Loading summary
Transcript54 lines
- [00:00]
A
Today's episode is a show and tell episode with none other than Kevin Rose. He takes us through his entire AI workflow and lets us in and a new product that he's developed that he hasn't shared anywhere. And we learn about how he thinks about building new products, Claude, Code, Vercel, all the tools he uses. So it's super, super fascinating. Kevin is just one of those iconic entrepreneurs and to have an inside look into his AI workflow and how he's building products in this AI age is absolutely fascinating. It got my creative juices flowing. I think it will get yours too. And if you stick around to the end of the episode, you will understand how to build products in the AI age, how to think about it, and what tools to be using. Enjoy the episode. We got the one and only Kevin Rose on the podcast. I'm super excited to have him. He has been going down the AI rabbit hole for a while now and I wanted Kevin to come on just to share, you know, what are the tools he's using, some of the AI workflows that he has, and he's just going to screen share. Kevin, by the end of this episode, what do you think people are going to get out of it, like, if they stick around?
- [01:23]
B
Yeah, well, I certainly. First, thanks for having me on. It's. I love the content and everything you're producing. It's so important right now, especially with things moving so fast. But I think at the end of it, you will see that things that are technically out of bounds for you, like things that you just think that you cannot do, are very much possible as a solo engineer, solo designer. And now we're finally at the point where I'm not even calling it slop anymore, like they call the AI slop or whatever it may be, or however you want to look at the code. It's damn good and it's getting better by the week. And I think that you'll just. I hope that you'll be inspired to go build something amazing because I'm going to show you something that is a little mad, sciencey, weird and all over the place. And you'll get to see kind of a raw version of my brain and how deep I go on some of this stuff. But I think that's the beauty of it is, is is this idea that we can go build anything and oftentimes when we do, it ends up being a little bit of a, I don't know, a sandbox that can be a little too big and messy and then you have to Refine it back to something that's actually usable. So I think that the future engineer and the future developer, the future product builder here, it's not going to be what you build as much as what you don't build, if that kind of makes sense. Because it's going to be so easy to build anything and everything. To pair it back to something that's really usable, I think is going to be a real skill. Yeah.
- [02:53]
A
The hard part is the clarity. Like, how do you get the clarity to know what to build? At least that's, that's, that's what I've been struggling with.
- [03:01]
B
Yeah. And then that's actually rolls right into the project that I've been kind of working on just for fun and. Yeah. So I'm excited.
- [03:09]
A
All right, let's get into it.
- [03:11]
B
Okay. So we can. I'll show you kind of this, you know, quote unquote, vibe coded, or let's just call it coded Project and I'll tell you the inspiration and then we'll get into some of the, the kind of dirty details and how it functions and what's possible. As someone that kind of walked into this thinking, actually just asking myself, can I pull this off? So let me go ahead and do a little screen share here. All right. You should be able to see tech meme up there, right?
- [03:41]
A
Yep.
- [03:42]
B
Okay, awesome. So first thing and, and I hope we don't in the edit. This has to make it in. I have nothing but a huge amount of respect for Gabe and what he's created in techmeme. What I'm building today is not meant to be a competitor. I don't plan on launching it in its like form. It was mainly a personal curiosity of can I build something that's on par or better than techmeme by myself and call it like a week and see what that would look like. And one of the nice things about techmeme is for those that don't know or for those that are lightly familiar with it, it's been around for a long time. Gabe's been building software since I was back in 2004, which is crazy. And we were both kind of, you know, experimenting in early social news and what that meant. And what you see here is an aggregator that pulls from RSS and other sources. I don't know all the sources that he pulls from, but then also pulls from social media as well. And when you gather momentum, whether it be multiple news stories coming in or multiple people tweeting about something or talking about it socially, that is considered signal. And then that is considered used as part of, you know, ranking and showing you what's prominent here. And so what, what, what you're seeing is actually visually, the higher an object is here, the more weight it carries with a user. Because if you come in here and you scroll down the bottom and you see something with very little tweets on it, it doesn't nearly carry as much impact as, you know, something with 15 or 20 different X posts underneath it. And so I like what he's doing here. I think it's a, it's a really cool way to say, you know, maybe you might recognize some names, especially in tech, given how small the ecosystem is. You can probably look through this list and say, oh, I know these two or three people, they're talking about it, what are they saying? And might I just hover over and see what their comment was about a particular story? So a lot to love here. I mean, Gabe's done a great job at creating this, but I have a slightly different area of interest. A lot of this is big tech news and I kind of want to dive into like more of what's going on in AI because AI is moving just so fast. You know, how can I slice, slice this in really interesting ways to find my version of this. And this is stuff that we're playing around with at Digg with the reboot as well. So it kind of goes hand in hand with some of the kind of exploration that we're doing in a lot of these different areas. And so my job these days, you know, with Alexis Ohanian on at Digg, is we have Justin as a CEO who runs kind of like the day to day and builds out, you know, the Reddit competitor version of Digg. And then we have, you know, me that I'm kind of more in the labs area where I'm pushing on the edges of ideas and saying what might we do? And does any of this make sense to roll back into the main product? And then Alexis is kind of just a overarching great idea. So much depth of knowledge around, you know, both micro and macro communities and how they work and how they scale and what to do and not do, and some of the missing tools and that's kind of how it all comes together. So, long story short, this is what I built. So first thing I want to show you is that there's going to be some errors here because it's been a minute since I have actually done the whole actually touches code base. But what you see here is I have 63 sources of information coming in. And so these are from just your classic RSS feeds. Yes, RSS still does exist. A lot of sites do break rss. And so there are actually ways to go and add in kind of scrapers that will create dynamic RSS for you, even