Tom Anderson (00:08)
This is Basic AF episode ninety one. Show notes for this episode and all episodes can be found at basicafshow dot com. Hello everyone, I'm Tom Anderson. Good to be with you again. And I'm joined by my co-host, Jeff Battersby. Hello, Jeff.

Jeff Battersby (00:23)
Good morning, Tom Anderson. Man, I almost forgot your last name. Is that n I didn't almost forget it. Right, exactly. We've never spoken before. So ⁓ it's nice to meet you, Tom. ⁓ and I I was getting a little ahead of myself. I was gonna say we're coming to you from the future because we're recording episode ninety one after we've recorded episode eighty nine and have not yet recorded episode ninety. So we're

Tom Anderson (00:27)
It's only been three days since we talked, but

Yeah.

Yeah, you too, Jeff.

Yes.

Jeff Battersby (00:52)
Yes, we'll record episode ninety tomorrow. So we're ⁓ we're a hot mess of dates and we're we are talking in the future, actually in the past about the future. So which is ⁓ look at me. I'm I'm very philosophical this morning.

Tom Anderson (01:02)
Okay, that got deep. Okay.

well, Jeff, I'm excited for this one. We have ⁓ a guest with us today. He's one of the co-founders of an app that we've talked about a good bit on the show. That ⁓ when we did our favorite things of 2025, last December, this was on the list. Also made the list of my essential apps and services that I sent out to my newsletter readers.

Jeff Battersby (01:17)
Yes, sir.

Tom Anderson (01:32)
⁓ Snipd so the AI based podcast app and we're happy today to be joined by Kevin Smith, co-founder of Snipd. Kevin, welcome to the show.

Jeff Battersby (01:35)
Mm-hmm.

Kevin Smith (01:43)
Hi, hi Tom. Thanks for having me. And of course also Mr. Battersby Jeff.

Jeff Battersby (01:48)
Look at you. Got that

name right. I can't get Tom's last name right, but you got my name correct. I love that. Yeah. Well done.

Tom Anderson (01:56)
First try. Mm-hmm.

Yes. So ⁓ Kevin, thanks for ⁓ carving out some time to hang out with us for this one. ⁓ I think it's going to be a lot of fun. Like I said, I'm a big fan of the app, and we're going to talk about it obviously a good bit. ⁓ but before we dig into it, why don't you give us a a little bit of background ⁓ of yourself and then we can kind of go into the journey of Snipd and how you decided to

Create the app, and and all that kind of thing.

Kevin Smith (02:30)
Yeah, I mean my backstory, at least as to the the extent that it's connected to snips, I think the best starting point is actually in twenty sixteen, where I started basically my first real job ⁓ after college. ⁓ so actually in New June University I studied to become what is called a quant. So I studied mathematics and economics, ⁓ all geared towards becoming one of these people who

Build all of these mathematical models for the stock market training, quantitative training, all of that stuff. But just at the end of my all of my studies, a friend of mine told me about this thing called machine learning. And it was still quite niche at the time. I mean, of course, today everyone calls it AI, but back then it was still called machine learning. And I got into it over a weekend and I basically completely fell in love with it and decided that this is what I have to do with with my life and my career.

⁓ so long story short, I ⁓ actually quit my my part-time job that I already had at a big bank ⁓ here in Switzerland where I live, and instead decided to join an early stage tech startup to build up the the AI team. ⁓ and that was in 2016. So there I was ⁓ my first day at this startup. I had no idea about startups.

And to get me onboarded into the whole startup culture and how to approach things in a startup versus you know like a big bank, a big corporate, the founder recommended this podcast to me called How to Start a Startup. Which coincidentally was ⁓ done by Sam Altman at the time, who in the meantime now very famously is ⁓ OpenAI CEO. Back then he like OpenAI hasn't hadn't been started yet when when they created that podcast.

Jeff Battersby (04:09)
Interesting. Right.

Kevin Smith (04:19)
And so here I was in in in Zurich in Switzerland on the other side of the planet to the Silicon Valley ⁓ San Francisco tech hub ecosystem. And I'm riding my bike to work every morning and getting this incredible advice from some of the most successful entrepreneurs that we that we've seen over the last ten years. Like people that you would never

Get the chance to talk to, or again, you would have to move to San Francisco and be lucky enough to get into the same room with them and then get enough time to get them to reveal some of their secrets to you. and this was really a moment for me where I realized how valuable podcasts are as a source of knowledge and and learning. yeah, and during that time, that was, however, also the time when I ran into all of these problems that I had.

You know, I'd be riding my bike to work and there there would be this amazing insight. I'm like, my God, I I need to remember this insight. So I'd get off my bike, take out my phone, open up the notes app, write down, yeah, you should do this and that. and you know, you do that a couple of times until at some point it's like, my god. I mean I I have to get to work. I can't stop every every two minutes. ⁓ so yeah.

Jeff Battersby (05:25)
Mm-hmm.

Tom Anderson (05:34)
Yeah.

Jeff Battersby (05:35)
Ha

ha ha

Kevin Smith (05:40)
At some point you you stop doing that with every insight and then you get to work and you want to tell your your colleagues about it, right? Because you want to share that the these insights that you've now learned. and then the things that you didn't write down, you don't really remember them anymore. ⁓ and some of the notes that you wrote down was you you're trying to communicate it to the other person, but it's like ⁓ I'm I'm not I'm not communicating in the right way. He said it in such an elegant way. Let me find that moment again in the podcast.

