College Dropout Raised $20M Building AI Tools | Cluely, Roy Lee

1. 0:00 – 2:07

   Cluely’s product direction: let users’ behavior set the roadmap

   1. AGAakash Gupta0:00

      How are you guys prioritizing what those engineers work on together? Are you guys building a roadmap together on some sort of cadence, deciding sprints, or any of those things normal tech companies do?

   2. RLRoy Lee0:10

      I feel like it's pretty obvious when you have so many users. Like, literally, when we started Cluely, we said, "Let's build the most general version of Interview Coder." We built Interview Coder for everything, and we've given it, like, a very controversial marketing flair and called it Cheat on Everything, and now we have, like, literally millions of requests coming in a day that literally tell us how people are using Cluely, and we have, like, hundreds of customer emails saying like, "Hey, this is what I hate about Cluely," and like, like, okay, okay, now I know what we need to fix.

   3. AGAakash Gupta0:34

      Roy Lee is one of the smartest and richest 21-year-olds you'll meet. With his company Cluely raising $15 million Series A last week, at typical dilution levels, he would be worth eight figures. What does Cluely make? AI that helps you cheat on interviews and tests, AI that can't be detected but can listen to your computer audio and read your screen. What do I need to know about your early years to understand the Roy that's sitting in front of me today?

   4. RLRoy Lee0:58

      I think the single most defining character feature of me from the second I gained consciousness was, um, provocative.

   5. AGAakash Gupta1:05

      So you've been kicked out of not one, but kinda two schools, with Columbia as well.

   6. RLRoy Lee1:10

      Yeah, yeah, yeah. After that one time, like, you get kicked out of school once, and it sort of, like, really toughens your mind and really makes you rethink, like, do I even need school in the first place?

   7. AGAakash Gupta1:18

      Talk about being at the top of the timeline. You guys are the main characters in the X For You page right now. How does that feel?

   8. RLRoy Lee1:24

      It feels like I have all these assumptions about virality, and I've spent months wondering, like, man, I wonder why X feels so different from Instagram, TikTok.

   9. AGAakash Gupta1:32

      One of the most interesting things I saw recently on Twitter was people claimed your system prompt was leaked. I have that up on the screen right now. Is this what was actually your system prompt? [beep] Really quickly, I think a crazy stat is that more than 50% of you listening are not subscribed. If you can subscribe on YouTube, follow on Apple or Spotify Podcasts, my commitment to you is that we'll continue to make this content better and better. And now on to today's episode. Welcome to the podcast.

   10. RLRoy Lee2:04

       Thanks for having me, brother.

2. 2:07 – 3:52

   Roy Lee’s origin story: provocative by nature, expelled, and refocused on startups

   1. AGAakash Gupta2:07

      What do I need to know about your early years to understand the Roy that's sitting in front of me today?

   2. RLRoy Lee2:13

      I think the single most defining character feature of me from, like, literally the second I gained consciousness was, um, provocative. Uh, like, literally since elementary school, kindergarten, there's been a camp of people who have been, like, avid Roy supporters and a camp of people who have been Roy haters. Every single time I've had a thought that's entered my mind, I've said it out loud publicly and, like, very brazenly, and this has, like, burned me and helped me many times. This has, uh, gotten me, you know, like, like, in school, leadership positions at clubs, and a bunch of people who love me, like a solid friend group, and it's also gotten me, you know, like, like, suspensions from school. I spent, like, every other Sat- every other Saturday in Saturday school and, like, a bunch of... a million detentions, and it e- even even- eventually got me, like, kicked out of Harvard. So, um, there, there, there's been a lot of, like, like, burns and a lot of gains from me being the polarizing person that I am.

   3. AGAakash Gupta3:00

      You were kicked out of Harvard?

   4. RLRoy Lee3:02

      Yeah, yeah. Senior year of high school, I, um, got into Harvard early as probably a result of me doing all this crazy shit, um, and later that... Like, like, I had a suspension for sneaking out of a field trip, um, and later that year, there was a camp of Roy, Roy haters who did not like a- all the brazen, crazy shit that I did. Um, they, they, they mass reported me, and, and it ended up getting me suspended, uh, rescinded from Harvard.

   5. AGAakash Gupta3:26

      Wow. So you've been kicked out of not one, but kinda two schools, with Columbia-

   6. RLRoy Lee3:30

      Yeah

   7. AGAakash Gupta3:30

      ... as well.

   8. RLRoy Lee3:31

      Yeah, yeah, yeah. I mean, after that, after that one time, like, you get kicked out of school once, and it sort of, like, really toughens your mind and really makes you rethink, like, do I even need school in the first place? And after that first time I got rescinded, like, my mind was locked onto, okay, I've always been unconventional, and I've always been successful at things that I've done. I need to channel this into, like, the biggest, boldest play ever, and that's, like, building a company, bro.

3. 3:52 – 4:25

   Cluely’s early traction: rapid ARR growth and “top of timeline” distribution

   1. AGAakash Gupta3:52

      Building a company indeed. And last time you shared the stats, it was at 5 million ARR. What's the latest for Cluely's growth?