though RSS is no longer around. But you can see here, it's a bunch of different RSS sources that are coming in here. Some of them are redis, some of them is TechCrunch, but all largely tech related. So because this hasn't been run in a while, we've got a lot of jobs that have to run in the background here to go and recrawl all of this. So I haven't run this project in a few days. And so it's essentially going out and hitting all these different RSS feeds, pulling in and ingesting all that information, bringing it into the engine and then expanding upon. And there's orchestrators and there's ways to actually go in and resolve certain chunks of information and find out more about individual stories. And I'll get into that in a second. But we're playing a little bit of catch up here. So if the news stories for some reason don't look exactly fresh, fresh, you'll know why. But they, they, they'll be pretty damn close here in a couple minutes. It just needs to go and, and do this, this catch up. So going back to the RSS feeds here. So, okay, great, we've got sources now what do we do? Well, sources lead to articles and when we go over to articles here, you can see now it's starting to pull in some of these articles in real time. So we've got Mac rumors 13 minutes ago. And then we've got a little status check here. And the status check is really what happens once the article is pulled in to the system. Like how do we process that article? How do we figure out what to do with that article? So you know, 21, 25 minutes ago 29 minutes ago the Verge 9 to 5 and so pulling in and saving all of this in a postgres database. And then we're also going in and doing things like figuring out who the authors are. And this comes down to author reputation. And so we can start taking a look at different authors that are here and I can click on, you know, any of these authors and actually, you know, see what articles they've actually contributed as well. So you know, this individual author here and what all of the articles that they published. But what's more interesting here is actually what we're doing with these articles. So what happens is as I start to pull in these articles and this is where it gets pretty crazy because I didn't have this. I didn't really know I understood technically how to pull this off in terms of the tools that would be required, but I couldn't write the code to do it. And so once we go in and we take an individual article, Let me just get into an actual. An article itself here. Let's see if that will take me into the actual. That might pop me out to. Okay, here we go. So here's a article right that now that just came in, talks about the new AIR tags that just got announced. And this came in from MacRumors. It's got an ID associated with it. So I was setting all this up within the database. It has a cluster membership associated with it, which we can get into in a little bit, but. But also it has a pipeline status. And so what happens is, is it comes in via rss. I hit iframely and iframely gives me back additional metadata about it. So like title, description, sometimes it can pull in even additional kind of deeper context about it, article data, things like that. And I want to store all of that. That's really important for me to have that rich data. And then I can see that Gemini has been run against it as well, which we can get into in a minute as to the reason why. And so the winner has been resolved, the TLDR has been generated, embedding has been done. So the winner is basically saying, okay, there are a bunch of different factors here that might come in. Like, for example, if I'm pulling this in with rss, I might not get a description there because RSS sometimes is like kind of truncated. I might not get a really good description. But if I go and I hit iframely, then I might actually get a longer and better description. Or if I go and hit Fire crawl, actually might get the full actual paragraph of. Or the full body of the article. So what we do is we have a judge that looks at it and says, okay, which is the best of these three that came in and pick a winner. And the winner is the one that actually gets stored to the database. So that's what winner means. We do that on a variety of different things. We. I don't know why we say we. I, I do that. I'm the one that actually ever coded this thing. So basically we can see here that for this particular story, it did not have. I think we got a failure we're coming from Mac Rumors. The winner was actually Gemini. So Gemini is used as a last resort. So you can see here, Firecrawl returned a 403. We had a hard air there. Iframely, we had some kind of low quality signals. We got a successful crawl. We had low quality signals. And then rss good. It's used as a fallback. Good amount of characters here. We can see what the actual summary of that content was when we got ours out of our ss. But yeah, as you can see here, that's actually pretty crappy RSS poll. There's not a whole lot of data there. You know, it just doesn't look that well or it doesn't look that good. So what we did is we hit Gemini and said, hey Gemini, turn on your ground truth, turn on your search, go out and figure out what is like what is this actually about? Right? And oftentimes they won't come out and say this, but they kind of crawl the article and give you back what the article actually says. And they might reword it a little bit, but it's largely what the article says. And I can prove that because they can. You can do that with Reddit articles and things like that where most all everything will fail because red so hardcore about blocking it. Does that Greg? Does that make sense so far?
- [12:58]
A
That does make sense. I just have one question around. Like why did you choose Fire Crawl and iframely especially? And can you just explain what the, what they do for people who aren't familiar with those services?
- [13:11]
B
Yeah, for sure. So let's go ahead and take the actual article here and so I'll show this tab instead. And so you can see this is the Mac Rumors article, right? Yeah. Now if I take this MacRumors article and I copy and paste it and then I go into iframely here I can paste in a URL, any URL. And obviously I'm doing this via API, but I just want to show you how it works. And I hit check URL and what it's going to do is going to give me this, this card back. So it's almost like those cards that you see on, on X, you know, where you get a piece of rich media, you get the best possible image, get a beautiful title, nice little description here. Sometimes it's a little bit longer and then I can just jump in here and look at the JSON and see like, okay, what did we get here? And kind of expand out and say, do you. Did we get a high quality logo? Like, you know Like I can look for and parse through all of this and try and find high quality data. And then also, I mean, it's called iframe largely because if there is an embedded piece of media, you also get all of those embed codes as well. So you get all of it. So I can actually take. And the reason I like Gemini is because they're tied in with YouTube obviously and I can get transcripts back from full YouTube videos, which is great for vector embeddings when I want to figure out what it's actually about. Okay. Does that make sense?
