Full Transcript
Noah: I was finishing up a blog. So normally I do a newsletter, but it was a blog on, I call it, "basics and buzzwords." Just kind of, what does each company do? What are some buzzwords? What does GPU mean? All that fun stuff.
Will: Yeah, that's helpful. I feel like sometimes you don't know the bubble you're in, and when it comes to AI, I think that's a serious problem. There's a big divide between the people who are implementing the technology, the people who are using it heavily, and then the people who are just hearing about it on the side. So, trying to level the playing field makes sense.
Noah: I agree. All right. Save that take for the interview because that's a good take. Are you ready to get started?
Noah: Yeah, sure.
Noah: All right, all right. So I'm here with Will Matz, co-founder of Superfan. Will, tell me about yourself, what do you do, and just kind of give me your general outline on the AI space and what you think is going on with it right now.
Will: Yeah, sure. I'm Will, I'm building Superfan. My friend Tommy and I are building it together. It's a social app for music fans. And I would consider myself a builder first. So, a very broad skillset, generally applied across a lot of domains. I'm not an expert in anything, but I think honestly at this point, that's not even necessary. There are a lot of tools around us that help fill in the gaps. And together, Tommy and I are scrappy and just doing everything we can to build a great product.
Noah: Very cool. Will says he is not an expert on anything, but I know you are talented on the coding side, in some capacity, Will.
Will: Even, even there, I wouldn't even consider myself a software engineer in some cases, because there are software engineers who are hardcore, who are just much more skilled and talented than I'll ever be. But I do feel like I've honed the ability to build products, and every time we run up against a tactical challenge, we're usually able to either brute-force our way through it or get really creative and solve it. So, I'll never be the best engineer, but I would consider myself a builder who's good at using some of these tools that we have access to.
Noah: Nice. So, tell me about Superfan. Just kind of give me a pitch on why someone watching this should use it. And then, do you guys use AI in Superfan, in building it, in the app itself? Tell me about that.
Will: Yeah, so first off, Superfan, it really started out as just a passion project and has been ever since. But I've been lucky enough to be full-time on it for about a year and a half now. I started about two and a half years ago, and the whole premise was, "I wonder what it'd be like if people could share their music really easily with friends every week." And so we started this basically weekly Spotify wrap. You'd sign up for this app, you'd connect your music, and then we'd track everything you listened to, wrap it up every week for you and your friends, and just build a social experience around that. And it turns out that there are plenty of people who loved using it early on and have stuck around with us for years now as we've continued to iterate and just try to build a great social experience around music.
In the future, we're really excited to build engaging ways for fans to get access to the artists they love and engage on that side as well. We do use AI. We use it in the product and then building the product. So, in the product, every week you get an AI vibe of the week, which is just this fun little snippet about, uh, sort of like your vibe that week. It's like your music horoscope in a sense. But AI gets really creative, and we try our best to make those fun and engaging, and that's been a real driver of growth for us over the years. People generally find them interesting and share them with friends, so that's useful for us in that way as well. Building the product, we try to stay up-to-date with whatever the latest coding engine or coding interface is. And yeah, we use Cursor heavily, some Claude Code, and are just always trying to build product faster, and it's a helpful tool in that sense.
Noah: Super cool. For the building, I spoke to a developer friend of mine in kind of our first interview here, and he talked about how he appreciated Claude Code a lot more than other options. And you started to answer this question already, but kind of give me a breakdown of, and it doesn't even have to be with Superfan, just personally as well, which AI tools are you using? And then we can step into a bit more in-depth with Cursor versus Claude Code.
Will: Yeah, I mean my stack right now is Claude Code, Cursor, and ChatGPT. Those are the three tools that I'm kind of jumping between, and I find that I'm using them interchangeably throughout the day for different things depending on what I'm working on. I will say I'm very new at Claude Code, so I don't think I'm the best person to speak about that, but my challenge in getting onboarded there and getting started was that we have a pretty large codebase, and it hasn't been doing a good job actually collecting the context it needs to really work effectively there. And I thought Cursor did a lot better job of indexing our codebase and actually navigating it reliably. And so in some cases, I find that using Claude 4 in Cursor has been a better experience than using Claude 4 in Claude Code, if that makes sense.
