Software Huddle - Finding Product Market Fit with Cassidy Williams of Contenda

Episode Date: January 9, 2024

Today, we have Cassidy Williams, CTO of Contenda. Contenda unbelievably started as a sticker distribution platform that pivoted into a product that converts podcasts and videos into various other forms of written content via AI. But in our conversation with Cassidy today, we talk about their latest pivot to a product called Brainstory, which is an AI based brainstorming application. We talked through some of their product choices around focusing on speech as the main input mechanism, some of the technical challenges they've had to overcome, how they're using multiple AI models in the backend to make all this magic happen, and where they're seeing initial product traction. If you're a founder or thinking of starting a company, we think you'll find this conversation super interesting. Check Out Brainstory: https://www.brainstory.ai/ Software Huddle: https://twitter.com/SoftwareHuddle Cassidy: https://twitter.com/cassidoo Sean: https://twitter.com/seanfalconer

We didn't really have all of the media that we have today on YouTube and stuff before. And so it was truly just experimenting. I remember even getting my first textbook on Java in high school. And it took me forever to figure out that I needed to install a compiler in order to be able to do anything that was in the book. As someone who's been in the space for a long time, where do you kind of sit in this world? Like there's some people who are very much like, hey, you need to suffer through, you know, memory leaks and C++ to like really get an appreciation for what's happening at the lowest level of the operating system. And then other people are on the other side. It's like, hey, let's give them the most accessible language to start with. I lean towards more the latter purely because I think it's very

motivating when you get something on the page and it works. Brainstory works for so many different use cases, but I feel like to do well, you kind of have to pick one use case and really lean into it. Yeah, picture battles. Hey, everyone. Welcome to the show. I'm Sean Faulkner. And on Software Huddle today, I have Cassidy Williams, CTO of Contenda. Contenda unbelievably started as a sticker distribution platform that pivoted into a product that converts podcasts and videos into various other forms of written content via AI. I've actually used this version of Contenda, pretty cool stuff. But in my conversation with Cassidy today, we talk about their latest pivot

to a product called Brainstory, which is an AI-based brainstorming application. We talked through some of their product choices around focusing on speech as the main input mechanism, some of the technical challenges they've had to overcome, how they're using multiple AI models in the backend to make all this magic happen, and where they're seeing initial product traction. If you're a founder or thinking of starting a company, I think you'll find this conversation super interesting. And if you do, please remember to subscribe, follow us on Twitter. And hey, if you're feeling particularly generous, leave us a positive rating review. All right, let's kick it over to my interview with Cassidy. Cassidy, welcome to Software Huddle. Thank you so much for having me.

Yeah, thanks for being here. I know it took a little bit of back and forth. You've been under the weather for different periods. I've been traveling and so forth, but you hosted me on Stack Overflow podcast a while back. So I thought it was only fitting that I returned the favor. Yeah, I'm excited to chat with you again. Yeah, absolutely. So, you know, I think you're pretty well known in some circles of the internet. You know, you have a very popular newsletter. You were previously director of developer experience at Netlify, and now you're the CTO of Katenda. Can you give a little bit of background about your journey in tech? Where did you start? How did you move into developer relations and then eventually to your

role today at Katenda? Yeah. Yeah. So I first started messing around with code when I was around 13. I knew about computers and stuff. But then in eighth grade, I was walking home from school, and I heard a neighbor say, check out my website. And I was just like, wait, you can have one of those because I thought only like, businesses could have websites. And when I realized that anybody could that kind of just opened my brain to this is awesome. And so I just started looking up how to make websites and stuff. And I've been in it ever since. And so eventually, I, after dabbling a lot, and just really being self taught, just trying to figure out how to build things, I ended up majoring in computer science at Iowa State

University. And then in my senior year, I junior and senior year, I started going to a lot of hackathons and stuff and interacting with a lot of companies. And what I really appreciated about that experience is I met a lot of people where they were representing their companies. And I was like, wait, this is your job. And I appreciated that they knew how to code and stuff, but they were also interacting with the community and teaching us stuff. And I really liked the idea of that. And so that's kind of how I got into the developer relations stuff from the beginning of my career, because I knew that I wanted to interact with developers. And I'm really passionate about just like teaching in general. And so in my first role, and my second role, actually, they were both combo roles where I

was the first dev advocate for the company. Plus, I was engineering for the company at the same time because dev advocacy wasn't as much of a thing back then. And so my first job was at Venmo. And then my next one was at a company called Clarify. And so I was representing the company at hackathons and conferences and meetups and stuff and writing a bunch. But then at the same time, I was working on the actual products themselves. That led to a lot of burnout doing two jobs. But, you know, that's what I did.

