Subscribe
Share
Share
Embed
Metacast is a podcast app for podcast lovers. We are building a powerful tool that helps you get the most out of podcasts. Auto-generated transcripts and chapters allow you to quickly skim episodes, search for content, and bookmark segments. Chapters, automated playlists, listening history & stats, and much more is coming.
On this podcast, co-founders of Metacast talk about building the company and the product.
Hello, hello, hello, and welcome to this November 7th edition of the Metacast Behind the Scenes podcast.
where we discuss how we build Metacast podcast app, share our business updates and what we worked on, what we are going to be shipping next, and maybe go on a couple of tangents too. So, all right.
Arnab, you want to kick us off with the business update? Yeah, so before we get to the, I think, the business update. Here goes the first tangent. First tangent.
Today, we won't get into this right now, but this is a hook for later in the episode, right? We had a couple of days where we were kind of mad at each other.
or maybe a day, but because we are in different time zones, it felt like two days. And it was all caused by cloud code.
in some way or indirectly so we'll talk about that in a little bit later today so stay tuned for that i'll have to make it stronger i was just pissed
I was just irritated. You were pissed, yeah. If you want to know what it is, stay tuned. Keep listening. Or you can use our chapters feature to jump straight there.
And then come back to the business update. Yeah, skip it altogether if all you want to hear about is the drama. So on the business update that you don't want to hear, we're creeping up to, I think, 250 paying subs. I think when we last recorded was...
Did we say about 200 or? I think so. Yeah. So it's 247 as of this morning. So very close to 250 paying subs. We crossed 1300 monthly actives for the first time.
And October was definitely slower than the kind of highs we were seeing in August and September. It's still better than like...
June, July and all, but it's definitely didn't sustain that kind of like crazy levels of, well, I won't say Mount Everest, but some other smaller mountains we were starting to see.
in between. So let's see what November has in store. While you were talking, I actually looked up our October 10th notes. We had 200 subscribers. So we grew by 20% in a month. It's not too bad.
Cool, cool. All right, so recently we released the 1.25 update for the app. You want to talk about that, Ilya? Yeah, that is...
Maybe one of my, I don't know if it's my most favorite update, but it is definitely one of my favoritest updates. We had not released anything for two months. I checked our release schedule. September, I think.
20th or so was our last release. I think it was more like September 15th or so, but yeah, almost two months. So because we've been working on something really, really cool, and that is three features that people have asked for.
And we personally have really anticipated. So it's listening history, continue listening and exporting transcripts. Let's take them one by one. So first of all, let's talk about listening history. And that is...
The thing that I've been waiting for for ages. So back in maybe 2017 or so, when the podcast app came out, the app that's called the podcast app, the iPhone app.
A super simple podcast player that since has evolved into more like a super app with all sorts of stuff in there, it had listening history. And that was probably the only reason, other than simplicity, that made me switch from Apple Podcasts.
to the podcast app and actually pay for it. So listening history has always been my thing. I track my books in Goodreads. My photos are meticulously organized in my hard drive.
But the podcasts have always been a fucking mess, right? And they have also been in Metacast up until two days ago when we shipped this new version.
Now you can go back and look at all of the episodes that you've listened or you have started. Basically anything you've touched, right? I've touched as in like started playing unless you removed it.
And yeah, that's really cool. And kind of the use case for me, first of all, it's just like ego boosting when I just go back and like I see how, oh my God, how many episodes I listened to in October. And that's because I listened to...
to like 25 two-minute Novales episodes. But you can go there and look at the bookmarks. It was a big issue for many users who would take bookmarks and then finish listening to an episode.
And poof, it's gone. You have to search for that podcast, for that episode to access your bookmarks. Now you can just go to the history. Well, unfortunately, right now you...
still cannot just see the bookmarks without starting to play the episode again. That's something that we have to improve on. But at least you can find it. And then if you start playing, you can go and access your bookmarks. For me personally, it's a game changer.
I think it will be for many of those ones who use the bookmarks feature. And yeah, I'm super excited about this. Yeah.