   2. RLRoy Lee4:00

      Yeah, yeah, yeah. We're, we're, we're almost at six, so we're, we're, we're going... we're growing quite quickly, and I think, um, one thing that people don't, don't realize and often forget is, uh, like, we wrote the first line of code for Cluely, like, 10 weeks ago. Um, we are earlier stage than, like, most YC companies of the latest batch. Uh, but it's, it's hard to remember that because it feels like we've been on top of the timeline forever. But, um, it's just a function of a really world-class amazing marketing team.

4. 4:25 – 8:26

   Virality on X/LinkedIn: controversial, digestible, and engineered for reactions

   1. AGAakash Gupta4:25

      Talk about being at the top of the timeline. You guys are the main characters in the X For You page right now. How does that feel?

   2. RLRoy Lee4:33

      It's pretty crazy, and it's really, really, I'd say, like, vindicating. Not... No, I, I, I, I, I, I think crazy is the most general and, and, and specific that, that this feeling can get. But it feels like I have all these assumptions about virality, and I've spent months wondering, like, man, I wonder why X feels so different from Instagram, TikTok. It feels like on Instagram, TikTok, there's creators who have cracked the algorithm. I wonder why the Tech Twitter feed looks so different. Like, why don't people apply these concepts on Twitter and just go hyperviral like they should? Um, and I've had all these assumptions, and as soon as I started implementing them, like, they started working immediately and to, like, a massive scale. Like, I've literally become the main character of Tech Twitter in the span of a few weeks, like a few weeks of pressing the button that I feel like is so obvious and clear to everyone. Um, and it feels crazy that nobody else has caught on. It feels like I genuinely am pioneering something.

   3. AGAakash Gupta5:29

      What have the millennials like me, what are we not understanding about your approach to virality?

   4. RLRoy Lee5:36

      I think what people don't understand is that short-form, like, these algorithms reward what is most controversial. And people on X, LinkedIn, like, these are nerdy tech bros who want to appear smart or like, like, like, thought-provoking, and they'll write these long essays that maybe, like, 200 people in the world can understand in, like, an attempt to appear smart. But in reality, the algorithm rewards things that are digestible, controversial, and force people to have an opinion about. And for some reason, people on X, LinkedIn just have not caught on about, like, what are the best ways I can, I can grow my company and, like, like, bring visibility to my company in doing this. And it's, it's just like, like, obviously, like, the nerdy tech bros that grow up and start tech companies are just not the same people that will be scrolling on Instagram, Instagram, TikTok and just be-Caught up on the latest in cultural trends, but, um, I think I am part of a new generation of founder that, that has. Like, I've grown up on, on, on this shit.

   5. AGAakash Gupta6:26

      Yeah, we grew up on Facebook and stuff, and I think it was always about, like, presenting your best self. Even on LinkedIn, it's, like, about presenting this curated version of yourself. But you're willing to present kind of a more controversial side. How do you correctly create content that's controversial but still drives to your product? Today's episode is brought to you by the AI PM Certification on Maven. Run by Miqdad Jaffer, who is a product leader at OpenAI, this is not your typical course. It's eight weeks of live cohort-based learning with a leader at one of the top companies in tech. OpenAI just doesn't stop shipping, and this is your chance to learn how. Run along with product faculty and Mo Ali, the course has a 4.9 rating with 133 reviews. Former students come from companies like OpenAI, Shopify, Stripe, Google, and Meta. The best part? Your company can probably cover the cost. So if you want to get $500 off, use my code AAKASH25 and head to maven.com/product-faculty. That's M-A-V-E-N.com/P-R-O-D-U-C-T-F-A-C-U-L-T-Y. Today's episode is brought to you by Jira Product Discovery. If you're like most product managers, you're probably in Jira tracking tickets and managing the backlog. But what about everything that happens before delivery? Jira Product Discovery helps you move your discovery, prioritization, and even roadmapping work out of spreadsheets and into a purpose-built tool designed for product teams. Capture insights, prioritize what matters, and create roadmaps you can easily tailor for any audience. And because it's built to work with Jira, everything stays connected from idea to delivery. Used by product teams at Canva, Deliveroo, and even The Economist, check out why and try it for free today at atlassian.com/product-discovery. That's A-T-L-A-S-S-I-A-N.com/product-discovery. Jira Product Discovery, build the right thing.

