Latest / The AI Ecosystem Podcast / Hackathons in the Age of AI

Transcript

Hackathons in the Age of AI · The AI Ecosystem Podcast

• Bob Moore: Hello and welcome to another exciting chapter of the AI Ecosystem Podcast. Bob Moore here with Lindsay DeFalco. Today we are talking about hackathons, one of my favorite subjects of all time. Relevant for decades now, maybe. I don't know when that word first appeared on the scene, certainly more than a decade, but kind of given new meaning and maybe a rebirth or a resurgence in the AI era. We've got one coming up here at Crossbeam. We're going to spend some time and talk about how we think about hackathons, how we've implemented our hackathon and have a little discussion about like, what should we actually be building in a hackathon inside of an established company in this era, in this moment in time? Lindsey DeFalco: ⁓ I it could get interesting too because generally ⁓ I love but I also have some spicy takes about hackathons that are probably directly a result of how you view hackathons, Bob, and ⁓ probably representatively ⁓ a lot ⁓ CEOs or in companies view hackathons. as a, mean, I know you also were a former engineer as a maybe more recently former engineer. and the person that wrote it. You've outpaced me recently, for sure. Bob Moore: We're recently reborn. I feel like I'm, yeah, was back on the wagon now or off the wagon, depending on how you look at it. Yeah. Lindsey DeFalco: Yeah, yeah, you can definitely claim it ⁓ probably a little bit more than I can at this moment. But yeah, I think we can get into some some cool topics because part of this from my perspective is always like, and I think this actually is even just more a thing with people being able to do a bunch of vibe coding and stuff with like create more things quickly with AI, which has been a theme of what we talked about is like, what's the difference between something you create in a hackathon or ⁓ vibe coding or whatever versus production ready ⁓ like what's process? ⁓ Bob Moore: Yeah, life is just one big perpetual hackathon at this point. ⁓ Certainly for the Vibe Coder ⁓ set, really kind of puts a very interesting spin on this question of what should these things even be? It would be helpful. I can give the speed round rundown on how we run these things, ⁓ which to get your daily RJ Metrics reference out of the way is exactly how we ran them in RJ Metrics. Or at least this is how I like to run them. And then maybe, ⁓ Lindsey DeFalco: Yeah, exactly. Bob Moore: You can layer your spicy takes on ⁓ why that's a deeply flawed system ⁓ as a follow-up, if that sounds good. Lindsey DeFalco: Yeah, I mean, we pull my bingo card and mark off that center square for your RJ mattress and then ready to go. Bob Moore: All right, ready to rock. So here's the idea with our hackathons. We shut down the company for a day and a half. One day is a full day that is dedicated. This is a nice cure for time zone issues. Like, doesn't matter when you wake up, when you go to bed, it's kind of like from when you get up to when you are not up anymore on that day, you are working. And if you want to stay up all night, more power to you. And the next morning, ⁓ we have presentations and everybody basically gets a five minute slot. It's three minutes to present what you did. And then two minutes of kind of buffer time for Q and a from the audience. And the next person is setting up. ⁓ used to do these at RJ in person and in in-person office at this point at Crossbeam It's a fully virtual experience, even if people are in an office, they're presenting on a, on a zoom meeting. and the core premise here is It's kind of a very pure free market system. Teams get formed however you want to form them. You can be a solo person, you can be half the company, it doesn't matter. ⁓ Ideas get sourced however you want to source them. There is usually some kind of very values level description of ⁓ good means and typically it's something as vague as you should be building on something that makes the company better. So if you want to do something about internal company culture, you want to do something about hiring, you want to do something about our financial model, you want to build a new product, you want to improve our existing product, you want to communicate with customers better, all these things make our company better. ⁓ If you want to go do some very random thing on another planet that you can't link to some downstream benefit to the company, it's probably, that's when you start to get outside the box. ⁓ We had one guy at RJ Metrics once build He got tired of the Keurig, the K-cup machine in the kitchen, always being empty and him having to fill up the water. So he built a system that works just like a toilet tank, like it had a little flapper ball in it. And when the level got too low, he did some plumbing work in the kitchen and actually hooked like a water supply into, he drilled into the K-cup reservoir. And that thing worked for like three years. It's just like the K-cup reservoir was always automatically refilling. Lindsey DeFalco: empty. Bob Moore: He didn't even place, but shout out, I think that was Sean McAvene. Shout out Sean McAvene. So, but anyway, it benefited the company. So it totally qualified and it was a little bit of a hardware hack, but it counted. So then after the presentations, we do rank choice voting. So everybody gets a ballot and you rank your top favorite projects and make the list as long as you want. If you have a top two, it's two. If you have a top 10, it's 10. And if you know how rank choice voting works, it's awesome. It kind of makes the consensus bet best results rather than just like looking at the project that got the most number ones. And we can, I'll do an hour on rank choice voting some other time. If you want to go beyond that. ⁓ Lindsey DeFalco: It's also a dramatic reveal too. It's a fun way to watch. Bob Moore: ⁓ yeah, it's really great. You watch like the burn down of like, yeah, did anybody get more than 50 % in the first round? If not, we eliminate the bottom most vote getter and go to what all those people who voted for that bottom most one put as number two and redistribute the votes. And you gradually, gradually get to the point where someone has a majority. ⁓ It's a fun process. And then prize wise, ⁓ we are typically, ⁓ My favorite version of the prize is what we call the Brewster's Millions Prize. ⁓ Brewster's Millions is a movie that's been made like five times throughout history. The