Just a little bit of, I think, technical details. So till now, since we launched last year, it's been about a year. All of your other information, like the podcast that you follow.
and your play like listen later following that stuff was stored on our like backend on our server side and the app would like kind of
fetch it from there and show it to you but listening history and what you're listening to that was only stored locally on your device we just didn't have time to like build all of this infrastructure yet right so we were only storing like a slim down local
data for that and so now when you install the 125 update if you haven't already to update it you will see a call out for like
something like let's go i'm forgetting the wording doesn't matter you tap that and that's when we'll figure out what's on your device and we'll kind of push it back to our servers and now this point on it's forever there unless you delete it of course um
And the other kind of effect of this is till now, because it was only locally on your device, if you ever uninstalled the app, that data would have been gone.
like gone forever there was no way to recover it right so now you can delete the app reinstall it or switch to another phone or sign out with a different account and sign back in and all of your history and continue listening all of that follows you
wherever you go. Only if you have an account. So if you use Metacast in the anonymous mode, do we call it anonymous mode? I think we do, right? Yeah, we don't call it anything, actually. So if you don't create an account, you're by default in anonymous mode, yeah.
Yeah, so in this case, the data is still stored in the cloud, but it is tied to an anonymous ID that we are unable to... If you sign out and you reach out to us and you're like...
I've lost all my data. Can you help me? We will not be able to. We will be absolutely helpless because we don't know who any of those anonymous users are, which is, I guess, great. But at the same time, if you want to...
be able to recover from installing a new phone, losing the phone, getting a new device, or switching between devices, maybe between an iPhone and an iPad. And in the future, let's say an iPhone or an Android phone and the web app.
you'll have to be able to sign in. And an anonymous user, by definition, cannot sign in on multiple devices.
So if you want to preserve the anonymity, you can create an account using Apple Relay if you're an iOS user. Then you'll create an account with an anonymous email that is provided by Apple.
I don't think Android has anything like that. No. If you really want to not be identifiable, you probably have to create a separate email account and sign up for that with Gmail. Overall, this feature is really powerful.
when you also have an account, because then all of your data is protected. Remind me, we don't charge for storing the history, right? But only premium users can access the entire history.
Yeah, free users get to see the latest 25 episodes just like in every other playlist. So it's the same as in playlists, yeah.
And then along with that, we shipped continue listening, which is also something we didn't have till now. What we had is the stuff that you have not completed and you have been listening to that was available at the top.
of the home screen as like a scrollable, horizontally scrollable.
piles it was kind of cumbersome but there was like no way to see that whole list anywhere there was no way to like mark it off there and stuff like that so now all of that is there uh there's a view all right at the top of the home screen or if you go to
playlist screen you can see the continue listening playlist there you can check episodes off of it which marks it as completed
And stuff automatically drops off if you haven't, like, if this is something that you picked up once, but you don't really care about it, after two weeks, it'll just disappear from it automatically.
But it will still stay in history, so it's not gone forever. It'll stay in history. Yeah, it just keeps the list clean. Oh, and I think we should also have mentioned that history is now a new tab.
So at the bottom of the app, there used to be four tabs, home, search playlist, and settings. Now there is also a history tab. And we do believe that...
History is such an important part of the nerdy podcasting experience that we made it a first-class citizen that you don't have to hunt for it. It's just right there. And continue listening like Arnab said in playlists.
Yeah, that's really cool. I think we've had a bunch of people tell us that they've accumulated too many episodes that they've started and never intend to finish or...
They don't want to continue listening to those. And there was, I think, one or two people who asked for...
If I understand correctly, they wanted to go to podcasts that they follow and sort of migrate their listening history to Metacast by marking a bunch of episodes as complete, even though they completed them on a different app. So that was...
pretty cumbersome before, because you had to start playing, then scrub to the end. Now you can just check, check, check, and...
Wait, even now, I think you'll have to add it to Listen Later first, right? Oh crap, you have to start playing it first. So, okay, so I think we've just done...
some discovery here but this is a really cool idea i like it maybe we should allow you to mark any episode as like listened yeah yeah maybe we should do it in the podcast view when you see a bunch of episodes and you can just mark them as complete.
Okay, so I see that this has actually already changed the way I use the app. So previously, as you probably know, people who are listening to it who have been using Metacast for a while, there is a following list, right?
Episodes that you follow, podcasts that you follow, new episodes from it automatically show up in there. And what I do every day, I follow like 150 podcasts.