- [14:35]
A
Crystal clear.
- [14:35]
B
Okay. And so you can consider Fire Crawl to be a very similar thing. It is a little bit more on the crawling side, so a little bit more fine tuned around crawling. They have some AI aspects to it as well where AI will actually try and go out and figure out how the best way to kind of scrape the content. And then. Yeah, lastly with Fire Crawl I think it's just they have some stealth modes that you can turn on. So some of these news sources, they get really picky about what they, what they're allowed to be crawled and so you can kind of like hide and turn on stealth mode and then get to the actual data. And it's not my, my intention is not to like, you know, get around their ads or do something like evil here. It's really just to figure where's the signal and, and start to cluster these things together. So back to. And then by the way, the working title for this is my, my little thing that I code on nights and weekends is called Nylon. It's my little incubator that I work on. So that's, that's why you see Nylon up here in the corner. All right, so scrolling back down we can see that. Okay, so here's the resolve content iframely. We got the picture, we got the author, we got the summary, and here's the main content from Gemini. So we use that Gemini one here. IFramely won the summary, Gemini won the actual main content. And, and then what I hit is I hit GPT5 mini, largely because it's fast as hell and cheap. And you don't. It's very, it's quite smart. Anthropic has a handful of models that also fall in this camp. I mean they all do. You know, it's really at the end of the day because I'm using vector embeddings from, you know, OpenAI. It's just when I have one model provider, I just stick with it. And unless I really need to bounce around, which I did for the actual Gemini crawl and understanding of other things. So I try to keep it as simple as possible, but sometimes you need multiple models. And I will say Vercel's AI gateway is a great way to kind of code once and just flip a model on on the go. So highly recommend checking out Vercel AI gateway is a way to quickly swap models rather than having to recode it. All right, so I want a TLDR that's important. I just want that human readable so you can see that. So I want a TLDR that is a vector tldr. And so for people that don't, don't know, vector embeddings are really interesting. I had never worked with them before, I'd heard about them, I understood technically how they function. But the point is that if you take a keyword rich and kind of deep understanding piece of content, you can create mathematical representations of that content and embed these with OpenAI and store them in Postgres by using a vector extension. So they're actually stored in your database as those pieces of math. And. And when you apply some of these clustering algorithms on top of them, they get really good at nuanced information where keyword search would completely fall down. So the old school ways back in the day in 2004 when I launched you know, Dig as a social news site, if you search for, you know, Apple releases, whatever it's, it's fine. It's looking the word Apple is looking for release and it's looking for whatever it was back then that they were doing like an ipod or something, right? And it would, it just does it based on just can I find that text in the database? And if so, show me back the article. The beautiful thing about what we have today with our understanding of linguistics and around using vector embeds and algorithms on top of that, is that you can say there is a difference even though they're both have the same type of keywords. But there is a huge difference between Apple sues Google and, and Google sues Apple. And that is impossible to do with keyword search because you're not understanding at a deep level what's going on here. Right? So anyway, this is a very rich, purposely rich, longer form version of the TLDR used just for, for vector embeddings. And then I also wanted to create some key points here that we can use to feed into other models later when we're comparing the difference between articles. When we see multiple articles starting to get clustered together. And then I don't use this, but I asked AI, like, hey, write me like a spicier title. Like a title that people might click on more or find more interesting and just to kind of rewrite the title. And Tech Meme does this too. Like, when you go to the front page of TechBeam, it's not the title of the article. It's like actually what their editors chose to write. So I just wanted to see how this looked, and then I wanted to put it in one of three different categories. Tech Core and a couple other categories, largely because there's a lot of stuff that come through these tech feeds, especially when you add in, like Forbes and some of these others, where it's like, you know, does not relate to actual core tech or AI or the things I care about. And I just want to put them in a bin. And so here's the embedding. I'm using the large model from OpenAI. You can see when it was generated, when it was done, and then just some information about making sure that we don't recall the article and have all that. So that's done. Okay. Any questions so far? We got one article into the system.
- [19:40]
A
No, keep going.
- [19:41]
B
Okay, so now let's talk about clusters. So clusters. Well, actually, let me show. Show off one other thing. So how does this actually work? So how does that. All that work on the back end? Right. And there's a couple different ways that you could do this. You could say, I want to kind of write this in next js and I want to have this just be a function. And if you want to get a little bit more fancy, you could say, okay, I'm using Supabase, so I'm gonna throw on some cron jobs there and fire off these things. And I don't know if they fail or, you know, there's. There's a lot of. It gets a little dicey here because oftentimes things, bad things can happen. Like, you can have a RSS feed that gets blocked temporarily. You could have timeouts. You could have a model that actually doesn't complete. And so I need durability around a lot of this stuff. Right, right. And so what I do for that actually let me just do a screen share again, is for the durability side. Okay. I use a service called trigger.dev and so I like trigger.dev because what it does is it allows me to create these functions. And they're all typescript that live in the cloud and that they are fired off either When I call them from my app or at a certain cadence. And so there are these orchestrators that will go in here. You can see the expansion orchestrator. That's when it wants to go in and expand a story and actually see, you know, more, more information about it. You can see clustering different things here. Here's my fire crawl, here's my Gemini, here's my iframely. So anytime I fire something off to be enriched, I actually create this new little micro instance that goes out and runs on its own. You can see here the ones that are executing, See the little spinners here? So these are all kind of running in real time. And then you can see the compute charge to go off and execute these and how many milliseconds they. They took. Now the thing that's interesting here is you actually get to see the whole chain in which everything went down here. So you can see, okay, I was looking to kind of resolve a story ID locked in candidates. I got, you know, one picture, one summary. Did it resolve the winners? Yes, Gemini was the winner here. It was published at. And then it finished the whole process. The nice thing about this is that if something fails, I get retries for free. And so I will automatically, you know, if an AI TLDR for a vector embedding, which is very important as we're building clusters, fails once, fails twice, it will continue to retry, as you know, and automatically spin up these instances and try again. And then will report back to me via Sentry or any other type of monitoring software that I have on the back end that, oh, I had a failure. Why did this actually happen?
