Latest / Bad Decisions Podcast / #57 The Founder who AUTOMATED Content Creation using AI with Varun Mayya
Transcript
- 0:00We covered the entire Apple event 15 minutes after it went
- 0:02live with. 15 minutes you can check.
- 0:04It's on Instagram 1530 minutes within 1530 minutes videos live.
- 0:07Wow, this is Varun. He's a builder, founder, and
- 0:12content creator that's using AI to create content at scale.
- 0:16Varun has developed The Secret Sauce to handle more than 1000
- 0:20videos every month, generating millions of views across
- 0:23multiple platforms by integrating AI into his
- 0:26workflows. 15 minutes from idea to distribution.
- 0:31We gotta break this down. So I'll tell you how it works,
- 0:33right? So we get.
- 0:33Can you explain exactly what you're doing on social media and
- 0:36exactly how you're using these tools?
- 0:38It simply wasn't the most consistent content creator.
- 0:40So I decided let me replace myself, just me speaking and my
- 0:43face and my hand movement with AI.
- 0:46A lot of people have access to the tools that you have access
- 0:48to. What are the steps you took to
- 0:50fine tune? Let's talk about agent first.
- 0:51We. Fine-tuned so many agent models
- 0:53to the point where we discarded this one secret that we figured
- 0:56out. Go to ChatGPT and ask it for
- 0:59the. What do you see the future of
- 1:02editing be with AI coming in? I think.
- 1:04Footage is going to be solved at some point on the editing part,
- 1:07your medium of instruction will change.
- 1:09It will become English. You still need to tell the
- 1:12computer what to do. If someone comes to you as a
- 1:14content creator and looks for advice.
- 1:16I don't know if I would actually give the advice for somebody to
- 1:19be a content creator to make money off, but I feel like what
- 1:21you guys are doing and what a lot of other content creators
- 1:23will do, which is. You and your team have been
- 1:26successfully using artificial intelligence and technology to
- 1:30help you create content at scale.
- 1:32How have AI tools changed the way you and your team produce
- 1:36and distribute content? Varun, welcome to the Bad
- 1:40Decisions podcast. Is this a bad decision?
- 1:44Well, we shall find out by the end of the episode, and you have
- 1:47to tell us by the very end if it was a bad decision or not.
- 1:50So I'm going to rate this on our decision scale, but thank you so
- 1:52much for having me. You know, this has been a long
- 1:55time coming. I should have met you in Dubai
- 1:56the last time I was there. Yeah, We couldn't make it.
- 1:59It was like I was. I needed a few more.
- 2:01Hours. No, the truth is Varun ditched
- 2:04us. OK?
- 2:04We, we, we told him to meet us and we were waiting.
- 2:07Yeah, we're waiting. We're waiting for hours.
- 2:09And Varun ditched us. And so we, we forced him to come
- 2:12on the podcast. So that's that's the truth to
- 2:14that story. But next time I'm going to be
- 2:16there. Next time I'm going to be there
- 2:17for sure. Do you have any plans actually
- 2:19on coming to Dubai anytime soon? No, but you know, we'll figure
- 2:23it out. Something comes for the other
- 2:24like there might be a client there that might I might end up
- 2:27coming like I need an excuse to be there.
- 2:29Or or by the. End of this podcast we become
- 2:32best friends and you come and visit.
- 2:33US or we come up with a business idea that you have to come to
- 2:35Dubai to discuss it. That makes sense.
- 2:38Yeah, I'm sure one of those three things is going to happen.
- 2:40The chances are high. Well, Varun, first of all, we
- 2:44appreciate your time. Thank you so much for coming
- 2:46here with us and making this bad decision to join the podcast.
- 2:50We have been wanting to speak with you for a very long time.
- 2:53We're big, you know, big proponents of content creation
- 2:57technology and the advancements in technology, and we cover that
- 3:01a lot on our channel. But you seem to do exactly the
- 3:04same thing at scale and with such incredible quality.
- 3:08So first of all, we just want to say congratulations because we
- 3:11see your face on Instagram every single day.
- 3:14You and your team are doing such an incredible job at creating
- 3:17content and updating the world on the latest in technology and
- 3:20official intelligence video games, for God's sake.
- 3:23And even when whatever happened with Lebanon recently, which was
- 3:26really sad, you guys covered it, but also the perspective you
- 3:30guys had on all the news is the craziest part about it.
- 3:34So first of all, congratulations on the crazy content that you
- 3:38guys are creating. But second of all, something
- 3:40that I'd like to talk about and Farhad and I have been
- 3:43discussing is how you and your team have been successfully
- 3:47using artificial intelligence and technology to help you
- 3:51create content at scale at amazing qualities.
- 3:54So I guess the first topic that we want to discuss today is the
- 3:58impact of AI on content creation.
- 4:01Varun, how have AI tools changed the way you and your team
- 4:05produce and distribute content? So.
- 4:11I'm actually a very inconsistent content creator as a person
- 4:14because I feel like life just, you know, it's life like a lot
- 4:16of people, you know this this one question that people used to
- 4:19ask me, right, Which is if it's short form content, if it's like
- 4:22one minute a day, that's 7 minutes a week.
- 4:25Why can't you donate 7 minutes a week and do it consistently
- 4:28forever? But no one manages.
- 4:30Almost nobody puts out a video every single day, right?
- 4:33It's it's effort because it's context with sometimes you're
- 4:35travelling and you just don't have the time to go to the
- 4:37studio and shoot. It's almost like you have a
- 4:39tether to the studio and you can't go too far away from it if
- 4:42you want to shoot content everyday.
- 4:44Plus you're going to come up with new topics, new ideas,
- 4:45you're going to do the research yourself, you're going to write
- 4:47yourself. It is very painful.
- 4:50So I think a couple of years ago I realized, I realized from
- 4:52other content creators actually that the secret to winning on
- 4:56YouTube and Instagram are consistency sounds odd.
- 4:59You'll be like, wait, it shouldn't be quality of content.
- 5:01Actually, not really. It's consistency because you
- 5:05can't tell which video is going to go viral.
- 5:07Like today we have some signals, but you can't tell and you gain
- 5:10subscribers based on videos that do really well, right?
- 5:14So unless you try thirty shots you don't like per month, you
- 5:18don't know which are going to enter the hoop, right?
- 5:21And which ones are actually going to add 20,000 subscribers
- 5:24to you this month? So it's that it's a game of
- 5:26that. And the lesser number of shots
- 5:28on goal, the lower the likelihood that you're, you
- 5:31know, you have these videos that get, we don't call them vital,
- 5:33We call them super reach, which is views more than the size of
- 5:36the, you know, the, the follow count that we have.
- 5:39So we said we need to be consistent.
- 5:41And I, I know this is the stupid way to say this, but I think AI
- 5:45really made us consistent because the minute you have an
- 5:47avatar and we use agent for the avatar, we, we fine-tuned so
- 5:50many agent models to the point where we discarded so many of
- 5:53them. And then we finally kept the
- 5:54good ones. And I remember before Hagen, we
- 5:57tried this tool called Wave to Lip.
- 6:00It's an open source model kind of St. stitch lips on you to
- 6:04sync it to any audio you want. It wasn't great, but when we saw
- 6:07it, we were like, yo, in one year I want to be using this to
- 6:10replace myself. I put a tweet about it February
- 6:122023. I put a tweet about this 1.5
- 6:14years ago. And we even tried a video on my
- 6:17Instagram with Wave to Lip. People could tell we were like,
- 6:21we'll just wait. But when this happens, when this
- 6:23opportunity is here, we have to take advantage of it.
- 6:25And then when Hagen came out, it's like actually before Hagen
- 6:28Synthesia came out, but it didn't really allow us to make
- 6:29our own avatars. We tried other things, but it
- 6:31looked kind of fake. But Hagen was great, especially
- 6:34with fine tuning. We fine-tuned so many models.
- 6:36It was literally brute force that we got the right one,
- 6:39right? And today we know a lot of
- 6:40things, right? For example, if you're moving
- 6:42your, you can't move your head more than like 45° on either
- 6:45side. A lot of things, right?
- 6:46You can't blink too much, you can't blink too little.
- 6:49You have to speak in a very excited tone.
- 6:50You got to like speak in an even more excited tone than normal
- 6:53because AI sort of brings you, you know, down.
- 6:55But the hard part was still audio.
- 6:57Because if you see some of the old videos we did that were
- 6:59automated, the audio was still bad.
- 7:01And we were using our own model at some point.
- 7:03Like the best model we found was RVC, which is retrieval voice
- 7:05cloning, where we were able to take somebody else's voice, like
- 7:08somebody on my team, and sort of replicate my voice.
- 7:10That was kind of good, but it still required effort from
- 7:12somebody to somebody. Somebody else was tethered into
- 7:15me, right? And I had to be sort of the same
- 7:17person. Otherwise, the tone is slightly
- 7:18different. So then 11 labs came out of the
- 7:21pretty good model. And even now people struggle
- 7:23with the 11 labs because they haven't figured out this one
- 7:26secret that we've figured out. I'm sorry, I'm just waiting
- 7:28everybody, but. This is the one secret.
- 7:31It's a very simple thing that we learnt with audio models, which
- 7:34is it wasn't getting all our words right.
- 7:37There's this one video where I said the word, but the AI model
- 7:39had to say the word devotee, right?
- 7:41Which is like this Indian thing where you know you're a devotee
- 7:45of something else and it just did it really badly.
- 7:47It said devotee or something like that.
- 7:49So we realized that it's all about coverage, right?
- 7:52It's all about can you cover most of the used words in the
- 7:55English language. So we go to chat GPD, it's very
- 7:57simple. It's a secret.
- 7:58Go to chat GPD and it's very simple.
- 8:01It's not really a secret. I don't, but I don't see anyone
- 8:03talking about it, so I'll say it anyway.
- 8:05You charge GPD and you ask it to output the list of all the
- 8:08commonly used words in the English language or all the
- 8:11commonly used words on YouTube. It'll send you a list.
- 8:14Just read them out and put them back into 11 laps.
- 8:17That way you have coverage over 99% of what you'll say anyway.
- 8:20And for certain specific words we sometimes do a top up on.
- 8:24We know that OK, these are the 10 words you outside of the
- 8:27range that you're going to be talking about in this week.
- 8:29So I do a quick, you know, read out of those words for the team.
- 8:32And that's why the model is getting better over time.
- 8:35It's genius. Wow.
- 8:36And for anybody who doesn't know, I'll just do a quick brief
- 8:39explanation about 11 Labs. So 11 Labs is an AI tool that
- 8:42can generate text to audio. It can do voice cloning.
- 8:46Basically is a library that you can create voice through it.
- 8:49We tried it, we recreated Mr. Beast voice using using 11 Labs.
- 8:53We tried Gandalf, we even tried our voices one year ago when it
- 8:56wasn't that great. But recently I think it improved
- 8:58a lot. Hagen is another AI tool which
- 9:01creates instant avatars. Basically you can create your
- 9:05life like avatar which can speak on behalf of you.
- 9:08Yes. And I think for context, Varun,
- 9:10because I know you, you're so excited, you went right into how
- 9:13you guys are using it. Can you explain exactly what
- 9:15you're doing on social media and and exactly how you're using
- 9:19these tools for people that have never, never ever heard of these
- 9:22tools at all? Sure.
- 9:23So what we do is, like I said, I simply wasn't the most
- 9:26consistent content creator. So I decided let me replace
- 9:29myself, the presenter part, just me speaking and my face and my
- 9:33hand movements and stuff with AI, which is what happened.
- 9:36And then we have an editing team.
- 9:37So the editing team actually edits it.
- 9:39We also have a scripting and research team that sort of does
- 9:41the scripting, does the research, where I give some of
- 9:43my inputs as well. And then it comes together as a
- 9:47nice little one minute video. And we can do it fast.
- 9:49We can do it in like 15 minutes after some news goes live.
- 9:52We covered the entire Apple event like 15 minutes after it
- 9:54went live with one full video 15.
- 9:57Minutes. Yeah, you can check it's on
- 9:58Instagram 1530 minutes within 1530 minutes.
- 10:02That is faster than any sort of media speed I've ever seen in my
- 10:06life. So you're scripting.
- 10:07So let's start the idea, right? The idea of course, you guys
- 10:10know Apple Keynote is happening, there will be something to talk
- 10:13about. There is some preparation, yes.
- 10:15So the scripting theme, are they using AI tools to generate the
- 10:19script? Yeah, so I'll tell you how it
- 10:21works, right? So we get, let's say 4 bullet
- 10:24points, right? We have, we have an actual
- 10:26analyst going through the, you know, sort of video the the
- 10:28actual keynote and saying here are the four bullet points.
- 10:31Here are the things that we think are important from the
- 10:33keynote. Then I add my insights to this.
- 10:35So I add two more bullet points, right?
