard that? Are you still worried? Are we still worried we're in a bubble or what? What are we thinking?

Dennis Mortensen: So, if any of us three knew, we wouldn't be sitting here. We'll be doing something else. So, we obviously don't know. We all have opinions, though. Yeah. I will say just as a as a footnote, I did have a pool of AI exposedish stock which I sold last week.

Ryan Eppley: Okay. Well, based off your success of and I, you know, exiting on April, what was it 11th of 2000? Like that's not a that's not a good thing. That sounds ominous.

Dennis Mortensen: So, we sold our first venture in April 2000. Then we did a venture that didn't work out. We can talk about that. That blew up in my face. Really expensive MBA because I financed it myself as well. Didn't needing VCs for that. Paid for the whole thing and then I like that expensive MBA. That's what it is. Colombia. No, I'll tell you. I'll tell you about a really expensive MBA. I'll buy a whole building. How about that? So the next one we sold in May 9th, 2008. Wow. More good timing. Had we not done that, I would have worked at least another 3, four years in that venture because the whole world just said, "Hey, how we put we're going to put everything on pause then is yeah, you sit here in four years. We'll have a chat again." So that was that was very fortunate. That was the venture we sold to Yahoo.

Rob May: All right. So sold.com right before the crash. Sold right before the financial crash of '08 and you just recently sold your AI exposed stocks the other week. So there you have it. All right. All right. Well, that's not not reassuring.

Ryan Eppley: You know, I'll sit definitely on the other side of this. I think what I liked about it is he went to his customers customers and he's really trying to find truth. And I just have not seen enough people not making real long-term traction with some form of serious AI. I think there's a lot of silly money thrown at silly idea AI, but there's a lot of real big billion-dollar opportunities that are still so early in their adoption. And even the early adoption is really transformative for real business outcomes. That makes me feel like how can that be a bubble?

Dennis Mortensen: I probably wouldn't go that far on the how can this be a bubble famous last words. But I do think the very people who ingest the most into the current story are folks where the money is somewhat available. So it's not just a pipe dream. It's Facebook who decide we want any number of new data centers or any number of new things where we have some capital ready and available and we've made the call and that will happen. And that's not a dream that's going to disappear in the next couple of months. That will take at least the whole next year, probably years before we get to the other side, should the whole idea turn out to be not as potent as we had imagined, but not next week. As in just like you, I can't I can't see next week. I did sell those stocks though, but there you have it.

Rob May: Well, and you have like like Meta's actually working on last I heard like a special ASIC just for recommendation algorithms, right? Because they have the size and the scale, they could make use of that. And I think there's a lot of a lot of the AI has been adopted at places that have the scale that they can push it out and even if it makes marginal improvements in their business at the scale they're at, that's tens or hundreds of millions of dollars a year a lot of times, right? Which is going to at least keep the cycle going like like look look at what Apple's innovation around the iPhone did for the whole ecosystem, right? Some of the stuff that they had to solve around like the glass and the you know their own chips in the phone and stuff like that.

Dennis Mortensen: So, and so there also seems to be some sort of backup baked into the investment, which is that much of what they invest in is compute. So, it's not like there's a version of the future for where you know what, we were wrong. We're going to need less compute. Yeah. It's kind of like sure, we might have, you know, laid that fiber in the mid '9s a little too early, but there was no version for where we didn't want that fiber. And I think I think it's going to be the same with that compute where meta might just have overinvested. Yeah, but it's not like they're going to sit idle those data centers. Sure, it'll take a little longer. We thought the kind of the spreadsheet would have played out differently, but that compute that'll be utilized.

Ryan Eppley: Yeah, I think it's a mature way to look at it. It's not like a crisis at a macro level. It might be a crisis at a quarterto quarter shareholders are upset because the P&L looks bad. I like NFTTS. Yeah, you're right. I don't think there's like a clip that someone's driving off of.

Rob May: Okay. Did Did you Did you guys ever buy any NFTs?

Dennis Mortensen: I did not.

Ryan Eppley: I bought my girlfriend an NFT. It was very fun. It was a picture that she really liked. I was like, "Oh, you like this picture?"

Dennis Mortensen: So, I u not NFTts, but my daughters keep sending me kind of once every quarter current status BDC. So, I bought all my Bitcoin at $250 or something like that.

