Did this episode not have chapters on Metacast? It actually had. It had 102 chapters. Okay. It still does. So that was one of the episodes where our LLM, I think, went a little bit crazy. Hello and welcome to episode 68 of the Metacast Behind the Scenes podcast. I'm your host, Ilya Bezdilov. And I'm Arnab Deca. This is the podcast where we build our podcast app, Metacast, in public.
So we discuss all of the things that go on behind the scenes. Both Arnab and I are co-founders. Yeah, this is just an outlet for us to talk about how things are going and share some of the challenges and occasional wisdom. And if you are listening to this podcast in any app other than Metacast,
We wonder why you are listening to us at all. So you should just go and download Metacast right away from the App Store that is installed on your phone, Apple or Google. Install Metacast and then search for Metacast behind the scenes and restart this episode. And also share it with five people and give us a rating. Obviously five stars. We don't need any low ratings.
Did I say everything I wanted to say? Yeah, let's get started with the episode. I think it's going to be a hodgepodge of lots of updates. It's been about a month since our last episode, if I remember, right? So lots of updates. And a lot of light stuff like the shirt I'm wearing as well as some dark stuff like the shirt that Ilya is wearing. So we're going to give you some of the app updates we have added recently. Very quick summary.
as well as our web app. We've been focusing on the web app quite a lot in the last month or two. Marketing, what's happening with the app, like how many people we have, where we are in like the lifecycle stage, as well as what we are struggling with. And then I think some of the distractions that we have faced, like this is a startup spring right now in North America, at least. So we'll give you like the Q1 update of what things took our time. Cool. So let's do a quick round on the app update.
So what is the amazing feature that we've just launched, Arnav? So we started using LLMs just like every other app, I guess, or platform. But I think we've done a pretty good job so far. It seems like what we do is we already had transcripts front and center in the app. Now we take the transcript and we kind of create automatic chapters out of that. So some podcasts already have chapters provided by the creator or by the platform. If so, we reuse those so you get the authentic experience.
as the creator intended to. But if it doesn't come with it, which is the vast majority of podcasts don't have chapters in them, then we automatically create the chapters and... There's a pretty nice UX in the app right now where you can skip through, switch to other chapters. Sharing of a chapter is not there directly, but you can jump into a chapter and then share the timestamp from there.
This has been, I think, indispensable for me since we've been trying it out in beta on our devices, what, maybe like a month back, I want to say. Because uninteresting parts, it has become very easy to skip over them as well as ads. So I've been using it pretty heavily. I think Ilya too. And we are hearing some great things from our users too about this feature. So pretty exciting.
Yeah, and they all sort of echo the same thing you just said. You can skip to whatever you find more interesting in case people are struggling to visualize what chapters are. So if you play an episode in Metacast... for the episode essentially. So you have 10, 15, sometimes 102.
the different topics. I'm not kidding. Sometimes LLM just freak out and produce too many chapters. But yeah, you have a list of topics that I mentioned in the podcast and you can just tap it and start playing from the topic. And so far, I've found the most useful use case for this is when I don't want to hear the intro. Even if it's 30 seconds, I just started playing, I don't want to hear because I've heard it multiple times.
So I just go to the chapters, just click on the second chapter and then start hearing the exact thing I came there for. But I think to your point, skipping uninteresting parts, sometimes it can be tricky because you still want to be connected with the host, with the guests. You listen to the whole thing, right? But if you listen to something transactional, like news thing, or if somebody shares something with you.
You don't necessarily want to spend an hour listening to the whole thing. You just go to whatever the chapter number six that talks about tariffs or something or whatever is why you opened this thing. You listen for 10 minutes and you close it and you're done. And it actually helps you to be very efficient with that. Yeah, a lot of my recent listening, I guess for a lot of our listeners also has been about tariffs and the world in general.
And I also listen to a lot of sports stuff and sports stuff tends to be mixed. That same hour-long episode will talk about a lot of players or a lot of things. And sometimes I don't have the whole time. I just want to listen to like one bit. It's become very easy to find that bit. Yeah. So chapters impossible to miss in the app. Yeah, go try it out. So since it's a Q1 recap, what is the other awesome thing that we launched earlier this year?
So we launched the following playlist. You could follow podcasts earlier. before this feature, but new episodes will show up inside those podcasts. And we had like a small blue dot indicating that there is a new episode in there. All of that is still there, but we also now have a list of all the new episodes. since you last saw them in the app.
in the playlist thing. That has become the primary way I triage my input queue or the inbox at this point, I think. I quickly go through it, add things to my Listen Later playlist with a small quick plus thing there. And then that's it. Then dismiss the queue. And we have blog posts about chapters and podcast inbox, the following inbox. So you can go read more about it. We'll add it to the show notes.