most famous one is Richard Pryor in it. ⁓ the storyline is roughly that this guy inherits $100 million, but there's a clause in the will that he has to spend like $5 million in a weekend or else he loses the whole thing. There's some convoluted reason why ⁓ he has to do this. ⁓ It's kind of a comedy of errors where it turns out to be really, really hard to spend $5 million without accidentally accruing assets or equity value or something else in a weekend. So the Brewster's Million prizes, we give you $500 and you have to spend it in a 24-hour period. It doesn't matter what. ⁓ You do, but it has to be gone in 24 hours. ⁓ back at RJ, we had some great ones. Buck, the co-founder of Crossbeam, one time got a tattoo with his Brewster's Million winnings. We've had people go to the casino and put it all on black on the roulette wheel. They lost by the way, ⁓ which I guess fulfilled, it satisfied the requirement. I don't know what they would have done if they won. I guess it counts. yeah, lot of people buy company swag or they'll go out to like a really fancy dinner ⁓ or they'll fly. If it was a team with people in different cities, they'll like fly someone to the other place so they can all like hang out. But there's a lot, a lot of good stuff has come out of that. Lindsey DeFalco: There was a, there was another. Bob Moore: But anyway, that's like the hackathon. ⁓ Historically, pre-AI, that's the hackathon. And before we get into the AI stuff, Lindsay, I'd love to hear like your, how you would tweak it in an ideal world or maybe how we have tweaked it at these times. Lindsey DeFalco: Yeah. Yeah. Well, first a couple other fun prizes because one of them makes me laugh a lot. Well, the first one, which does not make me laugh is me and Sean Blanda got the hundred dollar cheese steaks when we were in Philly. They're like they have like gold foil on them or something. Yeah. Delicious. Loved it. ⁓ then another which I loved is ⁓ Bob Moore: ⁓ yeah. ⁓ yeah, ⁓ that's a great one. That's a great one. At Dark Light Prime, right? Lindsey DeFalco: ⁓ Nick, Nick McAvoy, shout out Nick. he convinced his team to let him buy something that would like mill grain into flour to make bread. And then he delivered bread to his teammates ⁓ some regular cadence over some period of time. I'm not sure if it's still going, but, ⁓ if you guys are listening, you can let me know. ⁓ that one was creative. Yeah. I think the, ⁓ I, I love hackathons. ⁓ I think they were. ⁓ Bob Moore: very on brand. incredible. Lindsey DeFalco: admittedly a lot more fun when everyone was in per- in person staying late and you you like ordered food and you know people were up all night like that kind of old more classic hackathon structure was really fun but I still think they're great because it's it's a time to like put things down And maybe even just like churn through something ⁓ that's generally not something you'd churn through. So we did it. I think we had a winning hackathon project, you, and Mandy, in the very early days where we just created our help center. ⁓ yeah, it was. And we just literally made a help center and filled out all the content, created all the docs and all that. So the ⁓ tension always see with hackathons is ⁓ Bob Moore: Yeah, ⁓ that was great. Yeah, that was a winner actually, I think, yeah. Lindsey DeFalco: generally the perception that they're production ready and the amount of time that it will actually take to get them to be production ready. So, and it's generally conflict between go to market and then the product ⁓ design engineering side of the house. And I think it's tough to solve for. I think, ⁓ you know, that's if you end up in a culture where your engineering team hates hackathons. then you're probably doing that wrong. And there's probably some really big rift in the perception of something that comes out from a hackathon being like ready to go to production at scale versus it actually being a hackathon. So I think from the early days, something we've done well is said that they're about learning and that the point is to learn. But the reason I said I probably have a spicy take is that we also have a tendency to be like, okay, great, this can be done in this amount of time. And we used to ⁓ like an evaluation at the end of every hackathon of like, how long would it take to get this in production? And I think that starts to take away the fun, because it feels like you're just doing your work faster. And so I don't know if there's like a solve for it, but I think that's generally like the spicy thing. If there's ever tension around hackathons internally at companies, it's that. it's ⁓ to be fun. ⁓ But then people get excited and want it to be real. So ⁓ my spikiness. Bob Moore: Yeah. I like, here's my, the CEO, persona that is infuriating, uh, to product leaders, which is like, but like, you kind of can do it faster, right? Like there's like a, and I think there's a happy medium in this negotiation that exists, which is, uh, there is such a thing as like a proof of concept or a, uh, removal of Lindsey DeFalco: You are the problem. Bob Moore: the mental burden of all the reasons this is going to be hard that allows us to A, demonstrate why some things are worth doing the hard things to get to in terms of getting to the finish line and like once seen and felt and understood as a prototype are sometimes elevated in the sense of importance associated with them or like the priority associated with them A. And then B, there is this thinning of the knowledge transfer gaps that exist between like executives and product teams and engineers and designers that happened during a hackathon that collapsed those down, that I think actually do remove a lot of what makes these things slower, which is not the implementation details, but it's the surfacing of what the edge cases are. And it's the decision-making around, you know, the pixel perfect designs and ⁓ it's the handoffs between these teams and the... availability and meeting conflicts and all kinds of other stuff. Like if no one else is doing anything else across a very cross-functional set of people that would normally need to collaborate in a normal work environment, they just get shit done faster than they would have otherwise. so it's really like sometimes the lesson in you can do this faster is not like work harder, code better, design faster. It's like make your organization flatter, like, reduce the number of stupid meetings that people are going to like increase the ⁓ level of autonomy at which