Okay. So every day I have like maybe 10, 12 new episodes show up and I go through it in the morning, just like I do an email inbox, right? That was our whole idea for the following thing is it's an inbox. I quickly add like one or...
two of them to the listen later and then i clear that list off right now what i'm
What that means is the listen later list that I have is about 1100 items right now, ever growing, right? And I was reordering it, trying to figure out like, where do I want to start listening from today on my walks or drive or whatever.
what I'm doing is I'm using my continue listening part as the current, like, these are the things I want to complete.
Because anything I start listening to, if I'm done with the walk, I don't want to pick it up again. I just mark it as done. It's gone. So my continue listening, I actually have like three, four items at most. Like right now, I actually have only two.
Okay, so we've just shipped the feature and you already started sort of misusing it. Well, this is the use of it, right? It's like if you were in Netflix and you were like...
two, three things you're following at the same time, but something new, cool drops and you want to follow it, come back to this later. This is the use case for it.
fair enough for me because I start so many episodes and then I either get bored or my walk is over so I have so many unfinished episodes and sometimes I look at that list and I'm like whatever I'm not going to finish those now
they're just gone after two weeks. It's very nice. All right. So let's go to the third feature that we've shipped, which is the ability to export transcripts. I think that was probably the most popular request we've had.
over the last few months. So people have asked us if they can copy the transcript. And I think the use case have always been the same. It's like, I want to dump it into an LLM and do some more processing with it in ChatGPT or Cloud.
And I think somebody also asked about putting it into notes, but actually I don't know how useful it is to have.
transcript in the notes. And one person I think asked about translations or something related to it. Like maybe they're learning a language or like I do the slow news in Spanish podcast.
Right. So I'm learning Spanish. So this is a thing that goes through the daily highlight news, but in slow Spanish. So you can understand. So something like that. It's useful to take the whole transcript and put it in an LLM or something and say, translate this.
side by side yeah yeah we decided to go a very simple route we just added an export transcript button in the episode info page that when you when you tap that button
You have the choice of copying or saving it as a file. And the formatting of the transcript is actually a markdown.
So you'll have a link to the episode and you have all of the segments with timestamps, with speaker names, if speaker names are available. You also have the chapters as sort of level two headings, you know, as
sort of titles in Markdown. Also, if you have bookmarks, you will have a little star emoji on that segment. So actually, if you dump it into an LLM, you can say, oh, and I also bookmarked those segments with a star emoji. And you can instruct the LLM.
maybe to pay more attention to those or to extract context from around those bookmarks. Actually, I think maybe we should create a blog post with a prompt that really works in ChatGPT. I think we should do that.
The cool thing about Markdown is that even though it's like quote-unquote code,
I guess a typical person, if they see Markdown somewhere in GitHub, it will look like code. Monospace, like all those hash signs and all of that. But if you don't think of it as Markdown, if you are just presented with this...
It just looks like normal text with just a couple of hashes here and there. So it's a really nice format. Yeah, so the use cases for this, like we said, you can just dump it into ChatGPT and ask it to extract.
insights for you. And what I've done before, I mean, I had to go to our web app, copy the transcript, put it into the chat GPT. It was a nightmare. But I would do like, in this interview, they were talking about this, this and that.
So things that I thought were insightful for me personally. And I'm like, can you summarize those for me? And then it would give me something and I maybe saved it to notes or...
repurpose this into a LinkedIn post, rewrote those. But the thing is, you just forget about stuff that you talk about, but you may remember a specific thing, a specific queue, and you can use that to prompt LLM.
The other way I was using LLMs recently, which can also be used for transcripts. I have a repo with a bunch of markdown files and I have cloud code in that repo. And so you can just put the markdown file in there and you ask.
cloud code, whatever, like questions about that podcast or extract information. And I think actually that's very useful for the podcasters themselves. So if you have a podcast and you don't have your own transcript, you can get it from us.
And then you can maybe create social media content or blog posts. Actually, we do have a blog post. The last episode we recorded, or maybe two episodes back, the last long episode that we recorded, I think, I took the transcript.
And I asked the LLM to write a blog post from it. And this actually ties into the future topic that we will have on this episode. And it actually did a decent job.
creating a good structure and extracting good things from the transcript. I had to heavily edit it, but it was...
probably five times faster than me just sitting down and writing it. And the cool thing was also that I think I wrote this, like this blog post is like basically a replica of the podcast.