So then you open up the the the podcast app, you try to find that, you know, that 30 second, sixty second moment within a two hour episode. ⁓ you you hit that skip twenty second button like fifty times back and forth and no I'm too far. so long story short, you don't you don't find the moment. and the final piece of the puzzle is you you then recommend listening to this full episode to your friends or your colleagues ⁓ at work.

⁓ and a week later you asked them again, so hey, did you listen to it? And of course they they they didn't. because I I like comparing a podcast recommendation. It's it's often comparable to recommending a book. It's sort of like, that sounds incredibly interesting. I'll just add it to my reading list of five hundred and sixty-seven books that I already have on my aspirational ⁓ to-do list.

Jeff Battersby (06:51)
Yes.

Tom Anderson (06:59)
Mm-hmm.

Kevin Smith (07:06)
So yeah, these were all, let's say, the problems that that I encountered as a as a user and a fan of podcasts. but given that my background was in this machine learning, AI, ⁓ I did ⁓ various things with AI applying to text at the time. and then there was this moment where a new research paper had come out where a lot of the techniques that have worked quite well for text.

⁓ in the AI world, they were now being applied to speech. And it was still quite early, but for me it was like a sign that everything that we had seen work well with text would now also work well with speech, or at least in the next couple of years. yeah, and that together with some some other factors ⁓ in in my life, ⁓ I decided, hey, let's let's tackle this, let's start my own startup. ⁓

Just to just to give a bit of timeline, this was now five years after my first day at the startup. So I I I had a very nice very nice run at that start up, enjoyed it very much. ⁓ but it was time t for something new and so ⁓ I decided to to start Snipd together with two other friends. Yeah.

Jeff Battersby (08:25)
Excellent. So have you you're the one or the three of you I should say. I presumably one of 'em's doing the programming. Do you do the programming?

Kevin Smith (08:35)
We all do programming. Yeah, yeah, yeah.

Jeff Battersby (08:37)
Really? Okay. So you all you're

all part of that puzzle. the the machine learning, the AI portion of that I presume is is mostly you or is it is it the other two two people that that are working with?

Kevin Smith (08:52)
So the interesting thing is that this has changed entirely ⁓ over the last two years, I would say. ⁓ so back when we started, which just to give a bit of context, like the the first I think we first started playing around with it five years ago. So it's we've we've already been been playing around with this for a while. This was before Chat GPT came out. So before there were any APIs that you could just call. ⁓

Jeff Battersby (09:00)
Okay.

Mm-hmm.

Kevin Smith (09:20)
So when we started out, it was me and and and one other friend of mine who the two of us, we both our backgrounds was in in AI and machine learning. So we were doing most of that. But now, ⁓ today it's completely different. Like today we're not ⁓ training our own models ⁓ or fine-tuning them, ⁓ but we are using these APIs that are available. ⁓ so everyone has become an an AI engineer, ⁓ which

Jeff Battersby (09:39)
Okay.

Kevin Smith (09:49)
I believe is is the normal thing. There there's no distinction anymore, in my opinion, between a software engineer and an AI engineer. it's just part of the job.

Jeff Battersby (10:00)
Interesting. So you're using d do you mind saying what back end you're using for ⁓ machine learning for the what AP what AI or APA APIs you're calling?

Kevin Smith (10:12)
Yeah, so ⁓ I think you need to distin ⁓ differentiate between what we're using for building the app. So basically what we're using as tools, ourselves as workers, and what we're using inside of the product. So if you're a user of the product, what you're what you're using there. ⁓ so inside the app, we we've used all models that you can imagine from all providers and it keeps evolving. ⁓ it's really a question of

Jeff Battersby (10:27)
Uh-huh.

Kevin Smith (10:42)
What's the latest model that has come out? What is the pricing structure? How good are they for what for which task ⁓ to constantly optimize that? At the moment, however, ⁓ very interestingly, we're I think we have like 90% or even more running via OpenAI. but you know, ask me again in six months and it might look very different. And internally, there we we are using

I would say fifty fifty ⁓ Anthropic models and ⁓ open AI models. But this also is is like a ⁓ th that switches almost faster than what we're using in the product. Yeah.

Jeff Battersby (11:16)
Okay.

Tom Anderson (11:23)
Yeah.

Jeff Battersby (11:25)
Interesting. Yeah, I mean I think Tom can speak to that. I know Tom you've bounced around a bunch of these and currently at home you're Claude. You're using Claude at home and then and Gemini at work.

Tom Anderson (11:30)
Mm-hmm.

Yeah, yeah, pretty much yeah, eighty ninety percent

in Gemini at work. And been using more ChatGPT at work. We just got some ChatGPT businesses ⁓ accounts. So for us at work, it's we have to be careful with ⁓ the data that we put in to the AI tools. So we had to get the proper business licenses in place that put the data protections around Hey, we're not gonna train off of your chats and your data's not gonna go anywhere and that

That kind of stuff. And we just got that for Chat GPT recently. So I've actually started to use it more ⁓ than Gemini. I think Gemini's decent. It's it's not my favorite. Like I complain on here a good bit. Like the Google interfaces for some reason and everything don't really excite me very much. Like doc like Google Docs functionality, we use it, it's great, like for excuse me, for collaborative work, like with Google Sheets and everything like that. I mean, it's hard to beat. ⁓ but they're just not too good to look at in

my opinion. But ⁓ so question for you though, you d I'm just curious. So you said ⁓ you went from finance and then when you got into machine learning and you saw that that kind of really was like wow this is what I want to do. What about it ⁓ kind of drew you to become so like focused on it and wanting to use it? Like what what was that traction there?