Noah: Interesting.
Will: Yeah. I actually think there's a chance that that's just a limitation for my setup. I haven't really done a lot of optimization for Claude Code yet, and I think there's a lot of steps I can take in the codebase to help, you know, direct it and give it the context it needs to find what it's looking for. But I did a lot of that setup over the last couple of months for Cursor, and that has paid off tremendously. So, I think I just still have some work to do to set up Claude Code.
Noah: That makes sense. I was using Claude and Cursor a bit before Claude Code started up. But I did find Cursor was super useful in kind of looking at what I already had versus trying to go back and forth between Claude. I guess that's different with Claude and Claude Code integration, but that makes sense.
Will: Yeah, Cursor, I remember onboarding to it months ago at this point, but it was just like a magical experience onboarding to it, and it was, you know, they were the best models at the time, best in class. People have complaints now generally about how their system prompts are changing, basically changing the functionality and making it harder to, you know, execute on some of the code changes and stuff. I haven't found that to be too much of a barrier, but I just remember the early experience of using Cursor was just magical compared to VS Code, which just feels ancient at this point, honestly.
Noah: Yeah, I can definitely see that. So, with the AI space in general, I know you said you kind of primarily use Chat, Claude Code, and Cursor. Have you had a chance to take a look at everything else a bit that's developing? Have you taken a look at Gemini or just kind of some of the other models, XAI Grok, that are out there? And if you have, where would you fit those into kind of, you don't have to rank them, but just kind of give me your idea of where you see the space in general with models outside of just Chat and Claude?
Will: I would say I'm just not well-versed enough on those to talk about them. I've felt over the last couple months just a little bit of information overload. I try to keep up with what's going on, and I'm always interested when someone's telling me about a new model and stuff. But I haven't found a strong enough reason to go and explore those really heavily and actually build them into my workflows. So I don't really know. Like I've heard a lot of great things about Grok recently, some of the latest updates, but I don't find anymore that I'm jumping on new opportunities to go and test out all these new models, because I know that the interface is generally going to be the same. It's generally going to answer my questions. And beyond that, I'm not the person who cares to that level where I care about the specific way that it's interpreting, or understanding different logic, or context windows. I don't care about all these things. I just want it to work, and I'm already set up on some of our flows. Maybe at some point I'll, you know, have convincing reasons to try some of the new ones, but I do feel like it's generally just information overload at this point. So, yeah.
Noah: That makes sense. That actually keeps me up a little bit. Yeah. So, if you do feel there's information overload, definitely you can subscribe to No on AI. Yeah. So, talk to me about your efficiency. Say you were, let's put you at 100% efficiency back in, I don't know, 2020, 2022. If you had to look at your efficiency numbers now with the use of AI, where would you put that? I've had someone say 3.5 times more efficient and 50 times more efficient so far, so you've kind of got a big window to work with.
Will: I don't know about those numbers. I would say probably like, yeah, somewhere between two to three times as efficient we can ship, but it's also just, it's not efficiency, it's quality as well. I think if you're using these tools correctly, you can write much higher quality code for a lower amount of effort. So we might be building software twice as fast, but it's also higher quality. You're not just gauging code based on how fast it's produced. There are a lot of dimensions, and it depends what you care about as an engineer, as a company. But I've been really excited about how it just elevates the quality of our software, which was something that, I mean, I was learning how to code back in maybe like 2018 was when I first wrote a line of code. Back then, like before 2022, you would run up against a technical problem and just be banging your head against the wall for hours to try to solve it. You'd be reading Stack Overflow, and it was the worst experience ever, trying to look at the source code for the libraries you're using and referencing GitHub issues and all this stuff. And I just find that, and it's partially because I've become a better software engineer, but at the same time, these tools help you get past those technical blockers so much faster. So I don't ever find that I'm stuck in this cycle of, you know, not being able to solve a problem anymore, which is, I think the biggest win is just that I never feel that deep frustration. And that's where the efficiency comes from too. It's that you don't ever get stuck writing code anymore.