- [22:37]
A
Right?
- [22:37]
B
And so I like that because a couple things, if I'm developing this locally and I don't have this on production, my data continues to be enriched in the database. And so I can continue to develop before I actually deploy to production. So when I'm running this locally, like I'm showing you here, it's like it's, it's, it's functioning and continuing to build things out. There's a few things that are still tied locally that I need to catch up on to fire this back up because I just don't want to be burning through cache for, for no reason if I'm not using it. But. Yeah. So does that make sense?
- [23:09]
A
Yeah. I also think, if I remember correctly, Trigger dot DEV is open source. So we like that. I also think it's relatively cheap.
- [23:19]
B
Oh yeah, it's really inexpensive.
- [23:21]
A
Like it's something like 10, 15 bucks a month for 50 tasks. Or something. Last I checked, it's.
- [23:28]
B
Well, I mean, I, I think we. Look, when we looked on there, we were actually seeing the per. Per. I mean, I'm running thousands and thousands of these and you know, I. Yeah, it's like under a hundred dollars a month and that's, I mean, when I say thousands, I mean like per day.
- [23:41]
A
Yeah.
- [23:41]
B
So it's, it's, it's not obviously if they're going to be super, you, you can choose your instance type like with anything else. These are really lightweight, non computational kind of tasks to do. Some people use trigger for things like, you know, using FFMPEG to do encoding and you know, things of that nature that are going to require a larger instance are obviously going to be more, more pricey. So it really depends on like everything in cloud, on the workload, right?
- [24:10]
A
Totally. Yeah. I just don't want people to see this and be like, oh my God, this is probably thousands of dollars a month. It's like shockingly inexpensive for what it is.
- [24:17]
B
Yeah. And the other thing I was going to tell you is that Vercel now has something called workflows that they launched in beta that are basically, it's for free and it is trigger.dev for free. But it's part of the Vercel ecosystem, if you're into that ecosystem. But they're kind of these. It's a way to monitor, retry and have these long working tasks as you know, with edge functions. Like that's been the challenge. Right. Like you don't really get a whole hell of a lot of time and things get stale. So anyway, this is nice. I like it. You're right though, it is an extra expense, but I think you can. It is relatively inexpensive for this type of task. All right, so back to the clusters. So now what we've done is we've run an algorithm. On top of that, we've got all the vector embeds and we're starting to build clusters. So as you can see here, this is today. I don't know that everything's caught up. It looks like we've enriched. Oh yeah, we're close. 2,284 of 2,288 stories have been. So 99.8% has been enriched in the last 24 hours, meaning that they've gone through that entire pipeline of that ones that were more or less failures that we actually had to reach out to Gemini and get some extended data from them. 350, 53 stories were done there. And you can see here we're starting to get some actual weight. So 105 sources reported on this EU investigation into X over Grok generated sexual images. Nvidia invest 2 billion into debt ridden core weave. It's 47 stories and you're starting to get something that looks kind of like a tech meme ish type feed. And then obviously you can slice and dice this in a variety of different ways, but this is where it gets even crazier. So I promised you crazy stuff. This is where we go down the rabbit hole. So let's just take this idea of core weave and Nvidia investing 2 billion. So we'll click on that cluster. And now we have 47 stories. Nine were done via RSS and 38 discovered. So what does that mean? That means that when I see enough signal and I consider that three or more RSS stories talking about the same thing, I then hit Search APIs. You can hit Brave Braves Search API. I use Tavle Search API and you can say go out and find me other stories outside of my RSS ecosystem that may match this and bring them in so I can expand the scope and see is this a broader story that has more context that I'm just not seeing because of my finite set of RSS articles. Does that make sense?
- [26:59]
A
Absolutely.
