Subscribe
Share
Share
Embed
Interviewing indie founders about their journey and their products. itslaunchday.com
Dagobert (00:02)
Hey David, welcome to lunch day!
David Norman (00:04)
Hey, great to be here. Thanks for having me.
Dagobert (00:08)
Yeah, so you were one of the first people to buy a spot when I announced September. my god, it's almost one year ago. Crazy. And back then, and I want to start with this because I think that's interesting. You had this indie product that you were starting and then now I said, hey, you can have your spot now that I launched. But since then, you started raising a little bit of money, which is very interesting.
David Norman (00:33)
Yes, a little bit.
Dagobert (00:35)
Yeah, which is very interesting because I think you know at first I was preventing everybody who was not 100 % Indie because it was just simpler but I realized a lot of people like you can still be pretty much Indie and just have like crazy just a little bit to keep going and it's not like big VC or anything like that. It's just like some investment so you can you know focus on your product. how so how how what happened you know in the past nine months like what were you working on like what happened? I'm just curious about that.
David Norman (01:04)
Yeah, so ⁓ when it happened, was originally, I was working at another startup, a fintech startup, ⁓ and the AI agent explosion was kind of happening. So there were some kind of interesting products online like ElizaOS that you probably saw on Twitter. There was these kind of agents that were connected to a token ⁓ and they would kind of tweet at the same time. I got
really interested in that and started contributing just kind of on the weekends and stuff to the ElizaOS ⁓ project as a, it was an open source project. So just started contributing ⁓ and actually got invited to join the team as a core developer. So I started working kind of directly with them and really getting a lot of experience for how AI agents kind of work. ⁓
Dagobert (01:41)
Yeah.
So you have an engineering background, you're like a developer?
David Norman (02:01)
Yes. Yeah. So I've been a software engineer for about 15 years. I didn't start off that way. Um, but yeah, I've been, been in the space for, for quite a while now. I've worked at kind of small startups all the way to larger companies. was a software engineering manager at Capital One, um, you know, a few years ago, um, Cisco systems before that. So I, I, I grew up in the Bay area. Um, so it was kind of surrounded by that. Um, so that's kind of where kind of
ended up going down that path. ⁓
Dagobert (02:34)
Yeah, I
love behind you to see. Is it like a first Macintosh or something?
David Norman (02:38)
Yeah,
it's not quite the first one, definitely one of the early ones. This is the one I like kind of grew up with. It's the Mac, the Apple II, and then there's a Mac II SE on the other side. it's not the original ones. Yeah, not the original ones I had growing up, but I found some on eBay for a reasonable price. So I had to pick those up for fun. Yeah, so I...
Dagobert (02:50)
my god, legendary stuff.
That's beautiful,
David Norman (03:03)
During that kind of AI agents explosion, there was a lot of interest in that, especially on the kind of crypto side. So that's where I started applying. I was currently working on a startup. I didn't really have the bandwidth. I really wanted to work on a project. And at the time, the project that I submitted to you with originally was a podcasting app. I tried pitching that. And kind of at the same time, Google had come out with this, was it Notebook LM?
which was basically a way to turn documents into a podcast. so my original pitch for, I ended up kind of pitching to Orange Dow, which is a crypto incubator for a kind of AI podcasting slash crypto sort of a project. And so kind of just have been pivoting ever since then. I ended up kind of getting into that. It was a really, really good experience. I got some time to
to go to San Francisco. I live in Texas, Texas now, so I had to kind of go to back and forth to San Francisco quite a bit during the incubator, ⁓ but got connected to some really, really great founders through that program and ⁓ kind of pivoted a whole lot until we got to where we are now.
Dagobert (04:06)
Okay.
So it's like a small incubator, that's it. How do you get into these things?
David Norman (04:30)
⁓ So the way that I got in it, you know, just like the the standard ⁓ application process ⁓ they're kind of a ⁓ So you submit just the idea basically like I didn't have any VC connections I don't I don't know anything so I just You know there they you know this white Combinator does the same thing where they're ⁓ a 16 Z has a kind of incubator as well where once every quarter
they'll have like kind of a batch of companies. So you submit your ideas, they kind of go through these thousands of applications and they kind of pick through that. ⁓ find, know, based on all these VCs have kind of, you know, a thesis about what's happening in the market right now. And at the time that I was in, it was AI agents. ⁓ So, you know, having the kind of crossover there. ⁓
Dagobert (05:16)
Yeah.