people can operate or the amount of trust that exists between teams to like step into each other's territories a little bit when there's an appropriate amount of comfort. Like it's a little bit of like an organizational wake-up call of like this stuff is taking too long not because we're not working hard enough but because we're not getting jobs done as efficient way as possible. Yeah, that's a way of like me like throwing a middle finger at myself instead of at like the PDE teams, right? It's like you've built an organization where things could be getting done faster and they're just not because of a bunch of, you know, silly bureaucratic reasons. and this is a great way that. Lindsey DeFalco: Yeah, I think that's, that's probably why you start companies, right? Like the early days, the building phases, the, you know, being able to have an organization that's as flat as possible for as long as possible is, it's fun, it's exciting. Yeah, and I do, I think there's a lot of, and again, to bring this back to like, how I think this is what AI is doing in a really big way in organizations, is it's breaking down the, you know, not only like the Bob Moore: Yeah. Lindsey DeFalco: Let's say you have an engineer who's like, I shouldn't propose a product idea. It's not my job. I don't want to step on a product manager's toes. It's taking that away in the same way that hackathons take it away. And I do agree. A lot of it's just, think, a mental block where people are like, oh, well, that's not what I do. don't come up with the ideas. I don't come up with the design. There's somebody else that does that. And also, I don't want to offend them. And it kind of breaks all that down. And so that's actually what I'm most excited about with a like what AI is doing to the collaboration within product design engineering in general is it's giving engineers tools to write product docs and make designs really easily. It's giving designers ways to access code more easily and same with product teams. So, yeah. ⁓ Bob Moore: ⁓ So enter AI. ⁓ Lindsey DeFalco: Yeah. Bob Moore: So part of what it's like, where do I even start, right? Like hackathons, there's this incredible appeal of like, hey, maybe we can build like five things in 48 hours. And your mind gets blown that you kind of get to, you know, see something soup to nuts, get created that at least looks like it kind of works, even if the backend is like total garbage and nobody's paying attention. But like take away the word hackathons. And I think I just described like lovable. you know, like Lindsey DeFalco: Yes. Bob Moore: a rapid prototyping tool where really you're just substituting engineers being available to work on everybody's crazy ideas for 48 hours for an endless supply of commodity labor that performs at the, you know, X percentile of your typical engineer and that percentile is climbing. but like also available. like This is one of the things that I'm kind of, I'm excited and nervous and don't know what to expect about in this hackathon which is, ⁓ we gonna have a hundred, normally have like 10 teams. We can have a hundred teams and are they all gonna build ⁓ so much more than usually built that we're gonna have like every presentation is gonna run over? ⁓ Or will there be this self-consciousness about like, that the will be good ideas, ⁓ not, ⁓ ability to execute on those ideas and maybe we'll actually end up with fewer because everything that everybody ever wants to like hack on they're just kind of doing their day-to-day life and we're kind of blessing that it's acceptable way to spend your time in the company in the last couple of months Lindsey DeFalco: I'm ⁓ have no idea. Maybe it'll depend on whether or not they listen to this podcast before the before that happens. Not to how much how much pressure they feel around it. ⁓ I think there's ⁓ other thing I'm wondering, too, is like how much ⁓ how much A.I. slop will come out of it. ⁓ I say pessimistically, so. ⁓ Bob Moore: You mean like 98 % slop or 99 % slop? Is that what you're trying to decide between? It's gonna be a lot, yeah. Lindsey DeFalco: Pretty much. Yeah, it's funny because I've literally done hackathons and won our internal hackathons just on creating a workable front end of a new concept. it's like, was a killer hackathon. was like, show something that looks functional. And that literally takes one sentence prompt, and anyone can do it. It's both cool and then also immediately is it, okay, what's the next level? ⁓ It's like, it cool and also not cool that that's so easy to do now? So many hackathon projects that took the full 24 or 48 hours, I think will take like five minutes. Bob Moore: Mm-hmm. Yeah. Yeah, I mean, you're right. The hackathons used to be the old slop factory. And this is actually like maybe something we should have spent more time on in the original hackathon description, which is like, you mentioned those meetings we used to have about what it would take to bring these things to reality. And like the conclusion of those meetings is kind of self-evident, which is usually out of the dozen projects that got completed, one or two of them would end up. Lindsey DeFalco: Yeah. Bob Moore: actually making it out into production. And now those one or two might be incredibly impactful. Like if my memory serves correctly, Stitch, which was a spin-out business of RJ Metrics that ended up being very successful, was a hackathon, like started as a hackathon project. major, major parts of our product functionality, quote unquote, started as hackathon projects. Not to say that the ideas are always invented during the hackathon, but like the hackathon is the first opportunity to show the potential of something and it can end up getting kind of big and exciting and that's great. But the other 90 % plus of the stuff that comes out of a hackathon, you know, people put their blood, sweat and tears into it for two days and then we never look at it again. Maybe it changes our thinking a little bit, but that code never reaches production. And even if we like the feature, we rebuild it from scratch. And that is fine. And it's like an accepted parameter of these things. But it is funny that that's like... That is the non-hackathon norm ⁓ today. It's like the, we're all creating slop all the time, basically. Lindsey DeFalco: Like most of our, even our, you know, we do weekly, weekly reviews with the, like the PMs each lead a weekly review for their team. And, know, they're bringing ⁓ prototypes ⁓ all levels to the, ⁓ to