Kevin Smith (12:57)
Yeah, for me, so the part that I was studying, the the part that I enjoyed the most was always the modeling, working with data. in general the mathematics side of things. ⁓ that that's also where I had my let's say my natural talent and why I like how how I got into this. It was sort of ⁓ I was always very talented in mathematics and I enjoyed it. And then it was a bit of a question when when it came to what what should I study.

⁓ it was clear, like, yeah, let's study something with mathematics, but I wanted to have something ⁓ connected to it where I saw ⁓ a more practical application of it, let's put it like that. and when I stumbled across machine learning, it sort of to me took away the parts that I wasn't excited about in my studies and gave me much more of the things that I was excited about.

So working with data, understanding data, ⁓ getting the machine to learn the patterns for you, which was another part of maybe maybe I can go a bit deeper into into that part there. In the finance world, at least at the time, ⁓ the way that you modeled data was that you came up with a economic theory why like how stock markets should behave. And you have a lot of assumptions behind that. So for example, every

participant is a rational investor as it's called. yeah, yeah. ⁓ that's the first assumption, right? So you know. and every

Jeff Battersby (14:26)
Yeah.

Tom Anderson (14:33)
Ha ha ha.

Jeff Battersby (14:35)
Mm also mistake number one, but go ahead.

Kevin Smith (14:41)
⁓ yeah, and everyone has access to perfect information and and all of these assumptions that just like make no sense in the real world. ⁓ but out of that you build a very beautiful, I have to say I really use the word beautiful ⁓ on purpose here, really beautiful mathematical ⁓ theory and formula ⁓ that describes then a lot of the behavior and and it does get you quite far ⁓ on on certain certain stock market movements, etc. ⁓

But then once you try to apply these formulas in the real world, you realize, ⁓ you know what? It actually doesn't always fit in these moments or in that moment. ⁓ let's just add a little something to the formula. Let's just add one more parameter. Like one more parameter is fine, right? Come on. That's it's okay. So you add one more parameter. And of course you come up with some kind of justification and and and theory and argue about that. ⁓

And now you have to fit that to the data. So like you know, what value should this parameter have? ⁓ so you use again like statistical methods to to find that and fine tune it. ⁓ and I always wondered, like, but why don't we add another parameter? And why don't we add yet another parameter, another one, another one, another one? And why why should we not just go through all of these potenti like possibilities and see what fits best?

⁓ so I had this in the back of my mind and when I came across machine learning, that was basically that's exactly what machine learning is. Going through all of the different possibilities and learning ⁓ purely from the data, what the data empirically tells you, and only trying to be able to predict what will happen next, independent of of let's say certain theories or or biases that we that we have. ⁓

So that immediately ⁓ captured my interest and ⁓ yeah, I I I just loved it. It was also so accessible to me. ⁓ because I had all of the tools to to do all of these things, but it felt like there wasn't I I didn't have to come up with a grand new unifying theory of how people behave in the stock market for this to make sense. No, you could just apply it and see, hey.

D is is this better in predicting how the stock market will move the next day or not? so yeah, I I really love that.

Jeff Battersby (17:15)
Interesting. ⁓ a little side question. Are you still doing anything on the ⁓ the banking side? Are you paying attention to the stock market and how it moves? Are you building your own little tools to be able to take care of that kind of stuff? Or have you ⁓ kissed the banking world goodbye? Okay, all right, fair.

Kevin Smith (17:32)
I've kissed the banking world goodbye.

Tom Anderson (17:34)
Ha ha ha.

Kevin Smith (17:36)
yeah. No, I mean, of course I try to try to invest ⁓ my savings in a in a smart way and ⁓ I'm interested. The the interesting thing that has happened since entering the startup world is that I've become much more interested in the actual companies. ⁓ so this is something that almost the thing that that I was missing in my studies back then, or like that that

Back then in my studies, I wasn't really interested in the companies. I was interested in the numbers and the mathematics, but not in the companies. and interestingly enough, today I I have actually developed that that interest. ⁓ but coming from a completely different aspect, not from the corporate ⁓ kind of thinking, you know, if you study economics, they in in my opinion, ⁓ when you study economics, they they really don't teach you how to actually build a business.

Jeff Battersby (18:15)
Okay. Interesting.

Kevin Smith (18:29)
But it's all about these like c corporate structures, ⁓ which are sort of second and third order effects of if you have built a very successful business, then yes, you need to think about how to ⁓ structure your company in the right amount of equity and the right amount of debt. But like that's that's like far down the line. Yeah, like first, build something that somebody else finds useful enough that they're willing to pay you something, you know?

Jeff Battersby (18:47)
Way down the line, right?

Tom Anderson (18:49)
Mm-hmm.

Kevin Smith (18:57)
⁓ and that they really don't teach you. and and that's one of the things that I loved about the startup world and and still love today.

Jeff Battersby (19:05)
So back to the first iteration of Snipd. So you you had a very specific personal need, you know, not wanting to get off your bike and open up the notes app and and put in a note and and figure out where that was. What was the first iteration of of Snipd like for you? You know, what did you what did you create initially?

Kevin Smith (19:29)
Yeah. The very first hypothesis that we had was that we would solve two problems at the same time. ⁓ one was the how can we make more use of the information that we consume, which I would argue is a large part of what Snipd is today. ⁓ and on the other side, ⁓ let's solve how we can discover all of this great content.

So at the time ⁓ one of the big things that everyone was talking about was TikTok. ⁓ TikTok had had just blown up ⁓ on in on the world stage. And we thought, how about we have this concept where a part of our listeners then where while they're listening to podcasts, they create these little clips that we would call snips basically by highlighting great moments while they listen. And then there's another set of users.