Rob May: What year was this?

Dennis Mortensen: Like pretty early, probably. Yeah, there about. Yeah, feels right. Then I sold it all, invited everybody on a month-long endeavor to Barbados. I would travel back and forth. The family was there. We laughed and laughed all the time. This is on dad's kind of Bitcoin because I nailed it. I had a 10x. I think I sold them. Yeah. Whatever. Let's call it $2500 or $3,000 or something like that. So, that was kind of my little period. And now, of course, I can never go back. I have I close my eyes and I have a picture of some account where I paid 250 a pop.

Rob May: Your ego won't let you buy it again.

Dennis Mortensen: I just can't. I can't. Right. Right.

Ryan Eppley: Barely afford like one Bitcoin, right? With the amount that you had, though.

Dennis Mortensen: O Yeah. But I'm sure there's plenty of more dramatic stories. But that's my that's my story.

Rob May: Yeah. You know, I didn't put it on the on the agenda this week, but there have been stories going around where so a lot of people believe with the advances in quantum computing and AI that a lot of these tokens are going to be hackable and their cryptographic properties aren't going to be. So that's part of what's dragging on the price right now from the like computer science side of people that are looking at things.

Dennis Mortensen: That is just so that sounds evil. Not that particular corner of the story, but the whole idea of quantum computing is it's just so hard to wrap your head around, right? For where you read article upon article or paper upon paper and you think you somehow come I get what's going on? But even then I go back I have a diet. I look out the window,

Ryan Eppley: it's like sounds like I just watched like a, you know, a deep science fiction movie after I get done watching a documentary. I'm like, is that real still or like I just got a lot of mixed emotions where

Dennis Mortensen: well, it's real to the degree where we proved it. And oh, side note, so my my co-founder at x.ai I was part of that large team at CERN who proved the Hig boson hig boson particle he's in kind of on the side in the movie particle fever as well.

Ryan Eppley: Oh super cool.

Dennis Mortensen: So obviously would have endless kind of chitchats on you know what the future might bring in compute but it's just again [ __ ] hard to kind of just wrap your head around. And by the way, I spent my whole childhood optimizing for what 1 megahertz 8k of memory like I spent like weeks on some inner loop as in I did machine learning first 10 years programming right as in just nurturing small inner loops like oh yeah I think I can change this div I can do this mold kind of differently like wow just what does that look like when we get some some people believe. So I have Sorry, this is turning into a crisis for me.

Rob May: It's well what's what's interesting about what you just said is you know you have a whole generation of programmers that has grown up and like doesn't have to know anything about the hardware they program on, right? Which is like where I came up and like I learned C initially in like the late 90s and it was like you kind of had to know how much memory you had and how to access it. Like you had there were things that you had to do. You had to know some things about the device. There were certain blocks of memory you couldn't access because they were used for specific things sometimes and you had to take care of all that and all that's been abstracted away and everything and there there's a lot of benefits to that

but I have a friend who's a professor of electrical engineering at MIT and one of the things he thinks is going to happen is as as these AI workloads and there's so many different types of AI workloads and so you know he thinks you're going to continue to see chip innovation you know as as AI grows and everything else and he's like but you know the sort of like semos transistors we use now are so robust and so well done that a as you start to get in these other, you know, making them with different materials and stacking them through doing all these crazy different things with them. It's like they're going to get more finicky and you're going to go back to this world where you kind of have to know some things about the hardware. You kind of got to, you know, jiggle the door handle a little bit for lack of a better word while you're programming to get it to, you know, put your little hacks in there and things like that. So, it's going to be interesting to see if we do go back to that world.

Dennis Mortensen: Yeah, it is interesting. It might just I'm not so sure. So the vast kind of majority of kind of engineers, young engineers that I speak to now have lost a lot of that as in we had one of my courses were operating systems. Not as in hey let me tell you a few stories about it as in writing our own operating systems. One of our kind of courses were file systems as in ah we would write file systems as in that venture which we sold to Yahoo we ended up so this is I'm going to date myself but one of the biggest issues suddenly back in the 2008 9 10 11s when you had magnetic discs were that you had too much data but you couldn't pick it up from magnetic discs quick enough. So we ended up kind of writing our own file system because the way you pick it up is in a set of chunks, then you pick a chunk, you have to spin around, then you pick up another chunk. So we kind of create a file system that was the same size as kind of one spin on these discs. So I could just kind of keep picking up data and once we done that, we have our own file system. Then we write our own kind of column database because most analytics is really kind of adding up columns. SQL is kind of row based, right? Mhm.