David Norman (05:25)
was just kind of, think I just kind of got lucky with the right idea, with the rights at the right time. ⁓ and so yeah, I had a couple of calls and that's it. So there was some kind of conversation before I got invited to it and I didn't really have anything at the time. Like it was really mostly an idea and like a very simple MVP.
Dagobert (05:48)
And so initially you started with, okay, this is a podcast app to train, Google Notebook is it? Like that they do. So it's just, yeah.
David Norman (05:59)
Yeah. Yeah.
so the kind of journey to what we actually have now, which is a kind of a, uh, an AI workflow for, uh, that basically looks like a text editor. Um, Eliza OS was kind of an agent that, you know, a lot of people, even, you know, even the founder Shaw would kind of get frustrated sometimes. And I've seen you talk about it where it's just like, you know, if
people are using AI to kind of respond on Twitter in a kind of inauthentic way sometimes. And so they're actually, ⁓ you know, and that was kind of really the early version of my product, which like turned out ⁓ was just not the right way to go. ⁓ I think there's been a lot of kind of pushback against that. And ultimately it was kind of making the
Dagobert (06:33)
Yeah.
David Norman (06:55)
web a worse place, which is not at all what I wanted to do. ⁓
Dagobert (06:58)
Wait,
so how is it connected to podcasts and Twitter replies? don't get it.
David Norman (07:02)
Yeah, yeah. Well, that was the original idea. Like that it just pivoted so many different times it got to the. Yeah.
Dagobert (07:08)
Okay, so please take me through it. Okay, so you
mean you started with the Twitter reply or you started with the podcasting?
David Norman (07:15)
Well, it was originally ⁓ going to be the way that it's connected really is just in the way that AI kind of generates content is similar to the way ⁓ that an influencer would generate content. You're kind of like absorbing the information that's happening on the web. You're reading tweets, you're kind of reading books, you're reading information, you're reading the latest news and kind of commenting on it. ⁓ And so the original idea was like,
Well, we can kind of do a lot of that work for someone. And, and really this is something that I wanted to use. ⁓ you know, it's something that, know, if you could actually create something, you know, filter in all of the information that you, that I consume on a daily basis, I'm watching YouTube videos, I'm watching Twitter and kind of generate something from that. and then actually create something of value, not necessarily something that's going to be spammy.
Dagobert (08:07)
Yeah.
David Norman (08:13)
but something that ⁓ actually combines all the information that I would normally consume and finds the best of it, puts it together into one thing. I think a...
Dagobert (08:25)
Yeah, so
it's not just like some random comment, it's more like the brain behind it. How your brain thinks to come up with ideas for content.
David Norman (08:30)
Right, right.
Right. And so that's, that's kind of the long-term vision. I don't know that that's necessarily, ⁓ you know, possible right now. I think there's a, and that's kind of why the tool that I built now, there's a lot of human, it's human intervention first. There's a human in the loop at every single step and you're basically using AI tools to kind of create the context, pull the context together, ⁓ in the way that AI can do best.
Dagobert (08:53)
Okay.
David Norman (09:02)
And then you have the final say on what the actual content that goes out is. And so it's made for longer form to comment content rather than kind of just short. You're not replying to a hundred people with like these kind of quick comments. ⁓ you know, Gary Varenchuk has this idea that when you're on social media, you're supposed to do kind of this jab, jab, jab left hook. So you're the best way to kind of grow is to, is to do a lot of small things. ⁓ you know, write a lot of reply to a lot of comments, which is.
You know, one, one takeaway I took from your, from your Twitter course, ⁓ that I took like a while back. ⁓ but then you also need to have this kind of like longer, more thought out content. And that's really the sort of thing that I, I wanted to kind of create tools around, to make easier for, for creators.
Dagobert (09:43)
Yeah, that's it,
So just to be clear, it's absolutely 0 % related to podcasts now. OK, OK.
David Norman (09:55)
Yeah, now eventually,
eventually you might be able to create a podcast from it. But yeah.