You can either listen to an audio or in text. I don't think I mentioned LLMs. I don't remember that. Yeah, we should do a blog post. It'll be nice, I think. I think that is a really nice use case. Yeah.
Yeah, especially for folks like us, right? I don't want to retype what we've talked about in this podcast, but with the help of LLM, you can actually do a good job here. Can I tell you a quick... tangent here, but I had a really interesting thing happen with this feature using ChatGPT.
So for this episode, in this podcast, we always do our recommendations at the end, right? So stay tuned for those also. That's like what Ilya and I have been listening to interesting things we have heard. So for this...
Now that we have the whole history, I went back and I found exactly the episode that I wanted to talk about. And I knew in that episode I had bookmarked something already, right? So I went there and I copied the transcript or exported the transcript.
And I remember that it was a $300 trillion fat fingering incident that I wanted to talk about, right? So the prompt that I gave to ChatGPT is, here's the transcript of an episode.
I have bookmarked some stuff with a star, like the icon that we use for bookmarks, and get me like a two-sentence summary of the 300 trillion fat fingering incident from it, right?
And.
When you paste like large content into ChatGPT or Claude, it doesn't paste it in line, right? It kind of shows like uploading and then shows a sidebar that, okay, this is ready. Like when you upload a PDF or something like that, right?
did that. So I was like, great, it's done. So then I ran it.
And it came out within a few milliseconds, like, great, here's your two-sentence summary. The Bank of England had a $300 trillion fat finger incident. I'm like, what?
That's nowhere close to Bank of England. I clearly remember it was like a crypto startup that did it, right? And then, you know what happened here?
When I pasted the transcript, it actually only pasted the file name. It didn't paste the content. Okay. Either that or ChatGPT.
took the context just from the file name and not the contents of it. I don't know which one of it happened. And Bank of England wasn't the file name? No.
It wasn't. It just made everything up. The only thing that it had data about was the $300 trillion fat fingering. And it made up like a two-sentence story about it.
And thankfully, I knew that that's not what happened. I was like, okay. Yeah. Anyway.
Okay, so use LLMs at your own peril. Yeah, when you're using LLMs, be careful. We do have a warning at the bottom of the transcript, as well as when we do AI-generated chapters, that there may be inaccuracies. So you may have...
AI generated errors in the transcript, and also AI generated errors in the chapters, and then you can compound that problem by uploading this to the JGPD. Yes. Yeah. All right. Okay. So...
I don't know if we should talk about this, but I guess why not? Our site was down for a day because of a dependency upgrade that we didn't catch. For episodes and podcasts, not the whole site. Yeah, not the whole site. But basically, it was...
interesting incident where we made a change. It was like a seemingly harmless version bump in one of the dependencies. Cloud code, right? No, that was...
That was NCU. You did. Just an NPM package, right? And locally, builds just fine. Everything works. Tests pass. GitHub Actions, builds just fine. Tests pass. Vercel, builds fine. No tests, because we run tests in GitHub Actions.
So it goes to prod and then somebody creates a post on Reddit.
With the link to a podcast episode, we opened that page. Oh, unrelated. Unrelated. Like they had a question about some other feature. They posted a link. In that question, they posted a link to an episode. Yeah, which they must have copied from the app.
because otherwise they wouldn't be able to because the site was down. And then we click the site, and it shows a 500 error, basically server error. And we start looking at only the pages that were cached by Cloudflare in our region where we are show up.
Everything else, 500 error. So panic mode going in. And then, yeah, apparently, for whatever reason, we used the package that was an overkill for the functionality that we need.
And then we had to cut it out, replace it with something much simpler. But what really puzzle does is that Vercel must be doing something different from whatever they do when you run it locally, right?
So it was an interesting lesson learned for us, which is to actually, when we do the deployment, test pages in preview, like all of the key pages, just in cases like that. Yeah. Okay.
For the next update, 126, some people have been asking us to increase their private podcast listening hours. So right now you can have transcripts and chapters for the first 10 hours every month if you're a premium.
user and they have been asking for like more and more so and they're happy to pay more so we're adding a pro tier to the app the first
few benefits of that pro tier would be increased limits for private podcasts. But I think there are some really cool features that Ilya and I have in our radar that will eventually make it into the pro tier also.