Or at a different time during the day, other users would listen to these snips, similarly as you would go on on TikTok. And we thought that would be a great way to discover new podcasts that you can then dive into later when when you're back on your bike or on your commute. So that is sort of the first version that we built. And the expectation that we had internally was that.

It would be very easy to get people to listen to these little snips, like I don't know, one to two minute little podcast clips. But it will be very hard to get people to create snips. And then we released it out in the world and got users to test it. And of course, it was the exact opposite: that nobody wanted to listen to these snips, but everyone was creating snips all over the place.

Tom Anderson (21:21)
Ha ha.

Jeff Battersby (21:21)
⁓

Interesting.

Kevin Smith (21:25)
Yeah. ⁓ so that that was actually our first ⁓ first iteration and and that was then the moment when we ⁓ f started focusing more and more on the side of yeah, extracting the knowledge that you're listening to, making that useful to the user, ⁓ making it easy to to connect out, for example, to your notes app, ⁓ or to share it with a friend, and have then focused less on on

the discovery side or consumption of of of the snips.

Jeff Battersby (22:00)
So right now, if you want to explain to us th the basic way the current version of the app works. So I'm a new user, you you want me to, you know, to start using this. What are you what are you saying that the app does? How does it how does it go about? There are a couple of features actually in the new version that I want to ask you about. But give me the give me the rundown of how you actually use Snipd

On a day to day basis.

Kevin Smith (22:31)
⁓ okay, I'm hearing two questions. One is how do I pitch Snipd to to someone I come across? And the other thing is how do I use it personally? Okay. ⁓ I mean maybe we can start with the first one. so usually I would actually ask the other person whether they listen to podcasts. ⁓ so maybe I can do this ⁓ with with you, Jeff. Do you listen to podcasts? Great, great. ⁓ what what kind of podcast do you listen to?

Jeff Battersby (22:37)
Yes.

Yes, both of those.

Sure. Many. Go ahead.

it varies. So some tech podcasts, obviously I listened to this one several times, you know, the one that we're currently on. But ⁓ also like one of my favorite podcasts, a couple of my favorite podcasts. Through Line, which is an NPR podcast, Seen On Radio, which is another they've kind of bounced around. ⁓ I can I don't even know who's who's running that one. But both of those tend to be very deep dive and

⁓ honestly, particularly with seen on radio, they're like college courses. You know, every single every single episode is dense with with information and I often will listen to those over and over again, much like I would read a book that I love over and over again.

Kevin Smith (23:48)
That sounds quite incredible. The I mean the that sounds as you as you said, quite dense in knowledge. Do you ever have that moment that you hear an amazing insight on one of these podcasts that you would like to remember? What what do you do in that moment?

Jeff Battersby (23:59)
sure, all the time. Yeah, all the time.

often rewind it and listen to it again. You know, I I ⁓ I circle back and ⁓ i again I'm you know, of of the three of us sitting here, I'm the oldest in the club. You know, I've ⁓ so I've I ⁓ so I often like a like I would in a college course, I would, you know, write something down and then and then

Tom Anderson (24:07)
Ha ha ha.

Kevin Smith (24:22)
No, that's great.

Jeff Battersby (24:30)
you know, ruminate on it for a little bit and then come back to it again and kind of see if I can ⁓ I don't wanna say make it so it's something that's applicable, but I my brain tends to hold on to those things in a very generalized cloud. You know, I I hang on to that information and and and, you know, whether it's reading an article, for example, Tom knows this, but I've been ⁓ working on a novel for couple of years now.

and you know, I'll read an article, I'll read the whole of it, I'll put it into notes, but my brain retains, you know, I know what was in that particular article and the piece that I wanted to pull out of it. and that's much the way that I listen to a podcast. You know, I I listen to something and I'll read it and reread it. I'm a slow reader, so I'll listen and re listen in a podcast, ⁓ until I feel like I've

integrated that into that little cloud of knowledge that I have. God f God forbid that I get dementia. But

Tom Anderson (25:35)
Yeah.

Kevin Smith (25:36)
No, that's great. So basically, ⁓ you know, what you can use ⁓ Snipd for or what you could do if you listen to your podcast on Snipd is basically every time you come across one of these moments where you think, this is an incredible insight that that I'd love to remember or re-listen to in the future, revisit in in some kind of way, you can just tap your headphones. It works with any kind of ⁓ headphones that are out there. ⁓ it doesn't have to be AirPods.

But you just tap them, and what happens is that our AI saves the moment that you just listened to with the original audio, the transcript, plus it uses AI to generate a quick note for you, which now automatically gets synced to your notes app ⁓ or just saved in the app and you can you can revisit that at any time. So basically, when you come back in the future and you want to re-listen to it, you could choose to ⁓

Re-listen to the entire episode. You could also just re-listen to your snips as we call them, these highlights that you that you created, or you can go through the the notes that it generated for you just to because often that's already enough to like re ⁓ resurface the insight ⁓ to see the notes that that the AI made for you. yeah, so this this would be my ⁓ my pitch to you if I if I met you at ⁓ at a dinner party.

Jeff Battersby (26:57)
Okay.

Tom Anderson (26:58)
Ha ha.

Jeff Battersby (26:59)
Fair. All right. Excellent. Now, I know Tom, ⁓ you use this extensively. ⁓ like you've been using Snipd for a long time. I ⁓ as I mentioned, have not used it a lot. ⁓ I I am playing with it again, obviously specifically to have this conversation with you, Kevin, but also just to kind of get a feel for it. But Tom, you've integrated this into your readwise experience as well, is that correct?

Kevin Smith (26:59)
Mm-hmm.