5. 8:26 – 12:31

   Stop obsessing over funnels: iterate formats faster than the algorithm shifts

   1. RLRoy Lee8:26

      I think this is something that less people have to worry about. Like, there's way too many people that worry about a highly converting marketing funnel and all this bullshit, but in reality, there's no one else out there with more brand visibility than me, and there's probably not a company out there that would not die for, to have this sort of visibility on their product, especially at this early stage of being, like, two and a half months, um, in. And, um, I think it's really hard to put a monetary value on this because how exactly can you track all the conversions from literally being the, the main character of Tech Twitter? Like, it's impossible. More than that, the, the formats that will be viral, these change, like, literally overnight. Overnight, you need to know the new thing that will be viral, and you have no i- And if, if your content is changing literally every week, then how are you supposed to... Like, if you find a winning content format, it's not really like you can rinse and repeat it, 'cause in a month that's expired and it's not viral anymore. You just have to keep iterating and trying new things, and I think it's pointless to track conversions on bull- and then, like, bullshit like that because, uh, if you just know, "Hey, my product is visible here. If I was a consumer and I saw this pro- and I saw this video, I would actually be incentivized. There's an actual call to action here," then the video will be converting and you don't have to worry about it. Because even if you discover, oh, it's converting and it's like, like, roughly like a $2.73 second, like, like, like, like, it, it's, it's all, it's all noise, bro. Like, like, in a month, the numbers mean nothing and they're expired.

   2. AGAakash Gupta9:44

      The growth marketer in me died inside a little bit hearing that.

   3. RLRoy Lee9:48

      Perhaps, yeah. I mean, like, like, may- maybe I'm wrong, but it's really, you cannot put a dollar value on the, the visibility that we have right now.

   4. AGAakash Gupta9:57

      It's crazy. I would say you guys are easily the best distributing startup at the Series A or B range. If you were to break down the strategy for your friend who wanted to kinda copy your playbook, what would be the key elements?

   5. RLRoy Lee10:10

      A lot of people don't know, but we're averaging, like, in, in our first month of UGC on Instagram, we've done, like, we're probably on track to do, like, 50 million views a month. Um, it's not just X LinkedIn that we're viral in. We're, like, pretty much viral everywhere. And, uh, there's things... Like, the, the one thing that you need to understand, I think, is that since five years ago, uh, short-form content has surpassed long-form content in views and, and, and, and, and, and usage.

   6. AGAakash Gupta10:37

      Yeah.

   7. RLRoy Lee10:38

      And this has resulted in a world where the number of content creators has stayed the same, but the number of content consumed has about 100x'd. As a result, there is not enough viral content in the world for a consumer to consume. What does that mean? That means that as a s- as a consumer, you're scrolling on TikTok waiting for more things to see, but instead you're met with the same Minecraft Subway Surfers meme on a Reddit story. Like, you see that 100 times because there's literally not enough quality viral content out there for you to see. Meaning right now is a unique moment in human history where if you make a post or say something that deserves to be seen by millions, you will be seen by millions. And this is probably, like, like, the most unique, unique moment in human history. And I think, I think something that everyone needs to know is, uh, you need to make more content with the potential to go viral, and this is sort of like a, like a dual-threaded thing. Um, and this is, this is what I try and harp on every single time. Uh, the components of virality, at least on X LinkedIn, are, A, it has to be digestible. This is a problem that almost everybody gets wrong. Like, your content is not digestible. You cannot imagine that everyone... Like, you have, you have to imagine that if everyone in America saw your tweet, they would understand what you're talking about.

   8. AGAakash Gupta11:44

      Mm-hmm.

   9. RLRoy Lee11:44

      But almost nobody on Twitter does this. Um, all of their content is just like, like, like, like unexplainable blockchain, crypto, Web3, AI bullshit. Like, like, no- nobody will understand this, so it will not go viral. That's the first thing, and the second thing is you need people to have a reaction to it. You can't just post something and expect... Like, like, nobody wants to see just a regular post. Like, you, you have to really picture the world where someone, quote, posts your, your tweet, and the quote post is viral. What would the quote post say? If you cannot think of something, then you need to change your original post, 'cause it needs to be engageable. Um, and th- those are the components of the virality, and if you, you have that, those two things, it's digestible and it's reactionable, and you, you turn outEven, like, one of those a day, like, you will just end up on the top of the timeline, and that, that is what constitutes virality.

6. 12:31 – 13:59

   Stunt pipeline and content ops: daily brainstorms, UGC scale, and “reverse engineering” trends

   1. AGAakash Gupta12:31

      You're doing a lot of stunts along those lines, I would say. Like, there was a stripper somewhere involved. [laughs] There was a party that got shut down by San Francisco Police Department. How are you coming up with these ideas? What's your own pipeline for generating these viral, viral ideas?

   2. RLRoy Lee12:47

      I personally think I have quite strong viral sense. I spend a lot of time scrolling on Instagram, TikTok, like right now almost as a mandatory keeping up with the, the, the, the pace of the culture. Um, and you just need this viral sense, and when you have it, then it w- like, these things come obvious, obvious to you. Um, about every single day we have, like, a marketing brainstorming session where we'll just, like, like, unscheduled, we'll just spend an hour talking about what is the next crazy stunt that we can do. And because everyone here is an influencer with viral sense, they know what goes viral, like, everyone spits out these crazy good ideas. And, like, literally every day we come up with 20 ideas that have the potential to get 100 million views each. And, um, it's, uh, it's, it's sessions like this that you need, and it's sessions like this where, like, the whole, like, let's do stripper commercial, or let's do a, let's, let's do a, like, like, like, partying and getting, like, like that, that... These brainstorming sessions are where that comes from. It comes from a group of highu- a highly dense group of people with viral sense just brainstorming together.