Select first name from customer table and then you kind of pick that up. So we did that. Then we had to kind of write our own query language. These were all kind of very fair requests which we could go out and hire against. Not like some sort of esexteric I need to find some sort of PhD who did some research on something and I might be able to find two or three and pretty standard engineers which we hired to do that right and we talked about it all of us in kind of sessions and we all had some sort of exposure to it. Me trying to hire that today that would be some sort of hey we need kind of A+ superstars to come in we'll talk about kind of very lowlevel stuff I do wonder sometimes whether do some of this turn into kind of research endeavors where how many can write a compiler today as in who even writes that we wrote compilers as kids But nobody will write a C compiler today or even write their own little mini language just for the fun of it. It's just not really on your list of [ __ ] to do. I wonder sometimes whether ah well yes highly specialized individuals come out and only very few will kind of work on that and there's some papers and there's just a tiny minority that works on that.

Rob May: Highly paid.

Dennis Mortensen: Yeah. Yeah. Tim, Susan, yeah, they work on those compilers. That's it. Not 100,000 kids.

Rob May: Well, let's hop to the sort of last article we have today. I'm really interested in your take on because you were at Yahoo and one of the things I think that one of the biggest mistakes Yahoo ever made that a lot of people didn't realize is they licensed Google search engine for a while because they didn't think search was that important. And you know, big big big mistake. And now we, it just came out that, Gemini is going to power Apple's AI. And I'm wondering if you guys think like, brilliant move by Apple, models are going to be commoditized, no sense in working on your own, or wow, they're giving away the farm to Google.

Dennis Mortensen: So, I'm on the on that side of the fence, which is that that that feels like one of those own goals for where Tim Apple will eventually go. Not tomorrow, not next year, but it'll be a fumble to the degree for where Sure. You got that whole setup in Asia. Done very well. Hats off. And you 10xed over a decade. And we thank you very much, but you got to go. As in that was just a spectacular a moment of insanity where you show something that looks like oh yeah it's a simple partnership but perhaps this is the device like all of these kind of other little things that we have on our phone might just not mean anything the agent is the device I would not be comfortable as an Apple shareholder

Ryan Eppley: what are we doing as Apple is my only question over the last four or five years. This isn't like 2002 where like LLM is just hit the market like what's our AI plan. They have had almost half a decade now and still I don't understand how they're not a front strategy having a clear and they've had tremendous amount of executives be leaving their AI roles over the last couple years. So I'm just curious like why isn't this a priority when it clearly is a big economic macro opportunity over the next decades and they were early on it was Siri or at least thinking about it

Dennis Mortensen: now Siri wasn't that great but you made it was [ __ ] awesome back then not today but back then so I know the guys who who did that and they certainly did I think in my opinion spectacular work in the NLP space ahead of everybody else and by far the best agent we'll have a laugh about it now But many things kind of in hindsight looks kind of silly and toyish suddenly back then. But damn, I think as you allude to I think they had a head start. That should have that team that should have been OpenAI.

It's not like oh yes but they have infinite money right conceptually where you could have you could have run all these experiment their balance sheet. Just look at their balance sheet. The cash they have is just incredible. It could be multiple AI labs worth of just cash on their balance sheet today. Not even like capex, like just on your balance sheet of cash, cold hard cash. But then again, they might just I can just see how a pool of executives sit in some room. What do they get right now short of 20 billion a year, I think, from Google search? Some amount that rhymes with that at least. Yeah. Straight to the bottom line. Like it's not like, oh yes, let's look at the facts. No, no, no. Somebody sent me a check. And it goes straight to the bottom line.

Ryan Eppley: Well, you gave me such a interesting visual because there's all that, you know, commercials that we've seen about this calculator that was on your desk and the calendar that was on your wall and the, you know, blah blah blah. It's all like gone into these apps on your phone and then you're like, those apps actually don't mean anything anymore once the agent is truly mature. So, it's almost like this moment where they're giving away what made them once special is taking all these things and putting it into one device. Well, now all those apps on that device is becoming one agent and they just gave away that agent opportunity.