Tom Anderson (27:24)
Yeah, yeah. And you know, my experience before Snipd is very similar to Kevin's. Like I'd be commuting, I'd hear something, and this is going back. I mean, Snipd couldn't have existed back then because we were still listening to podcasts on iPods and stuff, but I would hear stuff and I'd be like, I need to remember when I get to work to write that down and I'd never do it. ⁓ and then I would try Siri, I'd say, you know, d Magic Word, do this and if it was Siri, so it didn't do it. So I think that's gonna get better with the new Siri, but ⁓ different

Jeff Battersby (27:37)
No.

Yeah.

Tom Anderson (27:53)
conversation. So with this, it's just been game changing, like fantastic because ⁓ I'm constantly hearing things in in shows that I I want to remember. And so, you know, I'm snipping, ⁓ I'll use the auto snips too if I'm just mowing the yard or something like that and listening to stuff. But I will use the tap the headset too. ⁓ but yeah I do have it tied to readwise. ⁓

I find that that's helpful with the daily emails they send or the weekly or whatever it is. ⁓ and so I'll use it that way. Now something I did recently just to test to see how it would turn out is I had Claude make a study dashboard of all of my highlights, which are basically just snips. and I had it give me the summary that's ⁓ Snipd creates, and then of course it gives you the transcript portion as well.

⁓ and then I had it detect the themes automatically. And so it breaks it out into productivity, business, personal development, whatever they are. ⁓ and then I'll review those and then it gives me like a quiz at the end to see if I've retain retained the information or not. ⁓ and so I just did that the other day just just to see if it would work, and it works surprisingly well. But I think that's the thing, 'cause a lot of the the the content that I'll be listening to, like

it it's it's an interesting snippet at the point, but it's really something that I feel like I may want to use later. Like for and what I like is that I can go back in, I can search it, I can listen to it again if I want to to kind of get even additional context because sometimes it's part of a three or four five minute discussion point in the podcast. ⁓ so it's nice to be able to go back to do that as well.

Kevin Smith (29:41)
Yeah, I mean this I I think it's a very interesting topic that you brought up. in my opinion, I really see the future of personal knowledge management being ⁓ these agents, these AI agents like Claude Code or whatever your favorite ⁓ agent is. And the role that apps like ours play, ⁓ and also ReadWise and and similar apps, is really making it as seamless as possible for you to capture

the context, capture the information that you are consuming and making it available to your AI agent. Such that in the future when you come up, ⁓ when you when you have a when you remember, hey, there was something that I heard on a podcast somewhere, ⁓ this was a great insight for this, that it actually has access to this information, can find it again, ⁓ and and resurface it. so I think that that's in in general a a a ⁓

a trend that we will see much more in our society. you can already see it much more in in in companies, right? Where now in a lot of corporates it's quite normal that meetings get transcribed and saved somewhere ⁓ such that you can access that information again. ⁓ and to me it's just makes sense to also do that with with the the content that you consume personally.

at least as long as you're consuming that content partly to to learn from it.

Jeff Battersby (31:13)
So one of the new features that you have in the ⁓ in it's in a beta version is is something called DJ. ⁓ which I played with actually this morning. I decided to ⁓ use the DJ to listen to our last episode of the podcast, which is about an hour and fifteen minutes. ⁓ way longer than we normally do, but a lot of coverage and we had

two other people on that current or we have two other people on that ⁓ current episode. it ⁓ synthesized the entire thing down to twelve minutes. So I feel very bad about wasting fifty seven minutes of everybody's time. ⁓ but w what ⁓ one of the things that always, you know, ⁓ that

Kevin Smith (31:54)
Well.

Jeff Battersby (32:06)
doesn't matter what AI it is or, you know, whether it's Cliff Notes on a book or something like that is, you know, so it took it took an hour and fifteen, it knocked it down to twelve. And ⁓ and it pulls out what it believes are and actually in listening to what the DJ pulled out, ⁓ you know, definitely relevant and key points from what the you know, the conversation that we had. But how is it making that determination? You know,

How how do I know it's not missing key points ⁓ of the conversation or maybe even a side conversation that is relevant or might be relevant to me. How do we miss that? So you want to tell us a l and again, I understand this is a beta feature, so you know, it's not it refined to the point where i it's perfect yet, but can you tell us a little bit about it's it's a great feature, you know. I I do like the idea of being able to to take

⁓ you know to basically filter out the garbage and come up with the key points. But can you ⁓ can you that that's correct, Tom. This is this is pure content. Every word is is relevant and important. But ⁓ can you tell me a little bit about, you know, what's what's going on behind the scenes? First of all, it listened to that podcast in short order and 'cause I imagine there aren't very many people unless it was Tom.

Kevin Smith (33:09)
Yeah.

Tom Anderson (33:11)
It's no garbage on this show, Jeff.

Kevin Smith (33:14)
Yeah.

Tom Anderson (33:15)
Ha

ha.

Jeff Battersby (33:32)
who have already run the last episode of basic AF through Snipd. So how is it how is it making that determination or those determinations?

Kevin Smith (33:42)
Yeah. ⁓ may maybe first I can give a little bit of background just on on the feature itself and then ⁓ go into how it how it works in the background. ⁓ so the idea of the the DJ feature, Snipd DJ, is maybe again starting, I always like bringing it back to how I think about it as a user myself. ⁓ I usually so I'm subscribed to I think more than a hundred podcast shows. There's so much content out there.

Jeff Battersby (34:10)
Yeah. You and you and me, baby. Yeah.