- 10:37Sometimes I send a voice note, sometimes it actually text.
- 10:39I'm like, hey, the implication of this is A, the implication of
- 10:42this is B. We take that entire thing and we
- 10:45ask AI to expand those bullet points right into cohesive
- 10:49script, right? Teacher, may I ask a question?
- 10:52Yes, in. The middle in the middle of your
- 10:53instruction, have you trained ChatGPT or any AI to also
- 10:57understand your way of talking about points?
- 11:00OK. So it's it's your own sort of
- 11:02opinions in the middle of those key points, yes, but.
- 11:05What we have done and what we're working on, we've tried a couple
- 11:07of videos of this in Claude. I wrote a book a long time ago,
- 11:10so we just fed that as context to Claude and we said talk like
- 11:12this first. It's not great, though.
- 11:15That's not great. It's not prime time yet, but I
- 11:17assume with one more leap in models it is possible.
- 11:19We think with O1, when it allows us to feed in context, this
- 11:23should be solved. So I think, I mean, our entire
- 11:27job now is to give those two insights.
- 11:29This thing has already learnt from my writing, which in the
- 11:32future we'll sort of put out, but for now it's ChatGPT or
- 11:34cloud. And then we have a script that
- 11:36comes out. Then we push the script to 11
- 11:39labs, which has been trained on, you know, my model, I mean my
- 11:43data. And then it goes into Hagen,
- 11:45sort of produces it. Then the editors do the job
- 11:47manually. The overlay of videos.
- 11:50Overlay the editing subtitles some.
- 11:53It's all AI assisted like all our B rolls.
- 11:55For example, we do flux, Lora plus cling.
- 12:00Right, Flux. Flux.
- 12:01I know what flux is. That's the image generation
- 12:03model that Grok is also using at this point, which creates these
- 12:07really photorealistic images. But you mentioned another thing
- 12:10I have not heard. Of yeah.
- 12:11So flux Loras are when you can train flux on your images.
- 12:16So now I can prompt flux and say generate modern as a cowboy and
- 12:20it'll actually make me look like a cowboy.
- 12:22And then we put that in cling, which is another image to video
- 12:25generator. We also have you sometimes use
- 12:27it on way. So image to video and that gives
- 12:29you B roll with me in the B roll.
- 12:31So if I talk about Mars, I can put me on Mars like smelling the
- 12:34flowers. So that's B roll.
- 12:36OK, so that's B roll. And then we have a subscription
- 12:38to Epidemic Music and a bunch of like sound services.
- 12:41So we're able to use sound, background music, all of it.
- 12:44We have like actually, we don't innovate too much on the
- 12:46background music. It's like sometimes we use the
- 12:48valid in background music, for example, right?
- 12:50So we're like anything that we feel is usable and relatable, we
- 12:53just use it. We put it all together and yeah,
- 12:57it's 15 minutes. Like we have 7 editors on the
- 12:59team at this point, right? Like on my team.
- 13:01So that's for short form. And yeah, we put out a video
- 13:05every day and over the last, like when Ricardo made that
- 13:09video, Not Ricardo, sorry. When RP and Roberto made that
- 13:12video, I think we were doing 25 million views a month.
- 13:15Now we're doing like close to 60 mil.
- 13:17What that video? That video is not made a long
- 13:22time ago. That's just like a couple of
- 13:24weeks ago. Yeah, a few weeks ago.
- 13:25And my Instagram just updated and showed me like 50 mil and
- 13:28YouTube were doing about 10 mil. We also do long long form videos
- 13:31and with long form we recently tried.
- 13:33We've always done it like with me in front of a camera and
- 13:36doing stuff because we feel like people will tell.
- 13:38But we made a video on Woocom, yes.
- 13:40Yes, yes, yes. And we were debating.
- 13:43Yeah, debating. Like is it moron or it's not?
- 13:45It's not you, and let me tell you something, as people who use
- 13:48Unreal Engine on a daily basis and create 3D avatars and know
- 13:52exactly what is real and what is not, we had a hard time figuring
- 13:55that shit. Out.
- 13:56But it's not you, right? It wasn't.
- 13:58It wasn't. Yeah, but the hand just sharing
- 14:00that video was very complex. So at one point I saw you were
- 14:03doing. This.
- 14:04Yeah, yeah, yeah. The reason you remember it, the
- 14:08reason you remember it? Because that frame repeated
- 14:10twice. That was our back.
- 14:11No, we paid a lot of attention. We did it.
- 14:14I mean, we watched the video and then after we were looking for
- 14:16it. We were looking.
- 14:17We were looking for it, I don't think.
- 14:18Anybody would realize that. Honestly, I don't think anybody
- 14:21would realize that. So, yeah, I mean, that's, we did
- 14:24a long found book on it, did the best on our channel, did like
- 14:26300K views. So, you know, AI avatars have
- 14:30brought me a lot of humility and like, I'm not really the useful
- 14:32person. It's the content that actually
- 14:34drives traction. I'm just like the medium.
- 14:37But it's, it's incredible. Like being consistent has been
- 14:40awesome. And it truly being consistent is
- 14:44a secret to being on YouTube or, or it's time, whether you're
- 14:46doing it with AI, whether you're doing it personally.
- 14:48By the way, AI avatars don't work for everybody else, right?
- 14:51Like, it works for me because I'm in the AI space and it's OK
- 14:54if I use an AI avatar because people are like, OK, get it.
- 14:56But if let's say you're finance creating, you start using AI, it
- 14:58comes across a little bit awkward until the model gets so
- 15:01good. But the minute people know, they
- 15:03get upset with me, I'm breaking news, right?
- 15:07So it's like, hey, I want to know what happened to the iPhone
- 15:09event. This is the fastest source to
- 15:11get that information from. I'm doing what the media
- 15:13technically does right, but doing it much, much faster on
- 15:16video. I don't think anyone's too many
- 15:18people are reading texts anymore.
- 15:19So I found that opportunity in video and I think it's doing
- 15:21well. I will, I will say one thing.
- 15:24First of all, we got to break this down.
- 15:25This is a very important conversation here and I think a
- 15:28lot of people watching and listening are hearing about this
- 15:31for the first time. I hope they are at least because
- 15:33it's very exciting. First of all, I want to say
- 15:36Congrats on you and your team because you guys are doing a
- 15:38phenomenal job. Once again, this is something
- 15:40I've never seen or heard of before and it's extremely smart.
- 15:43And we can talk about all the ethical problems and all that,
- 15:47and I'd love to discuss that. But what really matters to me is
- 15:51the value that you're giving people.
- 15:52So in that Vukong video, I don't care if it was you, if it was 10
- 15:56other people or 20 other people, I'm getting the same value out
- 15:59of that video. And you're talking about
- 16:01something that I'm interested in.
- 16:02So as long as I can get that value as fast as part as
- 16:05possible, in the best way possible, as your audience, I
- 16:08think the job is done. And for the Groupon video, it
- 16:11had your take and your perspective in it.
- 16:14I think that made it so valuable.
- 16:16It wasn't. I couldn't get the same news
- 16:18from other outlets because you were talking about how why India
- 16:21doesn't have AAA games. So it was it was your
- 16:24perspective. Yeah, so it's still like.
- 16:27I just wanted to shorten time from thought to output.
- 16:31And I think with AI avatars now I'm just able to do thought,
- 16:34which actually frees my time up because I can read now books
- 16:36like I'm reading this book about how Singapore was built, right?
- 16:39It's called From third World to 1st by, you know, the guy that
- 16:42built Singapore over decades and decades.
- 16:45And I get to read that and I get to compress it in a good video
- 16:47for people. By the way, here's an opinion.
- 16:49I think more people watch AI generated content in the future
- 16:52than real content. Here's why.
- 16:55OK, even now they prefer AI generated content.
- 16:56They don't. They don't know why, but let me
- 16:58let me articulate it now that we actually have the evidence.
- 17:01When you do AI generated content, you always start from
- 17:04the script, OK? When you do real content, it's
- 17:07you in front of a camera talking, mumbling, you know,
- 17:10going off track sometimes, you know, doing a lot of things that
- 17:13are not related to the video. And some people tend to like
- 17:16that, but most people, 90% of people just want you to be on
- 17:19point and run through everything.
- 17:20So when you write the script, you can be typed with the
- 17:22script, you can just be like, here are the five things that
- 17:24work. Here are the five things that
- 17:25don't work. You can just be very, very tight
- 17:26with the script. And because you're working at
- 17:28the script level, there's no arm, there's no butt, nothing.
- 17:30And then when you put that into Hagen or any AI avatar
- 17:33generator, you don't have the burden of now, you know, having
- 17:37footage to cut out. Like sometimes my editors used
- 17:39to cut out excess things that I used to say.
- 17:41So you that Vukong video, if I'd done in real life would have
- 17:44been a 20 minute video because I'd have ranted, I'd have gone
- 17:46on tangent directions like I'm doing right now, but with AI and
- 17:50because I have to write it or because my team and me had to
- 17:53write it and we have a really cool game that's helping me
- 17:55right now. It gets compressed.
- 17:58So humans prefer compressed information with great graphics,
- 18:03great narrative, great storytelling because people just
- 18:06don't have time. They don't have the attention
- 18:07span for a 20 minute video. So AI forces you to be tight
- 18:10because you're starting from a script and therefore more people
- 18:12watch AI generated content because you're getting the point
- 18:14faster. So the same reason short short
- 18:16form content works in for information, just rate of
- 18:19insights per second is higher. Way higher.
- 18:24I mean, 15 minutes is still baffling to me.
- 18:26I, I cannot I cannot believe that because I started making
- 18:30content if you remember in the beginning of YouTube when Ali A
- 18:33was making Call of Duty videos and all that and that was when
- 18:37when you wanted to make one YouTube video, it would take you
- 18:392 weeks, right? And now you're saying 15 minutes
- 18:42from idea to to basically distribution.
- 18:45It is we. Have we have another secret?
- 18:47We have another. Secret.
- 18:49I like that you were holding on to that, but now you're it's
- 18:51started to share it. Let's hear it.
- 18:53Yeah, my wife runs the largest editing school in India now, so
- 18:57she has she trains about 2000 editors a year and therefore
- 19:001000. Editors.
- 19:02Yeah, 2000 editors great. You should hide from them if
- 19:04you're interested. Yes, so so I get access to a lot
- 19:07of really high quality editors who already been pretend on good
- 19:11formats already know what to do. So we, we don't mind having a
- 19:15slightly larger editing team and we don't mind paying them
- 19:18reasonably well and sort of building our own sort of backup
- 19:21and so on and so forth. So I feel like that's a really
- 19:26big advantage because people struggle finding editors and
- 19:28editors leave very often. This churn and stuff like that.
- 19:31You got to run it like a company, you got to run it like
- 19:32a tight ship. It's very like a lot of people
- 19:34want to be remote editors and a lot of people are OK with hiring
- 19:36remote editors. Still there's churn three months
- 19:38later. Those that don't know churn
- 19:39means people leaving. But when you have an office that
- 19:42doesn't happen as as often because you can actually give
- 19:44them a career path, you can actually give them guidance, so
- 19:46on and so forth. So it's that that's our second
- 19:50secret, which is we have editors.
- 19:51So we're able to do this much faster at scale.
- 19:54You know, I plan to hire another four or five editors.
- 19:56But here's the thing about long form, right?
- 19:58I also want to do like in depth CG videos.
- 20:01Like we did one on the Slipstream.
- 20:02I don't know if you saw that, right, which is I did with Intel
- 20:05and we had two cars. We had an Aston Martin and some
- 20:09other car. And I think the Ferrari in the
- 20:11video, right? That was CGI.
- 20:13The entire scene was CGI. It was just me in the scene with
- 20:15these two cars. And then we wanted to go.
- 20:18The boundary, right, we're like yo, if we're automatic content
- 20:20here, we should be doing really crazy videos on something like
- 20:23for example, one of the things I want to work on is I5 years in
- 20:25the future and we actually want to do CG, right we want to go
- 20:28shoot some places in India and we're like what if there was a
- 20:30massive tower here right and it's not hard tracking is easy
- 20:34the iPhone does it with LIDAR. So it's not hard for us to use I
- 20:37don't know if you've seen tools like Similon or if you've seen
- 20:40tools like Wonder Dynamics. It's easy to put robots in so
- 20:43we're like let's try unlimited creativity.
- 20:47So in a way short form content is allowing me to focus my
- 20:50energies on doing really good long form content, right So I
- 20:54don't want to do too much AI generated long form content even
- 20:56if I can get more views, brand deals, whatever, I think it
- 20:59still has to be like very in depth content even if the view
- 21:01count is lower right so different games I'm playing a
- 21:04totally different game on shots and reels.
- 21:06I'm playing a totally different game on long form content and
- 21:11yeah, I think the burden of not having views allows me to
- 21:14creatively sort of expand myself.