And we haven't done like an update of dependencies and all that. So that'll be there in this release.
Okay, so for our main or the bonus topic today was this frustration, irritation, pissingness caused by...
miscommunication and a little bit of cloud code magic. Ilya, why don't you start this off?
Or you want me to give my, because I guess I was the main protagonist in a way, so I can, yeah. No, I was the protagonist. You were the perpetrator. Perpetrator, yeah. I was the main perpetrator, so.
Yeah, by the way, so just full disclosure, no hard feelings, even, you know, the next day. After I apologize, I'm like, whatever. I'm fine. I've known him for 10 years. I know he means well. But in the moment...
I think I also wasn't feeling super well. So basically I write this blog post about the V1.25 and I make a factual mistake.
in one of the statements based on my misunderstanding. I think I said something like, you only see the last 30 episodes, all the episodes for the last 30 days.
instead of you see the last 25 episodes in Listening History. Just somehow, I mix it up from our discussions. You give me feedback on the PR, right? And I read the feedback. And it reads something like, this is wrong. Where did you get it from?
Yeah. Well, there were other comments in there, too. But this was, like, chronologically, this was the first one that I read, right? Oh, okay, okay. So then...
Then the next one I read is something like, is this written by an LLM? Actually, I forgot what exactly you wrote there, but it was something like, is it written by an LLM? And then I have a flashback.
Like, way before that, I was working on something in the web app. And I asked Cloud Code to update the readme.
And it hallucinated some updates in README. They were mostly correct, but they were also quite significantly wrong in some ways, right? But...
I didn't have time to proofread it. I'm like, you know, I've done the code. The code is tested. I really want to push this PR to Arnab so he can review. And I'm like, whatever, like last minute. I'm like standing in the doorway. I'm like, generate the readme. It generates the readme, and I'm like...
Screw it. I just added to the PR. Didn't proofread it. Big mistake. And then Arnab spends like half an hour reading that readme file until he realizes that this is all just like junk.
And he should not be spending time on it.
But he did review the whole thing, and I tell him, well, actually, I should have mentioned that I didn't proofread it, so you should have spent your time on this. But he wasted a lot of time for that. So I have a flashback to that episode.
of you wasting so much time and the comment reads are like almost exactly the same in a way, right? And I'm like, okay, so he thinks this was written by an LLM. But then I have this feeling like, but still he shouldn't be talking to me like that, right?
No, there was another part of this is over the last, I think, about a month back, maybe, you wrote up a very interesting, like a...
workflow for cloud code to kind of understand the way the voice and tone that we use in our blog posts, the way we write blog posts and all that. And you had basically laid out a system for writing blog posts. I thought you were
using that system, I didn't know that you were like, that was just like a one-time thing and you were not using it. So I was assuming... I used that thing to convert that podcast episode transcript.
to a blog post. So but I never use it for like writing things from scratch. Okay, so that was the other
misunderstanding on my side so i got this pr and um like i'm reading through it and feels like there's like the cost was not a thing here so that was my first thing is like maybe this is llm so that's when i added the first comment
But chronologically, maybe you didn't see that first. The first comment I added was, is this created by an LLM? Because there seems to be a lot of inaccuracies here. And then later, I added the 30-day thing. Because the day before...
we had discussed saying okay this is going to behave just like other playlists like 25 episodes whatever but in the original issue description
for the history feature that we had created on GitHub. What we had said is we're going to have 30 days limit. So what I thought is cloud code basically just read the issue description of what the PR, what the feature should be.
and just generated something out of it, right? Anyway, so yeah, I added comments in like,
So this is the other part that I want to have a discussion on is, is this rational or irrational irritation? Like when I'm reading stuff that is generated by cloud code, right? Even code, for example.
I use it all the time now to do something, but I'm never happy with the result.
I don't know if that's just me or a lot of people or most people, I don't know. I am almost never happy with it. I feel like it generates...
Maybe just in Flutter, because you have been happier with it on the React and Next.js side more than I think I'm on the Flutter side. I feel like it gets a JavaScript part maybe like right 60% of the time and with some fine tuning.
and good prompting, you can actually get it to behave really well.
Right. So I, in the Flutter repository, I have like an extensive cloud.md that says, these are my preferences. These are, this is the way I write code, right? And it always seems to generate like...
something that an intern would write. For example, write very verbose code.