   3. AGAakash Gupta13:41

      How old are you guys? Is there anyone in the 30-plus range?

   4. RLRoy Lee13:46

      One person, yes.

   5. AGAakash Gupta13:47

      One person.

   6. RLRoy Lee13:48

      Yeah.

   7. AGAakash Gupta13:48

      So primarily younger people who really are understanding the heartbeat of these short form videos. They get together, you guys are coming up with 20 ideas, and then take me through idea selection, filming, uploading.

7. 13:59 – 16:29

   From Ghibli trend to “Cluely anime”: turning a cultural insight into a planned viral moment

   1. RLRoy Lee13:59

      Yeah. I mean, every day we're trying to figure out, like, like, we're almost, like, I, I, I would say what we're doing in these meetings is reverse engineering the algorithm. We'll talk about what we've seen lately that's gone viral, and then we'll try to dissect at, like, the core level why did this go viral and what can we learn from it. Um, something that's really interesting that will come out in the near future is a Cluely anime, and, um, this will start with a trailer, just, like, tease the waters. But, like, like, this conversation started with, you know, like, ChatGPT-4o image, and the whole Studio Ghibli trend took over, took over the internet by storm. We just kept on wondering, like, why did this happen? It wasn't super controversial. Like, what, what was the, what was the super viral component to this? And, and we think, like, among many things it is that the people in tech are people who grew up watching anime and Studio Ghibli, and this is, like, such a nostalgic, deeply emotional, provocative feeling for them that, um, there's something to be gained from this. And then we just talked down, "Well, what are things we could do?" Eventually we settled on, "Man, let's, fuck it, let's make a Cluely anime." And that ends up being a viral, viral moment. And, like, you can, like, keep this on record. This will do, like, over 100 million views, and I, I, I'd bet my life on this. Um, the Cluely anime will be, like, an extremely viral moment, and, uh, this is just one of many things that we have coming up.

   2. AGAakash Gupta15:09

      There's, like, a huge bridge between identifying a good idea like an anime, like even potentially I could do that, and then actually making it happen. So walk us through kinda your playbook of that. Obviously, it sounds like you're hiring people who are really good and have proven they've already gotten views. How are you, like, kinda further enabling them to make sure that they can create this good content?

   3. RLRoy Lee15:30

      Uh, yeah. I mean, we, right at this point we have an in-house UGC studio of over 60 creators who are on retainer, on schedule, and, like, like, that system is just running. Um, we have, like, an in-house influencer marketing reaching out schedule, and that, that system is just running. We have an in-house, and we're, we're building out this in-house, like, entire film studio where we literally have videographers, editors working to make, like, movie quality launch videos every fucking week, and that, that's up and running. And if there's ever anything that, like a content format that we don't know how to make, then we'll just, like, we all have agency here. We'll, we'll just look up a studio that, that can do it, and then we'll just, like, we'll, we'll find a way to make it happen. If you're convinced enough in the virality of a moment, then, like, how much effort would you put into a video that you know with certainty is going to do 100 million views? A lot of fucking effort probably, so it- it's, it's really not that much overhead. And if it turns out that this is super repeatable, then we might just, like, try and do it in-house, and we'll try and find an animate- animator studio to, to do this entirely in-house for us, but we're not there yet.

8. 16:29 – 18:08

   Cluely’s UX thesis: translucent overlays as the future AI interface

   1. AGAakash Gupta16:29

      So we talked a lot about your marketing. When I asked people what I should be talking to you about, everybody wanted to get deeper on the product.

   2. RLRoy Lee16:36

      Yeah.

   3. AGAakash Gupta16:36

      And let's start with that. I think that the coolest thing is your UX. I think it's actually a new paradigm for UX that I hadn't seen before. This started with your product before Cluely, Interview Coder, where you came up with liquid glass essentially before Apple.

   4. RLRoy Lee16:50

      Yeah.

   5. AGAakash Gupta16:51

      Talk us through your process of developing that UX.

   6. RLRoy Lee16:55

      Yeah. I mean, Interview Coder is sort of like the prototype of Cluely, and this started out as, like, th- th- this was the beginning of everything for me, and it was a tool to let you cheat on technical interviews. And I spent about two months just, like, really trying to narrow down, like, like, like, what the fuck is the, the most extreme, uh, UX that I can do so an interview w- interviewer that I'm in an interview with is not gonna be able to know that I'm using AI. And, like, I need AI to spit out an answer, and I need to be able to write that answer myself. So, like, like, like, like, like, br- like, bro, the, the obvious thing here is translucency. Um, you need to have code overlaid on top of your workflow and code that, that, that, that, that you can see below it, and this is, this is something new. Like, like, like, n- nobody's tapped into this. Um, like, everybody has, like, a separate app, app for AI or, like, a separate window, and you need to split screen AI, but that's not what AI should feel like. I mean, that's, that's not, like, the capability of AI is, like, transformative and, and, and the, the, the true graphical interface of AI has to feel seamless and integrated into every single workflow, and that's what, that's what the translucency comes from. Um, and I think that's, that's, like, obviously where Apple is headed with liquid glass. They're, like, prepping us for a world where AI is everywhere and, and, and the way it will be shown is this translucent overlay. Um, yeah.