Dennis Mortensen: And then they'll hang on obviously for the longest period of time. Then it'll be the hardware is still nice or still nicer than the guys with a better agent. Then the camera is still a little bit better, but slowly but surely they'll turn into any one of those camera companies that we might have remembered from the past that were kings.

Ryan Eppley: And to add to your your now 21 books of a backlog is there's a great great book that just came out last year called Apple in China. Yeah. It's all about Have you read it? Have you seen it?

Dennis Mortensen: It's in my pile.

Ryan Eppley: Yeah. It's a it's a really brilliant read and it's all about how the manufacturing processes migrated one by one to China and now there's this big concern because as we all know I don't think China's the most caring about IP protection and processes and trade secrets. So all these other manufacturers are starting to pick up on how that glass was curved and what materials you use and what processy at what time. And I think that's going to be a crisis. If you say Apple's gonna hang on because of that hardware technology slowly, I think that could go all at once because it's being undone pretty fast.

Rob May: Right now, a lot of companies get struggle sometimes to figure out what they're really good at, right? You just it works and you do some things and so like a question for Apple like you could argue they probably on their executive team probably sit around and say like we're a design company and a device company and we're not in but like you could argue they're an integration company right and they've done a better job of anybody else like they they moved off of Intel chips and did their own early anybody else they you know they've bought some of their suppliers they've been more vertically integrated like they you know ran their own they you know coming out with the phone they ran their own app store which was kind of like a little bit novel at that time. So they've you could argue that but like who knows how they see themselves internally. I don't know.

Ryan Eppley: Well, we're young enough to see it play out.

Dennis Mortensen: Yeah, that is true. So we'll meet here again in 5 years and we'll see kind of the lay of the land, but to me it feels like an own goal of gigantic proportions.

Ryan Eppley: Yeah. Yeah. Generational fumble.

Dennis Mortensen: Yeah. Maybe not a quarterly fumble, but feels generational.

Rob May: Well, before we wrap up, let's talk a little bit about what you're doing now. So, you've done a bunch of companies. What's your latest company doing? Like what attracted you to this problem space?

Dennis Mortensen: I run and I'll give you the advert of course that's my job. I run a list of hate on my phone so that I don't end up working on stuff because I think it's sexy. So, whenever I get exposed to [ __ ] in my universe which I don't like, I add it to my list. So, I don't do anything outside of my current venture. So, I don't sit on boards, do advisory, I don't angel invest. It's probably all kind of very poor decisions, but I only do one thing.

So, if I lose, I'll just kind of man up and say I wasn't good enough. If I win, I'll take a little bit of credit because it's all I did. Now, if I have an idea, and like entrepreneurs, we'll have ideas all the time, right, for a you know what, you could suck me into something tonight. Two pizzas, some diet cokes. I'll bring my laptop and we on it right away. Just the excitement I'm very easy and now you can code it tonight too with the we can get an MVP out tomorrow right so I'm very easy to kind of lure into things but I feel almost kind of unfaithful to my current venture if I think about other things as in it's kind of like almost why are you thinking about other women you got a wife then it's like so if I have something oh this weekend better I immediately kind of add it to my list so I can forget about it like I save some sort of sacred Maybe half the ideas, not sorry, 90% of the ideas are shite. Now, some of them are not. There's just things which I didn't like.

Now, coming back to your question, what am I working on? One of those things from that list for where if you make any piece of software B toC B2B and you put it to market you are positively forced to support it because there's no utopia for where your software is so intuitive for where anybody who will log in and use it knows exactly how to use it and will never have any questions right one day perhaps I don't think so nevertheless the way we solve for that today all of us is that we put some sort of help sensor in primarily your knowledge base. Here's 400 articles. Here's how you set up a new user. Here's how you add two factor authentications. Here's how you add segmentations. Here's how you use the filtering and so on and so forth. And each one of those articles are choke full of screenshots from the application itself. Hey, you go here, you click that. Here's the drop down. Here's what you drag and so on and so forth. Now, we've had tremendous progress on the engineering side, as in we can have the most junior engineer push [ __ ] to production this afternoon. As in things are probably happening to our app as I sit here because we have this very polished machine where engineering can move forward at very high velocity.