You don't need these 15 things. It's just our mutation of variables. Okay, we're getting a bit technical. Adding comments, unnecessary comments on top of a method name. The method name is clear that it's like...
Like you don't need to add a comment there saying that this method exports the transcript, right? Just unnecessary junk. So I have an irrational, or maybe irrational, maybe irrational. I'm just not happy with the way these things generate.
content i am extremely happy when i'm like writing letters or a document or something like that but not code or like technical stuff like let's say in a blog post right
So I think that's where this started off with. Yeah.
But I think we had a few, we discussed it over. We were also not seeing, like we didn't have any face-to-face meetings. So all of this was either happening on like GitHub comments or like we were messaging on WhatsApp saying, hey, I replied to the PR.
yeah yeah and and i was just like so upset like yeah you know some other things i'm like
Whatever, I'll just let it slip, because I know I'll forget about it next morning, right? But this one I just couldn't let go, so we had an extensive chat in WhatsApp, and I think we cleared it there.
And then we also had a very lighthearted chat face to face afterwards with no hard feelings at all on either side. And I think what I came out of it
was that you're not writing the blog. Even the first draft is not like... LLM generated and I should not assume that it is so that's the and I think what we came out of it was let's just clearly mark when whatever stuff is LLM generated will mark it so that it's clear and
Yeah. Either we kind of like...
Not like if it's a readme and it maybe doesn't matter as much, we just skip it saying, OK, you take a second pass over it yourself and I won't review it. Or we say that, oh, this is important stuff and it's LLM generated. So let's actually.
double click into this a lot more and see if there are any inaccuracies and stuff. Yeah, and it raises an interesting meta point, right? It's one thing when you're...
unhappy with LLM doing stuff for you. But when you work with other people, you start to assume things.
right uh you not not just you in particular but you as a person right uh like sometimes i would get a comment under my linkedin post and i'm like
Sometimes it's clearly generated by an LLM, one of those, like, they have a specific structure. But sometimes the comment is like, hmm, it's not clear.
Is the person a bit overly polite and a bit weird? Or is it actually LLM, right? And then what do you do? Because if that's LLM, I don't want to waste my time responding to this. But if that's...
a real person, I probably should honor the comment and respond to it. But I don't know. And it's like, how do you deal with this, right? And I think with strangers, I think we are just coming to the...
do this thing where we just assume, if it looks like LLM, then we just think it's junk and we don't deal with it, right? But in the team, that is different.
We're starting to figure out the right etiquettes of using LLM in our work.
You and me, probably like thousands of other people. For example, like if I was in Amazon or a tech company right now reviewing code.
And I suspect that this is like a large part of a massive PR is coming. So the other thing that's happening, I think, is because you can get things done faster, you're generating a massive amount of work.
It creates review like headache for others to go through all that work. And if this work is not reviewed by yourself first, it's like LLM generated, you're basically generating something and asking others to validate it.
It's an easy path. It makes you more productive at the cost of like other. It makes you more productive by some definitions of like productivity at the cost of.
Lines of code. Lines of code, stuff, features that you're delivering, maybe, let's say, right? At the cost of bringing down the rest of your team or your reviewers, or increasing their frustration more and more. So I think that's where it's a good idea to...
be super clear and crisp whenever you're producing work with LLM that others are going to review is like, mark it there. And then I think we ourselves are figuring out how best to do this too.
And I think the good thing is we've known each other for over 10 years at this point. Yeah. But this could be in a team of strangers, right? Or people who barely know each other. And then...
You know, you and I, we can just talk about it. Yeah. Right? Like, even when we're frustrated, we can't talk about stuff. And I think maybe more junior folks or folks who don't know each other very well, they can just swallow this and it can accumulate.
frustration, kind of mutual frustration, right, lead to behavior that is destructive to the team. So, yeah, actually, I'm really looking forward to some kind of code of ethics. Right. I don't know, like LLM...
coding manifesto kind of thing emerging, right? Where it just provides some sort of guiding principles for teams that help them stay productive while also using those tools. Another example, let's say in the past.
Somebody sent me a PR and I love like...