the now. And those historically for like a PM to create it would be a hackathon, like to create a full design, you know, you either need to know how to front end code and ⁓ go do yourself, or you need to know how to. use Figma at a decently high level. ⁓ My expectations of what people are going to generate are pretty high, I think, in terms of what the potential is. Bob Moore: Yeah, my biggest worry is that they're gonna have bad ideas. Like I really do, like I think there's gonna be great work done and the potential to build amazing things is gonna be very strong. And yeah, maybe my biggest fear is that people are just gonna work on the stuff. But, you know, the stuff to me is not necessarily the wrong stuff in general. And maybe that's kind of the point of like how democratized these things are. So I don't put finger on that scale. Lindsey DeFalco: Thanks. Bob Moore: I think I just wait to be, I wait to learn, ⁓ something. ⁓ ⁓ Lindsey DeFalco: Yeah. Yeah. So all right. So we have this internal channel where people can like drop ideas. What is interesting to you like thematically and do think there's been a difference in the ideas this time around versus previous times? Bob Moore: Yeah, that's a great question. I'm to pull this channel up and maybe like read some of them out. Because I think it will, ⁓ you know, for folks that are ⁓ following along here, these are like really good representative examples of this, right? So like a lot of this stuff, I'd say one theme here is we've adopted Claude internally as kind of our AI chatbot and authority and integration hub and everything else at the company. So there's been a... big movement toward people creating skills and projects and also creating either direct MCP connections or data integrations through N8n to give Claude access to certain classes of data. ⁓ So there's a suggestion here about giving Claude access to ⁓ feature flag data. ⁓ So, know, our end users and customer success can understand what customers have what feature flags turned on without needing to log into like our secure admin panel area. ⁓ this is really cool because it's like potentially a security hardening thing, ⁓ because it reduces this frequency and even maybe need for access to this more sensitive code area, just to get it like a narrow slice of data. that sometimes, ⁓ you know, might actually be easier to retrieve alongside other things and getting like a client report. So it's kind of a cool idea, but it does require data making it from like our production data sets out into Claude, which is something that is currently not. ⁓ kind of a blessed behavior. So there's almost like we need to create some kind of separated air gap layer for these non-sensitive slices of the production data that Claude's then able to see. So that's just an interesting, like very specific, but also maybe not that specific ⁓ thing that if this is true, what else is true, right? You can build, you can give Claude access to all kinds of other stuff. ⁓ Hop in if you've got any comments on any of these. ⁓ Lindsey DeFalco: Yeah. I think I noticed ⁓ there's like a theme of either replace this tool that we use for something internal or build a fully functioning tool for something that we do internally. And I'm trying to think like, because I feel like that was kind of a theme in the past. Like a lot of people look internally for, you know, like what can they do operationally to make their jobs easier or people around them's jobs easier. But I'm actually like, Bob Moore: Yeah, yeah. Lindsey DeFalco: really optimistic about us actually being able to do this for a bunch of these. And this is, I think we keep talking about this and this is like the big topic in the market right now is how AI has changed the build versus buy conversation. And so ⁓ think ⁓ I'm gonna be really interesting to see like if these actually end up being functional at the end of a hackathon. Bob Moore: Yeah, really. I love that. ⁓ There's a, ⁓ know, ⁓ area where I think, interestingly, ⁓ tend to create more ⁓ things is actually on like operations and go-to-market related projects. So you use example of that help center that, that you Mandy and I stood up, ⁓ you know, eight years ago. which I think is like the bones of our current help center still have like relics of the work done in that, in that particular thing, which is crazy. Like some of these things are really long lasting. One of our team members here proposed creating kind of a control center and like data aggregation hub for our AEO efforts, which is technically answer engine optimization. Although like it's really just call it AI engine optimization or Lindsey DeFalco: Mm-hmm. I didn't know that. Yep. Bob Moore: It's basically making chat GPT recommend your product. It's kind of what it boils down to, but in our, this is so funny. Yesterday, I had three, actually four completely separate meetings where AEO was the topic. We should probably just do a whole podcast episode on this because it's bonkers. But like I, in the morning I met with Sean Blanda who ran content marketing for us for years. And it's just like brilliant kind of. content thought leadership guy and it was a theme that like all of his clients are now asking for AEO stuff and like how can we create content that makes us stand apart not to the humans that want to consume our stuff alone but also like to the agents that might decide to recommend our stuff. And then I went over to the Franklin Institute, was a science museum in Philly where I do some volunteer work and we were ⁓ talking to them about their marketing website and like all of the like classic SEO style strategies that could very easily be employed that might then increase the chance that, you know, chat GPT recommends the Franklin Institute as a place to go on the weekend with your kids, you know, that has a parking garage and ⁓ ⁓ to make sure that that's incredibly visible to the robots that are scraping their site. And then I talked to Jake, my old business partner who runs a company called Common Paper. And we were talking about like, you know, the contracts that they create their ⁓ CommonPaper is a legal tech startup that works in the realm of standardized contracts and contracts as APIs. Those are really powerful content assets that might be kind of fascinating as an AEO strategy to get CommonPaper on the radar of folks that ⁓ are setting up a new tool that they're building in a lovable or in a code automation platform to have the contracts just built in. ⁓ always have