- 21:15So use a lot of Blender for the CGE stuff.
- 21:18We use a lot of Unreal Engine, although we've actually moved
- 21:21back to Blender and I have some reasons for why, right?
- 21:25I think that's what we're doing and it's fun and most people
- 21:29don't know that we actually run 8 channels. 8 channels. 8
- 21:32channels. Yeah, we actually have a lot
- 21:34more channels than just mine. Different faces, different
- 21:37names. There's no reason to link up all
- 21:38the channels together. Like they're all different
- 21:40brands. They're all under 1 banner.
- 21:42So the original channels sort of helped them grow, and then they
- 21:44help each other grow. Now it's very easy for us to
- 21:46start a new channel as well. We produce over 1000 videos a
- 21:49month because not just we do those eight channels, we also do
- 21:52several channels for, for customers.
- 21:54We have companies that work with us, some of the largest unicorns
- 21:57in India, some of the, you know, big billion dollar businesses
- 22:01outside of India. So, yeah, we, we, this is a
- 22:05solid business and all the advantages now really play out.
- 22:08Having that video editing school that my wife has, automating AI
- 22:11content here or the presenter here, I think has really given
- 22:15us advantages that we're sort of putting to play now.
- 22:18And yeah, my opinion is the best use case of AIS content.
- 22:21And you can't change my mind on it because I haven't seen
- 22:23examples outside of this. But it's been so valuable for
- 22:26anybody. Yeah.
- 22:28And it's what makes money. So the minute you start working
- 22:33with Google and Intel and all the other cool clients, whether
- 22:36you're doing work for them or you're doing a brand deal with
- 22:38them or you're doing channels for them, or you're even doing
- 22:40plain vanilla AI services, right?
- 22:41And I'll talk a little bit about that as well.
- 22:43But yeah, it makes money. It does well.
- 22:47That's amazing. I think I want to talk about the
- 22:50fine tuning part because a lot of people have access to the
- 22:53tools that you have access to, but you are the one who managed
- 22:57to scale it to this size. 8 different channels.
- 23:00We didn't even know about the 8 channels. 1000 videos that you
- 23:03mentioned is crazy. What are the steps you took to
- 23:05fine tune? Let's talk about Hadrian first
- 23:07and then 11 Labs later on. Is that something that you're
- 23:10working on constantly? So look, fine tuning is very
- 23:14simple, right? Like at least from what we did,
- 23:16it's a very, you know, ML engineers are going to hate me
- 23:19for this, but I feel like it's more of an art than anything.
- 23:22Like the way I find you. Like, I'll give you an example
- 23:24of how we fine tune my flux Lora, right?
- 23:27We just try a bunch of things until we get outputs that are
- 23:30close to what we want. And then we learned from it.
- 23:32Like the first time I learned this when was when with Stable
- 23:34Diffusion there was this equivalent to the Loras that you
- 23:36have now called Dream Booth, where you could make your own
- 23:39character look like you. And there were a lot of
- 23:41settings. So we used a tool that I'm
- 23:43called Automatic 111. There's this settings panel on
- 23:46Automatic 111 to train the model, everything from how many
- 23:49epochs you're running this on to what kind of position you want
- 23:53to run this on. So I basically learnt it by
- 23:57opening a new tab, Googling everything, seeing other
- 23:59people's results. There's like a ton of YouTube
- 24:01videos on best settings Reddit everyday.
- 24:03There's like a best settings for dream boot sort of thread,
- 24:05right? So I learnt what each setting
- 24:07does, each parameter does. Hyper parameter does 1 by 1.
- 24:11And then then it becomes like an art.
- 24:14You train it. Like for example with the Flux
- 24:15Lora that we made, which we actively use for thumbnails.
- 24:17The 16 thumbnail is made with the flux Lora, right?
- 24:20So I wear a mic and all the photos that we fed it, I won a
- 24:24mic in it. So it always generates a mic.
- 24:27I'm like, wait, no, I'm asking you to generate Varun.
- 24:28But yeah, there's still a mic in everything.
- 24:30Yeah, that's the input. Mic is part of Varun now.
- 24:33Yeah, so, so we learnt that. Well, you need to keep the
- 24:37subject whatever you want to replicate, you know, in every
- 24:41kind of image, but you need to change everything else.
- 24:43Again. The in dream booth, if you have
- 24:45me in every image wearing the shirt, it will always generate
- 24:47the grey shirt, right? So we we have like, what is the
- 24:50subject we want preserved this, this OK, fine, all of this we
- 24:54change in every photo. We use different photos from
- 24:56different times, right? So it's that, but you don't want
- 24:59the face changing too much. So if you take photos from a 10
- 25:01year period and I look very young in the early photos and
- 25:04very old in the data photos, it doesn't look like me.
- 25:07It looks like some hybrid between a younger and older me.
- 25:10So we learned these things. So I said it's an art and people
- 25:12hate me when I say that, but it's kind of like that and you
- 25:15just try a lot. And yeah, we trained on consumer
- 25:19GPS, we trained on my 4090. Sometimes we trained on, you
- 25:22know, online and it's getting easier.
- 25:24Like my only advice to people who are fine tuning and you just
- 25:27care about the outputs and you don't really care about learning
- 25:29too much about it. I think it's mostly about trial
- 25:34and error. And over time this is going to
- 25:36be one click because look, it's not rocket science for a level
- 25:39I'm sure that figured it out that you can just give them a
- 25:41synthetic script to read covering most words.
- 25:44They're going to do it. It's not rocket science, right?
- 25:47Same with, you know, fine tuning Loras flux Loras, right?
- 25:50You can do it online now and somebody will make it one click.
- 25:54So I'm like, you don't need to learn this.
- 25:56This is going to be one click and they're going to tell you
- 25:58exactly what to do and you follow steps, right.
- 26:01Agent removed fine-tuned avatars recently.
- 26:03I don't know if they brought it back, but they have now this new
- 26:04thing called instant avatars. We still tend to use the
- 26:07fine-tuned avatars, but they've they've removed the guesswork
- 26:12out of it, Right. And they just give you a bunch
- 26:14of ground rules that you use. But I think you get good at
- 26:17this. It's sort of like making a
- 26:19website, right? Like, or making anything.
- 26:21It's like paste. Anyone can make a video with
- 26:24Premiere Pro. Doesn't take you too long to
- 26:25learn it. But someone making a good video,
- 26:27bad videos work. Paste.
- 26:29Yeah. And going to the right places
- 26:31and choosing between A&B. This is the style I want.
- 26:34This is the style I want. It boils down to pure taste and
- 26:38I think we just knew exactly what we wanted.
- 26:40We wanted me sitting for our camera talking in exactly my
- 26:43voice. We had to just replicate this.
- 26:45And just so you know, this is not a new problem for us to
- 26:48create avatars. Many years before Metahuman, we
- 26:52actually made an avatar on me in Unreal Engine, a live speaking
- 26:56avatar on Unreal Engine. I'll tell you what we did.
- 26:58So we modeled it. We manually modeled it.
- 27:01We used some used Daz 3D actually at that time, right?
- 27:03We changed the face structure to sort of match and then we put
- 27:07some clothes on and then we exported from Daz to sort of
- 27:10Unreal Engine because we wanted it real time and then we just
- 27:13lit it. Well in Unreal Engine that is
- 27:15the live Cam. I don't know what it's called
- 27:18now, but the live Cam, the AR kit to, you know, like the blend
- 27:23yeah, where the blend shapes are sort of transferred over.
- 27:25And then I was just speaking and then what we did with the body
- 27:28is, you know, neck and below. We sort of made it run a mix MO
- 27:31animation of just, you know, talking right on loop and I did
- 27:36a live stream like that. It was still my voice.
- 27:39So I was like I was always looking for a way to sort of
- 27:41replace myself. This you can find this video
- 27:43from my YouTube from like 4-5 years ago, right?
- 27:45This is before Metahuman launched.
- 27:48So we did that and after Metahuman, there were ways to do
- 27:51photo scans of your face, like the photogrammetry equivalent of
- 27:55your face into this thing. There's still some tools to do
- 27:57that, but it gets a point cloud and then sort of puts it in.
- 27:59We did that. We didn't like the outputs.
- 28:01And then, you know, sometime later we saw we have to live
- 28:04with like, we can't fool people with a 3D avatar.
- 28:06It's hard. Fool is the wrong word.
- 28:07But we can't like sort of pretend like it's me.
- 28:10So it's better to do this because people don't like
- 28:12uncanny valley and below. Absolutely.
- 28:15If I put 3D avatar on, you know, on your feet, you'll just be
- 28:20like some cartoon shit. Yeah.
- 28:22And you move on. Yeah, right.
- 28:24So it had to be real to to gain real benefits for me.
- 28:29And I think we've crossed the on Cabin Valley.
- 28:30Yes, and listen, I, I do want to add to what you just said.
- 28:33First of all, a couple of things that you covered.
- 28:35One, every technology in the early days, there's so many
- 28:38things you have to learn that will eventually disappear and
- 28:41become much easier to use. So I agree with you.
- 28:43I think especially with AI, there's a lot of when you go to
- 28:46confi UI, there's so many nodes and so many settings that will
- 28:49eventually become a lot more user friendly.
- 28:51So yes, you can wait until it's easier, but then if you're
- 28:55trying to be competitive and stay on the the front lines,
- 28:59then you have to learn these stuff now, which is why we're
- 29:01talking about this with you today.
- 29:02But going back to your next point, you've been waiting for
- 29:06this avatar thing for a very long time.
- 29:08I can see that. That's why you're so early into
- 29:10it because you've been trying to do it for such a long fucking
- 29:12time and it finally happened and you managed to do it through a
- 29:15completely different method. I mean that 3D modeling,
- 29:19texturing, UV unwrapping, animating is nothing compared to
- 29:23what you can do with Hajen and create content in 15 minutes.
- 29:26But more so I I do want to touch on the fact that a lot of
- 29:31businesses have also been looking for this approach.
- 29:33So many people have contacted us to make their meta humans and
- 29:36then run the animation, but that almost never really works unless
- 29:41they're willing to invest a huge amount of budget.
- 29:43Because animating the meta humans every single time will
- 29:46take a lot of time and it will require the same person usually
- 29:49to come and talk. But now with Hagen and 11 Labs,
- 29:52this new approach is completely different and it's completely
- 29:55changed the game. So one thing that was
- 29:58interesting to me about your approach is you mentioned about
- 30:01the script generation, the ideas and the agent avatar, the voice.
- 30:05All of those seem to be a lot more automated now, whereas the
- 30:08editing part seems to be the most manual.
- 30:11And you mentioned that yourself, your wife having that school of
- 30:14editors, and you guys being in contact with some of the top
- 30:16editors. What do you see the future of
- 30:19editing B with AI coming in, the AI assisted workflows.
- 30:24I mean, right now Premiere Pro, yeah, Premiere Pro and Davinci
- 30:27Resolve is there. And these are amazing softwares,
- 30:29right? I'm an editor myself.
- 30:30I I love editing videos. But at the same time, if there
- 30:32are new workflows that will make editing much easier.
- 30:35Do we need to use all those buttons again?
- 30:37I'm just comparing meta humans versus hey, Gen. right?
- 30:40A completely new approach to making avatars.
- 30:43And it's more lifelike than a meta human can be, because it is
- 30:45actually you. Will we see something like that
- 30:48with editing itself? That manual approach?
- 30:53So my thoughts on this are a little bit nuanced, right?
- 30:57I feel like there are two parts to a video if you want to edit a
- 31:01video or make a movie or make anything, right?
- 31:03One is the footage, the base footage.
- 31:06And I feel Hadrian solves the base footage because we see the
- 31:09base footage as is. You see, it's not perfect,
- 31:12right? There's issues.
- 31:13Sometimes I'm saying things, but my hands, I'm not.
- 31:15Sometimes I'm pausing, but my hands are still moving.
- 31:17It looks weird, but we edited cleverly to remove those parts.
- 31:20It still, it still requires effort, right?
- 31:25I think footage is going to be solved at some point.
- 31:27Right now we can do it for a few seconds, like the flux slower.
- 31:30Plus runways of the world are not bad and I feel that's just
- 31:33going to get better. So the footage part is probably
- 31:35going to be solved. Consistency is still a problem.
- 31:38I would take two different runway outputs from the same
- 31:41image that I've sent or like from 2 sequences of that, you
- 31:44know, scene. It doesn't look like it flows
- 31:46very easily, and maybe that's a short term problem, maybe that's
- 31:49a long term problem, I don't know.
- 31:50But I feel like it feels reasonable to say that footage
- 31:54generation is going to be solved at some point or the other.
- 31:57I know this because we're trying to work on a project called SETO
- 32:00where I always wanted to do 1 take on philosophy, right?