I love having conversations in PRs and teaching people about like, OK, this is these are the best practices. This is think about this, think about that, that sort of stuff. Right. If I reply a very thoughtful comment, like I take 10 minutes to understand the code.
and write a reply and all that. And then later the person comes and says, oh, sorry, it was just LLM generated. I already know this. That feels really bad for both sides.
Yeah. Especially for the reviewer, right? So I think figure out how you can avoid those kind of scenarios is, yeah, the takeaway. Yeah. Cool. So let's go into our next and last section.
recommendations so you've been listening to some podcasts so yeah go ahead go ahead first okay so this one i just quickly did a glimpse of this earlier in the episode but
Tech Brew Ride Home podcast, October 16th episode. There's a segment on this. So there's this crypto stablecoin issuer called Paxos.
They did a fat fingering incident and they essentially printed, minted is what you would use in like crypto terms, stable coin, like coins worth 300.
trillion dollars okay this for context this is more than double the entire world's gdp and it costs them
$2.66 in transaction fees. And it was done within like five minutes or something. So they identified this mistake and kind of like burned the tokens down within about 20 minutes. But...
I think it's been raising a lot of questions in that crypto industry. Like if you can fat finger $300 trillion.
What's the whole point of like money anymore or crypto anymore at that point? I mean, can this be repeated? I mean, it was a fat fingering incident. So yeah, unless they've done some. So I mean, technically.
Probably not, right? Process-oriented? Sure, this one company could have done something that can prevent it. But also like $300 trillion is a mistake that you can easily see. What if it's like a $1 million?
Yeah, like some change here and there. And somebody's just siphoning, creating some extra money and taking it home, essentially, right? Yeah. Yeah, I would like to have those fat fingers myself.
I mean, if you're taking home $1 million just like that, you will end up having some fat fingers and the fat like all over the rest of your body. And a very quick shout out, Scott.
Hanselman and Mark. I want...
I don't know his last name. Sorry. I don't know how to pronounce his last name right now. Microsoft Azure CTO. They have a really good podcast called Scott and Mark Learn Something. And every episode is like they're learning something.
of weeks back they had a really good chat about ai productivity how it's helping like more senior engineers at the expense of junior engineers and what can junior engineers do to like keep up with this and learn from it and all that really good
On my end, I've not been listening to many podcasts. I'm going through the stage when I pick up an audiobook. I don't want to listen to anything else. I don't want anything to interfere with that.
Yeah, I've been binging on Neil Gaiman, so I've listened to the entire Sandman, whatever is available in Audible, the first three installments of it.
It was very nice to actually go through the whole thing the second time. And I also reread the entire comics collection also for the second time. And the cool thing, like when I read it the first time, like 2021,
some god would appear in the book, and there are multiple gods, like Egyptian gods, Norse gods, you know, Christian characters and stuff.
you just have to take it for face value. What I really like now with ChatGPT, I was just like, okay, so tell me about a bust, right? Like an Egyptian goddess. Or like give me a summary of all of the...
whatever, like Greek gods that were mentioned in his thing, right? And yeah, I was just, I was talking to my son and I...
He's too young to read this yet, but he's 13. But I told him about that part about Orpheus, which I think is also part of the Netflix thing. Oh, yeah, that was such a good series. Yeah, and he actually...
And he knows the entire story. So basically, Gaiman just retold the story.
with his own sort of twist and his own characters and stuff. But the core line of the story is basically from the Greek mythology. And I mean, I get goosebumps. So basically, like, he just... Oh, I didn't know that. He's just, like, so good at, like, mixing and matching.
those different myths together, myths and religion and sort of modern culture, culture of the 90s in this case. So I'm probably going to be doing it.
doing the third time of Gay Men's Thing. So, yeah, after I finished the segment, I started listening to American Gods.
It's my first time listening to that book. But you really get a deeper level of understanding when you actually know the mythology. And now having read The Sandman and having done some research for The Sandman, I understand a lot more of the American gods.
currents uh for like why they do certain things right like who they are um and um
Yeah, and kind of my next thing in line is The Mythos by Stephen Fry, which is a book about the Greek mythology. So basically now I really want to dive deep into the Greek mythology.
Egyptian mythology, Hindu mythology. So there was a Hindu character in American Gods too. And also maybe...
immersed a bit in the Bible and Quran too. But I'm probably not going to read the whole thing, but some kind of summary, right? Chad GPT, tell me about the Quran. Yeah, in one page.