- [27:00]
B
Okay, so and then I'm looking at there's average distance between articles and similarities. This is all part of this, this algorithm right here that I'm using for the clustering. So first story started four hours ago. Last minutes was 13 minutes minutes ago. So then I created something called the gravity engine. And so this is kind of like a editorial type score. And then it has actually it votes on it on how important this story is, generally speaking. And I rate these in a few different buckets here, like the impact. Another one called it's gravity, which we can get into the confidence ratio that this is pertains to technology and the things that I'm interested in. And then this is all the stories that it's evaluated. And then here's a little matrix here that I've had built out where you can see you have impact and then you have gravity. And it's in the high, high area here. Viral potential on how bubble size, like how big this will eventually get and how many people will be talking about it within tech is 60%. Early trend, like a growing trend is 75%. Intellectual gravity is 86% and impact is 88%. And so a total editorial vote of 85%. And then we can get into the, the the rubric here where it is. So these are the impact dimensions. So x axis drivers are industry impact. So it's 90%. So the massive $2 billion investment solidifies core Weave's position as a key new cloud provider, intensifying competition with hyperscale scalers. Critically, Nvidia is also launching its Vera CPU as a standalone product, directly challenging intel and AMD in the data center market. Consumer impact low 10% this is negligible for consumers, so it rated that actionability. Do I need to actually do anything here just for a general audience, Can I take action? 30% said some action for investors maybe, but that's kind of it. And then the risk and urgency, like do I need to act on this? This would be like, okay, there's a 911 iOS patch that you need to apply to your phone or something, right? That would be pegged at like 99 risk and urgency. And then the, the Y drivers here are intellectual gravity drivers, which is how non novel is this? Which is very important to me because I want high. And I'm not, I haven't fine tuned this in because it's saying how novel is this for today? But in reality what I want is novelty as applied against the longer horizon timeline. Because one of the things that's really fun is if you go back and you look at the first time on Hacker News when Bitcoin was mentioned, everyone was like, ah, this is stupid, blah blah, blah. But if I would have found that in here, the novelty would have been off the charts. And that's important to me because oftentimes the largest things that we do in tech and that we see evolve that turn into these blockbuster things over time seem very silly when we first hear about them or so odd or different. And that's the signal as both a builder and investor that I want to find is early as I possibly can, right? So that, that's, that's one that I care about deeply. Technical depth, you know, second order potential, builder, relevance. This was important to me on the AI side. I want to know how important, like as a someone that is building in tech, how much should I pay attention to this entertainment value which didn't have a lot of that. And then I've got some cross cutting signals here. Signal to noise ratio, viral potential and early trend detection as well. And then I added in some other like I could. These judges that just sit here and kind of walk through this, which is, you know, what's the PR fluff risk? Like how much of this is just a PR thing? Because a lot of this can be, you know, we'll see 20 or 30. And I can see this. It's. It's really crazy. You can tell I can detect when something is a paid sponsorship, even when sometimes people aren't calling it out as such, which is like really scary and illegal. Because I will see, I can detect the similarity and distance difference between the vectors of the published content, between the news articles. And they're all released with plus or minus an hour of each other. And they're all taught, they're all hitting the same major key points. And it's just like AI reworking the paid sponsorship and I'm just like pretty effed up, like, you know, and I'm like starting to find this stuff, you know, it's just like, wow. So this is, this is what I do at night, Greg. I'm embarrassed, but this is what I do at night.
- [31:31]
A
So just a question on this whole, this whole feature, because like, when I didn't expect it to be this full fledged, like on point, like, is this something that you sketched out on paper and built it, or are you like working with, you know, AI as your co founder to kind of help you come up with this? Like, walk me through the product management piece of this.
- [31:56]
B
Yeah, I mean, I'm very much someone that builds based on gut instinct. And for me, 99.9% of the features that I create. And don't get me wrong, there's a lot of stuff I would cut out here now, you know, I would consider like, I didn't like that, should probably cut that out. Like source distribution is a great one I wouldn't get into. But I started was like, okay, let's crawl rss. Okay, let's just do that. Let's throw images up, let's get as much rich data as we can. Let's do a very basic clustering algorithm even before we put in vector embeds. What does that look like? How does it feel? And then I just was like, well, what is tech me not doing that I really care about? And, and that led to my personal curiosity around novelty of objects. How can I detect these trends before they become big? You know, all these things where I was just like, I'm not seeing that anywhere else, so I should just build it. Right? And so it's one feature at a time. So what you're seeing here actually is like, I would say each of these things that you're seeing is probably a day or two of just being like, well, let's see if this will work and what it looks like and then putting it and then, you know, kind of using in today's tools. What I would do is I would just, you know, use compound engineering on Claude code and do a workflow and say this is the idea that I have. Actually, I probably would. Well, it depends on how big of idea. If it's a minor feature tweak, I would just do it that way. If it's something bigger that I needed to flesh out that had technology that I didn't know what the right choice was, then I would use AI as more of a sparring partner on that side. So I'll give you a great example. There are probably 10 competing clustering algorithms that you can use for news and hell if I know which one's the best. Right. And so I actually took the top two that it recommended for what I was trying to do. And you don't. What you see here at the top of this URL is see clusters v2. And the reason it says v2 is because the second one ended up being better than the first one. The first one no longer exists. And so it's a lot of kind of like just going down that rabbit hole and saying, well, let's. Let's try these things out. And. And sometimes I would immediately realize that was the wrong direction and just actually just uncommit that last whole GitHub repo or that commit and PR that I spent four hours on and just chunk it and throw it away altogether and be like, well, that was four hours lost. But at the same time I learned something new. And that' that's all of. That's all of building. All building is, is failure after failure and just. And that. And failure is awesome because it's just admitting that you've learned something new. So many people beat themselves up over failure. And I, I don't see it that way. I just see it as like, that's failure is like. It's the best part. That means the next time it's going to be a little bit better, you know.
- [34:52]
A
So anyway, I'm really interested in the. This whole concept of like testing products on synthetic audiences. I don't know if you saw this, but, you know, a few weeks ago, I think Toby from Shopify launched a feature where it was like, you launch your E commerce store, but like, based on synthetic AI audiences, here's how they would perform, here's how your conversion rate would look like, here's the products that they would click into based on these Personas. And I think that's like, that's sort of the direction we're probably going to head in. So before you, like, publish something, you know, before you publish an ad, you have some certainty that it's going to work before you publish, right, an article, you know that, you know people are going to be clicking into it, right?