Dagobert (10:02)
Okay, I see, I see, So can you show it? Because I'm just curious. to be honest, I really hate AI replies. That really makes me... That I block people basically when they do that. And I hear some arguments, like there are some people who are like... You know, I've noticed people who like have struggled to speak English. And so they say, yeah, but that's how I can express myself. And I'm like...
David Norman (10:10)
Yeah.
Mm-hmm.
Dagobert (10:31)
I think I prefer someone who speaks shitty English to be honest though, but yeah. Because like, you know, it's, I don't know what you think about that.
David Norman (10:34)
Yeah, yeah.
Yeah, there's an argument to be made for it for sure. I think it just depends on how you're, if it's literally just improving your grammar, that seems like that's, that's one thing. If it's, you know, if it's completely coming up with something out of nothing that, you know, didn't actually come from your mind, then it's just kind of inauthentic. So.
Dagobert (10:59)
Yeah.
Yeah, that's it. That's it. Okay, so show us this absolutely not podcast tool. I'm just curious.
David Norman (11:06)
Yeah, so this.
Yeah,
so this is Hyperfeed. And this is kind of the main interface where we have kind of pre-created a bunch of different blueprints that you can kind of choose from. And if you used other kind of AI content creation tools, one of the more popular ones is kind of like NADN or Zapier, where you're able to kind of...
there's kind of three steps to it. like pull in some data from some data source. ⁓ You run some AI prompting on it and then you kind of publish it out somewhere. ⁓ And a lot of time that's from like a Google Sheet to another Google Sheet or something like that and then it gets filtered out. ⁓ And you can kind of start with a template. ⁓ these are a number of templates that you can kind of start from different things from like ⁓
from YouTube transcripts to an RSS feed to...
Dagobert (12:12)
So it seems
to always be starting from existing content or existing thing. It's not coming up with bullshit. It's like starting from something.
David Norman (12:17)
Data, Right, yeah, it's connected,
grounded in truth is kind of the idea, grounded in some sort of actual live information ⁓ is kind of the goal. ⁓ So for example, ⁓ this is the main editor, looks just like a text editor. ⁓ You can use it exactly like a text editor. So this is the example of the... ⁓
crypto, you're basically taking a live stream of crypto data and it can be for any, you know, particular, whatever your favorite token is. ⁓ And it'll take that data. And so you have a selection right here. This is kind of pre-filled for you, but you could also kind of change this. So you've got all your data sources there and we kind of manage that all for you. So you don't have to worry about API keys or anything. We just, you just select the data that you want and it's there available.
Dagobert (13:03)
Okay.
David Norman (13:15)
That just kind of gives you an overview of what's the information that that got pulled in? See the current price here and then from there, you know, you can write regular text like Bitcoin data And then you can also add kind of compose so it's all composable so you can also If you ever work with like a template editor where you can add kind of ⁓ templates you this is a
Dagobert (13:20)
Okay.
David Norman (13:44)
a template that will run, take the context of the data source. So in this case, it's, you know, they get the Bitcoin price. ⁓ And we also have an image template, which is also taking that. And we have a prompt in here about how we want to display the information. And from there, you just kind of build that and it will kind of compose all that information for you and, and, and create some new content from it.
And so that'll take a second. have an example right here of when I ran earlier. And so you can see it, you know, generated the image with the current price at that time, ⁓ kind of summarize the information and then included a link ⁓ to the data. So that's like a very simple workflow.
Dagobert (14:16)
Yeah, OK. OK.
Okay.
And so you said you don't need to manage API keys and everything, what are other examples that would require external data? Because it's very interesting. This seems like just generating an image with data. It seems quite simple. Then I saw Google Search or a trend that seems like, can you show? mean, if you don't mind. Just because. OK.
David Norman (14:53)
Yeah, so I have another example here of I actually took
your latest YouTube video. ⁓ So it kind of pulled the transcript from that. And ⁓ so you can see kind of you can read through the transcript right there. ⁓ And then I've this is a little bit more of a complex template. ⁓ So this top prompt is just saying summarize the provided content.
Dagobert (15:00)
Oh my god, okay.
Yeah?