- 32:03I love like long form. I really want to do tech
- 32:05philosophy, even though I feel like it won't get too many
- 32:07views. I feel like if I'm already
- 32:09pulling views on short form, might as well experiment on long
- 32:11form, right? So for that, we really needed
- 32:14something like this to come out. So we waited.
- 32:16We waited almost a year. I've already written the story.
- 32:17It's a year ago. I have the Google Doc with me
- 32:20and I told my team the minute you know, we have accurate text
- 32:22to video at least for one 1/2 seconds because we're really
- 32:25inspired on that particular video by this channel called
- 32:28Melody Sheep. I don't know if you've seen
- 32:30Melody Sheep because this crazy like, you know, it's all about
- 32:33planets and it gives you this feeling of awe when you watch
- 32:39this videos, right? So we wanted to replicate some
- 32:42of that and still have me in it. It's still grounded to be human.
- 32:45The video was about is based on Max Tegmark's book.
- 32:50It's called Our Mathematical Universe.
- 32:52And then I wanted to flow in from that book to another book
- 32:55on philosophy by Douglas Hofstadter.
- 32:56It's called I'm a Strange Loop. It's, it's a very trippy video.
- 33:01It's supposed to be a very trippy video.
- 33:02We knew we couldn't do it without solid text to it and
- 33:05solid, you know, like width to width, sorry, and then solid,
- 33:08you know, image to width. So I think footage will be
- 33:12solved and we need it for this project.
- 33:13So we'll keep trying. On the editing part, I think
- 33:17it's unlikely, at least in the short term, because your medium
- 33:21of instruction will change. It'll become English.
- 33:24So for any of these tasks like coding and even editing to some
- 33:27extent, they become like English.
- 33:28You still need to tell the computer what to do.
- 33:31You'll tell it. So I can go and I can look at a
- 33:34video and I can be like, well, at 30 seconds, shut this off.
- 33:37OK, At 45 seconds, add like a nice banner that flows in from
- 33:41the top to bottom. At 60 seconds, do this.
- 33:45At 70 seconds, do this. But there are other parts,
- 33:46right? Like subtitling.
- 33:47It's going to be automatic, right?
- 33:49It's already automatic. There's some things like B roll,
- 33:52which is going to be 100% automatic.
- 33:53It's again power footage. But I feel like maybe editing
- 33:56will change to just telling the thing exactly what you want,
- 33:59Like then describing it to the extreme T.
- 34:02And I feel like then you have to go frame by frame and make sure
- 34:05it got the right thing. You still have to watch the
- 34:07video and be like, well, everything was good, but at 2nd
- 34:0953, I didn't like the transition on a different transition.
- 34:15So you just tell it and you edit like that.
- 34:17I feel like someone reasonably experienced with Premiere Pro
- 34:20can easily switch over to that form of editing.
- 34:24But we'll find it more convenient to go back to
- 34:26Premiere Pro because you can again now say if you want to
- 34:28move something from one place to another, you can just draw a
- 34:30spline, right? Like from place A to place B.
- 34:33So I feel like you'll just find it faster like to go edit by
- 34:37yourself. And even if it changes, it's
- 34:38going to change to I speak English and now it edits, which
- 34:41for people like me, I can do it right.
- 34:42I can speak and edit, but then I feel like for people like
- 34:46creators who are really busy, it still makes sense to have an
- 34:49editor and be like, you speak and you have that creative
- 34:51taste, you tell it what you want.
- 34:53So it might reduce the number of editors in the world, but you're
- 34:56still like hiring for taste. You're still hiring for somebody
- 34:58who knows what you like. It's just still hiring for
- 35:00somebody who doesn't make you look like an idiot in front of
- 35:03camera or cuts things completely out of context.
- 35:05You need somebody who knows the narrative and wants to set that
- 35:07narrative. So when we hire editors, we're
- 35:09hiding almost like people who can be editors, but also could
- 35:12potentially manage editors, right?
- 35:14If it gets automated tomorrow, we want them to be able to play
- 35:16one level up. But it's going to take a while,
- 35:19right? It's going to take several years
- 35:21for it to be English. The other approach is through
- 35:25something like a Lam, right? The thing that Rabbit came out
- 35:28with that they said it's a large action model.
- 35:30I don't know where Rabbit is right now, but the idea, if it
- 35:33worked, would be interesting. But just teach a computer to use
- 35:35the keyboard and the mouse and then teach a computer to use
- 35:37Premiere Pro so it has access to everything that's in Premiere
- 35:40Pro, right? So you say, hey, I want you to
- 35:44remove this part of the footage. It will go.
- 35:46It will use the blade tool, cut those two parts, remove it.
- 35:49You can train a model like that, right?
- 35:52I just feel like when that happens, you will still need
- 35:55somebody telling it exactly what to do.
- 35:57And I don't have the time for that.
- 35:58Like I know from my time perspective, and I know from
- 36:00your time perspective as well, you don't have hours, especially
- 36:04the 1000 videos, right? Producing 1000 videos, just do
- 36:06not have the time to sit on 1000 videos and tell them what to do.
- 36:08So we'll still have a team for it.
- 36:09So maybe less editors, but definitely you still need
- 36:13editors who have that taste, creativity, understand the tool
- 36:15because if it gets it wrong, you need to be like, no, no, you use
- 36:17the blade tool wrong or no, no, no, you need to use this, you
- 36:22know, plug in for the lens flare, whatever, right?
- 36:25Like I'm saying, you need to have enough context to know what
- 36:27to do. And it's the same thing with
- 36:28engineering, right? Code might die, you might speak
- 36:30in the English language to the computer, but you still need
- 36:33somebody to know what they're doing right if you're building
- 36:36an app. And more than that, you also
- 36:38need somebody who understands the product clear, you know why
- 36:41something would work for end consumers or not.
- 36:43And I think that piece is very, very important.
- 36:46So nuanced answer, there's no, is editing going to be automated
- 36:50or not? I feel like some parts will be
- 36:51automated, yes, some parts won't.
- 36:53But yeah, we'll still hire people.
- 36:55And I do want to add to what you said because you said it
- 36:57beautifully when I meant by automating, I don't mean like
- 37:00click a button and then the editing is done and you don't
- 37:02need any editors anymore. It's more about the editors who
- 37:05are editing. They have a much easier time
- 37:07editing what they are working on.
- 37:09So I'll give you an example. Me and Farrak suck a drawing.
- 37:12We're working on a Fortnite map at the moment, and usually you
- 37:15have to go to a concept artist and then try to convey your idea
- 37:19to them. And that timeline will take a
- 37:21lot, a lot of time. And not a lot of companies have
- 37:24concept designers, right, unless you're using them on a weekly
- 37:26basis. So for me and Farah right now,
- 37:28going to ChatGPT, which is even better than mid journey now
- 37:31because I get to talk to a model, tell them what I want my
- 37:34map to look like and they will spit out images in in 10
- 37:38seconds. And I can get new ideas that I
- 37:40didn't even have by just talking to the model.
- 37:43So all these things on ChatGPT, I imagine can be integrated into
- 37:48Premiere Pro or Davinci result when you're editing, when you're
- 37:51adding in footages to get ideas that you didn't even have
- 37:54before. So it makes editing better, more
- 37:56creative, easy here in the long run.
- 37:59Plus, I do want to talk about the Lamb thing because I had no
- 38:02idea about Lamb. You mentioned something like
- 38:04that the other day. If I was like why can't an AI
- 38:06just learn from our tutorials how to use Unreal Engine 5 and
- 38:10then we tell it what to do and create the scenes we want?
- 38:12To just go through our tutorials, 20 episodes, I'm sure
- 38:15you're going to learn everything.
- 38:16Then you can make everything for us after that.
- 38:19Yeah. So I think it's possible, like I
- 38:22feel like agents, the entire goal with agents is this, how do
- 38:26you teach it to use a computer? By the way, they did this a long
- 38:28time ago, right, with, you know, in the early days without a
- 38:32thought process. If you just gave a computer a
- 38:34keyboard and a mouse going back to machine learning, right, and
- 38:36you just give, you said, let's use reinforcement learning.
- 38:38The computer never gets there, even with millions and millions
- 38:42of vibrations. It just will never learn how to
- 38:43use Premiere Pro. But the minute you give it a
- 38:45chain of thought, the minute you allow it to reason the steps
- 38:49through and then build that out, it's a little bit better.
- 38:51So you needed something like O1 to do agents, right?
- 38:54GPTO one to do agents. And I feel like now that we have
- 38:57that, but sometimes, you know, the reasoning and chain of
- 39:00thought is what happens behind the scenes.
- 39:01The reasoning is what gets you expanded.
- 39:03You could sort of see it. I feel like it still gets it
- 39:06wrong very often. So it still understands even the
- 39:09reasoning of the thing. Pretty exciting to see how that
- 39:13plays out, but we are ready, you know, if something like that
- 39:16comes out, happy to use it. It's something that doesn't come
- 39:18out happy to play the old way. You just need to like what what
- 39:22I tend to do is I try to try all the new models.
- 39:24I tend to read the new papers that come out.
- 39:26I tend to try the most experimental models.
- 39:28You get a idea of what the future is going to be like.
- 39:30And then we're like, if this gets better, we should start
- 39:32using this in our workflow. And we want to be the earliest
- 39:34to use any new tool in our workflow.
- 39:36Well, like we will do this live in production, like Simulon live
- 39:39in production. We've we used it in one of our
- 39:41videos, right one second. So we want to be those people
- 39:44that sort of bring you the earliest previews of all this
- 39:48new tech, especially in audio, video, text.
- 39:50And honestly, AI really just solves audio, video, text,
- 39:53right? And now using computers and
- 39:56stuff. So yeah, I feel like show and
- 40:00tell is amazing and it's awesome to be a creator where you get to
- 40:03show and tell and then also speak about it. 100% and now you
- 40:08have a team and you have access to all these tools.
- 40:11If someone comes to you as a content creator and looks for
- 40:14advice, what is the first tool you ask them to use?
- 40:18Well, I don't think it starts with tools.
- 40:19I think creating content is this question of do you want to do it
- 40:23seriously? Because it's actually pretty
- 40:25risky to be a content creator. Seriously.
- 40:27Like doing 8 channels is even risky, right?
- 40:29It's craziest risk ever, but it's kind of scary.
- 40:32You don't know when it is going to die, right?
- 40:34You don't know when you're going to have a bad month or whatever
- 40:36it is any number of issues that that could happen, right?
- 40:40So I feel like as a person, I don't know if I would actually
- 40:49give the advice for somebody to be a content creator to make
- 40:51money off of it. But I feel like what you guys
- 40:54are doing and what a lot of other content creators will do,
- 40:56right, which is you create content to bring you a few more
- 40:59leads to be slightly more discovered.
- 41:01And if you get big during that period, if you get like you go
- 41:03crazy master during that period, then great.
- 41:06But don't start with that expectation.
- 41:08So I never started that expectation.
- 41:09My wife didn't start with that expectation.
- 41:11We'll just like we'll create content, more people will know
- 41:12us and then maybe you know, we'll get some more B2B deals or
- 41:15something like that. But you make money when it gets
- 41:18to scale yes, but you know, plan for that.
- 41:21And honestly, it was just me being lazy saying I don't want
- 41:23to shoot content everyday. So how can I automate it?
- 41:26So so you need to really just focus on having a business or
- 41:30having something to sell and then doing content as a way to
- 41:32generate leads. Otherwise it burns you out
- 41:34because still you get to like 200 three 100K.
- 41:37You don't make any money being a content creator, especially in
- 41:40India where it's kind of saturated as well, right?
- 41:44At least 100K to make any money, right?
- 41:46As a as a content creator is like doing a branded post.
- 41:50So I feel like people just give up.
- 41:52It's too easy to give up. You're 50K, you're not making
- 41:54money. You're like, fine, let me just
- 41:55give up, right? So you need something to keep
- 41:59you going. And the minute you realize it
- 42:00brings you leads to your business, it is better.
- 42:03And one thing I realized is it can bring you a lot of leads.
- 42:06Like the difference between you having like 20,000 subscribers
- 42:10and getting maybe 2 leads a month versus 700,000 subscribers
- 42:14and getting 50 times that is actually a very real thing.
- 42:18You can get hundreds of leads a month saying, can we do
- 42:19something cool together? Sometimes they don't have clear
- 42:21ideas. Sometimes they have very clear
- 42:22ideas. So the number of opportunities
- 42:25grows, and that's how we build a company.
- 42:26To be honest, we just got more people reaching out to us
- 42:28saying, can you do this? Can you do that?
- 42:30I said, cool, we'll do it. Yeah.
- 42:34So I feel like the first advice I would give is don't expect
- 42:37anything from it to a new creator.
- 42:39And second thing is from a tool perspective, I'd be like, get an
- 42:41editor. People say people think I'm
- 42:46saying this because I'm biased or whatever, but it's not really
- 42:48that. I feel like they're still
- 42:50contact creators today who edit their own videos.