Kevin Smith (34:12)
And I come across episodes, I come across so much content that I would love to consume. ⁓ and then there are episodes they come out and I see it, and I immediately know, okay, I will 100% listen to this episode from start to finish. ⁓ and and I will do so. And then there's sort of this second category of episodes that it sounds really interesting to me, and I'll add it to my cue. But if I'm really honest, I know I will.

most probably not get around to listening to this again two and a half hour conversation. ⁓ I had a I had a ⁓ one example I had recently was there was an episode about peptides. And it seems like everyone's talking about peptides now. And I should know about it. ⁓ but am I gonna listen for two and a half hours? Probably not. Like this is sort of in the second category. ⁓ so up until now I would add it to my queue and it would sort of die a slow death in my

Jeff Battersby (34:56)
Right.

Tom Anderson (34:56)
Mm-hmm.

Kevin Smith (35:10)
Hugh Graveyard. but you know, what I would notice is that every now and then I'd be on YouTube and there was a clip of the episode, you know, like a five-minute clip. And I was like, yeah, it's I don't know, it says called Okay, the the three most important things to know about peptides. And I'm like, you know, I can invest those five minutes. So I'd listen to those five minutes on on YouTube or watch it on YouTube. ⁓

So what we tried to do in the Snipd app is basically say, hey, if you for the second category of episodes that you otherwise wouldn't have listened to, can we enable you to invest maybe not two and a half hours, but invest maybe 20 minutes of your time and then just listen to the best moments of this episode. And once you've done that, you can still decide, hey, this was actually incredibly good. I'll go back and listen to the full thing.

Jeff Battersby (35:56)
Mm-hmm.

Kevin Smith (36:08)
Or say no, that that was actually really enough. I I don't want more. so that was the idea, and that's what the feature currently does in the app today. So if you come across an episode, you can activate the DJ, it will go through the episode, try to identify which moments are most relevant to you, and then we'll guide you through them. So you're still listening to the original audio. ⁓ we're we're not copying any audio. ⁓ the DJ is merely controlling the player plus

The DJ acts like a moderator. So it gives a quick intro to each ⁓ highlight moment and then before jumping to the next highlight moment again gives a little ⁓ segue or introduction to the next moment. Because in if you're just listening in audio and it's jumping from highlight to highlight, it's very disorienting. you know, like suddenly there's a new voice that comes on. It's like, Who is this now speaking? and

Jeff Battersby (36:57)
Mm-hmm.

Kevin Smith (37:02)
Wait, now they're talking about something else. Maybe you just drifted off for a couple of seconds, which is actually quite normal when listening to podcasts and doing something else at the same time. And suddenly they're talking about something entirely different. ⁓ so the DJ then acts as a moderator and and gives these these short introductions. ⁓ but the main idea is you're still listening to it's not a summary, it you know, it's you're still listening to the actual content. yeah, so that's that's what's live today. How does that work?

one thing I wanted to mention. ⁓ your experience, ⁓ I guess it shows that it's not yet fully refined yet. The idea is actually not to break down a one and a half hour conversation into just twelve minutes. That's very short. the goal that it has

Tom Anderson (37:46)
Okay.

Jeff Battersby (37:46)
Or we or we said nothing, you know, for f or for f which is which is

reasonable. That's a reasonable assumption on this podcast.

Kevin Smith (37:56)
The goal is actually more to identify like the 20 to 25% of the episode that are most ⁓ that are most relevant ⁓ or most insightful. The way that it does that is it's it's an AI model that looks at the entire episode ⁓ in in one single go.

And it tries to identify what are actually the core themes of of this episode, like really starting from the episode title. Usually there is one main theme that that they try to focus on. And based on that, tries to then go through the moments to identify, okay, these were the key insights that the guest wanted to convey or the host wanted to convey through his questions, ⁓ and then focuses on on those. this will inevitably always lead

To a loss of information. Our feature is not a promise that you get the exact same amount of information as listening to the whole thing. I believe there's always a trade-off in any of these things, and people should be aware of that trade-off. If you really want the entire conversation and the side question, the side quest that sometimes ⁓ gives you that.

little nugget of information that you otherwise wouldn't have found, then please do listen to the entire episode. ⁓ we encourage that. It's not in general our let's say ⁓ culture internally at at Snipd or how we see it is we we don't try to like we don't have an opinion on on ⁓ whether or not you listen to the whole thing or just a part of it. ⁓ in in my opinion that should be up to the listener ⁓ to to decide.

yeah, that being said, one of the biggest feature requests, or one of the biggest requests for this feature, let's put it like that, is to give the user more control over ⁓ steering it. So customizing it. ⁓ basically two sides. So one is personalization, and the other side is customization. So personalization automatically learning from your behavior, from your interests, from your

Jeff Battersby (39:50)
Mm.

Kevin Smith (40:15)
⁓ snipping behavior on other podcasts. And the other part is what I call customization, where you explicitly tell it, hey, I already I you know, I don't know, you listen to an episode with ⁓ I don't know, let's say Elon Musk, and maybe you say, look, I mean I've heard all of his Tesla stories. ⁓ I'm just interested in what does he have to say about now the big IPO.

So yeah, these are these are requests that are out there and and ⁓ hopefully we'll we'll be able to implement them in the future. And to basically it for me it always goes back to giving control to the user, to me, or the listener in our in our case. The listener should be able to decide ⁓ what what they want.

Jeff Battersby (41:04)
Excellent. All right. Yeah, that's great. I I do like the feature, ⁓ I have to say. And we can have a conversation around, you know, why anybody's doing a two hour podcast. That's a totally different totally different conversation, you know.