Yeah. And in here, I'm going to insert a clip from a podcast that we accidentally discovered that mentions our app back when it was still in beta. We just found in the Google search console that that other podcast linked to us. So we saw an incoming link. So I listened to that segment. It's called, I think, Old Grumpy... Grumpy Old Guys? Yeah, something like that. Grumpy and old, they're men. And the guy was saying, actually, we will just play that, right?
Today, there was a new app that's coming out. It's on Android, and it's in beta on iOS, and it's called Metacast. And I downloaded the beta. I've been playing with it. And it seems pretty nice. I'm going to tell you right now, looking at the screenshot on it, I'm going to hate it because at the very top, it says continue listening and it's a swipe. It's a card swipe thing. I just want a fucking list. And I don't want the podcast.
underneath with the podcast art of who i'm following i want a fucking list of the episodes on my home page all righty then well may perhaps medicast isn't for you it is not for me i don't want transcripts Stop wasting your engineering time on fucking transcripts. Nobody wants transcripts. Oh, I have data that suggests otherwise, but I'll go fuck yourself.
So what he said is that he doesn't want to see the greed of podcasts. And I think it actually resonates with what you just said, Arnav, because you primarily go to the list of episodes. So maybe actually we should follow what they said and have the list of episodes on the home screen. And then we can reach out to them and see, hey, guys, we heard you. Why don't you try again and mention us again and link to us again?
Yeah, or at least as an option, right? Do you want to see the podcast covers or do you want to see the episodes right there? Yeah, the podcast covers are kind of nice, but I actually am not finding it useful anymore because I'm using the following playlist all the time. All right. And then the other bulk of our work, actually, you have become like this awesome web app Next.js coder, Ilya. So you've been going super fast on the web app last couple of months.
Yeah, you want to talk about some of the features you've added? Yeah, I want to mention the meta thing first. I think it was back in October. My life was just too chaotic. Everywhere, like with kids, family, work, like everything was just like, I felt like I did not have a stable ground underneath me.
And then doing marketing is very destabilizing for me because there are no clear results. You don't know what's working. It's sort of spray and pray. It's very randomizing, sort of short tasks, many of them. So I was like, I need to do something else. How about I just code for a month? Do nothing else.
I mean, do obviously the necessary things, but overall do nothing else but code all the time. And it also coincided with me buying the ChatGPT subscription for 20 bucks, whatever they call it, the ChatGPT Plus. And yeah, I just started revamping the web app and I went deep. It allowed me to go deep into one stack, which is Next.js that we use, and not be distracted.
And it felt so peaceful to actually be able to do that. And yeah, so I've been coding full time ever since. So I think for about half a year now, I've been probably doing coding 80% of the work time. And I'm also doing a lot of coding. Like I would drop my son off at like a sports thing and then I code for an hour while sitting on the bench. It's like that enjoyable.
I found the joy and I rediscovered the joy. I kind of rediscovered the 14-year-old myself when I was first discovering Pascal and coding simple things for DOS. A couple of things I'll add here. First of all is our goal from the beginning was the mobile app. But the mobile app is sort of starting to become pretty stable. We are now adding features that no other apps have, right? That's where more of our time is going versus earlier on where we were like catching up to other things.
for the most part. There are still things there are in other apps that are not there and we'll get there. Stranger things. Stranger things. And a little bit later in today's episode, we'll talk about the problem of distribution that we've started seeing. And we did our first funnel analysis of how many people are discovering the app, how many people are installing it, how many are using it, and all of that. We'll talk about the numbers in a little bit.
But the main takeaway from there we have known since October, November is we need to increase the number of people who discover the app. And the way that I think we've decided to go on that is SEO. I know a lot of people feel like the SEO train has left the station.
But we are starting to see better and better results. And that's when I think, Ilya, you and I decided we need to go like all in and make our stuff discoverable by Google. So for example, if you search for like your favorite podcast or an episode. One of the first few links in there, if it is metacast, That massively increases the incoming volume of new users into our website, into our brand, and eventually into our app, because we are seeing pretty good conversion rates after that step.
We'll talk about that. That kind of coincided with your thing of like, I was only doing marketing and we were trying to do everything, right? Like LinkedIn, you were posting on lots of things. We were not seeing much success with all of that. So you also decided to kind of start doing more long form like newsletter, blog posts, that sort of stuff. And the website, we started focusing on it. And I think that has been a great journey so far.
I am starting to see more and more of our links pop up when I'm like searching for an episode. And over time, that's only going to get better. Hopefully Google doesn't mess us up in that. The last bit I would say about the coding stuff is... I just want to be clear that you mentioned that you got the chat GPT plus subscription. We're not vibe coders. No, fuck no. So there was a world in which we could have gone like cursor or something and just write the whole web app.
I've now heard it from very close people that I trust, multiple times who don't know how to code, who have started doing these things. And they almost end up invariably in a mess within about a month. Gigantic code bases where they don't understand what's happening where. Stuff is duplicated all over the place. They're not named.