- [35:40]
B
Yeah, I, I think there is so many domains where that makes a lot of sense. And Toby's obviously just a freaking genius and so brilliant. What I do, that's slightly different and I would, I'd be all for that type of thing, but is I build for myself. I think we're entering into this era of, of personal software, right? Like, if there's a workout app that you don't like because the buttons placement doesn't. Doesn't do it for you, or it doesn't track one key core metric, like, you just build your own, right? And like, that's going to be the norm in, you know, if it isn't already, it's going to be the norm like six months from now, right? And then the question is, how many people are there like you that also care about said thing, right? And so when I'm building this particular slice of the news or the industry and we haven't gotten into the understanding of who's touching things, because I think one of the things that, that, that has done really well in tech meme is that social touch of like, you know, when Marc Andreessen touches something with a tweet or, you know, how much more credibility and weight does that add to it? Even more so than Bloomberg or Wall Street Journal or Business Insider writing about something, right? So that's yet another thing that needs to be baked in here as well. But then, you know, if I enjoy it, the nice thing about this whole thing is this was, you know, maybe 300 in AI credits or something, you know, to go build this whole thing. And I could stand this up, cash the crap out of it, meaning, like, so that it's performant and, and just put it out there and everyone would be like, okay, if I'm also into the geeky things that Kevin is, I will use this too. And that might be a thousand people, might be a hundred thousand people, no one knows. But also at the same time, like, that's okay. Like, if you, if you have 500 people that really love what you've created, I get a lot of value and joy out of that. And it's like we, we think in numbers now on the Internet in terms of, you know, millions and billions of people. But in reality, if you, like, went outside your house and there are 500 people standing like cheering you on, you'd be like, I'm the biggest rock star in the freaking world. Right. So we lose this perspective of what it means, what success means. And so I just hope that we can all agree and just realize it doesn't have to. We don't have to swing for the fences. So, yeah, so I just, you know, that's. I think why I'm so excited about this is it's democratizing code quality or coding for everyone. And I was a computer science major. I dropped out and I didn't know why. And I was really slow and I could understand the core concepts, but I didn't know why I couldn't just do it as fast as my, my, my, everyone I was working with in my, my class. And just six months ago, I found out that I have something called aphantasia, which is this inability to have a mind's eye. So when, like, people close their eyes and they say, picture an apple or like, you know, the famous one of like when you're trying to go to sleep, like, like sheep jumping over a fence or some shit like that. I always thought they were joking. You know, I didn't know that you could actually close your eyes and envision things. And so because of that, you know, things like how to handle proper syntax and code and like all the things I was trying to beat into my brain, the retention just wasn't there somehow. The core concepts stick with me and the. In the creativity, I've always had that in abundance, which is great. But. Yeah, and now, like, the AI will fill in the deficiencies wherever they are for you, which is just beautiful.
- [39:15]
A
So I also dropped out of CS school. Only had three classes left. And I was the same way in the sense that for me, I wasn't. I couldn't get the last 10%, so like, I couldn't get the code to compile because there was like, I had, I had 92% of it there, but there was, you know, an integer missing here, a variable missing there. And what's cool is nowadays, you know, of course it's nice to know that stuff, but if you're just trying to get something out the door and to get feedback from people, you know, let Claude Code figure that out for you.
- [39:51]
B
Yeah, exactly. And the other thing too, I think that is lost on a lot of people that I see on social media around what, you know, quote unquote, vibe coding means, is they say, well, great and this has been the complaint for a while. Vibe coding is buggy. It's not performant. It'll fall over underweight, blah, blah. I would argue those are great problems to have. The hardest thing to do is to find something that somebody actually wants to use. Right? Like, that's the hard problem. If I have something that I vibe coded and I launch it, and if it crashes under the weight of 50,000 people beating my door down because it's the next best thing, I guarantee you I can find you engineers to work on that and scale it. Right? And so I don't think that should be a reason why we. We kind of like, I like it because you get more shots on goal, right? Like, I don't have to look at the code. It's not because, like, you. Yes, I can jump into a component in typescript and. And be like, okay, I kind of see what it's doing here. You know, I can. We can. We can do that. It's slow, but I can do that. But that's not the point. I don't care right now. I'd rather see actual humans using it and saying, yeah, Kevin, this is so cool. I want you to formalize this a little bit more, make sure it does scale, and if that comes in the form of usage, then I'll find the right engineers to. To take my kind of scribble code and make it real and performant. And honestly, compound engineering has already been amazing in that it's finding a bunch of stuff and making things more performant for me on the fly.
- [41:24]
A
So for this particular project, you're going to put it out like, success. I'm just trying to think like, success. Success looks like what?