your terms of use page, always have your privacy policy, and it's always common papers. Like, why isn't quad code just inserting common paper with every single project that ever stands up? And then ⁓ had a call internally here ⁓ about our own AEO efforts, which is what's being posted in this hackathon thing. ⁓ such a hot topic ⁓ no one knows what they're talking about. ⁓ ⁓ old SEO firms are making serious bank because they're just calling what they do AEO because it's not not true. ⁓ A lot of SEO has always just been cleaning up your crappy front end code so that it's very interpretable and discoverable and like consistently uses the right tags and attributes on objects. And you know, your servers are resolving down and forwarding to consistent URL paths and stuff like that. And then there's obviously like inbound links ⁓ like inbound links are kind of the same as inbound references. It's like get other people to... say things about you and point to your website that kind of deliver the message you want. And like that classic strategy works in AEO too. But all the crazy growth hacks and everything, there's this whole other vintage of startups that have started to build kind of what this hackathon project is proposing, which is like, I'm going to give you a dashboard where you can see how you're doing. And the way we tell how you're doing is that we, we run 500 different, chat GPT accounts, some of them free, some of them paid, all on different Docker instances, accessing it from different browsers with different perceived geographies. And we asked the same exact question and we see what chat GPT gives back. And we look for the incidence rate of your solution being mentioned in those answers, which is like cool idea, but very, ⁓ not necessarily super defensible, most people probably don't want to go through that headache. ⁓ And I think there are some really nifty businesses being built, but ⁓ I know at least a few startups that have those crazy rip roaring AI era ARR growth curves ⁓ doing like that because companies have budget to spend on AEO. And because nobody knows what they're doing, it's very easy to convince them to give you their money based on, you know, whatever story you want to tell. So I don't know, I'm a The importance of that area is enormous. The ability of anyone to execute on it well, I think is questionable beyond just doing common sense stuff, which I suspect does work over the long term, just like it did for SEO. But anyway, great hackathon project. I think ⁓ it's a great question of like, it's so, it could potentially be very broad, but I think in this case, the idea being suggested is just to... create some baseline monitoring and dashboarding and kind of attempting to attribute outcomes to these various projects we have going on for AEO. Lindsey DeFalco: Yeah, I wonder what the new like first page analogy is going to be for AI. Maybe just that it's returned at all. Bob Moore: Yeah, that's probably a great point. like somebody, this is where there's definitely businesses to be built and they sound like they'd be really fun to build for the first six months and then like absolute hell for the next 10 years, which is like, you could probably, if you're really good at like brand and you're really good at category creation, design an answer to that question that the whole world. decides it's the right one and decides to adopt and then like build the product that is the one that happens to know how to measure that and give some kind of score and like become like the standard. And that's all great until ⁓ Anthropic that they don't like your standard and does something intentional in their like harness around core Claude to make it make absolutely no sense anymore or be impossible to have that be the. the way in which you optimize and measure results. like, it's like classic SEO, right? It's like, you're always trying to kind of like chasing this ghost. Yeah. Lindsey DeFalco: Yeah, you're at the mercy of what all the models had to do, which I guess similarly, I guess you were always at the mercy of, you know, Google, but it felt bit more static than the current landscape. ⁓ I do. was, I was ⁓ to chat GPT about bags. I really like bags. I have kind of a problem like, you know, sling bags, backpacks, whatever. ⁓ and every time ⁓ Bob Moore: Probably, yeah. Yeah. Yeah. Lindsey DeFalco: it would tell me, give me a recommendation or give me an opinion on the recommendation. At the end of the message, it would then be like, do you want me to tell you about a super secret bag that's really popular in this niche community and nobody knows about it except them? And then I'd be like, yeah, why didn't you tell me that in your initial message? If it's that good. And then it would return like the same list of five bags that I was looking for before. And I was like, I don't know what this is. Bob Moore: Hahaha Lindsey DeFalco: And I don't know if this is, at one point I asked it if it was giving me an advertisement for something, because it was like really pushing. And it said no, but you know, ⁓ do trust it? ⁓ Bob Moore: Yeah. I, yeah, this is, this is like, chat GPT is so fascinating cause they are like, ⁓ even though it's not giving you an ad, think they are like soft testing where to insert ads and like how to like best, you know, likely drive ad conversions in the, in the moments. Like, so I wonder if like, well, strictly true. It's not an ad. You got put into some kind of like AV testing loop around like, Hey, this person seems like someone I might be able to push an ad to in a future state. Lindsey DeFalco: Yeah. Bob Moore: What happens if I position it like this versus something else? Lindsey DeFalco: Mm-hmm. yeah, it's funny from like an AEO perspective in this more like, ⁓ you know, generative AI world where like, would you rather be in that first response or would you rather be the teaser where they're like, this is better than anything I told you before ⁓ and just yes and I'll tell you what it is, ⁓ where it's you know, hyped up. It's like, ⁓ psychologically, am I going to be like, ⁓ Hell yeah. And now I'm more into this one than I am in the first four results that you gave me. So it's funny because it's also not like one to one with, can visibly see this thing on the first page of Google, or I have to click next page. There's different ways you can manipulate it and get the Bob Moore: Yeah. It's all pretty, it goes from being, it's kind of like a good metaphor for the whole AI era, because it goes from