So I think early on when you started coding, you would code and then I will do a very thorough review and then we'll do back and forth, lots of changes and all that. And over time, I think you've gotten really good at it. I have like more and more of this. Yeah, I don't need to review. At this point, I think that actually segues into something else. We'll talk about our siloing stuff, right? Where I primarily work on the mobile app now. You primarily work on the web app.
And we don't really review each other's code as much as we used to before. We kind of look at the UX, we give feedback on that. But only times we would review is if you point out that, hey, I'm doing something with crypto here or auth here. Or if I tell you that there's like a tricky bit of like UX here, that's when I think we'll go deeper and look at each other's code. Otherwise, we don't. And by crypto, actually, I have to make a correction here, right? Clarification.
We don't mean cryptocurrency. We mean the cryptography package to create authorization signatures. The OG crypto, yeah. Yeah, yeah. You know what I could compare vibe coding with? Many years ago, like 15 years ago or so, I remember I rented a BMW 1 in Europe. We were going from Belgium to Netherlands, and I rented this car for a couple of days. I got on the road from the place where I rented it from.
And then I needed to back up, and I couldn't figure out how to back up. Basically, I was lucky that wherever I parked, I was still able to go straight. And I went back to the rental shop. I'm like, how the hell do I actually back up in this car?
And then apparently you have to like press because it was a manual stick shift. So you had to like... press it and then like go put it left and I go up so basically wherever the normal cars have the first gear that's where the BMW had the rear gear and I couldn't figure it out for the life of me but I think this is where the vibe coders end up
I have heard from many people now at this point that it's a mess invariably. It's great for prototyping and starting something quickly, I think. But yeah, it's not nowhere near sustainable code quality yet. Anyway, so this is a good segue to actually talk about the thing that we wanted to talk 10 minutes ago.
So web app updates, right? We have added summaries, AI-generated summaries. So like you were saying, we are now using more of the LLM stuff. So other than chapters, we also generate short summaries for episodes. They are not in the mobile app yet, but they are already in our backend and we started using them in the web app.
So basically now, if you go to an episode page that has a summary, the summary will be shown above the description. Or if you look at the podcast, it has a list of episodes. If there is a summary, it will be shown instead of the description. The cool thing there is that the episode description might be too long. So the summary helps to just quickly tell you what the episode is about.
It could be entirely absent. So there is a podcast called Notes on Work that I listen to. It never has any description. What I heard about this guy, because I discovered this podcast from Jackson's recommendation on his show. He doesn't even produce the episodes. He basically just like records and publishes. That's what he does. He has all set up. So basically he doesn't have to do any of that. So he doesn't bother. He only creates a title and that's it.
If you look at the list of his episodes, it's hard to decide what to listen to. But if you have summaries, actually, it helps. And actually, because you and I are seeing the summaries in our beta shapes inside our database and all that, I think we're starting to see how useful they are. Because a lot of podcasts, even though they may have a description,
If it's like a guest who's coming in, they will just say, we talked to this blah, blah, blah person. The summary would actually tell you what they talked about, which is what I care about before I decide to like listen to the episode or not. Exactly. The episode I'm going to mention at the end of this podcast, right? The Josh Waitzkin interview on the Joe Rogan podcast. The description is actually very long, but it says Josh Waitzkin is a chess champion in martial arts.
champion and then there is like a page of ads lex's episodes uh lex friedman are also like that like he just mentions who they are and what they do like he's talking to put in the president of russia Sure. But I need to know like, what are the things they're talking about? So, yeah.
So people have been asking for us to have access to the full transcript of the episode. And it's kind of clunky in the mobile app anyway. And also there are some technical reasons, but it's kind of hard to give that to the users. So what we've done, at least as a workaround for now, we've added transcripts to the web app.
And while we were doing that, we added the transcripts not just for the AI-generated ones that we have created. We are parsing all of the common formats like VTT and SRT and JSON and even plain text and HTML. and we display those on the website So there are currently tens of thousands of those transcripts on the web app. Hundreds of thousands. It's just over 100,000, I think.
So what I found it really useful for is I was listening to that interview with Josh Waitzkin, right? And I wanted to pick a quote from there. And I know where the quote is because I bookmarked it in the mobile app. So I went to the web app. I found that segment by just searching for it on a page. I actually tried copying those few segments, pasting the chat GPT. I said, like, can you come up with a quote from this? And it was a piece of shit. So I basically manually made the edit.
and turn it into a shareable piece that I shared on LinkedIn, which was pretty cool because you can now just go there and copy, paste and grab and like edit. And you can actually copy the whole transcript, paste into chat GPT and ask it questions if you want to, which you couldn't do with mobile app. we will be improving more and more there so we'll also make it deep linkable and you would be able to like start playing from a specific segment if you want to
But yeah, but this is the first step and I'm really enjoying it already. And I think this is also one more thing. The web app is already ahead of the app where we don't have support for these other creator provided transcripts in the app yet. We only support our own AI generated ones. That'll come soon, too. I don't remember where I read about this, but it was a long, long time ago, where I think somebody was saying that the web is a perfect innovation platform.