- [41:32]
B
Well, it doesn't look like anything. I might never launch this. I might put together a little one pager that's just like the best AI stories and shown by the things that I care about, like novelty and impact or whatever it may be. Or, you know, there's just. I realize that you've got Product Hunt, which is kind of people that have already launched things. You've got, you know, Tech Meme, which is fantastic at overarching big news, you know, core weave, $2 billion. Like, that's a tech meme story all day long, right? And then you've got X, which is just a lot of stuff that's coming at us so hot and heavy. And it's just like, okay, well, where do I decide to spend my time? Like, we were just talking about this before the podcast started. You're like, oh, if you play with this, you play that. And I'm like, ah, you know, because it's like you get. So there's so much coming at you. I want this thing to eventually, if it ever sees the light of day. What I wanted to do is to say, Kevin, this is important to you because it maps to you. These very important people have touched it and said it's worth your time and it's past that threshold to where now you should go install it, play, have fun and learn about it. Because otherwise I'm just going to be in. Sadly, because of my adhd, I'm going to be bouncing around too much to even get anything done. So that's, that's my hope. But yeah, that's so you can see there's a lot of other features we can get in here that we won't have to. But it's, you know, this is me playing to see. Because a perfect product here actually would be to cut 90% of these features and just find the 10% that really means something to me and a lot of people and launch it as a standalone single page website. Right. And that's what all. But this is the messiness of it all. And I, I just want to show people that like, for me what I do is I put everything on the table, all the stuff on the table, all the stuff that I would never show anybody, like the, the, the distance and similarity scores between two stories, maybe I would show that. But get it all out there and then decide to cut and go in there with that kind of director or editor's cut and start making, and start trimming, trimming, trimming, trimming down, down to where you eventually get something that's really usable and useful to folks.
- [43:53]
A
One feature which I would love to see on something like this would be, I'm just thinking out loud, like if I was a PM on this product, I would be like, what is the mechanic to bring people back? So if you go to go to idea browser.com. so I built this, it started off as like a lead magnet, actually.
- [44:17]
B
Yeah, I've seen this by the way. This is like a very famous thing now that you felt.
- [44:24]
A
And so basically the idea was the lead magnet started off as here's a database of 30 ideas I would build today. Yeah, and then it was like put in your email to get the access to the database. And then it was like, okay, how about what we do to make it more fun is almost like, you remember Groupon, let's do instead of a product today, an idea of the day and then the mechanic is the email. Every single day, you know, we get a 50% open rate, people open up the email and it's grown, you know, very, very fast through that way. So I wonder, you know, if I'm building what you're building, you know, nylon. If it does see the light of day, it's like, what is the mechanic to get people back to the website?
- [45:13]
B
Yeah, I mean, in my mind, the, at least for news in general, it is relevance to the end user. You know, if it is discovering stuff that you are missing or you have overlooked and it's saving you time and energy because it's saying, you know, it's presented in a visual way. Like I'll give you an example, if you were on a conference call today and you had, you know, let's call it like just like the, some leading minds in tech of Sam Altman and Mark Andreessen and like a handful of other people in AI said, hey, you know, Greg, I think you should go check this out and play with it. There's a good chance by that afternoon you would be like installing it and messing around with it. Right. And so visually I would have to say there, here are a thousand signals that came in today. How can I show you the five things that have launched or that are in beta, that are in GitHub, that are worth your time and map to your interests as well? Right. So I would probably go in and pull your last 100 or 500 exposts and, and look at how you interact and create, you know, vector representations of who you are as an individual and then also try and get a little bit more custom so that it maps to you. So if you're very much into robotics, you would see a bunch of amazing kind of robotics information being presented to you across a variety of different fronts. So, you know, you would see it both in terms of things that are being talked about on X and also things are being talked about on Product Hunt or things that are be talked about in Hacker News or you know, Reddit or any number of sources of the new dig or you name it. So I don't know, I really don't. I think at the end of the day for a consumer app, you know, it has to come down to am I finding something on useful from this site or service that I don't see anywhere else and is it helping me save time and energy? And the answer may be no. And then guess what, it's, it cost me $500 and I flush it down the drain and onto the next thing you know. Yeah, like I've got another one I could show you if we have time. Really don't have time. Do we have two minutes?
- [47:24]
A
Yeah, let's do it.
- [47:25]
B
Okay, so this one is, is like December 15th. So December 15th of. So 12 years ago I posted this idea that you could have a blog where you can actually see the person in the background in real time, but it's blurred out, but it gives you a sense of kind of presence that they were actually there. And I was, I'm still kind of enamored with this idea, especially because we're entering into a world like there, there I am, like inputting in information or typing as I'm typing. But we're, we're entering into this world where we don't even know if there's another human on the other side of it. I think that's just going to get worse and worse over time. And so, you know, I like full on built this in Claude code and now I have it, you know, completely running and I can, I can just show it to you real quick. Let's see here. Just pull it up.
- [48:26]
A
While you're pulling that up. It just, it strikes me that there's a bunch of good ideas from, for example, 12 years ago that could be recreated today.
- [48:34]
B
Oh, 100% right. Yeah. I mean there were things that were just. Either so much of this is just right idea at the right time, you know, and, and some of it was not technically possible, these crazy ideas that we had, or some of it is even richer now that we can bolt on AI and make it, you know, cooler and different in some unique way. So there are a lot of things that I feel we'll see the light of day again, but slightly modified because it, the cost to do so is, is, is next to nothing, you know, which is, which is great. So what wasn't possible back then was real time video compression in the browser. And the reason I say that is the prototype I built 12 years ago, the issue would have been that you uploaded RAW video to the site and then you blurred it after the fact. And which meant a user could go, just trim away the CSS and actually see that person in a very awkward kind of position or whatever and like maybe an environment where they were like looking for a little bit of blur privacy. And so now this is like all done in real time. So what you're seeing now is that it's actually recording me. It's like, hello world, this is a test. And then I can just kind of move My arms around like this a little bit, get a little movement and please, like, this is the v01 alpha of this whole thing. And then hit submit. And it's doing some horrible broken math here. But you see that, that's actually me in the background and that's not me now. That is the blurred version of me. And it's a horrible interface. I would never release this interface. But see, that was me when I was waving my hands a second ago. But I could do all of the compression now on client side so that I do preserve privacy and put this up in the background. And so you can imagine these little slivers and this little visibility into people's world as they're kind of blogging. And in my head I'm like, okay, well, maybe this should just be my blog and I'll release this and I'll just open source it and give it away. And there doesn't need to be a business model because this is just fun, you know? So, like, that's, that's what I love about the time we're living in, man. It's the best. You can just have fun.