9. 18:08 – 19:38

   Design without designers: brute-force iteration and founder-led product taste

   1. AGAakash Gupta18:08

      How did you personally kinda design that? 'Cause I think the design is amazing. Like, did you have a lot of iterations? Did you get it right pretty early on? How did you get there?

   2. RLRoy Lee18:16

      Yeah. There were probably, like, 20 to 30 versions of Interview Coder that we tried before we landed on this final version, and, um, this took, like, a lot of, uh-A lot of back to the whiteboard, but, I mean, this is like just ended up being the best, most seamless version that, that, that, that we came up with. Um, yeah, it was, it was literally just brute force. Brute force towards the best product.

   3. AGAakash Gupta18:36

      And do you guys have a designer on staff?

   4. RLRoy Lee18:39

      No, no, no. The design... Like, like the product design is just led by, by us, like the founders.

   5. AGAakash Gupta18:44

      By all crack engineers primarily. Is that how you're gonna continue building the company?

   6. RLRoy Lee18:50

      Yeah, I mean, the, the, the thing about company building is when you're really good at one thing, the other things sort of come. If you have really, really good engineers, then the distribution will come, and if you have really, really good distribution, then good engineers and good product will come. Um, and this is sort of what we're seeing, like, like our inbound is, like, pretty massive. It's crazy. Like, we have people from every company in the world. We have the best engineers who wanna come, um, work, work with us just as a function of our distribution being so good, and this will in turn lead to our product being amazing and that will in turn like, like funnel all the other things. And, um, generally, company building is a positively reinforcing cycle of success, and when you have your first big win, this will just... Winners affect you into having more and more big wins. Um, and yeah, like these engineers, they, they know what they're doing. They have great product sense, and they, they, they, they, they help us iterate closer and closer to the end state of Cluely.

10. 19:38 – 21:17

    Early-stage execution: lean engineering team and user-driven prioritization (no formal sprints)

    1. AGAakash Gupta19:38

       How are you guys prioritizing what those engineers work on together? Are you guys building a roadmap together on some sort of cadence, deciding sprints, or any of those things normal tech companies do?

    2. RLRoy Lee19:48

       Bro, it's like I feel like, I feel like it's pretty obvious when you have so many users. Like, like literally when we started Cluely, we just said, "Let's build the most general version of Interview Coder, m- launch it, make it massively viral, and then see what people are using it for, and then we can like iterate towards like the actual product that we wanna build." And that's like, that's like exactly what we've done. Like we've, we've, we've, we've built Interview Coder for everything, and we've, we've given it like, like, like a very controversial marketing flair and called it Cheat On Everything. And now we have like literally millions of requests coming in a day that literally tell us how people are using Cluely, and we have like hundreds of customer emails saying like, "Hey, this is what I hate about Cluely," and like, like, okay, okay, like now I know what we need to fix, and now we know where the stickiest use cases are. Um, yeah, I'm, I'm not sure if that answered your question.

    3. AGAakash Gupta20:31

       I think the tough part is just like letting anyone do whatever they want, 'cause there might be four people trying to solve the same problem differently and trying to ship, you know, conflicting ways to solve it. How are you kinda organizing your guys' work to make sure it's on the highest priority stuff?

    4. RLRoy Lee20:48

       I think we're... I mean, the, the, the team is so lean. Like, we're so aligned. Like, like we literally have 4 engineers right now, so I mean just-

    5. AGAakash Gupta20:55

       Oh, wow

    6. RLRoy Lee20:55

       ... every product decision makes... Yeah, it's just, just getting made by the team. Yeah, and I th- I think it's, it's easy to, to forget, but we are really, really new, really, really early. Um, but we've just been on top of the timeline for so long. This is a function of my distribution strategy being correct that everyone thinks we've been around for forever, but in reality, like the first line of code was written 10 weeks ago, so it's, it's extremely new.

11. 21:17 – 24:07

    Technical architecture: audio capture, screen context, compression, and latency tradeoffs

    1. AGAakash Gupta21:17

       So the technical challenge I see you guys solving is how do we get kind of long context screen recordings plus audio transcripts that are two ways to a model to get like a pretty quick response. So what are the kinda innovations you guys have made on that front?