Customer support, customer success, anything to do with how we support our customers, not so much. As in they have a version of the past which they document it because they simply can't move forward at the same velocity. So what we built is this automation machine that can log into your piece of software, take a screenshot of all your features, log into your sendes or intercom, scan all your articles, replace all of the outdated images, run that as a job every Sunday night so you don't have to. Now that's what we're doing today.

What we really want to do once we kind of get hold of all the kind of challenges around that is I want to really just have you add a pixel to your app so that the very use of your product becomes the documentation as in the way people use your product as in ah yes I had 100 other users set up two factor authentication yeah I track that I'll create documentation for it because that's obviously what they do and their very use means that oh engineering launched something at 2:00. Some users got exposed at 3:00. Dennis picked it up at 4:00, 5:00. We already have an article on how to use that because it happened automatically. So that's the space I'm in or very short if you ask me in the lift. What do you do, Dennis? Some very nice PGs. That's really it today.

Rob May: Yeah. I mean, does that make sense?

Ryan Eppley: Yeah. Because we just, you know, we're less than a, you know, year or so startup and yeah, just created our first docs for customers because like someone couldn't log in. I was like, "Oh, how do you not know how to It's like a CLI tool." So you have to download this and you have to run a terminal command and so here I am taking screenshots of it.

Dennis Mortensen: You probably I'll tell you what your process looked like. It's probably like to begin with very early in our journey there. There is no kind of customer support. We're just founders trying to make people happy and I'll do anything to make you happy. Then at some point there's a little bit of scale. We hire our first few people. Those first few people their instructions are not McKinsey style. And they're just, hey, we got 100 incoming requests. You solve all of them and you don't go home before you solve all of them. Then after a while, hey, I'm drowning here. We should probably kind of put some documentation in place. Yeah. Okay. Then we sign some sort of contract with some vendor. We start to put a few articles in place and kind of say, hey, I hope they can do things without asking us. Yes, yes, yes. We'll change the product, but I still need to put this in place. And that's where we arrive as well. Hey, how about I help you automate some of that?

Ryan Eppley: Yeah. No, super smart. I love the way that you make a list of things that you work on is I really hate this.

Dennis Mortensen: Yeah, at some point I'll get to fixing it now. Yeah, it's just things I hate. Which is always nice because then if you work on things you yourself hate, you don't have to kind of assemble some sort of distant empathy from having talked to users. I'll just close my eyes and say, you know what I hated the most as a founder was when I looked at Oscar. First of all, as a founder, I live in the future. So I don't even talk about the app we have. I talk about some version of an app we might have in four months and somehow hey the world in my mind it already exists. It doesn't. Then of course reality is four months behind. Then I look at the documentation that's another four months old. So there's like an eightmon delta between the universe I live in and what people see. Like not [ __ ] cool. So yes.

Ryan Eppley: Yeah. Eight months isn't that bad. Could be eight years.

Dennis Mortensen: You know for some it is. But the cool thing is I've become so cocky when I speak to people which is that there's not a single product where without any research I can't immediately say hey let's jump into your documentation let's see if it's up to date. Yeah guaranteed 100% of the time let's just pick anything. Oh yes [ __ ] Yeah we rebranded. Yeah you did. When was this? 18 months ago. Yeah. Come on.

Ryan Eppley: I always like how you'll launch a new website and like a year later you're getting traffic to some page. You're like, "How's that that page should still be there? How who's going to that? And where does it live?" Like, that's really funny.

Dennis Mortensen: So yes, launch Brightly. That's what we're working on. We're early like you guys. So, we are just 10 guys in a in a basement. This is my first venture doing that fully remotely. So, everything I've done in the past have been onsite 0800. Here's the address. Yeah. You guys on site or remote?

Ryan Eppley: We're a combo. We have We're nine people and like five of us are here in New York.

Dennis Mortensen: So, but do you have an address?

Ryan Eppley: We have WeWork.

Dennis Mortensen: Yeah. Yeah. No, I love WeWork. We work is kind of like the I don't know the AWS of space. You can just spin up a new room when you need it and delete it when you don't need it. We used it for all of XI as in I probably had 25 rooms in the end. Wow.