Kevin Smith (41:16)
Well I'm glad

Tom Anderson (41:18)
There's some long

Kevin Smith (41:19)
they

Tom Anderson (41:19)
ones.

Kevin Smith (41:19)
I'm glad they're doing that. ⁓ I I you know I love I would say ⁓ most of the podcasts that I listen to are one to three hours long. ⁓ so I love long podcasts. I'm also a big fan of the Acquired Podcast, if you're familiar. I mean some of their episodes have gone up to six hours. ⁓ and I love it.

Jeff Battersby (41:37)
Wow, no. I would never

in a thousand years

Tom Anderson (41:42)
Jeff and I wouldn't be friends anymore if we did that.

Jeff Battersby (41:42)
listen to us Yeah, yeah. No, no. Six six hours much much too long for ⁓

Kevin Smith (41:46)
Ha ha ha.

Tom Anderson (41:50)
So Kevin, we're ⁓ for this. we're honored. Let's keep going. That's funny. something I I wanted to to talk to you a little bit about. I know we're getting kind of close to time there, ⁓ eight hours aside, that joke aside, but s so like

Kevin Smith (41:51)
I've blocked the rest of my day off or I thought this was going yeah, I thought this was gonna be a like ⁓ eight hour episode, marathon. I ordered lunch

Jeff Battersby (41:56)
Na shoot.

No, great.

What did you get us?

Mm-hmm.

Tom Anderson (42:19)
You know, we're seeing a lot of apps these days ⁓ that are kind of bolting on AI just to say they've got AI added to them. ⁓ some of them doing it in ways that are more thoughtful than others. But I think ⁓ if you could talk about your approach here, because this with Snipd is like AI is essential to the application. Like it's it's weaved throughout the the whole listener experience.

Kevin Smith (42:20)
Yeah.

Tom Anderson (42:48)
⁓ so could you talk to us about the I guess one we've we've not really talked well we've talked about DJ and a couple of other you know, the snips of course, but maybe touch a little bit on some of the other ⁓ tools or features in the app and how AI works with those.

Kevin Smith (43:07)
Yeah, ⁓ happy to do that. So so maybe we can start with when a new podcast episode comes out. So what what we actually do is we ⁓ transcribe the podcast, we identify what who are the speakers that are speaking in the podcast. we try to find a bio about them, ⁓ we try to find an image of them on the internet.

⁓ then we identify when exactly in the podcast they are speaking, such that we really have good data and knowing, okay, this statement was made by that person. ⁓ and afterwards we create chapters for the podcasts, we identify all of the books that get mentioned ⁓ on a podcast, such that we can show that to the user in a nice way. Also, for the books, we we go out into the internet, we research the book, we try to find the cover, the title, the author.

⁓ all in a in a nice way that it that it looks beautiful in the app.

yeah, so we do a lot of lot of processing. ⁓ we do a lot of processing to help the user know what this episode is about, and then for for our AI internally know ⁓ what this episode is about, who was on it, ⁓ and what w were they talking about.

When it comes to features afterwards, some of the features that we have, like the main feature we we spoke about, the snipping, the ability to to save this moment just by tapping your headphones. ⁓ but we also have other features with which one of them is is chat with episode. So you can actually find any moment within an episode or or look something up or create a custom ⁓ notes ⁓ for you just by chatting with the episode, similar to ChatGPT.

We have the DJ feature as as we mentioned. ⁓ So all of these features have AI incorporated in them in some kind of way. But we have at least always tried to not build a feature just because we can, just because AI is now around, but because we're actually trying to solve a user problem. ⁓

And I think that's that's often where some products go go wrong, especially maybe from like larger corporations. I think there was a moment maybe two years ago when AI came out or when ChatGPT came out and everyone felt like they had to have a chatbot. ⁓ everything just had to have a chat window somewhere. and yeah, we always tried to bring it back to the user problem.

The fun thing for us, also the great thing for us, is that especially if you compare us with other podcast apps, other podcast apps are very neutral with respect to the target audience that they're speaking to. So we've been very opinionated from the very start and saying, hey, we are the podcast app for people who listen to knowledge rich podcasts. if you listen to true crime podcasts, Snipd is not going to do much for you.

And that is totally fine. but given that we have this clear persona of person that we're that we're targeting and speaking to, it's actually becomes then very easy to know ⁓ what can be features that we can build because we know their problems, we know what they're trying to trying to do. and then AI is is a great help that now, or a great enabler ⁓ all of a lot of these yeah, to solve a lot of these problems, especially when it comes to

information and knowledge, right? Because that's one of the big things where these LLMs are extremely good at, ⁓ taking large amounts of information and and repurposing it, reformulating it or identifying things, extracting things, adding things. yeah. ⁓ did that answer the question?

Tom Anderson (47:05)
No, it did. Yeah.

Jeff Battersby (47:06)
Yeah, that's great.

Tom Anderson (47:07)
Yeah, it did. And you know, we were talking there ⁓ a couple of minutes ago about, you know, I was using ReadWise to do some things with the snips afterwards. I just wanted to go back to that to say y it's not required for that. Like you've got perfectly you know, suitable, you know, chat with your snips or review your snips built into the app too. So you don't have to have anything beyond that. Like you can go right into the app each day, talk to the snips you've already made, ask questions, go back to listen to the

The Snipd itself, the audio clip. So all that i is built into.