    2. RLRoy Lee21:35

       Yeah, yeah. I mean, we have our own custom audio engine that, that takes in system audio and microphone audio from the Mac. As far as we know, only Granola has, has something like this. Um, even like OBS Studio doesn't have this implemented yet. Uh, that's, that's one of the things that we have. Um, the other thing is we, we, we have like a pretty refined me- like, like for, for, for understanding the context of your screen in particular, uh, it... The models are not intelligent enough yet for you to just pass in like an ever-recording screen recording of what's going on in your screen and have it expect it to reason over that. Instead, we s- we, we take a screenshot at the time of query and pass that in with the rest of the token, and we have a specific image decompressor to like, like w- like the native Mac screenshot is like, is like decompre- is like, it's, it's, it's super HD PNG, and, and we have like, we have like a specific image compression to, to make that decrease the token size and, and sort of like decrease the input. And there's, there's, there's some things we do with, with servers and some things we're playing around with right now for, um, decreased lat- latency on the server request handling side. Like, there's a lot of, uh, variance with OpenAI. Like, sometimes you'll, you'll, you'll, you'll make a route when it's busy, and it'll take five seconds, but when it's not busy, it'll take one second. And there's workarounds to this, which is like host your own models on your own servers, then you won't have to deal with that variance. Um, th- th- those, those, those are all things that we're exploring.

    3. AGAakash Gupta22:53

       And right now it sounds like you guys are mainly built on top of OpenAI?

    4. RLRoy Lee22:57

       Uh, yeah, yeah, yeah. GPT 4.1 is, is the main model provider right now.

    5. AGAakash Gupta23:02

       Okay. So it sounds like you guys are gonna end up following a pretty similar playbook to like the Cursors and Lovables of the world, where you start to build in multiple different models for different use cases and start to eventually personalize these models for people.

    6. RLRoy Lee23:16

       Yeah, of course. I- inevitably, yeah.

    7. AGAakash Gupta23:19

       And when you look at that future, what is the, what is the type of investment you're gonna be need to be making? 'Cause I think like these model companies like Anthropic and OpenAI, they are like burning cash, like billions of dollars. Is that the type of investment you foresee making?

    8. RLRoy Lee23:34

       I don't think so. I think the application layer is just fundamentally different from the model layer. The race to AGI is competitive and lucrative enough that the, the model providers are busy building AGI, and in the meantime, there is like literally trillions of dollars of economic delta to be swept up by application layer companies, and that's exactly what we're doing. And as a result of us just being like application layer, we don't have to take upfront the costs of, of, of racing towards AGI, and we can afford to just have traditional SaaS margins, and it's just like the best place in the world to be right now.

12. 24:07 – 26:33

    Enterprise pull vs “AI for everyone”: features shipped and the product’s wedge

    1. AGAakash Gupta24:07

       What's the most successful sort of upgrade you guys have shipped since the initial version of the product?

    2. RLRoy Lee24:13

       Yeah, I mean, we've been focusing a lot, like, like we, we just signed a huge enterprise contract, so we've been focusing a lot on shipping features specifically for them. But pretty much every single enterprise feature, likePost-call summaries is, is, is, is, is probably one of w- one of the big ones. Every time you use Cluely in a meeting, you can end the meeting, it'll give you a summary of what happened in the meeting, and it'll also tell you times you should have used Cluely when you didn't use Cluely. And this is, like, the ultimate training tool for sales reps. Like, like, as a result, like you can, "Hey, customer had an X objection. You should have, you should have used Cluely to, to answer why, but you didn't, and as a result, like, you lost the deal." And this, like, literally spells out for you, "Here's why you lost the deal, because you didn't use Cluely at, at X moment."

    3. AGAakash Gupta24:53

       Yeah. I feel like the sales tech market is gonna be huge for you guys. I was a VP of product at Apollo.io, if you know that company.

    4. RLRoy Lee24:59

       Of course. Of course I know Apollo.

    5. AGAakash Gupta25:00

       Had CEO of ZoomInfo on the podcast. Like, all of his growth recently has been through these AI tools for salespeople. Do you envision kind of building, like, a Cluely for sales, or are you gonna continue to build it as kind of a product for everybody with maybe some specific sales features?

    6. RLRoy Lee25:17

       I mean, yeah, like, like, ChatGPT for sales is just ChatGPT, right? Um, that- that's like, it's like, sort of wh- what, what we're angling for. Like, there are enterprise, really specific enterprises use cases that are highly, highly profitable, but at the end of the day, we don't wanna just only capture the sales tech market. We want to be the default way that you use AI in the future. And a- as you've said, like, the most interesting thing about Cluely is the UX. I think this UX is not just good for tech- technical interviews and salespeople, like, it's good for everybody. And if it is good for everybody, if there is potential market fit for everybody, then we might as well try and capture the biggest audience that we can.

    7. AGAakash Gupta25:50

       Do you think that all AI apps are gonna have that kind of UX in the future?

    8. RLRoy Lee25:54

       Well, almost inevitably. I think, like, chatbot is just... it's, it's just wrong. The models will get better. Like, literally, the model you use today is the worst of mo- is the worst model you will ever use for the rest of your life. These models are inevitably going to improve, and when they do, they will be intelligent enough to take knowledge of your screen and be able to u- use it to give you, to give you a better answer. Like, like, it's just, it's just so obvious that chatbot is not the future. And like, Karpathy is over there, like, like, at Startup School talking about, "Bro, like, the, the new GUI for AI needs to be developed. It still hasn't been developed. Like, we, we, we, we need it. We're waiting for it." And Apple's coming out with Liquid Glass. Like, it is obvious that this is the bet that every single reasonable, rational human being is, in the world is making.