So if you have a web app and a mobile app, the web app is much easier to change. You don't have to deploy it, as in users don't have to install it, right? And just generally, it's easier to make changes. And also there's more real estate in terms of the screen size. And I feel like that's what we're doing now, like having gone through all of those creator transcripts.
I've gained an understanding of what we have there. Now, when we approach this on our backend and in mobile app, we don't start from scratch. We have already done the experimentation in a safe sort of web app environment. I feel like we are doing something similar with summaries too now. The web app gives us the window into what we can do on the mobile app. But in the web, it was like literally... three lines of code to add them. So on mobile app, it will be more elaborate.
Well, there was more to support like the VTT and all those other formats and all, but yeah. But there are packages that parse this, but I've looked through probably like at least 100 transcripts now of different formats. I've gotten a sense for quality at this point. And then I think you also massively changed the social sharing.
including our URLs look much nicer now, the open graph. If you take our URL for an episode and share it with iMessages or text messages or Slack or whatever with your friends, which you should, by the way, you should always be sharing Metacast URLs. Now you will see like a nice picture. The URL will be much nicer than the opaque kind of IDs we used to have.
And most importantly, the thing that I discovered is when I share our URLs with friends on Apple's iMessages, on text message basically with others, You can play the episode right inside the text message because we have an audio player in the web app now. which is pretty cool. Yeah, and those images that are displayed when sharing in messages or in social media, they used to be just, I think, dark background with the square artwork in it.
So what we've done is we use the artwork to create the backdrop for the cover. It kind of looks nice and blurry all in the same color palette. It just looks pleasant and good. It's something that you feel inclined to share with friends and tell them to download the app and give it a good rating. Which you should. Five stars. You've also changed your kind of approach to work and also the implementation of the blogging.
Yeah, I think that probably warrants a separate blog post. I think when we first started with our blog, we installed the package called OutStatic. which uses, I think, like 360 other packages. I remember this because when I installed it, we got rid of like 360 packages. Yeah, I think we all need to write a blog post about the technical parts of it. So you're saying a meta blog post about blogging on the Metacast website? Yes. Cool. All right.
So now we moved everything to sort of our homegrown markdown generation pipeline. So basically now to create a blog post, I just create a file in the file system. and it just creates a blog post. Like, I don't need any other extra things because we've already done the plumbing that makes it all work. And actually, that's something that eventually we might open source, I think. It's very simple, easy to use. I like it. At the same time, I again want to clarify that
You've made it easier to create blog posts, but that doesn't mean we are vibe blogging. I know a lot of people are starting to like use AI to create blog posts and all that. We are nowhere close to it. If anything, Ilya, I think we review and... edit the blog posts even more than the website code sometimes.
i guess i'll just have to clarify so without static what it did it had a very nice interface it looks like medium where you can like type in a very nice editor and paste images with like command v and it would store it as a file But it pushed directly to main. And I just couldn't live with that. I need to have my post and draft. I want to edit everything locally. I want to name the files the way I want to name them because it's good for SEO. And also I don't want the mess in my images folder.
So it's just like these little things. And I didn't want to write because of that. Now I feel like because it's easier, I just feel like I remove the friction for myself to actually do more writing. In the last few months, I've been writing a lot on LinkedIn and my posts have become shorter and shorter. And also I started writing them faster and faster. And I called it the writing slop. And I don't use AI to write those. I don't use LLMs. I type everything myself.
But I think what it led to is these are just quick bursts of thoughts with no really deep analysis, with no craft in it. And when I sat down to write a newsletter last week, I felt a bit of a sort of blank page fear, what do you call it, like a blank page, anxiety, writer's block. I'm like,
I've lost that skill to write long form. It atrophies. It's like if you don't speak a foreign language, you start to forget. If you don't drive stick shift, you can no longer use a clutch after a few years because you just forget it. So I know what you did when you felt like this. You went to LinkedIn and posted about this.
I went to LinkedIn and posted about it, yes. And then I actually made the commitment that I'll write something. And then I sat down and wrote and I went to a coffee shop and with no distractions, I actually wrote the post. And it was very satisfying to actually sit down and write a very long post. But bringing us back to marketing, just because my life has been a bit of a mess in the last year with all of the moves and it's just been challenging.
Our podcast was no longer bi-weekly. We didn't have a Builders Gonna Build episode for almost a year. The newsletter used to be bi-weekly. It became sort of bi-monthly at best. Yeah, we just moved into a kind of better place and I have my own sort of corner here. Kids don't distract me all the time. I think also the energy of the place is actually much better. I feel like the place where we lived for the last 10 months
I think there was something wrong with the energy in the place because like sucked out the energy out of me. I think that sort of led to that whole sort of collapse of my mental state. It's thinking about supernatural. But anyway, so I feel like I'm in good shape. So we already recorded an episode of Builder's Gonna Build. I've actually sent out a newsletter last week and I'm sending another one this week. And I'll just keep writing long form.