- 42:52I'm like, do you know what a monumental disservice that is to
- 42:56you and even the team you get tomorrow?
- 42:58Because what they need you to do is focus on making that Channel
- 43:00big or go out and crack B2B partnerships or whatever and
- 43:04hand off the editing role to somebody or also creating a job
- 43:06in the process. It saves your time and your time
- 43:09becomes very, very important at scale because you have so many
- 43:12inbound opportunities. You have to evaluate them.
- 43:14You have to be like, should I do the one with, you know, NVIDIA,
- 43:16or should I do the one with something else Like that becomes
- 43:18important, right? So it's better to hand off to an
- 43:23edit team so you have the time and freedom to think about
- 43:25scale. Otherwise, if you're editing,
- 43:26you can't really think about scale and how the channel grows.
- 43:28There's a difference between working in the channel and on
- 43:32the channel. I don't know if that makes
- 43:34sense. It's better to work on the
- 43:35channel and think strategy and be like, how does this scale six
- 43:37months from now? How does it scale a year from
- 43:39now? What do we need to automate it?
- 43:41And you need somebody with free time and free energy to read,
- 43:44learn, talk to other people because there are lots of
- 43:46people. One thing I learned in the last
- 43:47two years is there are lots of people with secrets.
- 43:49There are lots of people who know stuff where if you just sat
- 43:52down with them, you'd learn more in a conversation and you'd pick
- 43:55up, you know, you'd find the secret to actually doing well.
- 43:58But they will never say publicly because either they'd get
- 44:00cancelled for saying it or they're just like, what's the
- 44:01incentive for me to say this publicly, Right.
- 44:04And nobody knows them. For example, there's a guy who
- 44:06sells blankets, right, who makes millions of dollars a year and
- 44:11he's just like, here are my 10 secrets.
- 44:12Easy. Why aren't you all doing this?
- 44:14You have so much distribution. Why aren't you selling blankets?
- 44:17And I guess not really our thing, but interesting insights.
- 44:19I could apply this to something else.
- 44:23So yeah, it's it's, it's mostly that.
- 44:25I guess, no, I agree with you. I think what's really
- 44:28interesting for me and Farad is based on your response, we used
- 44:33to watch this guy. I, I'm sure a lot of people
- 44:36probably know him, Casey Neistat, he was one of the best,
- 44:39one of the biggest Youtubers. Now, he's a very unique case of
- 44:43someone who loves the editing process, so much so that he
- 44:46never wanted to hire anybody to do it because he wanted that
- 44:49part to be done by himself. He was a storyteller and his
- 44:53goal was storytelling on a daily basis.
- 44:55So if you're a creator or an aspiring creator watching and
- 44:59listening and that's your goal, then I guess that's where you
- 45:01can just keep doing the editing and everything yourself.
- 45:04But I guess what Varun was referring to and what we're
- 45:07trying to talk about is more about people who are trying to
- 45:09scale the business, grow the business.
- 45:10They have other interests and those interests are being
- 45:14hindered or prevented because they have to spend a lot of time
- 45:18on what can be done by other people or now be automated in
- 45:21the process. So again, just going back to
- 45:23your workflow, it's, it's crazy how you are looking at this.
- 45:27Your perspective is incredible. I, I do want to talk about the
- 45:30ethical point of view as well, because in, in one of your own
- 45:33videos, you were talking about the time 100 people when it
- 45:36comes to the most influential in terms of artificial
- 45:38intelligence. And there was a Bollywood actor
- 45:40there. I forgot his name, but he has
- 45:44been against the use of AI for using his liking or his voice.
- 45:49Now there's people like that, a lot of people like that who are
- 45:52against AI, against, you know. He's not against AI, he's
- 45:55against the use. Of the misuse of AI, let's put
- 45:58it that way. And then there's people like us
- 46:01who are, you know, using AI. We're gladly feeding in our
- 46:04information, our liking, our faces to AII want to know your
- 46:07thoughts on this and this whole thing.
- 46:09I'm sure your audience has their own point of view.
- 46:11They're commenting on your videos all the time.
- 46:13You probably talk to a lot of people who have different
- 46:15opinions. Where do you stand here?
- 46:17Because clearly, just like us, you're gladly testing out,
- 46:20experimenting this technology. Permission and consent.
- 46:24Great. All of this is we do the same
- 46:26thing. We're so excited.
- 46:28Meaning. For agent, yes, you can use
- 46:29anything you want. My voice, my face, my audio.
- 46:32So I have a line, I have a line. The line is if you did fake
- 46:36somebody else or if you clone somebody elses audio, that can
- 46:39sometimes be a problem. And maybe a few years ago we've
- 46:41done that. It's some of that, right?
- 46:42Like just to experiment because the models weren't that good
- 46:44back then. It wasn't believable.
- 46:46And a lot of Indians are shy. So my team would always take a
- 46:50celebrity and, you know, try cloning their voice or whatever
- 46:52because there's just like, I don't want to use my own voice.
- 46:56But I think the line is you can't clone somebody else's
- 46:59voice or somebody else's face without their permission.
- 47:02It's as simple as that. But that's my line and I feel
- 47:07like the legal line is not going to be that across the world,
- 47:09even though there's right now people have different opinions
- 47:12and thoughts because it falls in the same line as caricatures in
- 47:15a newspaper. I can make a caricature of
- 47:16celebrity and it's OK because they're public figure, right?
- 47:20So I feel like AM or the sort of the same, but we let the legal
- 47:22system, you know, sort of comment on that where my line
- 47:26and the other side of the line is, it's OK to defrag yourself
- 47:28or clone your own voice model. It's fine.
- 47:31Like and some people like tells this guy and I got really pissed
- 47:34there's this guy who's like in the comments like you should
- 47:36tell the world that this video is AI generated otherwise it's
- 47:39causing people harm. I was like, yo, number one, this
- 47:42video is not fully AI generated 7 seconds of this video has a
- 47:45presentation with me in it. That's the only part that's AI
- 47:47generated. Everything else has real
- 47:49tangible human effort. The editing is manual human
- 47:53effort, right? I'm only there for seven
- 47:56seconds. Don't give me that much
- 47:57importance. And plus it's my own voice, it's
- 48:00my own face. What do you want me to do?
- 48:01Come there and stand. Do the exact same thing.
- 48:03It's the exact same thing. And plus, if you think, if most
- 48:06people think there's no effort done in that.
- 48:08I spent 30 years of my life building a real world brand.
- 48:12I spent 31, action 31. So I spent 31 years of my life
- 48:15building a real world brand. I spent 31 years of my life
- 48:18honing the voice that I have today, which is an attribute of
- 48:20me. I spent 31 years old, 31 years
- 48:23creating videos, right? So my hand movements, the way I
- 48:26talk, the animated nature of who I am, it's all captured in the
- 48:30bottle. But I made it what it is in 30
- 48:34years. A lot of people without that
- 48:35personality, and you'll find them at 40 and they don't have
- 48:38that personality. So I'm not taking too much
- 48:41credit for the personality. I built my environment shaped me
- 48:43into that, but that's still work.
- 48:45That took a lot of time that we captured in a very quick form
- 48:49through these models. So it's a snapshot of me that
- 48:52wouldn't have worked if it wasn't me, right?
- 48:55And the other people who do AI generated content, the outputs
- 48:57are like, you know, it feels like a robotic zombie, right?
- 48:59So I'm like, credit me for the amount of time, energy and work
- 49:03I've put in actually making this, you know, getting this to
- 49:06this point. And then yeah, I'm just
- 49:10automating myself. Like I don't think there's any
- 49:13harm if I'm anyway going to be sitting in front of the screen
- 49:15doing the same script. There's a model doing it who
- 49:17gets that's my opinion on it. But yeah, my hard line is we
- 49:21don't be fake. Anybody else anymore?
- 49:23We don't do any audio of anybody else anymore without their
- 49:26consent because we have a subsidiary of AOS, the company I
- 49:28run AOS called AOS Labs, where we also do AI services for
- 49:31companies. We work with Amazon Prime, we
- 49:33work with Netflix where sometimes we do, we, we make
- 49:36trailers where we put, you know, you inside a trailer with an
- 49:39actor, but we don't do that without the actors permission.
- 49:42And of course there's Amazon Prime and Netflix involved.
- 49:44So they have to get permission. It's a big deal for them.
- 49:47So we do stuff like that. And I think with consent
- 49:51everything is fine. So that's my line.
- 49:52It's a very nuanced line. But you know, today people are
- 49:54like either pro AI, anti. I'm like, that's so stupid.
- 49:58I'm like, everything is nuanced. Yeah, it's like it's.
- 50:01Not white or black at all. Correct.
- 50:03It's like saying, oh, are you pro nuclear or anti nuclear?
- 50:06I'm like, yeah, if you use nuclear as a bomb and drop
- 50:09around people's head, it's bad. But if you use nuclear for
- 50:12power, it's not bad. So so everything's nuanced and I
- 50:15feel like social media has stripped the nuance out of
- 50:17people. One minute videos have made
- 50:19people either do this or this all the time for everything
- 50:22they're just used to a like or dislike.
- 50:25And I think it's also done something to people's
- 50:27inhibitions, right? Like people just comment
- 50:28whatever, whatever comes to the head, right?
- 50:31And they don't think it through. And then and then they don't
- 50:34realize people take decisions on those comments, especially
- 50:36politicians and stuff. They actually take decisions on
- 50:38those comments. They're ruining society by just
- 50:40commenting the first thing that comes into the head without
- 50:42realizing what the 2nd order effect of something is, right?
- 50:44Like banning nuclear, it just sounds like a bad word because
- 50:48of Chernobyl or whatever. And then you made that decision
- 50:50and then you're going backward in time.
- 50:52It's not as bad as it was. And, and those were accidents.
- 50:55You can't, you can't. Anyway, you get what I'm saying?
- 50:58And another important aspect is the amount of time and effort
- 51:01you put into fine tuning these models, creating the avatar
- 51:05multiple times, trying different tools, 11 labs, the other AI
- 51:09tools. So you spend a lot of time
- 51:11researching the outcome is that 7 seconds, but the effort is all
- 51:15the time that you put to fine tune these models, find the new
- 51:18tools. It's we have done a lot of R&D
- 51:20in everything that we do and doing R and DA lot of times lead
- 51:24to a dead end. There is nothing at the end.
- 51:26You try for three days, you get nothing out of it.
- 51:28You have to try two or three different software to get the
- 51:31result that you want. So it's not just looking at that
- 51:347 seconds, it's looking at all the time that you have spent on
- 51:37on the entire process. Yeah, I guess it is that I just
- 51:42feel like sometimes those comments don't mean well in the
- 51:46sense when people say, oh, this, this you're creating air
- 51:49generated content, you're harming society.
- 51:51I'm like, no. Well, eventually you're going to
- 51:55get harming nature, global warming, you're going to get all
- 51:58of that in the comments. Oh, you're.
- 51:59Using GPU's but. The thing is, the reason why I
- 52:02also say it's great because a lot of those people commenting
- 52:05anything negative about AI, they don't even know they're using it
- 52:08already. I mean, technically anybody
- 52:09buying the new iPhone series will be using AI.
- 52:11It's like integrated into it and eventually sorry.
- 52:15Every filter you've ever used? Yeah, I mean not every filter,
- 52:18but many filters. Many of them, right?
- 52:20And there, there's a lot of AI in our life already.
- 52:23It's just people pick one out and then they call AI bad just
- 52:27for that one. So I, I completely agree with
- 52:29you. I do believe it's nuanced.
- 52:30And I think this is going to be an ongoing discussion as AI
- 52:33develops really rapidly. And there's going to be
- 52:36companies that do it ethically and companies that won't because
- 52:40at the end of the day, it is a capitalist market or everybody's
- 52:43just trying to make profit at the end of the day, because
- 52:45there are Shell shareholders involved.
- 52:47And that's when it gets really nuanced.
- 52:50But the point is that I think, as you mentioned, as long as
- 52:53you're not using other people's work or other people's faces and
- 52:57voices to try to trick people, then that is fine.
- 53:01And I, I do know people who are doing that today.
- 53:03So on social media you do see a lot of deep fakes of, of of
- 53:06people. I want to talk about a video
- 53:09that you did and you talked about a developer doesn't care
- 53:13about storytelling and design and an artist doesn't care how
- 53:16their ideas are executed. First of all, are you a
- 53:19developer or an artist? And 2nd how does one solve this
- 53:23problem? So I'm a little bit of both.
- 53:25I feel like the minute you do any sort of game dev, you guys
- 53:28have done a lot of Unreal Engine.
- 53:29Where do you put yourself when you use blueprints?
- 53:31So for example, if you use blueprints, right, you're doing
- 53:34object oriented programming. I studied object oriented
- 53:36programming in college, right? I studied computer science for
- 53:38four years. I built my first business was a
- 53:41recruitment platform where we built it on Ruby on Rails.