- [50:55]
A
You can just, you can just put out things. And, and sometimes, though, it's sort of interesting. It's like when, when there's not. When you just put out projects for fun somehow. I don't know why, but they end up being the, the projects that could end up becoming the biggest businesses.
- [51:14]
B
You are 1000% right. It is the weirdest thing that I do not know how to explain. Like, when I, when I did dig back in 2004, it was just like, I just want to see if people can vote on things and what it looks like when the best stuff hits the homepage. And then, you know, a year and a half later, there's 38 million people a month using the site. When I made Zero, the intermittent fasting app, I was like, I just want a way to track my fast and like, have a silly little timer. And then, you know, the company's doing double digits, millions of dollars in revenue off of a little like, timer app with some other content on there. And I'm just like, how is this even. These were just for fun. These were, you know, and it's, it's very, it's very strange how that works.
- [51:56]
A
So before we wrap up, you're. You're working on Digg and you're kind of incubating projects. Like, I'm listening to you and I'm like, how can I work with Kevin? I'm sure people are listening to this being like this. This sounds cool. Like, how could people support. Get to work with you?
- [52:14]
B
Yeah, I think there's a couple different ways. I appreciate you saying that. There's. I have a. I'm sitting in an office right now that's completely empty. And so I have an incubator here in la, the studio that I'm going to kind of open up and there's just going to be free desks for people to come in that are jamming on really cool stuff. So if, you know at Reply me or DM me on X, if you're building really cool AI stuff and even if you're just coming through LA and just want to come in and jam and the point is like, let's just compare notes and talk about what's cool over lunch. And like, you know, if you need an office, you need a room to take a call, you got it for an hour. Like, no big deal, right? No fees or any type of, like, I don't want to charge for desks or anything like that. So I'm going to surround myself with those types of people here in Venice, out in la, and I want to do that. So please, like, hit me up and let's, let's figure out a way to connect if you're building really cool stuff. And then, you know, I'm a venture capitalist still over at True Ventures. And part of what I think, I think VC is going to evolve dramatically over the next couple of years because I believe that people don't need to raise capital. And oftentimes most ideas, especially like just great lifestyle businesses that get to, you know, 2, 4, 5, 10 million in revenue, like, own that shit 100%, don't sell it to VCS. I'm not supposed to be saying that because I'm a vc and so, but like, don't do it. And so, but that's why I hope people will realize that, like, that that's why I. There is a time and place for vc, like when you really get to scale and you're like, damn, I need, I need a couple million bucks to hire because the growth is so outrageous, you know, like, and that's, that's kind of what I do. And that's why I want to play. Also, I think VCs, the, the era of VCs, just being these like, MBAs that sit there and like, try to tell you how to run your business, I think is so boring to me. I want a VC that is playing, that's building alongside me, that's like pressure testing my ideas. That's a thought partner on this stuff. And so that's kind of like why I don't even want to call. I hate the even word venture capital. I think it's like evil. It feels evil these days. I just, if I, if someone does. First of all, I try to talk everyone out of taking money. And if they do need money, especially hardware companies need a lot of money, then they should find someone that's also building and that's built stuff at scale. Not just because somebody has a, a great, a great pedigree or you know. Yeah. Does that make any sense at all?
- [54:44]
A
It does. And I just want to add like I'm the first person to be like, you know, people listen to me on the pod, know that I think that you should your MVP your business. You know, you probably don't need venture to start and there's some businesses that shouldn't raise venture. But I also think that in this world where you can build software, MVP it. If you want to build hardware, you're like, you build this cool piece of software, it starts to take off and you're like, you know, oh, there's this logical extension to hardware you're probably going to need to raise money.
- [55:18]
B
Yeah, 100%. We invest in a company called Sandbar that's doing this AI ring. I don't know if you've seen the prototypes for it, but it's really cool. And I can't remember how much you put in. It was close to 10 million or something like that. But it turns out to do the tooling to go and get that to scale. Those are real dollars still required to pull that off.
- [55:40]
A
Totally. Kevin, it's been an absolute treat having you on the pod. I hope you come back on again and show and tell more because you're up to some really cool stuff. You're a legend and I'm gonna take you up on that. Next time I'm in la, I'm seriously.
- [55:56]
B
Yeah, done. Like you got an office here, so come, come hang. And thanks for having me on. I appreciate it. And thanks for all the work you've been doing in this space, man. It's like I, I know we don't talk that often except for like the random DMs and stuff, but I see see you all over X and the content and stuff that you're putting out, it's absolutely fantastic.
- [56:13]
A
Thanks, man. Just trying to create some signal out in the world of noise.
- [56:17]
B
It's great. Well, maybe the Nylon app will identify that and put it in front of more people.
- [56:21]
A
Exactly. That's what I'm hoping. All right, Take care, Kevin.
- [56:24]
B
Take care.