And I've also done a bunch of like blog posts for new features. I've created the changelog on our site. And I'm approaching all of that from the perspective, not just like writing for the writing sake. How can we create content? that is either indexable by Google and good for the long term. One of my blog posts about OPML. I wouldn't say it's doing like massive amounts of traffic, but it brings us some traffic.
But it's this kind of post that we want to invest more in. So yeah, like longer form reusable content, indexable. I think this is where we want to experiment more now and see how it pans out. This sort of stuff helps keep the engagement going with our long-term users as well as new users, right? New users see that you are doing all this stuff all the time.
And long-term users don't feel like we're going off into the sunset or something like that. Yeah. What I also found really helpful by having a blog post about chapters, for example, on the website. If I wrote about it on LinkedIn, it's gone the next day, pretty much. But like blog posts, like a year from now, I can still send it to someone or like mention it in some other newsletter or some other place.
Still, distribution is a challenge for us. Last week, when I sat down to write a Q1 review post, I ended up writing another blog post. I ended up writing about the distribution being a challenge for us. And we've tried multiple different things. And I still feel like that's the problem with marketing.
pray and pray until you find what works and for us we haven't really found something that we want to double down on that's been a real sort of i think emotional challenge do you want to share some of the numbers The listeners will get an idea of what the challenge is with distribution. Sure, yeah. So we have about 500 active users using the app. That includes free and paid. And we have, as of this recording, April 10th, we have 86 paying users. And about 15 to 20-ish in trial usually, yeah.
I think give or take, we have about 20 in trial all the time and about half of them convert to paid. So actually the conversion from trial to paid is pretty good. I think 50% is actually a pretty good conversion rate. What we need to do more of is like the paid users can only come through trials. There is no other way for paid users to come into being paid users.
So we need to extend higher in the funnel, right? So we need to extend the number of users who sign up for trial, but we also need to just drive more users into the app. And I think one of the bigger challenges that we have right now, obstacles, I would say. On the Apple App Store, if you search for a podcast app, We are somewhere number 45. I don't even know the number. You just keep like scrolling until your thumb starts to hurt. That's where we are.
You can only really find us by the name. So unless people search by the name, they will never be like, oh, what are the podcast apps available in the app store? And they just like install the first five or 10. They will never get to it. We don't know yet how to actually tackle that. Maybe it's because we don't have enough reviews. Actually, one thing we do know... is right now, whoever you're listening to this podcast, right?
You should try the app. If you like the app, leave us a five-star rating. Share it with people that you know who listen to podcasts. Ask them to do the same because I think our assumption right now is that we don't have enough ratings and Apple probably uses that heavily to show the order in which they show app.
Yeah, they must take this into account, the number of reviews, but also maybe the number of downloads, maybe the number of uninstalls, because they have the data. Yeah, it's sort of like a self-fulfilling cycle. Yeah, it's like if more people give us ratings, I guess, it's a rich get richer sort of flywheel.
The more people that download the app, the more people who will see and then download the app based on that. You know, it actually kind of made me think about that concept of cold starts. But I always thought about cold starts as a marketplace problem. So if you start a marketplace, like the Uber, for example, so the cold start problem is that you don't have enough drivers because you don't have enough customers. And the other way around, it's a vicious cycle there.
And you have to sort of kickstart one of the sites for the marketplace to start working. But I think here we also have a call-style problem because our distribution depends on other parties. So when we are in this sort of cold...
storage state, or cold turkey state, you know, you don't get the exposure that we need to be installed. Because I remember, I think it was somebody on Android. We asked this person how he discovered us, and he said, I just installed, like, whatever, 10 apps. So that happens on Google Play, but it doesn't happen on the Apple App Store. I feel like we are losing because of that. And I think our primary way that people find us right now is through Google search.
If you search for podcast app with transcripts, we are actually one of the ones way more higher up than in like, say, Apple App Store or Google Play Store right now. and that's what we need to get to in the stores also and you have a great blog post about it so we'll link it to the show notes if you want to dig deeper into the numbers and all you can go read that Yeah, and we also actually came at the plug here that we will not mind, well, we will actually greatly appreciate any kind of help.
If you read the blog post and you see our struggles and actually if you have ideas or if you have expertise or if you know someone who has expertise, just send us an email at team at metacast.app. Both of us get it. And actually just before we started recording this podcast, I talked to somebody we used to work with at Amazon who reached out to me and he's like, oh, I have some ideas for you.
He actually did have one really good idea. I told him that we've got, let's say, 100 customers because we can't segment our user base, our target audience by demographics. So it's not like we just can go to one place and target those users. It doesn't work that way. He's like, how can you turn those into 1,000? Can they help you find people like them?
It's the same approach as you have referrals in big companies. If you want to hire good software engineers, you ask your software engineers to recommend their friends, right? This just doubles down on your point. If you haven't shared Metacast with your friends yet, please do. But it also brings the point of like, maybe we should build something in the app to help people spread the word.