13. 26:33 – 31:41

    System prompt ‘leak’ and eval-driven iteration: prompt isn’t the moat

    1. AGAakash Gupta26:33

       One of the most interesting things I saw recently on Twitter was people claimed your system prompt was leaked. I have that up-

    2. RLRoy Lee26:38

       Yeah

    3. AGAakash Gupta26:39

       ... on screen right now. Is this what was actually your system prompt?

    4. RLRoy Lee26:43

       I mean, probably. Like, like, this, this gets changed, like, every three days probably. Like, like, l- the, the, the prompt is not the mode. We're gonna open source this anyways. Um, the specific system prompt, I, I mean, like, like, in the end state with the intelligent enough model, like, the system prompt will be, "Here's the audio, here's the screen. Help this user the way they need help." That will be the, the end state of the system prompt, and the model will be able to reason. Like, right now, we're just, uh, we're, like, or, like, adding in things to, to, to, like, like, like, like, try and make it better. But I mean, like, we're not selling the system prompt here. We're selling Cluely, and we're selling the specific user experience. Like, like, this, this has nothing to do with the product.

    5. AGAakash Gupta27:14

       Okay. That's really interesting, 'cause I'm a bit of a prompt engineering nerd, so I wanted-

    6. RLRoy Lee27:18

       Sure

    7. AGAakash Gupta27:18

       ... to geek out on this. But you're saying this is changing so fast, and it's not really that important.

    8. RLRoy Lee27:24

       Yeah, yeah. I mean, um, we, we have, like, custom evals in-house, and we're using them to, like, like, iterate towards a, uh, a better prompt. But at the same time, like, even our use cases, the main use cases for Cluely change every day. So, I mean, like, like, it, the, the, the system prompt also changes every day. Like, this is by no means, like, perfected or analyzed or, or, or iterated on at all. This is just, um, like, something we just came up with that seems to fit the evals and metrics of usage that we see today.

    9. AGAakash Gupta27:53

       One of the things I noticed experimenting with the product myself was that you have this personalized component.

    10. RLRoy Lee27:59

        Yeah.

    11. AGAakash Gupta27:59

        And I think that that really unlocks a lot of the value in Cluely. Are you encouraging people to use that?

    12. RLRoy Lee28:05

        Yeah, yeah, yeah. I mean, we're playing around with the new onboarding right now. Um, like, like, that will show them personalized, 'cause in reality, the, the general AI assistant, the tech is not there. But the tech for, "Hey, you are a sales assistant, and you, your express purpose is to help them make the sale. Here are common objections. Here's product knowledge," like, that works really fucking well. And even more so, like, "Hey, I am Aakash Gupta. I run this newsletter on XYZ, and, and I want you to respond XYZ way," like, that works really well. And again, like, like, the thing we're selling here is the, the UX, so I mean, we're, we're, we're playing around with the new onboarding to get people really used to the personalized. But yeah, you're right, that is the, the value unlock right now.

    13. AGAakash Gupta28:40

        Yeah. [laughs] As a former head of product growth at a SaaS company, that's what I jump to too. It's like the onboarding and then the information architecture just to make sure that, like, if people are using it regularly and they're paying, see, and they haven't personalized, to keep sending them [laughs] back to, like, personalize. Or maybe even, like you said, kind of getting the model to update its personalization on its own.

    14. RLRoy Lee29:02

        Yeah. Yeah, those-

    15. AGAakash Gupta29:04

        I think that'll be-

    16. RLRoy Lee29:04

        ... those are both things we're pretty interested in. Um, that, that, that will be default. What, what we're really hesitant to do is push out a feature that will get solved by a smarter model, and that, that, that's, that's why we're really hesitant about the personalized, 'cause in reality, if there was a smarter model, you wouldn't need to, like, push people to personalize. You could just have the smarter model just reason. Um, but, but, but we're not there yet, and we're, we're, we're still debating internally on what to do with that.

    17. AGAakash Gupta29:31

        So, if you're building cheating software for everything, how does a product manager cheat with Cluely?

    18. RLRoy Lee29:38

        Uh, yeah. I mean, there's like... Imagine you're on a customer call with someone, and you're trying to sell them your enterprise product. Uh, like, there, there's, there, there's many way, like, you're on a, you're on a call with a new person, you've just back-to-back meetings, you know nothing about this person. With Cluely, you can just, like, Command + Enter, "Who is this guy?" And it will just immediately tell you, surface the information, and every single time they say something that you don't know, um, like for example, "Hey, I spent the last eight years at Adobe Force," and you have no idea what that is. All of a sudden, you can just Command, enter, enter, and now you know exactly what Adobe Force is. And it's like, "Oh, yeah, yeah, Adobe." Like, like, you know everything about this person. You are now the default expert in, um, anything that he talks about, and this is, like, this is what cheating looks like on meetings. You all of a sudden have access to this breadth of information that you never had access to before thanks to AI, and the other person has no idea. This is cheating. Um-And, and th- th- this, this is, this is, as you can imagine, like, every single enterprise call or even just, like, a regular call in general, like, how many times do you say, "I don't know," or how many times do you say, "Hey, I, I forgot that last question"? Like, like, every single instance of this is solved when you have this AI that's listening and can surface relevant context.