being this extremely deterministic, consistent thing to probabilistic. So like that metric can no longer be that I'm number one on the first page always. It can just be that like I've gone from 21 % of the time to 27 % of the time. I ended up getting, getting included. Lindsey DeFalco: Okay. Yeah. Yep. Yeah. Yeah, because you could be, I mean, let's say you measure, OK, you're actually in the initial response, you know, from whatever ⁓ then, OK, great. But what if they're like, hey, here's ⁓ really lame option? ⁓ OK, not great that you're in that first response. And then it's like, tell me if you want cool options. ⁓ then that's the second, you know, it's ⁓ is funny because it's like. Bob Moore: Mm-hmm. Mm-hmm. Lindsey DeFalco: you have to be context aware more so than just the context being the number of page that you're on. Bob Moore: Yeah, I don't envy all the SEO firms or internal SEO people who are gonna get like the DM from the CEO at two in the morning like, just asked chat GPT who's the best and they gave an answer that's not us. ⁓ Because they're in the 1 % of times that chat GPT gives that answer. Yeah, being at the mercy of like ⁓ every single searches. Lindsey DeFalco: Yes. Bob Moore: somewhat of a roulette wheel, especially when it's subjective answers is, it's gonna be tricky. Ads are gonna make that trickier too, but hey, I see one in here that I love because it's the craziest case study that's blowing my mind. So remember we talked about hackathons used to be like the original slop factory and they're like the lowest bar of like getting something live and seeming like it's working. You put in here, Lindsey DeFalco: ⁓ yeah. Bob Moore: This is a suggestion of yours, which is to build a rep to rep mapping, potentially not an account mapping matrix, right? Which would be like in Crossbeam, it's like, here's how many customers I have in common with this partner. But what a lot of people want is I am sales rep Bob at Crossbeam and I want to see how many customers that I own. I have in common with Lindsey over at ⁓ trend kite. ⁓ And we want to collaborate just on our books of business and where they overlap. You can do that in Crossbeam. but you have to go to, you have to build a report and you have to filter down by the two reps. Wouldn't it be cool to just have like a giant matrix of like all the reps versus all the reps and you can just drill right in. Like we do with the different deal stages. There's a comment, you posted this as an idea and then RAA said, hey, Lindsay, I thought you already vibe coded that. And you did, right? So like you have like a UI that seemingly works in our UI that has that exact matrix. And now, ⁓ You said, yeah, but it was all demo data and pure front end. But like all demo data and pure front end is how you just described the way to win a hackathon. So like now we're in this mode where like on a whim, just to like prove a point during the normal workday, you can spend an hour and whip up something that would have won a hackathon in the past. And the hackathon is actually becoming the place where you implement like the first version of the potentially production ready thing. Lindsey DeFalco: Yeah, exactly. Bob Moore: Because in a full day, that's what you can do. So like, this is a crazy sign, maybe the perfect sign of the evolution of what this thing is becoming. ⁓ And maybe changes the standards around what you think about what a successful hackathon project looks like. are you bringing a new mind-bending idea to bear or are you making something that feels like it should take longer, literally actually be done, done and production ready in that time period? ⁓ Lindsey DeFalco: Yeah, it's like maybe the new, next, no, maybe later this year if we do another one, it's the end of the hackathon, you're launching the feature, if it's a feature, or like you're launching this thing. And that's the threshold. Yeah, it is funny because I created that, I like vibe coded that to demo at a standup that we had gotten the product team access to Claude Code. Bob Moore: Yeah, ⁓ Ship-A-Thine. Lindsey DeFalco: yeah that probably would have won me a hackathon like two years ago ⁓ now i'm like yeah now i'm like it's slop it's ⁓ it's using fully fake data i forced it to do a lot of things in a way that i know is bad and ⁓ yeah it be great to like clean it up ⁓ and i ⁓ i have a tendency all my hackathon ideas are like build this feature but in two days so everything i said i hated about hackathons i also ⁓ am a part of ⁓ ⁓ But I actually think it's possible now in a more realistic way than it probably was before. Bob Moore: really true. All cool. Any ideas on that list that jumped out to you or like any other stone unturned here? Lindsey DeFalco: think that's pretty much it there. I think it's like ⁓ I thought was cool. So there's a lot of stuff from the engineering team and ops, which is pretty standard, think. ⁓ team, ops, and product. The thing I want to see ⁓ more that I think AI makes possible is a ton more from go-to-market. Because now we are in a position where go-to-market teams fully create working things from an ops standpoint internally in a hackathon time period really easily. And I think that's also usually historically the go-to-market teams feel left out of hackathons because they can't do anything. They ⁓ ⁓ I don't mean that they're not capable. mean, generally if you think of a hackathon ⁓ ideas, it's like you need somebody from ops or you need someone from engineering. And so it feels like you don't have as much that you can actually do yourself. Ooh, a good example. It's not a hackathon idea because she just did it, but Mandy, who is our CPO, Chief Officer, we had this tool that we use that basically introduced people to have coffee every week. ⁓ And we're for this. Sorry, I hope this product isn't listening. ⁓ And she and recreated the workflow in n8n ⁓ And I it took her less than a hackathon amount of time went in and now... Bob Moore: or yeah, yeah, Lindsey DeFalco: It's our own workflow that's like introducing two people to set up ⁓ a coffee chat every week. And that's like a fantastic example of somebody not on the product side the house at all, who literally set up an n8n workflow and used ⁓ Claude to this app that for us is not worth paying for. Bob Moore: And I think like, I don't know what we're paying for that app that we're not using anymore, but I bet it was on the order of like $5 a month. And like, there's one lens you could look through, which is like, okay, we have a C