We also haven't added really like call outs to rate us and all that. There's a lot of stuff to add in the app. We should, yeah. So talking about things that we haven't been able to add, this is a great segue to the things that distracted us in Q1. Because we did, I think, produce a lot of great things, but we also had a few unforeseen time sinks that we spend a lot of time into these things. So let's quickly talk about those.
The first one, what did you notice, Arnab, when you looked at our charts in Google Cloud Console? We actually qualified for like a Google startup credit thingy because of which we don't pay $2,500 or something and capped at monthly at $300. So we never saw any invoice. Like in email, when we see the invoice from Google, it's just zero. They don't tell you about like, if you didn't have the credits, how much it would charge you.
But I think at some point, maybe in February or end of January, I noticed something somewhere we started looking. It was like $160 or something just in Firebase. And BigQuery was a big cost sink for us. So we were like, OK, I mean, at 100 paying users, 500, 600 active monthly users, if it's costing us 160, we're in a sea of bother if our scale increases very fast and we run out of these credits. So we decided that we need to tackle that.
We already knew that we had a lot of like things we were redundantly doing over and over again. And we knew that, okay, someday we'll spend some time and add more efficiency there. And so we went ahead and did that. I think that was what, about two, three weeks of work, I think, right?
I think so, yeah. There's been a lot of optimization going on there. But I think it cut down our costs at this point to at least 10x reduction, probably even more. And we'll know more in April once we see the cost. Yeah. We had to rebuild our BigQuery schema, BigQuery tables, backfill them for our biggest tables because we had a bunch of unnecessary data updates.
And because of that, the BigQuery just kept receiving those updates. And we had the data set that, just to query the data set, it was 100 gigabytes, 200, 300 gigabytes. It was an enormous scan. And now I think we reduce it by like 20 or 30 times. And the cost is like 20 or 30 times less. We got to the point where to run the one query cost $1 to us. Yeah. How many episodes do we have in our catalog now? Couple of million. Couple of million. So.
Each query that we used to run on that episodes table used to cost us one freaking dollar. It was insane. Which means like you make a typo in a query, you like run it one more time, two bucks, gone. We had like a scheduled daily extract. from that cable and probably more than one so it was like That and data structures itself, we changed quite a lot of stuff. We added to support your nice URLs. Anyway, we knew we had to do more work to change the data structure. So we did a lot of work.
This took two, three weeks. That was one of the major distractions in Q1, yeah. And then we also have a similar thing going on on the web app. because of the bots. So last episode, we talked about the attack of the bots. So now we actually get 300,000 requests per day from Meta, OpenAI, Anthropic. and then sort of bots disguising themselves. We already spent a lot of time like optimizing the web app so that we don't overspend on it. And I think now we are in sort of okay shape.
But we probably have to put Cloudflare in front of it. So we are not done with those sort of cost optimization things. I would rather be building features, but we also don't want to go broke because of all of those bots. And we have a blog post about this. We'll also link it in the show notes. The other thing that happens in Q1 to, I guess, all companies, at least in North America, is
Yeah, because we are a C-corp. I don't think we could have done anything differently. Well, I mean, if we were both in the US, Then we could have done a LLC and it would have been much simpler in terms of like the documentation and taxes and all that. But we could not because I'm in Canada. I think we've talked about this in the past in one of the episodes, but it would have complicated our taxes quite a lot if I was directly a LLC partner in a US LLC.
Yeah, we have to do what we have to do. This explains why we have the complicated corporate structure. We didn't have any other option, really. But this year, we actually did it ourselves. And we paid, I think in total, outside of all of the government fees, I think we paid only about $300 or so for the software, like TurboTax, which I had to use on AWS workspaces in a Windows machine because nobody has a Windows computer and they only have this product.
as a windows.exe file that you have to run on a Windows computer. So we actually wrote a blog post about that too. I kind of like that strategy where we can just point people to a blog post. You used TurboTax to prepare the tax preparation. You had to file what? Federal U.S. C-Corp and then Delaware? Yes, the Florida C-Corp. Okay. Two. And Florida C-Corp has to be sent by mail. So then Delaware franchise tax that you file on the Delaware Department of Treasury site or something. Three.
Then you have to file the Florida annual report, which is different from the corporate tax thing. Four. And then in Canada, we had to file the Canadian taxes. And obviously our personal taxes that also take into account the corporate thing. Yeah, just on the corporate thing, five so far. And then because we are a BC and Canada registered, we have a holdings company that owns part of the USC Corp.
We had to file paperwork like declarations and all that for that. So that's about seven different things we had to file, I think. Yeah. And all of them cost money. The Florida report cost, I think, 150 bucks. The Delaware thing cost, I think, 450 bucks. The federal tax is free, but then you have to pay for the software to do that. Well, the biggest issue with the tax was the mental tax, the mental toll, because I just kept thinking about, I have to do this.