And then over time, as I moved from New York to Seattle and then now to Chicago, I've done a bunch of different roles at both startups and large companies. And there was a point about, I think it was around the pandemic or like end of 2019. I started interacting with people who were startup advisors. And I thought that was really cool. I liked the idea of being able to help companies without having to be full time at the companies where I could, I could just talk with them,

give them advice and see where they take it. And that's how I got involved at Contenda where I'm at right now, where I have a Patreon group where it's mostly where I review resumes and we kind of just play video games, talk about memes and stuff like that. And I always mail stickers to people and I was doing that all by myself. And then Lily, my CEO, she said, if I built something that would help you mail stickers, would you use it? And I was like, yeah, that'd be great. And thus Contenda was born. It started as

a sticker mailing company. And so they were mailing stickers for creators on the internet, like Twitch creators and stuff. And they ended up actually breaking a world record for Ludwig, who's like the most subscribed to Twitch creator for having the most subscriptions because he did this whole sticker campaign with Contenda, which is a very cool start, but stickers are not very sustainable. So they ended up saying, okay, we're software engineers. What if we built like digital tools for helping subscribers be retained for different creators. And many, many pivots later,

we started doing a bunch of generative AI stuff, which was very cool until the GPTs of the world started getting particularly popular. And we're just like, Oh, now everyone's doing this and we can't really differentiate in this market. And then we ended up building some different tools on our API and kind of doing a little bit of consulting for different companies to be like, what can we do

with what we know in terms of creators, in terms of how people think, in terms of just turning media into text and going from there. And that's kind of how we came up with BrainStory, which is our current flagship product, where what you do is you talk to the browser, basically, and there's like a digital coach and you say, okay, I have an idea about a podcast or a blog post, a conference talk, a side project, what I want to eat tomorrow, and it'll talk you through your idea. And then at the end, you get kind of a summary of what that idea is. And it's been going really well. That's amazing. It's quite the journey. A lot to unpack there. Actually, going back to the very beginning when you were creating those websites, when you were 13, what platform were

you using for creating websites at that point? Was this a GeoCities era internet or were you beyond that i guess dr j you or anything yeah that and like freewebs.com which is now webs.com and then neopets of course um a lot of it was just like using notepad on my computer and and like i would view the source of websites and be just like okay OK, that's what this HTML tag does. And then I would like write it down and kind of just experiment with it. There wasn't there weren't a lot of tutorials out there back then. And there was one I think it was called like Lisa Explains It All, where she explained certain HTML tags.

And that's how I learned a bunch. And Neopets taught me a bunch. But we didn't really have all of the media that we have today on like YouTube and stuff before. And so it was truly just experimenting. Yeah. That's very much how I started as well. It was, I think the web, especially in the early days, it was when you didn't have all the stuff available to you to like, really like understand, like how do I actually build stuff is, was great because you can actually go view the source of a website and get an idea of like oh okay that's like how they made uh you know this uh globe spin around or something you know something like that yeah yeah or the marquee tag yeah or even like um being able to see like the javascript and stuff

of course now with uh with React and one page service or client-side scripting, it's kind of ruined all that because you can't really view the source and gain anything. And everything's minified and stuff. It's a bit harder to do that now. Yeah, but of course,

now we have YouTube and a million tutorials that actually learn these types of things. I remember even getting my first textbook on Java in high school. And it took me forever to figure out that I needed to install a compiler in order to be able to do anything that was in the book. Yeah, man, I'm kind of like, I'm jealous of people learning today. But at the same time, it was good for me to probably like learn everything from the ground up. When I was first learning Java, I was also in high school,

and I was writing it in Notepad++, compiling it myself, and then running it. It was so, so manual. I remember when I went to college and I learned Eclipse, and I was just like, wait, it has autocomplete? And just little things like that. I've been just in the stone ages, handwriting everything. So yeah, it was probably good for me to learn that way. Yeah. At the same time, like, as you mentioned there, there's a certain amount of value to kind of like struggling through those things. It does kind of like become a forcing function for learning. Now it probably on the,

on the negative side is probably a deterrent for some people to get started. I guess like, where do you, as someone who's been in the space for a long time, you know, really has been involved, like, in educating developers and so forth, where do you kind of sit in this world? Like, there's some people who are very much like, hey, you need to suffer through, you know, memory leaks and C++ to, like, really get an appreciation

for what's happening at the lowest level of the operating system. And that's going to lead to you being a better engineer and so forth. And then other people are on the other side. It's like, hey, let's give them the most accessible language to start with, remove the compiler, use something like Python or whatever so that they can get up and creating something. And then that'll lead them to naturally asking the right questions and have them deep dive into the details as they progress in their journey.

Yeah, I lean towards more the latter purely because I think it's very motivating when you get something on the page and it works. And you're just like, wait, okay, why? And I kind of go backwards from that where once you figure out how something works, you you made a component in React or something like that, then once you say, okay, I can actually make something that works on the screen, then you should learn why does it work that way. And so, for example, in a lot of the workshops I used to teach with React, I would say like,

okay, this is how useState works. You run it, you see how it works. You have a counter on the page or something like that. Then I would go into let's rewrite use state from scratch. So you can understand exactly why hooks have to be at the top of a component, for example, or this is how it's stored. Granted, it was like a very basic version of it. But I think understanding how things work is really, really powerful for your understanding of a system in general. But at that same time, I would never go back to C++ if I can avoid it.