level manipulative at the company spending a day or I don't know, maybe it was an hour, but basically replacing this thing. And what's the point at which you actually get real returns? But I think... Lindsey DeFalco: I Mm-hmm. Bob Moore: We're in this moment where I don't know or necessarily care if Mandy builds another N8N powered like internal application that gets used by the company herself personally. Again, what I do deeply care about is that every single person on our leadership team has a core end to end understanding of like the mechanics of how this stuff comes to life. Like, and what the life cycle of... ⁓ from idea to implementation to deployment out to the company to the maintenance of it if any is necessary over time like looks like because I think that allows for this like diffusion of Responsibility assignment and expectation setting and all kinds of other things to ⁓ single other person on the team who's potentially able to ⁓ know make everybody's lives better or make their jobs easier or whatever else by using this stuff so I do think it is really important that everybody's got at least one of those things. And it is very cool in our business. Amy, our general counsel, has built so much really, really neat stuff in terms of cap table models ⁓ financial forecasting models. And they're doing some stuff right now on the way that prorated contracts work. it can get really complicated with revenue recognition and legal language around that stuff. ⁓ a big engine that can kind of reconcile all that and generate the legal language automatically for contracts. pretty sweet and it's all being done in an AI enabled way. And that's exactly what I want, right? It is this weird thing where our senior most executives are like tinkering around in this stuff, but they're not building slop, at least in all cases, right? There are some things that look like they're gonna really persist, which is super cool. Yeah. ⁓ Lindsey DeFalco: Mm-hmm. Yeah. Yeah. I think that's probably the biggest opportunity with hackathons now that we're in the AI era is the fact that everyone at the company can contribute. Like, I don't think I've ever seen Kelly post a hackathon idea. And so here we have her official title. Bob Moore: a, let's call her head of corporate culture or something like that. Yeah, she's ⁓ among things, my E.A., but also ⁓ orchestrates our internal company events and does a of it, runs our standup meetings in her all hands and. Lindsey DeFalco: Yeah, she runs everything. The like, critical piece of keeping cross me running and happy. Yeah. And, ⁓ you know, not, not technical by nature. and she dropped a, an idea in here about, ⁓ helping her with for events in terms of flights and hotels. yeah, there's a bunch of apps that do that and they're all, you know, varying costs and qualities of course. but that to me is so cool. Cause I'm like, Bob Moore: Yeah, a real glue person. Yeah. ⁓ yeah. Lindsey DeFalco: I want to see more people that are not on the design engineering side of the house in here being like, ⁓ I can actually build stuff and I can build stuff that's valuable. Bob Moore: Yeah, one thing I hope to see a little bit of, and I think this is probably where I'm gonna end up focusing, and it's a little bit of a throwaway project, but I think can open a lot of doors, which is just looking at bleeding edge ⁓ ⁓ applications ⁓ ⁓ core ⁓ models. ⁓ a good example of that is my project. about a year ago was all related to video generation. was like the very first thing that I ever did with AI video was like those Crossbeam commercial, fake commercials. That was about a year ago. And it kind of sent me down this crazy video rabbit hole. But how far all that's come in the last year, it just, suspect that whether it's video or not, or kind of media related or not, there are probably just absolutely wild. ⁓ ideas and tools and technologies that have come to light that none of us even have the attention span to be aware of. And I think I'm going to do a little bit of research and digging and like look at the current batch of Y Combinator companies and just like go through the stuff that only a hundred people have tried out yet ⁓ and pick one or two of them and see if there's something like pretty profound that can come out of it ⁓ to try to understand a little better like what's coming next versus what a redundant idea. Lindsey DeFalco: Nice. Yeah, I actually, I'm curious how you do that because I have done something similar for like, you've got your standard product problems that people are always trying to solve, like feedback collection and analysis and, product documentation and, you know, ticketing and ⁓ all of that. And I always pop my head up and see if there's new tools that have solved stuff ⁓ in a really way. And for some of them I found them and for other ones I haven't. ⁓ but I struggle with finding like the super early AI tools in companies. ⁓ And like, where do you do that, Bob? Are you in like special groups? What's like your path if other people are like wanting to see what's out Bob Moore: Yeah. It is a lot of like surrounding myself with humans who are very experimental and intellectually curious and accepting upfront that 90 % of the time they don't know what they're talking about or they're spending their time on something that won't be that useful to me. And then still caring enough to pay attention to that last 10 % and indexing it against, and by the way, I'm also wrong, 90 % of the time. So that leaves 1 % behind that's actually really interesting stuff. that is like, ⁓ it is like, user groups or email lists that all the CEOs of a given venture capital portfolio are on. Or it's like I'm on a text thread of like founders who have exited companies that live in the East Coast that are just like talking about the craziest shit all the time. ⁓ Or it is a little bit of, you know, there's a lot of people that have worked for companies that I've founded in the past that I've just stayed close with that they themselves have either started new businesses. So like I'm an active angel investor. So doing that gets me in front of a bunch of people that are trying to build out new stuff and then forces me to get at least like mildly smart on a bunch of different areas. ⁓ And be wrong a lot, ⁓ ⁓ of stay, ⁓ in the realm of where people are experimenting. I think that's really it. It's like staying close to the bone of experimentation, I think is really, really the theme. Lindsey DeFalco: Yeah, it's funny because