Yes. But actually, you can use that. If you listen for cues and you can ask the right questions, you don't have to read the whole thing. But I'm just interested in the mythology too. Yeah.
Yeah, absolutely. I have a quick question for you and a recommendation for you. So American Gods, the audio book is that I think I have read that a long time back. I don't remember. Is that the theatrical version also?
Sandman or no? It's like a narrative. So it is a narrative format, but it's a 10th anniversary edition. And so we read the preface by Neil Gaiman. So what they've done for the audiobook.
It's a narrative, but it's a full cast production. It has probably at least 10 different voices in there, male, female. They probably omit some of the he said, she said, because you can just hear it by the voice, right?
But the cool thing about the audio production is that Gaiman talks about it in the intro. He took the unedited version of the book. So the original version was heavily edited for...
whatever reasons. And that's the book you probably read that got all of the awards and stuff. So that one is like, he took the unedited thing and then he made it, I think it's longer, I think it's 12,000 words longer or something.
And yeah, basically it's a different cut of the same story, which is kind of interesting. I'll try it too, yeah. I have not...
Like I have found that I find like the theatrical multiple voice audio books very jarring, but otherwise I love audio books, right? Because of the way I listen to it. So I'll give this a try to see how it goes. Yeah.
Yeah, this one doesn't have any sounds, any music like the Sandman. Okay. This one just has different voices. Right. Okay. I'll give it a try.
My recommendation, I think I talked about this a long time back on this podcast, but there is a really cool show on Netflix called Chaos, K-A-O-S. It's the Greek...
mythology retold for like modern times. So it's still the same gods, but they're kind of like how they are inside our society right now. So for example, Orpheus is a rock star. Okay.
And the river sticks that you cross after you're dead and get to the gates. Yeah, to get to Hades. And all of these are like real places.
Okay, so you actually swim through it. And there are the instructors, I'm forgetting what their names are, who help you swim through that. They're actually like putting on life vests and all that on you. So it's a really cool, fun story about...
Greek mythology in modern times. Give it a try. And it's only six episodes. And I think because they got so many big actors for it, like Jeff Goldbaum and all, I think they didn't renew it. So it's not... coming again so only one season but really good
You know, the reason why my son knew about Orpheus is because he read Percy Jackson. Ah, yes. Which I'm sure your daughter probably also read. My daughter also is a lot, like, she knows a lot more about, like, the Norse and Greek mythology and all.
that because of Percy Jackson and that sort of stuff than I knew at that time. She should try or anybody who's listening should maybe try The Norse Mythology by Neil Gaiman.
I have it, yeah. You have it? Yeah. So, did you listen or did you read? I read it, I think a while back, yeah. If you listen to it, actually, it's read by Neil Gaiman himself.
And I could not speed it up. It's like I listened to the One X and it's such a great narration with so much humor. And my son also, he read the book. We have a paperback.
Such a great book. Incredible, I think. I wish he did the same thing for all of the other mythologists he writes about. Right.
Okay, cool. Well, this was our long, short episode. So good luck editing it quickly and all that, Ilya. Yeah, but now we don't do a video and...
That's true. We also are not perfectionists anymore, so we don't cut pauses or stutters and stuff like that, unless it's super long. So it actually doesn't take...
too much effort to produce this episode. Although, you know what, maybe I should have mentioned to you first.
before talking about it on the podcast last three four people that have told me how they found out about me or how they researched me is youtube youtube yeah so like
I'm meeting somebody and they're like, oh yeah, I went to YouTube and searched for you. And like, I saw your podcast and I think we used to do.
maybe podcast earlier on on youtube so that's what they kind of oh i know about the google and amazon like story and all that so like okay maybe we should be putting these on youtube again no you know but you know what like if it
if it's like in a more professional context, right? So like they search your name and I'm pretty sure the first episode that will come up will be the episode with Jason Fried because it's the most popular one. So, and then you go, okay, so Arnab knows Jason Fried.
So, okay. You know, sometimes in discussions about Scrum, I would name drop our conversation with David Deish Thomas. And that's such a huge...
reputation sort of boosting thing. Right. Maybe we should actually, maybe we should remove all of those episodes where we don't have famous people. Right. Only the, yeah. Cool, cool.
Okay, cool. Alright. Okay. Bye. Alright.