That was painful. So you mentioned this recent pivot to launching a product called BrainStory. So can you talk a little bit about, I mean, you've already kind of mentioned some of the origin story for Canada, but maybe you can give a little bit more sort of context and background on why the switch to BrainStory and then also what is sort of the vision of that product? Yeah. Yeah. So pivoting is kind of scary, but exciting at the same time where I think, again, it's a weird economy right now. And then also, we just were kind of struggling with figuring out how do we get to that next step of getting people to keep using the product, getting people to actually pay for the product

and stuff like that was our previous version, which was Contenda Studio. Contenda Studio, it still exists, but only as a paid version now, because we're just like, we're going all in on BrainStory, where you could put in a podcast or a video or something, and it could then be turned into a blog post or a LinkedIn post, Twitter thread, various things like that. And it worked really well. And I actually still use it for my own things on occasion. But again, there's so many generative AI things that were just like, oh, we'll write a blog post for you. We'll do this for you. We'll do this for you. That first of all, on one hand, it felt a little bit icky that we were doing writing for people where I think a lot of the summaries and certain posts were really, really valuable. But we didn't want

to take that away from anybody. And then at the same time, once again, the economy is weird. We were doing so many sales calls and a lot of people would be just like, wow, this looks awesome. We just did layoffs, so we don't have a budget. So we'll talk next year. And that can only happen so much before we have to be just like, okay, we're not making enough money to continue in this direction. What should we do? And so we had a team offsite this past August, and we were kind of all just chatting about like, what should we do next with studio? How should we make it better? Should we make it better? And we had a developer API that people weren't really using a lot. We had some people using it.

And I had suggested, what if we just build stuff on our own API? Because on one hand, it could show people, hey, look at what you could do with our API. And then they might buy a key of some kind. Or we can end up building something that's pretty cool. And so the first iteration of Brainstory was actually an app that coached high school students to write their college essays, where it was, we just thought it'd be interesting. What if you just talked to this digital coach that were, as you said something, we put in like the common app college questions and it would say like, tell me about your family. And as you talked about it, it would say, oh, okay.

So it seems like you're really close with this person. And like, it would just continue to prompt you. And then at the end, it would give you advice for writing your college essays. And we particularly focused on not writing the essays for students because we wanted them to do it themselves. And the results of that were so good. We're like, what if we made this for everyone,

not just students? And thus Brainstory was born. And when we released it, we had so many people signing up and using it and being like, whoa, this helped me with my D&D campaign or with my comedy set or with a blog post I'm writing or for an argument I'm preparing. Just so many different things that we were just like,

should we just go all in on this? And I think we just kind of had a team discussion and kind of reflected and we were like, well, I think this is an obvious choice because it's actually going somewhere and studio is still kind of struggling to get these sales. Yeah. So I have a tremendous amount of empathy for decisions around like a pivot. I went through a lot of this with my own company years ago. And I think one of the smart things that it sounds like you're doing

is really kind of like going full force on the decision to focus on brainstorming. One of the mistakes that I made as an entrepreneur was kind of realizing that the old business model wasn't really working, but trying to keep the lights on with that and continue to serve it while also trying to explore. And you just don't have the resources to be able to do that. And you spread yourself super thin and you just do a crappy job across everything that you're doing rather than doing hopefully a good job with the one thing that you're focused on. Yeah. It's kind of like that sunk cost fallacy where like we were,

we were definitely like, Oh, we put so much work into studio and we're just letting it die. But at the same time, we're like, if we just go all in on this one thing, maybe we could do really well with it. And again, we we've been getting enough customers. So we're just like, okay, we should, we should just do this and let it be. Yeah, I think one of the things to be successful in a startup, you have to be kind of comfortable, regardless of what your role is, with the idea that something you put a lot of effort into, you might just throw away and then go to the next thing.

And it's all about, especially in the early days, getting the product market fit. And that's going to take a lot of iteration and cycles. So you might be chucking stuff constantly and you kind of have to be comfortable with that. Maybe creating work that maybe never sees the light of day or gets killed in some capacity. Exactly. Yeah. And I think we've, as a team, have gotten really, really comfy with that. Because again, it's been through quite a few different iterations and stuff. And I think it's, it ends

up being a strength to be willing to be just like, okay, this was cool, but clearly this other thing is working. Let's just, let's just switch. And I think it's been a very good learning experience for everyone too, where it's just like, okay, well, we learned with that one. We don't want to build it this way. Let's build it this way. And it's, it's gotten us to build a lot faster where we actually built all of braininstory in less than two weeks. And like now, of course, we're refining, we're talking to more customers, we're figuring out how can we make it better. But the core product we built really fast because we had all of that learning

experience. Yeah. And also you need to get to a place where you can actually like try it out with people because you're never going to really know whether it's valuable and worth putting the full engineering resources into making it super scalable and so forth until you actually get it in front of people. Yeah, you need to just ship it. Yeah, exactly. Like the Reid Hoffman quote about no product survives its first encounter with real users or the Mike Tyson quote of everyone has a plan until they get punched in the face. He got to get punched in the face to learn how to dodge that the next time. You said that when you launch the product, it generated a ton of interest and you're seeing way more momentum for that versus the former

studio product. But how did you actually build that interest? How did people find out about that you're launching Brainstory and actually sign up for it and start engaging with it? Yeah. So when we were just like, okay, we think it's ready for humans to see it, it was actually just on a long weekend. I actually forget which day it was, but we had that Monday off. And that Friday we were saying, okay, I think we're done. Let's go. And we had just tweeted and stuff, and it was hardly anything.