- 53:44So I actually transitioned from being, I don't like using your
- 53:47transition. I still do a lot of engineering,
- 53:49right? Like I still have an engineer's
- 53:50brain. I feel like these are all
- 53:53arbitrary things that we have created for ourselves to study
- 53:55things that, OK, this is engineering, this is content.
- 53:58If you come to our studio, we have these giant LED screens
- 54:01where it's just the backgrounds are beautiful, right?
- 54:06Because it's an LED screen, it's not green screen, so it actually
- 54:08emits light. So you have the light wrap
- 54:10effect on you. So you look good like you look
- 54:12like you're actually part of the scene.
- 54:14That's technology. It's hardware.
- 54:17And the images we put on that hardware are generated by
- 54:20ideogram or mid journey of flux, right?
- 54:23So but the end output is a video like when Disney does Disney has
- 54:27Imagineering, right? Where they do, where they do
- 54:29these little robots, they're eventually doing it to be in the
- 54:32next Star Wars movie or whatever, right?
- 54:33Or in their amusement parks. But are they a tech company?
- 54:36Are they a content company? Are they an Arctic Arc company?
- 54:40I don't like any of these distinctions.
- 54:41And that's why, if you ask me, of all the things I've ever done
- 54:44in life, was my most favorite. It's game dev, love game dev.
- 54:48I built this demo many years ago called Mumbai Make Hatsa.
- 54:52It's about this incident that happened in Mumbai.
- 54:54I want to make a very, very short game, something that you
- 54:55can finish in 10 minutes, right and tells you a story and you
- 54:58can play the game. It's it's on hitch.
- 55:00Just search for Mumbai Make Hatsa, you can play it.
- 55:02Very recently, a year ago, we made this game called
- 55:04Overpowered the Card game. We just want to make a card game
- 55:06fun, not very commercially successful because we never
- 55:10really intended them to be like, you know, money making,
- 55:12whatever. It's just like we'll get back to
- 55:13the community, cut our hands a little bit because how can we
- 55:16talk about game dev without shipping things, right?
- 55:19So, so that's why we made them. But I enjoy that.
- 55:23Getting the assets ready, you know, doing the mocap if
- 55:27necessary, combining the two and then the writing code to make
- 55:31sure that, you know, there's some events that are triggering
- 55:32things that happen in the game. I love it.
- 55:35It's the most fun thing ever, right?
- 55:37So, but I never go there thinking, oh, I'm either an
- 55:39artist or an engineer. I'm just like, this is what I
- 55:42do. And even today, like people
- 55:46don't know what to call me. They're like, well, you also run
- 55:47a company with people. Are you a founder?
- 55:49Are you like a manager? Are you like an artist?
- 55:52I'm like, actually, I don't know, I don't care.
- 55:55Like call me whatever you want. Because here's the thing, right?
- 55:57It's it's delegated on both sides.
- 55:59Artists say, oh, bro, you're an engineer.
- 56:01You're like disrupting art. This is not true art.
- 56:04You know, you go sit there in front of the camera and speak.
- 56:06You're you're spoiling this freedom of expression also thing
- 56:09and you know the other side, like you don't know anything
- 56:11about engineering, blah, blah, blah, because the researchers
- 56:13doing stuff, I'm like, you know, the synthetic thing that we came
- 56:16out with that came from technically a process very much
- 56:18like research, right? So, but forget about the
- 56:22research part, right? There's like you're not a real
- 56:23engineer, blah, blah, blah. I'm like, I've built an insane
- 56:26amount of stuff, right. But anyway, both sides like you
- 56:31become like this pariah where you're like no side likes you or
- 56:33no side relates to you because they feel like the engineering
- 56:37side is better and you have a little bit of arts.
- 56:38You're an outsider. It becomes a tribal thing, you
- 56:42know, an engineer in India or actually across the world, be
- 56:44like, I'm part of this tribe called software engineer.
- 56:47We the mighty tribe. We've always made the most
- 56:49money. The artists are like, oh, I'm
- 56:51part of this tribe. You know, this is how I'm
- 56:53supposed to be, this how I'm supposed to dress.
- 56:55Here's my flag. I'm like, you know, I don't know
- 56:59what drive a part of. It's like I just want to create
- 57:02cool stuff, right? And I feel like it should never
- 57:04be a limiter, whether you're an engineer or artist to create
- 57:06cool stuff. But also what we have realized
- 57:09is because there's so little competition.
- 57:11When you stop thinking about tribes and you're doing this
- 57:13thing where you're combining 2 really weird sites.
- 57:16And we will always use the engineering side to get an
- 57:18advantage. We will always use the art side
- 57:19to make things cooler than the engineering side would have done
- 57:22it by themselves. I feel like this looks there's
- 57:26very little competition and when there's very little competition,
- 57:30you know, money flows towards you and then you become a
- 57:32capital allocator. So it's weird, but I start
- 57:35calling myself internally somebody who allocates time and
- 57:38capital, the most boring side, right?
- 57:42But I do want to create even more awesome work.
- 57:44Like I feel like, you know, some people will go on their entire
- 57:48lives knowing me only as an artist.
- 57:49Some people will go on, especially people who met me 3-4
- 57:51years ago would go on knowing me only as an engineer and some
- 57:54people who probably work with us know me as a manager.
- 57:57Whatever, it's OK. And if you look at.
- 58:00You have a wait, sorry. Please continue.
- 58:02And you're with Elon Musk, right?
- 58:03There are a lot of people who like, oh, this guy just hires
- 58:05people and makes them work. Or another guy would be like,
- 58:07oh, you know, he's the guy that designs everything goes to
- 58:09space. So nuanced.
- 58:13Yeah. It would have been like 1000
- 58:15decisions. And, you know, whatever.
- 58:17People have their own people make their own decisions.
- 58:19People will make their own assumptions over time.
- 58:22And I think that's OK. One thing that is clear, you
- 58:27work really hard. I mean it, it's pretty much
- 58:29obvious from the stuff that we see on social media.
- 58:31But also speaking with you, we also figured out after inviting
- 58:35you to the podcast that you have a crazy time per day that you
- 58:39work in. Like the timeline of your day
- 58:41starts at 2:00 PM till 5:00 AM, right?
- 58:44It's a crazy night owl like schedule.
- 58:47Why did you decide to take on something so intensely different
- 58:52than what usually people would be taking out?
- 58:53Usually earlier morning sleep earlier, but then you go into
- 58:56this crazy hours every single day.
- 59:02I don't know, like I just feel like I got stuck at this weird
- 59:05sleep cycle and it never ends. I do like the quiet of the
- 59:09night. I do like, you know, I don't
- 59:12like the mornings. I feel the noisy.
- 59:14There's too many people moving around, scars, there's traffic.
- 59:17The night is just better. Short answer, I just like the
- 59:20night. Less distractions.
- 59:22Less distractions, less people, you know, asking for things.
- 59:25It's just better in the night. I don't know, I love the night.
- 59:29And then he goes to US. No jet lag, 0 jet lag.
- 59:33He's with the US live every single day and.
- 59:36You, you do read a lot of books as well.
- 59:38Is there any books that you would recommend for people who
- 59:40are interested in starting start-ups or generally content
- 59:44creation or anything inspiring to you that you've read in the
- 59:47past one or two years? OK.
- 59:49So we've added a lot of employees over the last one
- 59:52year, right, Because we're also doing content fathers, we're
- 59:53also doing AI services. Business has done well in the
- 59:56last year or so and it's almost started becoming like company
- 1:00:02culture management. So I'm reading this book called
- 1:00:04Inside Apple. It's about how Apple's culture
- 1:00:06is built. I'm reading this book called
- 1:00:08from from Third World to First it was about how Singapore was
- 1:00:10built. I told you guys about this
- 1:00:11before the board. Maybe these books I'm reading,
- 1:00:15it takes ultimately boils down to how you work with other human
- 1:00:19beings and the rules and the, you know, things you set right.
- 1:00:23For example, yeah, I'm a big fan of Gerard's memetic desires,
- 1:00:27right? Which is?
- 1:00:28Everything that we've ever desired in life, it's because
- 1:00:31somebody else desired it, right? Except for basics like your
- 1:00:34food, water, shelter, everything else, like you want a Lambo, you
- 1:00:37want a Lambo because other people find it cool, right?
- 1:00:39So it's stuff like that. So really understanding how
- 1:00:42human psychology works requires time, effort, because you
- 1:00:44unraveling what you think about humans already in your brain and
- 1:00:47be like actually with social creatures driven by simple
- 1:00:49rewards and just accepting that requires you to destroy your ego
- 1:00:54and requires you to have the humility.
- 1:00:55You're just a monkey running around like I'm just a monkey
- 1:00:58running around sort of playing monkey games with other monkeys,
- 1:01:02right? Which is awesome if you're OK
- 1:01:04with it, which is not awesome for people who think that the
- 1:01:06center of the world. So books really gave me that
- 1:01:10because when you read from somebody who's like 40 or 50 and
- 1:01:13done a lot of stuff, they come to the same, especially people
- 1:01:15who've accomplished a lot of things in their life.
- 1:01:17They come to the same conclusion that ultimately we're just like,
- 1:01:21like this animal here with very little control over things.
- 1:01:23And it's about how we work well with other animals and setting
- 1:01:25rules and also setting rewards, like a punishment reward system
- 1:01:28for ourselves. Yeah, it's, it's, it's mostly
- 1:01:32that. And the minute you say that on
- 1:01:34Twitter or Instagram, you get cancelled because people like,
- 1:01:37you know, life's about get super rich by 25 and, you know, travel
- 1:01:41all over the world. And I'm like, yeah, that's one
- 1:01:43life. And yes, that's still part of a
- 1:01:45head. Like if you travelled everyday
- 1:01:46like I now travel like 3 * a week not not for fun but even
- 1:01:50interestingly. I hate trial.
- 1:01:52It was like, it's too much of it, I.
- 1:01:53Just want to be home. I just want to be home, right?
- 1:01:55So everything changes depending on what you have.
- 1:01:58If you have too much of something, you want something
- 1:01:59else, right. The billionaire who's, you know,
- 1:02:03extremely wealthy, but probably you know, is is doesn't have
- 1:02:07help will really want help, right.
- 1:02:10So you always want what you don't have.
- 1:02:11And social media is actually this, this bottle that captures
- 1:02:16things that like, it's basically that, right?
- 1:02:18It's a bottle like a transparent bottle that captures what
- 1:02:23everyone wants to have but doesn't, right?
- 1:02:27And it's that like even in my content, it's kind of that,
- 1:02:30right? When I make a video about the
- 1:02:31future, Indian fibers in the future and I show technology
- 1:02:33something I want but I don't have, it's the same thing that
- 1:02:36some 20 year olds saying, oh, you here's $1,000,000, here's my
- 1:02:39laboratory. It's the same thing.
- 1:02:40It's something that they want, but they don't have.
- 1:02:42So they're showing the projecting that future.
- 1:02:44I'm just doing it slightly differently with what I want.
- 1:02:46I said only needs to be like they're doing it with what they
- 1:02:48want the status and their lifestyle to be like.
- 1:02:50It's, it's just it's still the same thing.
- 1:02:53So social media is not reality. It's like an extended version of
- 1:02:58people's expectations. So but when you read a lot, you
- 1:03:01realize this is not new. This has happened forever as
- 1:03:04this happened 50 years ago in the era of, you know, post
- 1:03:08colonial, you know how the world recovered when suddenly there
- 1:03:11used to be this leader. This should be this
- 1:03:12authoritarian figure. Leader is the wrong word,
- 1:03:14authoritarian figure. And that figure left and then
- 1:03:17you have to now figure things out yourself.
- 1:03:20And then you realize all these expectations that you had with
- 1:03:22them. Now you have to reset those
- 1:03:23expectations, new set of people and reform it.
- 1:03:25It's the same like when with the era of kings and Queens, right?
- 1:03:29So like the reason, the only reason you'd want to have the
- 1:03:35structure of you only become a king if you have noble blood.
- 1:03:38If the minute you remove the criteria of you need to have
- 1:03:40noble blood to be a king, then all the subjects decide they go,
- 1:03:44they want to be king and they fight and they create, create an
- 1:03:46insane amount of chaos and they all try to be king.
- 1:03:49So you have to make it families and even in families, like even
- 1:03:52if it's a noble blood thing, brothers like 2, two young
- 1:03:56brothers or sorry, two sons or siblings of a king will always
- 1:03:59fight for the throne. Imagine that fighting among
- 1:04:02millions of people. So you want to limit, you want
- 1:04:05to limit the fighting to a specific criteria.
- 1:04:07And the best criteria is like royal blood.
- 1:04:09So you're reducing the amount of damage created.
- 1:04:12So it's stuff like that. And when you read all that, like
- 1:04:15this still happens today because we have not our human brains
- 1:04:17have not evolved from 200,000 years ago.