Jeff Battersby (47:41)
Alrighty, well Kevin, really grateful that you took the time to ⁓ talk with us this morning. this afternoon in Switzerland, I suspect. ⁓ but we're ⁓ we really appreciate it. I I know Tom's a real lover of the app and as I said ⁓ earlier, I ⁓ accidentally have a year long subscription. I So I'm gonna I'm gonna play with it, ⁓ I'll play with it more

And ⁓ and have the opportunity to use it. But I will say well designed app. ⁓ really ⁓ really well designed, really thoughtful. And I do like the features that you that you have in it. ⁓ so I'm I'm looking looking forward to giving it a year long run and seeing And and I think specifically for specific podcasts, like I I do listen to Criminal, which is a true crime podcast.

Kevin Smith (48:12)
Thank you.

That

Tom Anderson (48:25)
⁓

Jeff Battersby (48:34)
as you said, not not likely to get any ⁓ any insights other than how to be a criminal. ⁓

Kevin Smith (48:40)
I mean you

can use the chat to ask who did it. no, but the the maybe just just maybe one one side note for the listeners. Of of course you can you can use Snipd for all of all of the podcasts that are out there, even if the features are not ⁓ relevant for them. So you can listen to true crime podcasts. We do have all of the other standard features of a podcast app.

Jeff Battersby (48:44)
There, there you go. Yeah.

Kevin Smith (49:06)
⁓ another thing that I wanted to mention is that I'm actually very happy to to provide you guys with a ⁓ with a link that you can share with your audience, ⁓ with which ⁓ anyone can try out the premium version of Snipd for free for a month. so happy to happy to do that. And with respect to ⁓ you, Jeff, ⁓ one thing that I actually personally always find important to to do is, you know, you mentioned you have an accidental ⁓ one year subscription. So ⁓

Jeff Battersby (49:32)
Ha ha ha.

Tom Anderson (49:35)
Mm-hmm.

Kevin Smith (49:36)
Like for us it's very important. Like if someone actually gets a subscription accidentally, you know, they they they they got a trial and they forgot to cancel. It has happened to me. It's it's completely normal. ⁓ like if anyone's listening and that happened to you, you can just reach out to us, we will refund it. Like we don't want anyone to pay us any money if they actually don't want to. Like, ⁓ I think that's just a is a very sensible policy to have, ⁓ and prevents, you know.

Jeff Battersby (49:52)
Yeah.

Kevin Smith (50:05)
All of these kind of dark patterns in in in trying to get people to accidentally subscribe. So you can you can reach out.

Jeff Battersby (50:11)
Yeah, no. I I intend to use it. Yeah, I appreciate I appreciate that, Kevin. But I

I will I will make use of the subscription. But I I do appreciate that and that is good to know, right? That that you're not you're not just trying to, you know, grab people's pocket change. ⁓ so tell us ⁓ where we can find you, Kevin, on, you know, social media and places like that, as well as ⁓

Kevin Smith (50:21)
Great.

Jeff Battersby (50:39)
weekend. I presume you have a web page for the Snipd as well.

Kevin Smith (50:43)
Yeah, so me personally, ⁓ you can find me on on Twitter or X. my handle is Kevin Ben Smith. ⁓ so pretty straightforward. ⁓ to find Snipd, the actually the easiest thing is to go to the App Store or the Google Play Store and enter Snipd. ⁓ it is spelled S-N-I-P D. ⁓ that's the only tricky part, getting the spelling right.

⁓ otherwise you can check out our website first, ⁓ just Snipd.com snipd. and yeah, or just click on the link that that you guys ⁓ hopefully will provide in the show notes. Yeah. And I mean maybe on top. ⁓ we're like in general, we we love building this app, we're actively working on it. ⁓ we regularly come up with new features and things that we try.

Jeff Battersby (51:21)
Yeah, yeah, we'll have one in the show notes.

Tom Anderson (51:22)
Yeah, definitely.

Kevin Smith (51:38)
So if you if someone's listening and says, Hey, I have always had this feature request, ⁓ reach out to us. ⁓ and we have some already some cool things in the in the pipeline for the next ⁓ couple of months. So ⁓ yeah, give it a shot and see whether you like it.

Jeff Battersby (51:53)
Excellent. All righty. Tom I'll I'll give the closing notes and then you can shut us down. So Tom ⁓ Tom, where do we find you?

Tom Anderson (52:02)
Sounds good.

Tomfanderson dot com is the website. Tom Anderson on threads will do it.

Jeff Battersby (52:12)
Got

a nice newsletter in addition to this podcast. So and Tom's for those of you that have college bound kids has a ⁓ a really great website for how to select the right computer for your kids going to college. So that's ⁓ that's a a good thing. ⁓ that Tom, you've been doing that what, three, four years now?

Tom Anderson (52:34)
This is third year.

Jeff Battersby (52:35)
Okay. Yeah, it's a great it's a great resource for anybody that ⁓ that has a someone that's tripping off to college and ⁓ so highly recommend that. show music, psychokinetic, Celsius seven, always grateful to them and ⁓ Randall Martin Design for our show logo and ⁓ Tom, I think that's us. Kevin, thanks again, man. Really appreciate you taking the time to to come on with us and

Look forward to having people hear about you, your app, which is really quite good. And ⁓ thank you.

Kevin Smith (53:13)
Thanks for having me. I enjoyed it. It was fun.

Tom Anderson (53:16)
Yeah, enjoyed it. Really do appreciate it. And we'll definitely share that link for the trial to get some people to try it out. I think they'll they'll enjoy it. So that'll be good. All right, well that's it for this episode. Thank you everyone for hanging out with us for fifty four minutes and some change. And so until we talk to you again in a couple of weeks

Jeff Battersby (53:22)
Yeah, great idea.

⁓ d synthesized down

to three.

Tom Anderson (53:35)
Have a great rest of your day or your nights.

Jeff Battersby (53:38)
See ya.