And so I kind of went rogue and I was like, I'm launching this on Product Hunt on Monday. And we kind of just all started shouting that Monday, even though we weren't necessarily working. And it went pretty well. I think we got like number seven on Product Hunt, even though we launched like Monday afternoon when none of us were really online. And we just started shouting in like Discord groups and again, posting on like the Twitters and the LinkedIns of the world. And then we're also just trying to blog more and just post on LinkedIn more. LinkedIn has been a really good source for us. And then we're also just trying to blog more and just post on LinkedIn more. LinkedIn has been a really good source for us. And I think just with all of the social media

platforms being different nowadays, as people figure out where they should be, LinkedIn has been a pretty steady ship. And so we've, we kind of just posted a lot and it worked well enough that we, we were really happy with it. And so yeah, Product Hunt and probably the LinkedIn postings were probably the most effective. Yeah. It's kind of funny how sometimes you can read about the best practices in terms of which day and what time to make certain announcements and stuff like that. But sometimes it's just like going rogue and just doing it when you feel right. Like just doing it.

Yeah. Like I remember I posted something on like a Saturday night and it is, you know, not the ideal time to post it, but it's like one of the most like viral things that I've ever posted. And I don't know if it had to do with Saturday or maybe just the content, but it was like completely unexpected that it just happened to like take off that way.

So I think, you know, you just have to kind of go with the momentum sometimes. LinkedIn, I agree, is also a very good platform for my company, Skyflow. We see a lot more value there than we have in other types of platforms. I think there's just a lot of noise in different places. Right. I think a lot of places are getting their footing still. And once again, it's a steady ship. And kind of like what you were just saying,

I remember I took a writing class, it was a while ago, and I put a ton of effort into one piece. And then the next piece, I kind of just threw it together because I was like, oh, I think this would be a good idea. I'm just going to write it. And I didn't have time to put in as much effort. And I consistently got so much better feedback on the second one more than I did the first one. And people were saying, we could tell that you cared a little bit more because you just wrote it and you didn't overthink any of it.

You just put it down. And I feel like that was what really worked for a lot of our brainstorming shilling, where we're kind of just like, you know, we actually care about this. We actually think it works. So let's just say what we're feeling and not have a fancy strategic campaign around anything yet. I want to get into some of the, you know, the technology that's involved with the development of brainstorming, kindory and what's going on behind the scenes. But maybe before we get there, just to give further context for the product, can you walk through what is a use case as a consumer or essentially someone who's using BrainStory? What use case or problem are they solving with it? Yeah. Yeah, it's, I think the best use case we've seen, or just like, oh, this could be like a really sticky thing is using it for meeting prep, where a lot of times, when you go into a meeting, there's a lot of the meeting is context sharing, figuring out like, where everybody's heads at, and then doing action items based on that. And something that when I was at Amazon, and we have some other people who have experience with this, we would often at Amazon, we would write what are called

six pagers where you put in all the details. At the beginning of the meeting, everybody would read this gigantic document, and then they would start chatting. This can kind of replace that where you can start an idea. And something that we've started doing to dog food it is we have like a weekly team memo where our CEO, she'll go to Brainstorm and say, I want to talk about the weekly team memo. And again, there's the coach that kind of prompts you based on that. And so she'll say, okay, I think the team is doing great on this. This is what we need to improve.

This is what we should focus on this week. And then when you get feedback or when you share it for feedback, people can respond with brain stories saying, okay, I want to go through this point by point or overall, I agree with you or something like that. And then it'll aggregate all the feedback. Everybody has the shared context. So when we go into the meeting, after going through this brain story thing, our meetings have gone from like, sometimes hour and a half situations for our weekly all-hands sort of things to less than a half hour because we kind of just already know what to expect. So it's been really, really helpful for that kind of async collaboration in particular. Yeah, and probably even more relevant today with all the hybrid team setups that people have.

Right, exactly. Yeah, I mean, this is something that's super interesting to me. My prior life at Google, I felt like most of my meetings in the last two years there consisted of a meeting where we were figuring out what the next meeting we were going to have. Yeah, gosh, I've been there. It's exhausting. It's very frustrating. It's like, I just spent 45 hours in meetings this week, and I've like accomplished absolutely nothing. And I feel completely mentally drained. It's like not a good feeling

to essentially have. So in terms of the like in the meeting prep use case, is everybody simultaneously or maybe asynchronously collaborating within the same like a brain story instance. Like how does that work? Yeah. Yeah, exactly. And so, so usually you can create one and, and I often use it as like a personal productivity tool where again, I plan various things for myself, but for a meeting prep sort of thing, Lily might say, Lily is our CEO, but Lily might say, okay, this is, this is the team memo. It's, to you. Go for it. And then whenever it's convenient for people, some people do it like first thing Monday morning. Some people where it'll say, okay, do you want to talk about Lily's ideas overall or hit it point by point? What do you want to go through?