Noah: What type of workflows are you using within your system when you're building with AI? Are there any specific workflows, or is it essentially just you're working in Cursor or you're working in Claude Code or working with Claude?
Will: Uh, yeah, mostly just that. I'm just working in the interfaces at this point. Claude Code, the thing that I like about it, and actually Cursor just pushed an update that is similar as well, so I'm going to be trying that out too. But you can kind of run multiple queries at a time now. And so as long as you can maintain all this context in your head, you can have a few different work directives going at the same time where it's crawling your codebase, making changes, and then you're basically just monitoring them as they go.
So yeah, it's basically just like, it's also changed the way that I approach software as well because you have to make a lot more detailed specs before you begin a project or before you begin a code change nowadays, just to basically direct Claude or Cursor to go make the right changes. Claude seems to be better at that; they kind of do that naturally. And with Cursor, you have to prompt it to do that a little bit more. But yeah, the workflow is just like, create a spec for whatever change you want to make. Try to make it as detailed and specific as possible. And then you basically just let it rip and you monitor it as it's going. Like, try to catch errors early, because sometimes it'll just completely go in the wrong direction because you realize you underspecified the software, so then you have to go back and adjust the spec and try again. But yeah, that's kind of the workflow. Nothing magic. I know there are some companies that have put in a lot more sophisticated, like, autonomous reviewing and other things. We don't necessarily have that infrastructure yet, partly because of cost limitations, but also just, I don't think there are convincing numbers yet on just like the efficiency gains unless you spend a ton of time really investing in building that infrastructure, and we just don't necessarily have that at our disposal right now.
Noah: Yeah, that makes sense. Have you used Copilot at all? I know you talked about VS Code a little bit, but have you used Copilot at all?
Will: I haven't used Copilot recently, no.
Noah: Okay.
Will: No, but I do think, obviously as a developer, all of your workflows are centered around GitHub, and so it's probably a really natural interface, and I think after some tuning they're probably going to find something that works there. I don't think, I mean they have some embeds in the editors, but I think Cursor and Claude are probably going to make a lot bigger strides there initially. I do think, obviously, Copilot, they have access to all of the code repositories and all of the workflows. Like GitHub Actions is a huge surface area for AI agents, and I think they'll probably be much better embedded into those initially as well.
Noah: Do any personal projects with AI?
Will: I wouldn't say necessarily personal projects because they're all sort of like Superfan-oriented, but when we need proof of concept for new ideas, sometimes I find it's a lot more powerful sometimes just starting from a blank slate. And so if you have an idea and you're like, "we need a mock-up." A recent idea is like we had to create an admin dashboard. Our old one was just super outdated and wasn't serving our purposes anymore. And I was like, "let's dream big. What does our ideal admin dashboard look like?" That actually gives us the capability to, you know, see into our user base and manage the things that we need to on the app. And just giving it some pretty good directive there, starting from scratch, it absolutely knocked it out of the park. Just like creating a really engaging, I mean, engaging admin dashboard is kind of, maybe that's not a thing, but it really, just like the mock-ups that it was able to generate in code were absolutely insane. And it was just like exactly what we needed, and like surfacing some ideas that we hadn't really thought of yet, but were totally possible to build. So we didn't finish that project yet, but yeah, I find that starting from a blank slate sometimes it's surprisingly good.
Noah: Interesting. Cool. Interesting.
Will: Yeah. So, I don't know. I use it in limited ways, and sometimes it'll be just like a good sounding board, but I don't ever rely on it to produce really novel insights at this point. I'm excited for the day that it does though.
Noah: Yeah, for sure.
Noah: Definitely. If you had to, if you were talking to a senior in high school, entering their first year of college, or someone who just graduated college, what would your advice to them be in terms of how they can leverage AI in order to maybe get a job or to just in college specifically when they're going in?
Will: I think I would've said this same thing before the AI boom as well. But go build stuff, just like find problems or opportunities and stuff and just build, you know, build stuff. It's the best time in history to do it. It's easier than ever. It doesn't have to be businesses. I think we kind of have oriented our lives around startups and businesses and stuff, and so that's our default. But I think just like the act of creative expression through software or design or other things is just like, it's easier than ever. And my advice would just be to, yeah, stay on top of tools and try to use 'em to build stuff for fun or for, you know, any other outcome that you're looking for.