David Norman (15:22)
And then we're actually breaking it out into a thread. So you can see it's kind of separated here in these different areas. And the context kind of flows through the thread. So you can see these two pieces are connected. So this image is connected to this part of the thread. This image is connected to this part.
Dagobert (15:39)
wow. Okay,
that's advanced shit. I see. Okay.
David Norman (15:42)
Yeah,
and so it's advanced, but it's also kind of the best way for AI to work. And ⁓ it also abstracts away a lot of things. Because if you were to look at this on a Zapier or an AN template, it's going to be like ⁓ this node connects to this node connects to this node. ⁓ Whereas this kind of is a text editor. And there's a few more.
Dagobert (16:10)
Yeah, no, no, that looks
actually quite simple from a user perspective, but I meant in terms of how much work you put into this. ⁓ OK.
David Norman (16:15)
Yeah, right, right. And so
when you generate it, so ⁓ the image pulls in the context from the transcript, generates an image based on that, kind of breaks it out into the different pieces. And so the particular image should kind of match to the content in each one of these places. I have a better example of how that. ⁓
Dagobert (16:29)
Wow.
No, that's actually quite good already.
David Norman (16:45)
And then, yeah, if you wanted to post, I don't think this isn't necessarily something you would post. ⁓ But if you wanted to, you've got kind of the direct connection here. And it'll show how it's broken out into a thread for you. And so you can kind of go from, if you find a template that you like for creating threads or blogs or content, you can kind of reproduce that by just changing. And you can.
you can easily just kind of change the data source and then you get all new content.
Dagobert (17:17)
Yeah.
That makes me rethink maybe... Yeah, I'm really against it in theory. Not against, but like, I wouldn't post this in theory, but it looks quite cool actually, so... Can you show like from Google search? I'm super curious about Google search one.
David Norman (17:22)
⁓ This is another one of my favorites.
Yeah.
Yeah, yeah, I mean, we can do one live if you want to. Here, let's go find here. can just.
Dagobert (17:44)
Yeah.
David Norman (17:52)
Let's just change one of these existing ones. So we've got.
Dagobert (17:58)
I think you know maybe the key because you know the people I don't block who reply with me with AI is the ones who identify themselves. Like they say they're AI and actually don't mind. What I mind is like the bullshit but I don't mind because sometimes AI says interesting things. And maybe the key maybe is to have like a launch day account not my account launch day account that is like kind of like half a bot or something.
David Norman (18:12)
Right.
Right, exactly.
Yeah.
Dagobert (18:26)
and
then it's clear that it's that. ⁓
David Norman (18:28)
Yeah,
yeah, I think there's definitely some value for creators who are like, you know, they're putting out long form content and they want to transition that into other formats too. So you want to take like a long YouTube video you did and then transition that into, you know, share that on Twitter. And here's the thread of it because it's not going to be, you can't just take the transcript from the YouTube and post it on the Twitter. That's. So yeah, Google search.
Dagobert (18:43)
Yeah.
Yeah, no, you can't. You definitely can't.
David Norman (18:59)
You want to try Launch Day? It's Launch Day, possibly.
Dagobert (19:03)
Really good.
Everybody's shitting on the fact that I have very bad SEO, so I'm scared, but go ahead.
David Norman (19:10)
Hahaha.
⁓ So we can see what that comes up with before we run it. If it doesn't seem like it's coming up with good results, we can try something else.
Dagobert (19:24)
Yeah.
David Norman (19:26)
SpaceX, Space Edge. Yeah. Let's go and run that.
Dagobert (19:29)
yes, SpaceX, interesting. No, it's fine, go ahead, do it.
David Norman (19:39)
OK, yes, ⁓ at this point here, so you can kind of see, this does take a little while, but it kind of gives you an idea of what's happening on each step. So it'll go and fetch the data from the data sources. It'll then kind of go in and write the text for you if you have text templates in there. And you can see we have two AI images in there. So it'll kind of generate images for you based on the context that you've kind of connected it to. ⁓
Dagobert (20:04)
I think that's much more interesting than just
AI writing the tweet for you. Like it's one level deeper because it's finding ideas, but actual relevant ideas that you still need to lead. You still like give directions. So it can be like, okay, this is a YouTube, this is a whatever. I like it.