And as you talk, it'll say, okay, you really hit on these points. You forgot this one. Do you want to touch on it or are you done? And it's really good at reacting to you. And you can even say, hey, you should know that you just spelled this wrong. This is actually how you spell this. And it learns based on how you converse with it and yeah it it's just kind of one brain story instance of like a node and then all the feedback is attached to that node and then it's all aggregated into one thing okay it's kind of like you have a um like a it's like a juniper

notebook version of like your your meeting uh prep and everything that's going to go happen in the meeting so you made this choice i guess with having uh it speech be sort of the main modality of input versus typing so why would do you think that was the right decision to make? We've gone back and forth on this so much, but honestly, the speech thing lets people speak so much more off the cuff where we did have text in there. We're probably going to add it back in there in case someone's like in a loud room and they can't talk. But so many of the things that I came up with, I realized, oh, as I'm talking to Brainstory and I'm thinking about X, Y, and Z, I didn't think you. I feel like when you are verbally talking about these things rather than writing it down,

there's a little less thinking involved when you're speaking. And I think that that leads to better ideas because you are talking it out, you're thinking about all of it, and it's just responding to all of your raw unfiltered thoughts and you can't say oh scratch that i actually don't want to go in this direction and it'll react really well to that um and i think when you write things you never write as fast as you're thinking you're you're you're kind of carefully putting down what you want to say you might fix a typo you're in there you might edit edit it before she'd begin off. But by just speaking and treating it as if you're talking to the rubber duck on your desk or to a person,

it lets you kind of just have a bit more free-flowing conversation and idea generation. Yeah, I can see that. I think people too, sometimes when it comes to writing, they can get a little bit hung up on how do I like perfectly word this thing versus when you're just having a conversation with people, like I might not be really like caring because I can adjust it on the fly if I see them make a face or something like that. And I also think, you know, to your rubber duck analogy, I've certainly experienced that with, you know, you pull someone over to help you debug something. And then as you're explaining what's going on, you're like, oh, I know what's wrong. It's just like, wait, I got it. Okay. Yeah, you can go away now, essentially. So but is there a unique technical

challenges with having sort of audio as the speech as the main modality of input? Yes, but luckily, that's where we were able to use some of our learnings with our Contenda Studio stuff. Because, for example, when we were generating LinkedIn posts based on podcasts, a lot of podcasts are people looking at their own belly buttons going, so here's what I want to do. And cutting out a lot of that noise and everything is something that we got really good at already. And so when someone is talking to the brand story and being just like, okay, so actually,

yes. Okay. So I want to talk about this, but I mean, and like they, they fumble over their words, we can kind of cut through the noise and just keep it organized as if you're talking to a person.

And I think again, that's, that's just something that we got particularly good at with previous iterations of the company. Can you walk me through sort of the data pipeline that's happening behind the scenes in terms of, you know, I go, I'm interacting with Brainstory.

I speak out what I want to accomplish in my meeting. And then it's going to prompt me for another question of like, what are the main topics or something like that? And it's going to prompt me for another question of like, you know, what are the main topics or something like that. And it's going to walk me through that process. Can you, what's going on actually behind the scenes to go from that to actually putting together an agenda that then other people can collaborate on. Yeah.

Yeah. So a lot of the actual like interactions are client side, but then it's hitting our API. And so when you, to really get to the nitty gritty, you log in. We're using Fusion Auth for our authentication. You click new idea and that opens up. It's kind of web sockety, but not all of it. It's kind of streaming, but not all of it.

And it asks for your microphone permission. And in the browser, you start talking to the app. And we chunk the audio a little bit so that we can very quickly transcribe as much of it as we can. And as we transcribe it, then once you finish speaking, we take that transcription, we cut through all of the noise, and then we send it to the AI. And the AI is a combination of a lot of different models, stuff that we worked on, some GPT, some clone, various things. And we use that to get a very condensed response to what you just said.

And then we push that to the browser and the browser's listening for it and it gets it back. And it basically does that repeatedly until we have a full transcript of the conversation. And you can see your transcript happening as you go. And when you say leave conversation, which you can stop at any time after you answer one or two questions, it's just like, okay, if you want to stop, you can. Otherwise, I'm just going to keep talking to you until you want to be finished. When you hit leave conversation, that's when it starts streaming in the summary of everything that you've said. And so it tries to condense it and break it down into like a headline point and then all the sub points within that headline as condensed as possible without losing details. And then from there, that is what is given as the prompt for anybody who might want to give you feedback.

What's the limits on the audio recording? Can people kind of drone on forever or is there a hard set limit? They can drone on forever. We have like a manual browser thing where it's just like, it's been four minutes, please stop. And so we usually, I'm one that can drone on for a really long time. And usually to answer a question, I can be cut off around the two minute mark, or I can cut myself off around the two minute mark and it doesn't break a sweat. I've never gotten the little pop up saying like,

please stop yet. So take that as you will. We usually, it's around the four or five minute mark where it's just like, you should move on to the next question because you're just going to get caught in your own talking loop. And then in terms of essentially taking the audio, transcribing it, and then sending it to the AI to summarize. How do you handle essentially the limits in terms of how much you can pass like token wise to the AI models? Do you have to figure out like a subset of that text to send?