And then I look at this form and the cost of doing this with a CPA. And I'm like, I really don't want to do it. Yeah. Last year, how much did it take us? Because we did it with a CPA, right? I think it was $1,800 plus all the fees. So I think we ended up paying like $3,000 or so. But I mean, at our scale... And with our extremely simple P&L, right? Like we know exactly what is coming into the company and what we are paying for.
This is not like a massive corporation with lots of employees and stuff like that. It just doesn't make sense that you pay $3,000 to like prepare the simple P&L and all of those balance sheet and all that. Last year, that was our biggest expense. So I'm glad we're done with it. I'm really glad we did it for like 10x cost reduction there too, I think, right? Like $3,000 to like $300 almost 10x reduction there.
and that we don't have to think about this for another 11 months. Yay. But also next year, since we already know how to do this, we could do it sooner and just forget about it in January. So we wanted to talk about this, how we started to silo more. I feel like it's an important... subject. And maybe we leave it for the next time. Sure. I don't know exactly what you want to talk about, but there's probably at least 15 minutes worth of talking.
in that area, if not more. Well, I guess just to prime this, right? Maybe it's a quick subject. Maybe I'm overthinking this. Once we started to silo more and code review less, I feel like we started working a lot more independently and do things a lot faster. And part of that was because Jenny left, because you and Jenny were mostly doing code reviews of each other, right? I wasn't really touching the mobile app that much. But you were looking over my web app code.
But now we just push code without review most of the time. Let's talk about this in the next episode because we did plant the seed for this in this episode already by talking about how we started siloing. But a big part of the code review, no code reviews, super fast or like very shallow code review approach.
is also because of the desperation, right? Distribution problems and desperation tied to it. So I think there's enough to talk about why we are headed that way, why the stars have aligned to like go that way. Yeah, let's talk about it in the next episode. We should call the next episode the Desperate Silo. Yeah, actually, I don't know when we will record the next episode, but the very brief sort of PSA is that I'll be coming to you and we will work in person again. When are you coming? May.
End of May. I'll be going first to the Wharton reunion in mid-May, and then I'll fly to you. Let's do another episode in a month from now, like beginning of May, remote. And then when we are here, we'll do more. Let's do an in-person one. Yeah. And also my family will most likely come over too. And my younger son really looks forward to meeting our CEO, who actually we have not heard from much on this podcast. He's upstairs. He mostly sleeps around. We should fire him.
If we had a department of government efficiency here, he would be fired like right now on the spot. He's a typical boomer. The joke is that his name is boomer. I've been talking a lot. So I'll pass the baton to you. So what have you been listening to? I mean, of course, the world has been crazy the last few weeks. I do subscribe. I do follow a lot of daily news and opinion podcasts, but I almost never go through them unless there's something crazy going on.
And something crazy has been going on over the last few weeks. I have been listening to a lot of like the opinions and different takes on these things, latest updates, because like I said earlier, podcasts are the main way I keep up with what's happening in the world. Not through TV or I don't read newspapers or things like that, either on the web or anywhere else. So I don't have any news apps and also podcast it is. And I spend about my morning like walk usually doing that.
So there's a lot of that already. I won't talk about all of that. Everybody knows what their favorite go-to daily kind of things are. The two things I will mention. First one is... Planet Money did about a 30-minute episode on basically coming back to the terrace. So most of the news and daily shows are what's happening with the tariffs, what's being paused or not paused or changes and retaliations and all that. And all of that you can get from wherever you want.
Planet Money did an episode on what trade deficits are in their typical style of like a 101 show of what actually is a trade deficit. And why does the US have so much of it? And what are the repercussions long-term and short-term of it? And I really liked that episode. is less of like, what is the right thing to do? What is the wrong thing to do? Rather than a more knowledge about like what it is and why does it matter? So that episode is literally called Do Trade Deficits Matter?
It's a good half an hour listen, I'd say. And then I landed on this Everything Everywhere, Here Daily podcast, I think about a month back. And I didn't think that I would enjoy this kind of podcast. It's a 15 minute every day. And they take an episode like Roman-Punic War. Maybe the next day it's like black holes. Maybe the next day is the ancient Pangaea that we had. That's sort of like every day it's a completely different thing. I didn't think I would enjoy it at all.
But I have been binging that podcast a lot because the quality, even though it's just 15 minutes, it's really good. I think the creator has figured out in 15 minutes, it's very hard to talk about like a giant topic. But he does a great job of not going too deep into each and every little thing, which is what I would have fallen into the trap if I was doing something like that.
At the same time, giving you enough information and most importantly, getting you excited to know more and more about them. The latest one that I heard actually two days ago was the Kuiper Belt and the Oort Cloud. So we all know our solar system. We have the nine planets and all that. And then I think most of us probably have a vague understanding of what the Quipper belt is or cloud beyond that is. But I didn't know the size of how much further these things are.