Ryan Eppley: But just it gets expensive when you get to 20 30 people I feel like

Dennis Mortensen: but sometimes particularly if you don't know like because people want you to sign at least a three-year lease if not five for a business lease. So yeah we had a very I was very happy with it. I'm not on commission from we work but it just worked very well for us. So we spun up our whole labeling operation inside we work. Then we moved that to Manila later but I just spun it up. I only had 30-day contracts on anything. So never cost me really anything to spin it up.

Rob May: We work $1,000 for this episode. Next sponsor.

Ryan Eppley: Well, they won't actually give it to us cuz I hate WeWork. I can't do it. We bought We bought an office and you know, we're only paying a couple grand a month for it, but it's what we need. It's a four person team about to be five and I just You came by a little before we started hanging up my posters, but I have all these like band poster. I'm really big into like heavy metal and rock, so I have all my band posters up. We play music and I don't care about like having neighbors and like a wei work. So, you know, it's also nice to have your own space sometimes.

Dennis Mortensen: Yeah. Yeah. Yeah. We ended up getting our own floor at some point as well, which made the whole thing a little easier and, at some scale you get a little bit of a discount, but you're right.

Rob May: So, so let's use bands then. So, we always end on a question about New York. And I, so I think maybe the question for today should be like, what's your favorite band that originated from or spent most of their early days in New York?

Ryan Eppley: Yeah, that's a tough one because there's so many bands that didn't grow up in New York, but they became like a New York band. But it's an artist that I really really have fancied for a long time. But Spotify Wrapped told me they're my number one listen to last year. So, I have to give them a shout out. Yeah. Rob Zombie.

Rob May: Oh, wow.

Ryan Eppley: Yeah. He went to a couple art schools here in the city. And I never knew that until recently when I went through my rap and it was giving me like the bio. I was like, I didn't even know this was a New York artist. But yeah, he he went to I think Pratt. He went to another art school. Dropped out of both. But yeah, I'm a huge, you know, Rob Zombie fan. White Zombie, you know, the whole thing.

Rob May: Mine would be Yeah. Some great great music. Yeah. Mine would be the Beastie Boys. Like they were like so prime in the 90s when I was like in my like really, you know, formative music shaping years. And so Yeah.

Ryan Eppley: Have you seen their recent like live show, not like music, but they do like a documentary of their band and it's live. It's really good. You should watch it. It's like an hour long and they tell the story, but they do jokes in between it and they have like clips of it. Yeah, you should go. I think it's on Netflix or something like that. Okay. What about yourself? What do you got?

Dennis Mortensen: Have one. Good question. JC perhaps.

Rob May: Mhm. Okay.

Dennis Mortensen: If I was to kind of Yeah, cuz he's just old enough where he was young when I was youngish. Yeah. So yeah. And then perhaps a little bit more because I live so the JC is in my building. So I perhaps I've we I move closer because of that because I look at it every day like oh yeah.

Rob May: So I will say my my favorite current band that plays in New York is called Just the Tip. And they, you know, really funny name. These guys play. I saw at Red Lion, which is one of my favorite live music places down by NYU. And they yeah, they've played in there and they they do all kinds of classic rock covers. They're a lot of fun. And they've got like six or seven or eight people in the band, so it's a big band, which is fun, too. A lot of the guys play lots of different instruments, so yeah.

Dennis Mortensen: Okay, I'm going to look that up. Yeah. All right. Next Spotify. If you write that on Google, I'm not sure I'm going to get

Ryan Eppley: You're not going to be careful putting that in your search on Spotify and see where it takes me.

Rob May: Amazing. Cool. Well, Dennis, thanks for coming on today. And if if people want to learn more about your company, just launchbrightly.com.

Dennis Mortensen: Launchbrightly.com. I'm at Dennis Launchbrightly and everywhere in the internet under Dennis Mortensen. Feel free to kind of ping me.

Rob May: Cool. Well, thanks for coming on and everybody out there, thanks for listening. Give us a thumbs up, rate us if you like the show, otherwise don't. And please go watch all the other ones and if you know somebody that should come on as a guest host, talk about stuff in AI, please reach out with an email. Thanks. Heat. Heat.