Or are you doing it like in chunks? Like what's that look like? It's a little bit of chunks and then it's a little bit of our own decision tree behind or under the hood of which models to use if there's just a ton because we're because we're using a combination of a bunch of different things depending on what's going on there's a few different avenues where the text could go where we can condense it

pass it to something or condense that and pass it to something. And that just happens on the back end. And luckily, you don't have to worry about it seeming different if you talk much longer, much shorter about anything in particular. Where you're using multiple models, how do you handle sort of the session dynamics across those models so that the context of, you know, essentially each conversation is kept consistently across the model interactions. Yeah, that's where we have like our entire transcript of the whole conversation coming into play where you can, it's really interesting how you can navigate it via voice where you could say like, okay, I hit out all of these three

points, but I want to go back to the second point and it'll kind of remember, oh, the second point you made was X, Y, or Z. And it's kind of using that entire conversation part as the system of this is the conversation, like the assistant and the user and stuff. The assistant has this role, the user has this role, and it takes the context of all of it as it grows. And then in terms of the decisions that you're making on the engineering side for basically build versus buy, how do you think about, okay, this makes sense to leverage an existing tool, we're going to pay for it. Or maybe it's an open source project versus like build something from

scratch. We're doing it as much as we can, as cheap as we can, because startup money where we are paying for certain things. Like we don't want to build our own auth, which is why we're using a tool like fusion. Yeah, absolutely.

We smart move. Yeah. We we've certain security things. We are making sure that we have some tools that handle a lot of it for us. We use AWS for certain storage and certain streaming things. But if there is an open source tool or if there is something that we can use for free

and mess with, we will use it. We have some really interesting FFmpeg stuff under the hood for transforming audio and stuff and getting details from that. And a lot of it is done with open source tooling that we've built. And luckily our team, again, has had a lot of experience through the various iterations of the product where they can say, oh, we've built this before. Let's just rewrite this part to handle this. Okay. And then as you work to improve the AI models,

the product, make it more useful, make more magic happen for the user, how do you know that you're actually making things better? I think one of the big challenges when it comes to building AI applications is it's hard to test if I'm running something through a unit test because the actual generated response is going to have a certain level of variance in it.

So how do you know that the response today is better than the response that you had two weeks ago? Yeah, we're doing a lot of that right now. One of our teammates, he's an NLP specialist, and he does the most magical things with AI I've ever seen. I could go off on some of the stuff he's done outside of work for days. But anyway, a lot of what he does is like improving the feedback aggregation or improving

how the feedback is received or getting like sentiment about the feedback and figuring out like one of the things he's diving into is if you don't want to read everybody's feedback on your idea could you give like a general sentiment like emoji where it's like thumbs up or I'm not so sure or something like that so that you can get a general idea if you don't want to read too much that sort of thing and so it's a lot of him generating a lot showing us and then we say this is good this is not good but also just a lot of him generating a lot, showing us, and then we say, this is good, this is not good, but also just a lot of talking to users. And so we have a Discord community where we have some pretty active users who are very passionate about their thoughts of things, which that's always a blessing and a curse when you have users saying what they want. You have to kind of balance which ones you listen to or which ideas you listen to.

But anyway, we've been able to kind of lean on our community and build in public a little bit and say like, okay, this is what we think we might change. What are your thoughts? We're kind of going with our gut, but we want to see if it's interesting to you. And luckily, that's been going pretty well as well. So in comparison to trying to do something like this, just natively using like ChatGPT, besides the collaboration component,

what are some of the other things that are kind of differentiate the product from just using the chat interface of ChatGPT directly? Like what sort of secret sauce do you have that makes this so much of a better experience for people? Yeah. And a lot of our customers have a lot of thoughts on that because they went to using us from using GPT for similar brainstorming sorts of things. And I think the biggest secret sauce of all of this is that we don't give you the answers. A lot of times when you use something

like GPT, it'll be just like, oh, you should just use this. Bye. It's very solution oriented because that's just kind of how it's designed. And we specifically don't give you the answers where if you say, I want to learn more about X, Y, and Z, and I'm brainstorming a product around X, Y, and Z, something like GPT will explain what X, Y, and Z is. And then say, and this is the product you should build or something like that. And you could, you could try to massage it and not do that, but that it ultimately leads toward, leans towards giving you the answers. We very specifically keep prompting you to think better about your idea. And that's like how we refine it and how,

how you converse with it, where it will very purposely not seek outside resources or hallucinate like certain websites you should look at or anything. It's very specifically focusing on what you're saying and how you can think about it a little bit better. And I think that that's been a really good secret sauce because, for example, one of our customers is a stand-up comedian. And she was saying, I don't actually think that AI can help me with this, but I'll try it and we'll see. And less than 10 minutes later, she messaged me and she was just like,

I don't know how you did this. I thought AI was supposed to be bad at humor. This is incredible. And as we were talking through what she talked about, I was just like, you realize it didn't actually tell you what your joke should be. It was just you talking to it and you were coming up with the joke as it prompted you. And so it's very much like we have like a little slogan that we're kind of workshopping where it's like, think smarter, not harder, where we're just trying to help people get to certain thoughts that are in their head somewhere. They just need to kind of draw it out and refine it.

Yeah. It's really interesting. One of the primary use cases that I use chat GPT and BARD for today is brainstorming, you know, whether it's, you know, thinking of a, you know, talk topic or title or, you know, podcast thing or something like that. And I agree like a lot of it does take a certain amount of massaging to get it to do the things that you want. Even then I'm kind of like throw my hands up in the air and I'm like, okay, I guess I'm just on my own to do this.