- 1:04:19So every story captures from the past, captures a little bit
- 1:04:23about of how I would interact with other people.
- 1:04:26And you know, some of the ground rules have been the same, like
- 1:04:30when you fight with somebody. I'm sure you guys fought with
- 1:04:33each other and fought with other people, no?
- 1:04:34Never bro. What did you say about?
- 1:04:36What are you talking about? It's gonna happen right out.
- 1:04:41It's gonna happen right out of this podcast.
- 1:04:44So when you guys? I don't know what he's talking
- 1:04:46about. So when you guys fight, half the
- 1:04:48time it's not even about topic that you're fighting about.
- 1:04:50It's like that underlying emotion that you'd have right?
- 1:04:53So, for example, you you might be arguing about how something
- 1:04:55is implemented, but you would have been angry with Farrah
- 1:04:57because he said something about you in front of somebody else
- 1:04:59that you didn't like how we'd effort to you whatever it could,
- 1:05:01it could be setting to a. Therapy session.
- 1:05:03Therapy session for me as far you're right, No, I'm sorry.
- 1:05:08Yeah, so. So people don't like hearing
- 1:05:10that. Actually, now that the company
- 1:05:12has scaled, 90% of my job is no longer hacking.
- 1:05:14It's enabling new hackers in the team and the company.
- 1:05:17Yeah. And it's by preventing all these
- 1:05:19stupid things from bubbling up so people can focus on work.
- 1:05:22So now that you're running a company and we talked about how
- 1:05:25AI impacted your content creation journey, but now that
- 1:05:28you're running company managing multiple people, has AI helped
- 1:05:32you in that way as well? Yeah, I don't think we'd be able
- 1:05:34to. The unit economics of my company
- 1:05:36would not make sense without AI, without when we do 1000 videos,
- 1:05:40it would not, it would not be profitable if we didn't have AI
- 1:05:44like everything from B roll to background music to it's not
- 1:05:47just AI, but it's also other tools that have now existed.
- 1:05:49I don't think it would be possible as the presenter.
- 1:05:52We, we replace presenters in a lot of channels that we do for
- 1:05:54the people. So I think it just not be
- 1:05:56possible without AI, even AI services like we wouldn't be
- 1:05:59able to sell AI services without AI like Jen, AI sort of really
- 1:06:02kicking into high gear. So we're really thankful to
- 1:06:05everyone that makes like folks like Lama that make open models.
- 1:06:09We're really thankful to this ecosystem existing with like a
- 1:06:12lot of people saying terrible things about Stable Diffusion.
- 1:06:15I wouldn't be here without Stable Diffusion, without all
- 1:06:17that early experimentation, without just having it on comfy
- 1:06:20UI and having that my screen open in front of me, just trying
- 1:06:23a bunch of random stuff. I won't be here all the
- 1:06:25researchers who contributed, so no chance this would happen.
- 1:06:32Wow, what do you think about AI encoding for software engineers?
- 1:06:37How do you think it would shift the way they work?
- 1:06:40So right now if you look at the software engineers in the
- 1:06:42company, they're using cloud large generating code.
- 1:06:45They still have quite a bit of input.
- 1:06:46I think all software engineers are going to move one level up
- 1:06:48when the agents work. The agents are not great yet.
- 1:06:51We built this tool called Auto Code Pro to try out agents
- 1:06:53before they for anyone knew what they were.
- 1:06:55It's a long time ago before the Devons of the world.
- 1:06:57But I think with reasoning, with chain of thought behind the
- 1:07:00scenes, I think the agents are getting slightly better.
- 1:07:04And at some point the agents are going to do the work.
- 1:07:05But you still need somebody on top sort of telling it what to
- 1:07:08do because the CEO does not like, I don't believe in the
- 1:07:10idea of one person Unicorn because you can really scale.
- 1:07:13And when you scale, when you have so many things coming in,
- 1:07:15you need people to job some of that make those executive
- 1:07:18decisions because you don't have the time yourself and you want
- 1:07:20some freedom to think about the next steps.
- 1:07:22So I think agents will solve some of code, but you will move
- 1:07:26to a higher level role of describing exactly what you want
- 1:07:30and using English to sort of create software.
- 1:07:32But I also feel like software is getting commoditized for a
- 1:07:37slightly different reason. We learnt this with our last
- 1:07:38business, which is when you're building software and there are
- 1:07:40lots of other people building similar software, the features
- 1:07:43don't matter anymore. Like if you can build something
- 1:07:45and somebody else can build something in a week, like, then
- 1:07:48what's your vote? Right.
- 1:07:50And then we realized the hard way, it's about distribution
- 1:07:52getting in front of the user, which is why, which kind of
- 1:07:55dictated why we're doing what we're doing now.
- 1:07:57But yeah, we saw there's diminishing value in adding new
- 1:08:00things. And there were also a lot of
- 1:08:02APIs and SDKS that were appearing on the market that
- 1:08:05sort of added new features to us.
- 1:08:06So we took, to give you an example, we solved real time
- 1:08:10audio calls over 8 months in my last business.
- 1:08:12We took eight months to solve that.
- 1:08:13And then there's another business that came out that sort
- 1:08:16of just gave you one SDK to sort of solve it.
- 1:08:19And then a few months later, there were like 5 of them and
- 1:08:23we're all competing for price. And we're like, why did we build
- 1:08:25this? We spend so much money, we spend
- 1:08:26so much engineering time. We should just be able to
- 1:08:28predict that this would be out. And this time we didn't make
- 1:08:31that mistake. We could have spent the time,
- 1:08:33energy and money sort of figuring out how to solve, you
- 1:08:36know, lip stitching and a lot of other people have done that.
- 1:08:40But we're like, let's just wait for somebody else to do it.
- 1:08:41Let's capture the value. There's a difference between
- 1:08:43value creation and value capture.
- 1:08:45And we have created a lot of value.
- 1:08:47In our last business, we weren't managing to capture all the, we
- 1:08:50weren't able to capture the value.
- 1:08:52So this time we were like, we just want to play the value
- 1:08:54capture mode for as long as we can.
- 1:08:56And then once we have enough money, then we'll start creating
- 1:08:59brand new types of value through R&D, right, Which is what we're
- 1:09:01doing now. So yeah, it's, it's, it's that.
- 1:09:06So I think software, you'll go one level up, but I feel like
- 1:09:08software in general is not as useful as it was in the past
- 1:09:11because anyone can create it. So it's commoditized.
- 1:09:13So I'm not really confident about software jobs in the
- 1:09:16future in a way, or I feel like there are going to be less
- 1:09:19number of jobs. And in India, we've seen some of
- 1:09:21those effects. I don't think it's AI related,
- 1:09:23though. I just think it's like the zero
- 1:09:26interest phenomenon era is gone where people are trying random
- 1:09:29new stuff with like unclear benefits and they want to go
- 1:09:32back. And that's why, you know, in a
- 1:09:34crisis. Weirdly, businesses spend more
- 1:09:37on marketing than product development.
- 1:09:40Why would you say that? Because think about it, right?
- 1:09:42Like let's say you're a company. Let's say you're a car
- 1:09:44dealership, OK? Or let's say, forget a car
- 1:09:46dealership, let's say you're a new, you're a company that sells
- 1:09:48something and you have two choices.
- 1:09:51You have a product that works OK, that can make you more
- 1:09:53revenue if you advertised it better.
- 1:09:56Or you build a new product where you're not sure if it works.
- 1:09:58There's unclear ROI. You don't know if tomorrow the
- 1:10:00products actually going to work make you money.
- 1:10:03What would you do in a crisis? You have $1,000,000 to spend.
- 1:10:06Would you take a risk and build a new product, stop advertising
- 1:10:08the first one? Or would you just say, I already
- 1:10:10have a product that works, It's not the most efficient, it's
- 1:10:12probably not the best product, but if I spend $1,000,000 in
- 1:10:14advertising, I can make that 2 mil.
- 1:10:16What would you do? You choose the first one.
- 1:10:19So we have seen especially during periods of competition,
- 1:10:22especially during periods of lots of, you know, like funding
- 1:10:25crunches, people focus on their core products and spend more on
- 1:10:28advertising and getting it to the market.
- 1:10:30That's a discipline way to do things actually.
- 1:10:32So I think we're doing better as a content company now because it
- 1:10:36falls cleanly under the department that people will
- 1:10:39spend on when there's a crisis. And we see especially with
- 1:10:42content, when you're creating content for a company, it's like
- 1:10:45it's organic. So you know that the spends are
- 1:10:47slightly high in the early days, but if the channel works, your
- 1:10:50spends are lower in the later days.
- 1:10:52You want organic distribution. So you could go after organic
- 1:10:54distribution just right place, right time.
- 1:10:56I guess that's got a little bit lucky because we just followed
- 1:11:00this AI curve to the point where Mark Zuckerberg once said this.
- 1:11:03He like don't, don't follow the local minima, a local maxima.
- 1:11:06Like just because you got one idea I don't go double down on.
- 1:11:08We built this tool called Alpha Ctr. to solve thumbnails.
- 1:11:11We got to double down on. We could have said, hey, we're
- 1:11:13going to be the best AI thumbnail tool, but we're like,
- 1:11:15no, no, no, good idea. But let's keep checking.
- 1:11:17Let's keep seeing the surface of the wall for where the highest
- 1:11:20value here is Alpha Ctr. and a bunch of other apps we built.
- 1:11:24They're doing well. We've got more than 1,000,000
- 1:11:26users on them. Really.
- 1:11:27Like, this is going to die at some point, right?
- 1:11:29Or this is not where the end value, it's the guy who uses
- 1:11:31this to make content is going to make the most value.
- 1:11:33So we should be the person, you know, taking advantage of making
- 1:11:36most value. Yeah.
- 1:11:37Wow. So you're right.
- 1:11:39Now you're looking at the products and services that are
- 1:11:42out there that are providing that, that are solving a
- 1:11:45problem, and you use those to provide value.
- 1:11:47But then eventually, as you're experimenting, you would like to
- 1:11:50solve problems yourself. Yeah.
- 1:11:51And it's replacement. More than anything, it's
- 1:11:54replacement, right? For example, face swapping in
- 1:11:56the early days, we should use other tools.
- 1:11:58Now we're sort of building our own model for face swapping just
- 1:12:01like we're not doing much except just capturing.
- 1:12:04Like there was a company in India, not going to tell you
- 1:12:07their name, but they started with printing that they realized
- 1:12:09there's a business in ink and they realized there's a business
- 1:12:11in publishing that they realized there's a business if you're
- 1:12:14publishing and there's a business in schools if you're
- 1:12:17publishing anyway. So stuff like that, it's just
- 1:12:20value chain after value chain picking up new stuff.
- 1:12:24And it comes from experimentation because until
- 1:12:26you experiment, you don't know if it works or not, right?
- 1:12:29So we don't have a lot of experiments.
- 1:12:30We fail on like 8 or 10 things we do.
- 1:12:32We're only hearing of the successes.
- 1:12:34Lots of things we've tried that have not worked and we just
- 1:12:37scrap it quietly and we're like, this is part of the game.
- 1:12:40How do you handle failure? At what point you say that oh
- 1:12:43this doesn't work, let's go to the next thing.
- 1:12:45It's actually very quick. The cool part about having
- 1:12:47distribution is I can tell you right after putting out a post
- 1:12:51within 30 minutes from the website activity or whatever
- 1:12:54we're selling activity or how the comments are, I can tell you
- 1:12:56whether this is something viable or not.
- 1:12:58It sounds odd. That's how you're testing your
- 1:13:01ideas then? Yeah, we just put out a video.
- 1:13:03We see what audiences think. Like, for example, I know, let
- 1:13:06me talk about content first. We know people don't like
- 1:13:08content about robotics and people will not buy robots
- 1:13:13unless it's solved, like a lot of issues.
- 1:13:14They wouldn't buy like simple robots because we've tried that
- 1:13:17a few times. Content, we've tried some
- 1:13:18experiments. They don't care.
- 1:13:19There's no engagement on it. It's not an interest area for
- 1:13:22people. It's not hot yet.
- 1:13:24Games as well, right? Like like games.
- 1:13:27We know there's no market, right?
- 1:13:29You talk about AAA game in India, a lot of people get upset
- 1:13:31just that the minute you put a price tag on it, it will leave,
- 1:13:34right? We have a free demo that people
- 1:13:35can play. Nobody plays it right.
- 1:13:37So interesting. So, so stuff like that.
- 1:13:41And with businesses, if you have a landing page, you make, you
- 1:13:44know, video, put it on all your channels.
- 1:13:45And we have a lot of channels. We'll know very quickly if it
- 1:13:47works or not. Yeah, it's, it's mostly that
- 1:13:52it's like distribution is really valuable.