David Norman (20:13)
Right.
Yeah. And because
it doesn't post immediately for you, it's not like a scheduling tool where you can go and schedule 10 of these out. You have to be involved in the process. So you have to read it and go like, yes, I approve of this. We also have like editing tools built in. So if it doesn't get something in the image right, you can kind of edit the image right there. ⁓
Dagobert (20:31)
Yeah.
cool. And you said
there was a human involved. Do you mean just that? Just the user?
David Norman (20:48)
Yeah,
just mean that you're the final, when it actually generates this template for you, ⁓ like for this template right here, for example, you can go in and edit any of this content. If it didn't get something right, you can go in and add links and edit any part of this. it's not locked. You can use it as you would any other text editor. And it's just starting you off down the path.
Dagobert (21:07)
Yeah, yeah, yeah.
David Norman (21:17)
to content creation based on live, real information.
Dagobert (21:23)
I guess there could be a bridge to also make, you know, TikTok reels and shit like that. I see it's trendy.
David Norman (21:27)
Yeah.
Yeah, I may have that done before this video is out. We'll see. But like going from a dead to a real, it's not that much.
Dagobert (21:33)
yeah?
It's a new level
of completely different work to do for you.
David Norman (21:42)
Yeah, but you can kind of put the pieces together. You've got the content, you've got the images, and you could either, you're just kind of putting those things together. So, yeah.
Dagobert (21:51)
And how you think you can put them together if you don't mind sharing? Like what tech can you use for that? I'm just... Or if it's SecretSource, you don't have to say.
David Norman (21:58)
no, no, yeah, it's not secret. ⁓ Generally, what I've been kind playing around with for an MVP for that feature is just basically like, there's a command line tool called FFmpeg, which everyone uses, which is like the lower level of video encoding abstractions. So if you want to have, you want to combine like five different images and you want to put text over top of it and then export that as whatever video format ⁓ and kind of
Dagobert (22:11)
Yeah.
Yeah.
David Norman (22:27)
do some video editing, you have a, that's kind of the command line tool that everyone uses and you can kind of build a lot of really cool things just on top of that.
Dagobert (22:36)
I see, I Man, that's cool, wow. I like when meeting people makes me think differently about things. Now maybe I will think differently about AI. Yeah, because I would like to have like a launch day account, but I don't really want to manage it, so that makes sense. I'm just not sure people would like it, but on TikTok, and even me, you you can consume AI content. I'm getting used to it and it's fine. The problem is when there's like this, when it's somebody tricking you.
David Norman (22:45)
Awesome.
Yeah.
Dagobert (23:05)
That's what's annoying when you think it's not AI. But when you know, it's
David Norman (23:05)
Yeah.
Yeah.
Dagobert (23:13)
Cool man. Man that was good meeting you.
David Norman (23:19)
Yeah, I appreciate it. is a, yeah, hope, yeah, glad to hear I changed your mind a little bit on AI content that there are use cases.
Dagobert (23:25)
Yeah, you know, I'm like, know, yesterday I
sent an email to someone I said, no, I don't you seem really cool, cool product. I like you. But it was an AI reply tool and reply AI is a bit too much for me still. But maybe I will change my mind. But yeah.
David Norman (23:38)
Yeah. Yeah.
Dagobert (23:42)
Yeah, sometimes I'm just like... I'm quite actually... I'm in tech, but I take a long time to go with new things. I'm not excited by new tech usually. I'm really not. I'm more like waiting until things settle. So that's why it's slower maybe for me to adopt things.
David Norman (23:58)
Mm-hmm.
I,
yeah, I think that makes a lot of sense because I mean, there's new things coming out every day. And so like, you're just going to waste half your time if you're just like jumping on every new bandwagon and you know, you really got to use what works for you and have a real reason to switch. So yeah.
Dagobert (24:13)
It's true, yeah. yeah, 100%. Yeah, 100%.
Yeah. Well, thanks,
man. I hope you can get some sales from Launch Day because you have a really cool product. It's more deep than what a lot of AI tools are. So awesome. And I hope you have some sales.
David Norman (24:31)
Yeah, I appreciate it. Yeah.
Looking forward to launch day. Thank you very much.