And it kind of goes back to some of the things that we're talking about earlier in terms of, you know, debugging and sort of you realizing the problem just through explanation. And if you can kind of productize that experience, it sounds like it could be really valuable for people. Yeah, yeah. That's really what we're focusing on, which also makes it significantly harder to sell. Like figuring out how to explain, no, no, just try it. You'll see how it's different. It's definitely challenging. And that's kind of what we're figuring out right now. How do we get people to try it and see what they think and everything. But we've had such different use cases. And it's been really interesting where it's people where,

for example, one person, she's studying to be a pastor, actually. And she was saying, normally, as I'm preparing a sermon, I talk to my husband saying, OK, what do you think of this point? What do you think of this point? And it's less him saying, you should talk about this. It's just him saying, have you considered this angle or something? And when we showed her brain story, she was like, wait, this totally makes that entire process so much faster. Because he usually

has to think for a while before he comes up with these questions to ask me, this just does it. I could save him time and me. And it's, that's such a different use case than all of like the tech people that we normally talk to. And so it's, it's different because yeah, once again, it's, it's, we do the massaging for you rather than you having to do a bit of prompting with like the bards and the gpt's of the world right yeah i mean it's kind of like and very much uh somewhat of like a new category of like brainstorming probably there's the mind map products but that they're they're just kind of helping you like organize thoughts they're not sort of directing or steering you down a certain path but then there's sort of the

on the other side of the spectrum there's the chat gpt that's going to tell you what to do but that's not necessarily what you want either and this is sort of you know bridging the gap in some fashion between those two worlds or like an ai assistant that's really focused on helping you you know think better as you mentioned yeah we were saying it's as we tried to figure out like what products are out there we we said it's almost like a combination of like little bit bureau and that there's like the async piece but then it's also like the planning piece and so figuring out how to market that is once again it's been an interesting challenge but it's also really fun because

it's something that we all genuinely use for a lot of things. Is there sort of unique tech or like what are some of the like edge cases that you've had to solve for when building up this product? Like one of the things you mentioned earlier was that you learned a lot from the prior product in terms of like how to cut out the noise around like speech. Has that been one of the hard technical problems that you had to solve for? Or is there other things that you've also kind of had to figure out along the way? Yeah, luckily a lot of the AI stuff

is something that we have experience with. So it hasn't been as bad. We do have to do a lot of refining around hallucinations where we don't want it to ever say like, I agree with you when you actually said, okay, I see where you're coming from, but I actually disagree and kind of respond that way. And so that's been interesting because AI is very non-deterministic. And so sometimes it'll just make things up. And luckily, we have experience with that. So when we see it, we know how to address that sort of thing, or how to generate multiple responses, then grade which

one is better, and then give that response, that sort of thing. I think a big challenge for us is UX in general, and figuring out like, how do we just get someone to making their first idea? Or how do we get someone to share their idea first? We don't have a designer on board. Because again, startup money, maybe someday. And so a lot of what we do is we experiment, we try something, we reach out to our community saying, what do you think about this? And luckily, there's some designers in it where they're just like, you know, this screen should actually do this or this or the other. And I feel like that's honestly been our biggest challenge because we're a lot of engineers where we are okay with just a button

that does the thing right. And as long as we get the output, that's what matters. But that's not what gets users. Okay. And then this is your first time as a CTO. What have been some of the unexpected challenges with moving into the role? Dealing with the politics of investors.

That is probably the biggest challenge for me, where a lot of like the engineering management stuff, I've done that before. A lot of like the dev rally type of things or where I'm building tutorials or writing or something. I've done that before. And even just or something. I've done that before. And even just the coding, I've done that before. But I'm not very good at doing the delicate dance of speaking with venture capitalists who are just like, where's my money?

How are you making my money grow? Which has been a particular challenge. We have good investors. They are supportive in various ways. But at the same time, it's a very weird economy. And learning how to navigate how to speak with them has been a really interesting challenge because you can't just be cut and dry necessarily, because they're not always technical, and they don't always speak the same language as you do.

And so learning how to navigate that, and then also how to navigate conversations where investors are just, like investors you don't have are just like saying, so who invested in you again? Oh, okay. Where they're like, a lot of times they just get intel, and you have to take those calls because it's good for you for the future and stuff like that. It's been a learning experience. I admit that's not my favorite thing, but I've learned a ton just in figuring out how to navigate those kinds of conversations and do the dance.

Yeah, absolutely. It was a huge learning curve for me to going through my own founding experience as a CTO and raising money. It was just something completely like foreign and understanding kind of, you know, I guess like putting yourself in their headspace, like what is it that this person wants to hear from you in terms of how you talk about your business and so forth is hard. And it's also can be difficult, I think, just psychologically dealing with a lot of the negative feedback that you can get through that process. There's great investors out there, but there's some pretty harsh feedback investors as well

that develop a tough skin though. Yeah, where it's just like, it's not good until it's good. Otherwise, it's just like it's it's not good until it's good otherwise it's just bad and and and yeah we we can say these are all these are all the paying customers we've had this is all this this is all that and they're just like okay but how are you going to 10x this this isn't good enough yet and and so yeah dealing dealing with that has definitely been tough emotionally and and everything but but again it, it's a good learning experience. So there's a ton of concern in the AI world right now, especially in the world of, of course,