- 1:13:54Once you have distribution, you can test things, you can hire
- 1:13:56people because you just put up a hiring post.
- 1:13:58People join you. You're able to get new leads and
- 1:14:02new business. I mean, I don't know why for so
- 1:14:05many years of my life I thought content creators were idiots
- 1:14:07when I was much younger and I thought I was stalking on screen
- 1:14:11or whatever. I just didn't realize how
- 1:14:12valuable it is. You have a.
- 1:14:14Great ecosystem of try and error using all of your channels.
- 1:14:18You basically can have an idea and then just try it
- 1:14:21immediately. That's that's priceless.
- 1:14:23Yeah, I mean, think about the math, right?
- 1:14:25Like some some of you, you've probably done this in the past.
- 1:14:27Buy a domain, make a landing page, pull it up, put a stripe
- 1:14:32thing on it. If it's a service, you can
- 1:14:33actually build the product. If necessary, build an MVP of
- 1:14:36the product. We can do it all super quick and
- 1:14:38the better AI gets, we can do it even faster.
- 1:14:41Like I want to go make an MVP in like, like I want agents to get
- 1:14:43to the point where I can make an MVP in no time.
- 1:14:45Make a video about it. Even if the product feels it's
- 1:14:48OK because we have whatever, we'll have like at least half a
- 1:14:51million people see it and at least you know, 23,000 people
- 1:14:53try it. We'll know immediately if it
- 1:14:55works. Just don't price for the first
- 1:14:56this thing, because you set up the wrong expectations to price
- 1:14:58something and you're killing it and it's not cool.
- 1:15:01But yeah, as long as you set expectations of the audience
- 1:15:03that this is a trial, they're OK with it.
- 1:15:05But the audience is completely OK if you tell them this is
- 1:15:07trial and I'm just figuring things out.
- 1:15:11You know, clearly a lot of people get their sorts of news
- 1:15:14from you. Where do you get your source of
- 1:15:16news? Twitter Reddit, Reddit's become
- 1:15:20like horrible now, so we don't get any unbiased news from
- 1:15:23Reddit's already biased, but Twitter, sometimes we look at
- 1:15:26research papers archive. There are a bunch of different
- 1:15:31places that you know, we, we, we get our news from Sometimes it's
- 1:15:35other people's YouTube channels. Sometimes there'll be a YouTube
- 1:15:38channel where they do a deep dive on like I was reading, I
- 1:15:41was watching this deep dive on EACC right, which is effective
- 1:15:45acceleration. And there were lots of things
- 1:15:48for and against it. And I was like, should I make a
- 1:15:50video about this? Maybe my own video, but topic
- 1:15:54like my thoughts on because people only hear my thoughts on
- 1:15:57that topic. Sometimes it could be that,
- 1:15:59sometimes it could be, you know, a lot of Twitter noise around
- 1:16:04something. So in a way, we bucket content
- 1:16:09in different places. Like sometimes it's Twitter
- 1:16:12import, export in a way, sometimes it's paper import
- 1:16:16export. I mean, I'm simplifying what we
- 1:16:20do. It's a lot of effort.
- 1:16:21Just love to do inside. But yeah, we need to look for
- 1:16:24interesting sources sometimes in my own opinion, sometimes just
- 1:16:26like, yo, I have thoughts, I'll make a video about this.
- 1:16:29So it's mostly that sometimes it's the reactions to stuff that
- 1:16:34have happened on Instagram. They have a very diverse, like,
- 1:16:37I can't tell you next week what content we'll be doing because a
- 1:16:40lot of it is time sensitive. How far ahead you plan for the
- 1:16:44content? Because we were able to make
- 1:16:46videos in 1530 minutes, not very far, right?
- 1:16:49So not very far. I we have an interesting idea
- 1:16:52today. It can be done today, which is
- 1:16:54incredible. That is, that is a new way of
- 1:16:57looking at things. I, I, I've worked in companies
- 1:17:00where you have two weeks worth of content in advance.
- 1:17:03But I guess for you everything is so time sensitive and the
- 1:17:05fact that you guys can do it so quickly.
- 1:17:07Now, do you apply that same method to the companies you're
- 1:17:10working with? Because I understand for other
- 1:17:11companies there's that delay of you guys reviewing things with
- 1:17:14them and, and also them understanding the technology and
- 1:17:17how fast you guys work. How does how does that
- 1:17:19relationship work with the companies?
- 1:17:21So with external companies, they love us like in the sense that
- 1:17:25they often just give us the steering wheel after a couple of
- 1:17:27months, they're like, this is working, just keep you to you,
- 1:17:30right? And there's only a problem if
- 1:17:31you put up something that they don't like.
- 1:17:33But they've never done that, not even once, right?
- 1:17:36So I think it's good. My dream is to take that
- 1:17:39services company and make it a pureplay product company at some
- 1:17:41point in life, right? And I think it's possible,
- 1:17:45right? Every year with AI with every
- 1:17:46new advance it's possible and even if you don't manage to make
- 1:17:49it a full like product company, even if you have a little bit of
- 1:17:52services, that's OK. Like Salesforce does have a lot
- 1:17:54of services right even in the product, even though they're a
- 1:17:58product company. So we just want to make sure
- 1:18:00that we're able to do it very lean, very efficiently, and that
- 1:18:03should be good enough. I.
- 1:18:04Love that Varun, thank you so much for your time.
- 1:18:08It's been a pleasure speaking with you.
- 1:18:09Honestly learnt a lot about what you do and how you and your team
- 1:18:13have managed to create content at scale but with quality, which
- 1:18:17is something that is unheard of, especially about the times that
- 1:18:19you're talking about the speed. And we're looking forward to see
- 1:18:22what you guys create. We're very interesting in the
- 1:18:25way you approach not just leading the team, but also
- 1:18:29looking at how you want to manage different problems in the
- 1:18:32world using the technologies that you learn about.
- 1:18:34So we're very excited to see what you create.
- 1:18:36We cannot wait to see everything.
- 1:18:39Absolutely. In fact, I know, I mean, I
- 1:18:42actually want to hear a little bit about you guys, right.
- 1:18:43I'm sure of course you're doing, you're doing a lot, lot of stuff
- 1:18:46on 3D. What's the business behind the
- 1:18:47scenes? I'm sure there's a business
- 1:18:48behind the scenes. Oh yeah, 100%.
- 1:18:51Bad Decisions is twofold. 1 is the content machine that they
- 1:18:55create tutorial series. This is the podcast that we are
- 1:18:58doing and we do a lot of tech updates and then we have Bad
- 1:19:01Decision Solutions. Then we deliver Immersive
- 1:19:04Experiences. We created recently a fortnight
- 1:19:06map for can we say the client name?
- 1:19:09I think we can say the client name for Dubai Police.
- 1:19:11We are creating Immersive experiences, 3D immersive
- 1:19:14experience for clients here in UAE.
- 1:19:17We are now in Dubai and in Canada as well.
- 1:19:20Wonderful. So one of the cool ideas I had,
- 1:19:22if you're open to it, is we should do, we should do a video
- 1:19:24together like we should do a proper CGI video together
- 1:19:27because I know you guys are great at Unreal Engine.
- 1:19:29We have a good team as well. We can collab and do like 3 or 4
- 1:19:33crazy videos. Let's do it.
- 1:19:35I think that's a fantastic idea. I honestly believe that we can
- 1:19:39do something really sick together because what you guys
- 1:19:41are creating is awesome. We work a lot behind the scenes,
- 1:19:44like you said, when it comes to Unreal Engine, but also Blender,
- 1:19:46because you did use the word Blender earlier on.
- 1:19:48I do have to mention our team is Blender, Unreal Engine.
- 1:19:51We do explore other software as well.
- 1:19:53We do touch Houdini a little bit, but Houdini is pretty damn
- 1:19:56complex. But when it comes to the
- 1:19:59business side of things, I mean, me and Farah were discussing
- 1:20:01this. The beauty about Unreal Engine
- 1:20:02is that it can be used for so many different worlds, right?
- 1:20:06You can talk about real estate, it's useful.
- 1:20:08You're talking about game dev, it's useful.
- 1:20:10Fashion because it's real time, events because it's real time.
- 1:20:14So many different industries are using Unreal Engine and I can
- 1:20:18see that for the foreseeable future.
- 1:20:20Again, I don't know how AI. Will change things in 10 years
- 1:20:23time, but at least for the next few years, I know Unreal is
- 1:20:25going to be the king when it comes to a lot of different
- 1:20:27industries. And so that's been our main
- 1:20:29focus. We do work with a lot of
- 1:20:31clients. Most of them use our Unreal
- 1:20:33Engine services. But at the same time, the reason
- 1:20:35why we were discussing content with you is because
- 1:20:38understanding social media is not easy.
- 1:20:40I mean, everybody can make videos, but like you said,
- 1:20:43there's taste to it, but there's also skills of understanding the
- 1:20:46trends and what people will watch and listen to.
- 1:20:48So a lot of the times in, in the clients we worked with, we do
- 1:20:52offer social media services as well, because naturally they
- 1:20:56will see the content that we create and they're asking us,
- 1:20:58hey, why don't you guys create content for us?
- 1:21:00We're like we never, we never thought of ourselves as a social
- 1:21:02media agency, But I guess we can do it better than most people
- 1:21:05that you can work. With actually the funny story is
- 1:21:07the first time that it happened, we were delivering our normal
- 1:21:11services and the client asked us why don't you guys do social
- 1:21:15media? Your social media is bigger than
- 1:21:17our social media agency. You guys should do it their.
- 1:21:20Social media agency had 2000 followers and we had like over a
- 1:21:23hundred 100K. So they said why don't you guys
- 1:21:26become a social media agency? We said we never that was not
- 1:21:28our goal but we can do it. I mean, if you guys.
- 1:21:30Want us to do it together? Let's give it a try.
- 1:21:32And we did in our video perform much better than their own
- 1:21:35social media agency. So we do offer that service as
- 1:21:38well. It is, it's really fun.
- 1:21:39I mean, understanding the psychology behind social media
- 1:21:41is essentially understanding the psychology of people, the the
- 1:21:45overall psyche of society at a certain point in time.
- 1:21:48And that's always been. That's why I read so much.
- 1:21:50That's why I read so much. Yeah, exactly.
- 1:21:52Exactly, that's why I understood what you said about reading
- 1:21:55books because if you understand the psychology of people you
- 1:21:57will understand what works on social media.
- 1:21:59Why are people watching TikTok? At a certain point in time, it.
- 1:22:03Also extends to business. It's also extends to business.
- 1:22:05What people would buy. It's also related.
- 1:22:08Yes, yes, it's quite important, understanding the psychology of
- 1:22:12humans because we're fucking dealing with them every fucking
- 1:22:14day. We should do a video on that.
- 1:22:15Yes. If you want, we can do a really
- 1:22:18cool CGI video on this. You let me know how we should do
- 1:22:21it. Happy to try an experiment.
- 1:22:24If it works. That's awesome.
- 1:22:25Which makes it serious. I, I think we have to come up
- 1:22:28with a bunch of different ideas and then see which ones are the
- 1:22:31most interesting to all of us, and then see how we would
- 1:22:34produce those ideas to be something unique, something
- 1:22:36different that one of our teams couldn't do on their own.
- 1:22:39We'll have to work together to collaborate to create that.
- 1:22:42And I think that video will be fucking awesome.
- 1:22:44Lovely, I'm excited to hear your ideas.
- 1:22:47Like my inbox is open and you have access.
- 1:22:50I freaking love them, man. Dude, thank you so much.
- 1:22:52I I would, I would sit down with Farrad and our team and we would
- 1:22:56definitely go over some ideas with you.
- 1:22:58And I guess for everybody watching and listening right
- 1:23:00now, they probably should be aware that some cool video.
- 1:23:03We're only covering. Videos are going to drop between
- 1:23:06our channels. Awesome.
- 1:23:08I love it, man. Varun, so I got to ask you, was
- 1:23:11this a bad decision? Because, I mean, you didn't
- 1:23:14mention it in the beginning. I got to ask you.
- 1:23:16I mean, not really. I think you know it.
- 1:23:18It was a good decision. But, you know, I was joking
- 1:23:20about the intro. Yeah, you did.
- 1:23:22You did. You did something that kept
- 1:23:23retention up for the intro. So I said bad decisions.
- 1:23:27Amazing. But yeah, this is awesome.
- 1:23:28This is fun. We got to go just the right
- 1:23:30level of technical for the audience watching.
- 1:23:31And yeah, you guys are great hosts.
- 1:23:34Thank you so much. That means a lot.
- 1:23:36And when you're in Dubai next time, we're definitely meeting
- 1:23:38up. Done.
- 1:23:40I love it, Vern. Thank you so much and for
- 1:23:43everybody watching and listening as well.
- 1:23:44You guys are amazing. We will see you guys on the next
- 1:23:47episode. Until then, Ciao.