Noah: Yeah, that's good advice. All right, so I got one more question for you, and it's a two-parter. First part is, how do you see Superfan growing over the next few years? Is there anything specific you guys are aiming to do or just kind of keep building what you're building?
Will: Yeah, we have some very ambitious plans. I can't necessarily talk about all of them, but directionally we see big opportunities in consumer. So we do want to keep investing in our consumer experience, and I think AI plays a big part in that as well. It's opening the door to build experiences for people that just weren't possible before. And that's like us builders, that is the most exciting thing for us. And so we're honestly just having fun with that and investing heavily in that. But at the same time, we're working in music. We are working there because we have so much passion for it, and our users do too. And you can only go so long building like fan-to-fan experiences in our case. The fans are here because they love music and they love artists. And so it's a natural progression for us to start working with artists, and we see that as the next big phase of the business. And we're not going to do that until we're ready for it. And I don't feel like we're ready in this moment. We still have a lot of work to do to build engaging fan experiences, but yeah, that's the most exciting thing for us at this point. How can we actually impact the careers of artists? That'd be awesome.
Noah: Yeah. Yeah. Okay. Where do, and then where do you see kind of the future of AI in general? You can use a specific use case or just kind of talk in a broad overview in the next five, 10 years.
Will: I was actually talking recently with some other founders about this, and we were debating and agreeing on some things, but in general, we think there is this window right now where it is incredibly easy to build software, and the people that already have some software chops are just going to, they're going to capitalize on this opportunity and just build a ton of really cool software. I think a lot of them are going to make money in this, but 95% of these companies are going to go to zero at some point because the software is just, like AI is just pushing so quickly the boundaries on what it can handle and what it can't. And I think any businesses that, you know, are businesses that don't build defensibility in this period are just going to get crushed at some point. But I think they can make a lot of money. And so I think, you know, we're going to see a lot of small teams trying to, trying to do that and just chasing opportunities. It's hard to say long term what I think will be the most durable businesses, but I don't think it's going to be like B2B SaaS. I think at some point, the cost of building that sort of software is just going to go to zero. And we see this internally as well. We'll get quoted by a vendor, you know, thousands and thousands, tens of thousands of dollars to send push notifications to our users or do these things that are effectively just like we could just spend a few hours or a few days of our time building an 80% solution that works well for us in our workflows. There's no need to pay these SaaS contracts anymore. And so like, yes, I think there's a window to build certain software, but at the end of the day, a lot of this B2B connective tissue software is just not going to have any value long term. So we're excited about consumer experiences. I think there's a lot of opportunity there, and I think some very durable businesses will be built there, and we're just excited to experiment in that sense.
Noah: Yeah, that makes a lot of sense. All right. Before I let you go, I will ask you, since you are building an app in the music space, who are your top three artists?
Will: Top three artists? So recently I saw an artist called Perfume Genius. We saw him up in Cleveland, which was just an incredible show, like one of the most engaging shows I've been to recently. Nice. I don't know, some others I've been vibing with. We love Richie, Mitch, and the Coal Miners. They're actually, like, they've been, they were some of the earliest supporters of Superfan, and they also make killer music. Like they started in his mom's attic or something, and they've just been making great indie folk music for a couple years now, and they've blown up and had some really solid hit songs. Past that, like Ohio Boys Camp, absolutely love Camp. My brother actually lives, yeah, they live pretty close to us here, but for years, we've been huge fans of their music, and I've gone to a bunch of their shows, and they just seem like good people making good music and have to support 'em. So, yeah, I don't know if that's my official lifelong top three, but I absolutely love them.
Noah: Cool. All right. Is there anything that I didn't ask you that you wanted to say so far?
Will: No, I don't think so. I will say, again, I'm not an expert on any of these things, and I can always talk about my own experience, but I know there are so many people getting so much more value out of these tools than we are at this point. And, you know, we're just going to keep having fun, like using what we can and building where we can.
Noah: Excellent. All right, well, I appreciate your time.
Will: You got it.