I kind of knew the comets are basically visiting us because of this, but didn't know the relationship between these things and the comets that regularly or irregularly visit us. So it was a great episode. That's actually very interesting. Yeah, I'd love to try it out.
While all of that stuff is going on, my approach is very different. I don't keep track of the news as they happen, almost never. In six months, I'll just read the summary and I'll be fine. I mean, unless it starts to affect me personally.
I usually just don't follow the news. You know, I get anxious when I do that and I just avoid that. So I try to listen to the content that's sort of evergreen. Speaking of which, you've always been making this example of Neil deGrasse Tyson with his Roman calendar thing.
as a way for Metacast to, like, how can you find this content? This is one of our North Stars. Like, how do we help find that? My thing was always Josh Waysking. He's a chess champion, martial arts champion. He had not been very public.
So Tim Ferriss had him twice on his podcast. The second episode of the Tim Ferriss podcast, back from, I think, 2013 or so, is with josh basically i think the first one was with arnold schwarzenegger maybe it was the other way around so it was like one of the very first episodes so here of this book called the art of learning which i read three times and i've not heard anything about weight skin ever again so i've been sort of
hungry for his stuff. And then just recently, he popped up on the Rogan podcast. I'm like... What? Really? Like three hours of interview with Josh Waitkin. So I immediately jumped on that. And what I found interesting... Tim Ferriss, you know, he has a very specific style. He would really try to dissect the meta things. So because Rogan is a martial artist, if that's the right term, the fighter, and Waitskins, he won the Tai Chi Push Hands competition in Taiwan.
20 years ago. And he also has been doing Jiu Jitsu and foil surfing. So for the first, I think, 40 minutes, maybe an hour, they just talk about different Jiu Jitsu fighters, their styles. how like they grapple. I don't understand this stuff, but I've still persevered. I've struggled through that. I was just like listening to it in the background because like eventually we'll get to the good part. Did this episode not have chapters on Metacast?
It actually had. It had 102 chapters. Okay. It still does. So that was one of the episodes where our LLM, I think, went a little bit crazy. It needs a bit fine-tuning. Because it was just not useful. It was like for every new paragraph you have a chapter. I really respect Wayskin for is learning. It's sort of meta skill of learning, how you teach kids. I just want to quote that, what he said about children. He said that the moment a kid...
starts thinking that he looks bad, that's when he gets locked up. He was talking about chess, I think, but it generalizes across everything. That's actually pretty commonly seen in all sports. the term we commonly say in sports they're in the zone
That means they don't care about whatever is happening outside their world. They don't care about who's looking at them and they're playing. And in that zone, you perform much better. As soon as you start noticing that other people are watching you or observing you. Yeah, that was just a great interview. And I'll probably re-listen to this because I have lots of bookmarks from that already. I mean, weight skin is just...
amazing. And if you're in the martial arts too, I mean, that would be just like a great episode to listen to. But if you're not in the martial arts, skip the first hour. And then I really enjoyed an interview on the Tim Ferriss show with Seth Godin. He's a marketer and I won't go into the details, but it was just a very good, inspiring interview and he's a really good speaker.
It's just one of those sort of easy content that makes me feel good. And I don't remember much, but I remember it made me feel good. Oh, and I got a special mention. I finally watched Lord of the Rings. the direct trilogy, and we started watching Hobbit with my son. And I'm actually enjoying it. I feel like I was stupid 20 years ago that I didn't watch it when it first came out.
Nice. Yeah, that's an evergreen thing. And I think the quality kind of lasted. Thankfully, they didn't make like seven sequels after it, which is when things start to go south. That's because Disney didn't buy them. Although Amazon, I think, did buy and they have started making some...
I don't know what the quality is or anything. But the three original movies, they're like beautiful. You can watch it again and again. Oh man, I also watched The Terminator. Both of the original ones, 1984 and I forgot when the second one was shot. These are very good movies. Talking about Terminator, taking a segue from there, Arnold's son, Patrick Schwarzenegger. He's now an actor. He is in the latest White Lotus season.
I don't know if you like White Lotus. I love that show. The last season just concluded. It was really good. All right, let's wrap this up. We'll see you in about a month, hopefully with lots of new updates in the web app and in the app. And we'll tell you all about it. We'll talk a little bit about the silo. Until then, Ilya, where can people find us? So people can find us at metacast.app. So the homepage has two huge links.
to get the app from the App Store or Google Play Store. So go download the app if you haven't already. That's also where you can rate it. And also that's where you can share the link with your friends and to recommend it to them. Right. Is it where they can rate it or is it where they must rate it? Five stars. Also, we are restarting our newsletter. I'll try to be more systematic about it. So metacastpodcast.com is where you can subscribe to our newsletter. And yeah, if you have
anything you want to say to us, it's hello at metacastpodcast.com. That's the email you may contact us on. That's a wrap. And that was a very long 30-minute episode. So yeah, we'll see you in a month. Time is relative. Yeah, I'll try to speak faster next time. Cool. Alright. Okay, bye.