everything, most of things you just said are like, you have access to groups that most people don't have access to ⁓ because you're in the founder community and have these like direct ties to VCs. ⁓ I found like, cause you mentioned you were just gonna go look at the most recent group of white combinators. Like for us folks that don't have your level of celebrity dumb, What do you think the most accessible way to find out like what all the new tools that are coming out of the AI world are? Bob Moore: I actually think like go to Hacker News and go to like Show HN, like filter by Show HN. ⁓ They often do, ⁓ don't know how on like the members of a ⁓ YC batch actually get and disclosed, but like the batches happen so large and so often now that even if it's not the current one, you wanna go look at the most recent one, you can definitely see like all of the companies and find roundups on them and. If you know one person who has ever gone through YC, they have this thing called Book Face, I think it is. ⁓ It's basically like Facebook for all the YC founders. So everyone who's ever been through YC has like a private social network thing they can get onto. And I think you can look at all the companies that have been through and all the people that have been through and all that stuff. like, if you can just like network your way to one YC alum, it might be really interesting to just like sit with them in like a guided. Lindsey DeFalco: on this. Bob Moore: session and kind of go through like what's what's cooking lately and what the discussion are about. I think it, I think it is. It's a lot of this ⁓ kind of to human network based stuff combined with leveraging technology in some way. I'll tell you what I would not do. I would not ask ChatGPT And I would not, because like ⁓ they're, first all, trained on data that's out of date. And then second of all, they're not that great at distinguishing fluff from non fluff when they go and search the internet. And that's probably also part of why it thought these five bags that, ⁓ You you could buy at Target, we're like this best kept secret of the industry. ironically, the last person I would ask about what's going on in AI is AI. Lindsey DeFalco: Yeah, yeah. I ask, and maybe this can be like a future topic, but like finding information in this age, I think we touched on this a little bit a while ago, but like how quickly things change. I've been hearing more and more that people are like, oh, I'm in this WhatsApp group, or I'm in this like Discord. And it literally is just people who are either building it or involved in some way. But it is funny because you can't just like go, you can't like go to Google and search like product feedback AI tool. Bob Moore: ⁓ yeah. know, Google, ⁓ my God. We should spend an episode on like what a wasteland Google is. you can even make it past the sponsored results, which I challenge anyone to do these days, ⁓ like what comes after it and why that's what comes after it when it comes to, yeah, like a bleeding edge research, like information discovery standpoint, it's just, ⁓ I'm a big believer in like the dark internet theory. Like it's all there's. Lindsey DeFalco: And. Mm-hmm. Bob Moore: the incentives to create content for humans on the internet are rapidly approaching zero. And most of ⁓ novel knowledge and information is being created in these like walled garden, closed access networks. And whether that's a social network that millions of people can join or whether that's a WhatsApp thread that you're either on it or you're not, the internet itself as we know it, like what Google can see is... It's like the coral reefs, right? It's like, it's basically where it's a skeletal system of its former thriving self that's just gonna sit there forever and people are gonna go, you know, they're gonna go scuba diving around it and ⁓ that's gonna be the extent of the interaction. Lindsey DeFalco: I had this interaction with Google the other day where I searched for something and then it wasn't what I wanted. And so I like followed up and I realized that I was trying to like talk to it. Bob Moore: ⁓ yeah, I tried to talk to Google. Yeah, To Google's credit, the one thing that is doing kind of well is like the insertion of like the Gemini answers at the top of the search results. Yeah, and then you can kind of go, yeah, you like go back and talk to it and it like kind of works. ⁓ Lindsey DeFalco: And it That's what I tried to talk to. Yeah. Yeah. Bob Moore: But yeah, it's hard for them to jam ads into that. So they must not love it too much. Well, anyway, I do plan on winning the hackathon. I forgot to mention that. If you hadn't heard, that should have been my description up front is that like anybody can win. And I do talk a huge amount of trash every single hackathon about. Lindsey DeFalco: ⁓ Sweet. Bob Moore: think I've won once maybe, but it is good. Somebody's gotta care deeply. Lindsey DeFalco: anymore. actually We have a running joke. We call it sabotage because Bob also likes to hackathon projects. one time he reset ⁓ entire company domain in Google from get Crossbeam to Crossbeam and it broke everyone's logins into every tool that they were using in their hackathon. Bob Moore: ⁓ yeah. Probably about rip, this is like the startup stuff that I love. were like 10 people at the time. We had been get crossbeam.com for a year and I finally, finally negotiated the acquisition of Crossbeam.com as a domain. And I looked around the room and I was like, we can spend a year like painlessly, softly migrating ourselves and living in like this dual confusing world, or I can just break everything and whatever. needs to be fixed is gonna get fixed pretty damn fast. And then anything that doesn't need to be fixed, we didn't need in the first place. So yeah, I just like switched it over and a lot of stuff broke, a lot of stuff broke. ⁓ we had no customers yet at the time. So it was mostly like self-inflicted ⁓ self-experienced pain. But yeah, I think it was pretty much where everybody else's hackathon, which you're welcome. Lindsey DeFalco: Yeah, because everyone's hackathons were ruined. Yeah. I don't even know if he won that one, so was it worth it? Bob Moore: I should have won, but I think out of spite, think I got very few votes out of pure spite in that cycle. There's a bridge too far. Lindsey DeFalco: all right. Bob Moore: ⁓ All right, well we'll catch you next time. We'll report back on ⁓ the hackathon and what we learned, what won, and more to come. Lindsey DeFalco: Bye!