gender of AI around like privacy and security and, you know, what happens to my personal information or my core intellectual property and where you're, you know, essentially, potentially creating like a, or you have a tool essentially that helps people brainstorm. They're going to be sharing potentially stuff that is core IP or personal. Like, how do you think about, you know, navigating the privacy and security challenges that might arise there? Yeah. We've been, we've been talking about that a lot, just for, first of all, like privacy security in general of like the application make

sure people can't see what you want that there's you they can't see what you've created unless you share it or things like that um but in terms of that part we've been working on like our privacy policy and figuring out like okay we want people to be able to say i want you to delete everything and it'll just delete all of it we want to be able to say like I want you to delete everything and it'll just delete all of it. We want to be able to say like, okay, the, what I've said in this conversation is just in this conversation and I don't want it to ever get into another idea or something like that. And so we, we build in a lot of silos in terms of how the conversation flows and stuff. We don't like have memory quote unquote of a previous conversation per idea or anything like that. And we've, we're working on policies where we can say, like,

we only retain logs, like for X number of days before we delete them. And that's purely for debugging purposes. So we don't ever sell your data, stuff like that. It's, it's been, it's been an interesting thing to navigate. Luckily, it hasn't been too challenging yet because we just, we, we could kind of answer those questions, be like, this is, this is what we're doing. This is what the, these are the services that we use and these are their privacy policies and stuff. But it's, it is something that I'm sure is going to come up more and more. We've had a lot of conversations about that recently. Yeah, absolutely. I mean, it's good that,

I mean, it sounds like it's something that you're very aware of and that you're starting to put the right sort of guardrails and systems in place to make sure that you're not at risk of essentially leaking someone's personal information across conversations or between accounts or anything like that. Yeah, exactly. In terms of what's next for Contenda, where are you thinking that this goes after this kind of initial launch? Yeah, good question. I think we're kind of trying to figure that out ourselves because, again, Brainstory works

for so many different use cases, but I feel like to do well, you kind of have to pick one use case and really lean into it. Yeah, pick your battles. Yeah, we're leaning into the meeting prep one more currently, just because that seems to be a bit stickier with folks where it's, again, it's collaborative and not just a productivity tool and stuff.

And that's been interesting. We have so many things that we want to build with BrainStory, where again, because we have an API underneath all of it, like we can build so many different things where I want to build, for example, a blog commenting system with Brainstory or like a Chrome extension where you can comment on an article with Brainstory or things like that. And where you can hit certain things point by point and it'll just work. I would love to build all of these things. But at this point, we are actively trying to get as many users as we can so that we can get to that next step where we can take it and run with it. Because right now we're definitely a lean team trying to just use our runway responsibly and get users and I don't want to say survive because that sounds kind of negative. We're doing well enough, but again, the economy is just weird enough that we want to be as cautiously optimistic as possible. Yeah.

I mean, the good thing with when these cycles happen where fundraising is not as easy as it was at one point or the economy is in a downturn is the companies that do survive those periods are some of the best companies that have ever existed because they're forced to essentially focus on the right things. Nothing focuses you like a lack of resources. Exactly. And that was a thing that I really learned from my time as an entrepreneur was, we did our dumbest things when we were flush with capital.

And we did our smartest things when we were basically really backed up against the wall, didn't have a lot of money because it forced us to focus. I think focus is like, or lack of focus is like the kryptonite of a startup. You're much more likely to die from indigestion of too many ideas than you know starvation from uh too few ideas so i think you know leaning into something that's working and kind of really pushing on that makes a ton of sense while is still having maybe this sort of larger vision of where you could go right yeah exactly and so we are very focused on just kind of seeing the different use cases that people use, finding like teams that might want to use it for their team meetings and, and communities that might want to use it.

A book club was using it recently, which we thought was interesting where someone just talked about the book and people were able to give their thoughts, which I thought was really cool. And so, yeah, figuring, figuring out like what is that market that will give us a good foothold to be able to build more rather than focus specifically on just making everything a bit better. Yeah, there's a land and expand strategy there. If you're able to really nail meetings, I'm sure there's other performance things that

you could do in a business that is analogous to that. Yeah, exactly. Well, as we start to wrap up, is there anything else you'd like to share? And how else should if someone's interested in BrainStory or interested in you, what's the best way to get in contact? Yeah. So if you'd like to stalk me on the Internet, I'm Cassidoo, C-A-S-S-I-D-O-O on most things. And if you'd like to try BrainStory, you can try it for free at brainstory.ai. It's spelled like it sounds, brainstory.ai. If you do use BrainStory and you

have questions or you want to check it out, you could go to brainstory.ai.discord and we are very actively chatting in there all the time with different users. We'd love to hear your feedback. Awesome. Well, Cassie, thanks so much for joining. And I really appreciate you picking a name for a product that I can spell and remember. So kudos to you for doing that. That's not always the case. And I'm definitely looking forward to trying it out. I think there's a lot of places where I can use it. As I mentioned, brainstorming is one of the common use cases I use these tools for. So if I can do that in a more efficient, effective way, I'm excited to do that and excited or happy to even pay for it. Well, that is music to my ears. I'd love to hear what you think.

And thank you so much for having me today. All right, cheers.