    19. AGAakash Gupta30:41

        AI Evals are one of the most important skills for PMs, and I know you know they matter. The question is, are you doing them right? Most teams are winging it with basic metrics and hoping for the best. Meanwhile, the teams that actually ship reliable AI, they've cracked the code on systematic evaluation. Today's episode is brought to you by the AI Evals for Engineers and PMs course by Hamel Hussein and Shreya Shankar. This live Maven course will teach you the battle-tested frameworks from Hamel and Shreya, who are the engineers behind GitHub Copilot's evaluation system and 25-plus production AI implementations. Four weeks, live instruction. Next cohort starts July 21st. Start shipping AI that actually works. Enroll at maven.com with my code AG-PRODUCT-GROWTH for over $800 off. That's A-G-PRODUCT-GROWTH. There's this rumor

14. 31:41 – 36:47

    Big bets: fundraising preempt, revenue mix targets, and extreme end-state visions (CRM → brain chips)

    1. AGAakash Gupta31:41

       online that you raised your $15 million Series A in two days. Is that true?

    2. RLRoy Lee31:46

       Uh, it was a preempt. I mean, we didn't even do a proper fundraise. Like, they just kinda came to the office as soon as we got to San Francisco. Uh, they saw what we were doing, they liked it, and they just immediately just offered a term sheet pretty much. Um, yeah, like, like two days is just a, ju- just something we said, but in reality it took, like, zero days 'cause we didn't do a fundraise at all.

    3. AGAakash Gupta32:05

       Crazy. So that was basically the power then of your distribution and a16z wanting to make a bet on you guys?

    4. RLRoy Lee32:12

       Yeah, yeah. I mean, I mean, it's, it's, it's, it's pretty clear that, that we spike really, really hard, and I think if you're wanting, like, gigantic fund returners, then you don't bet on the guys doing the same thing everyone else is doing. Um, it's, it's very clear that we have a world-class distribution and marketing team. So, um, yeah.

    5. AGAakash Gupta32:28

       So you're already making more from enterprise than consumer, and you wanna reach 100 million ARR within the year. What's your forecast of how that revenue split will look like?

    6. RLRoy Lee32:39

       Yeah. I'm, I'm, I'm hoping that at our end state it looks like 30% enterprise, 70% consumer. Um, and I think that's just, uh, like, we have a strong consumer product, and, and something like this is, is, is PLG to the moon. Um, and, and I'm, I'm... I think there's, like, a very strong consumer use case, and obviously we have the best, probably the best consumer marketing team in the world right now. Um, s- yeah, but, but, like, there's more enterprise inbound than we can handle at the moment, and, uh, yeah, like pretty interested in exploring that as well.

    7. AGAakash Gupta33:07

       The tough thing about consumer is they churn so badly versus enterprise. You know, they just stick around [laughs] forever.

    8. RLRoy Lee33:13

       Yep.

    9. AGAakash Gupta33:14

       What's your vision for Cluely even beyond 100 million ARR?

    10. RLRoy Lee33:18

        Yeah, I mean, we really think that we've got... We've, like, been first to move on the UX of the future, um, and I think everybody is just, like, confirming that we are correct about translucent overlay being the future. And, um, right now it's just a land grab for us. The enterprise end state, I mean, this is a tool that will be able to pull from your CRM, use that knowledge to assist you real time, and then push back to your CRM. The end state for enterprise, like, we literally kill Salesforce and we become the new CRM of the future. It's AI native, it's real time, and it takes in all the context of all enterprise knowledge and feeds it to you directly. That is the enterprise end state. And the consumer end state is tr- like, like, I've been on record saying this a few times, but, like, brain chips. That is the ultimate form factor for AI. When we have AGI and super intelligence, we do not want to be, like, humans versus AGI. We wanna be, like, humans plus AGI, and that means brain chips. And, um, that is... I truly think that is the end state of the company of Cluely. Right now we're distributing and we're going hyperviral distributing the future form factor for AI. And if we're going to double down and project this out for a few years or a few decades, this means that we're going to stay hyperviral, we're going to distribute the brain chips.

    11. AGAakash Gupta34:21

        You, Alexander Wang, Elon, you guys all seem to be on this brain interface-

    12. RLRoy Lee34:27

        Yeah

    13. AGAakash Gupta34:27

        ... pathway. What are the m- rest of us missing on this?

    14. RLRoy Lee34:32

        I mean, I feel like it's just inevitable. Like, tech is exponentiating right now. If you just look at the number of, like, the amount of software by lines of code being written, it's literally hockey sticking right now thanks to AI. Every single technological innovation is going to come faster. Research is going to be significantly developed. I mean, like, if technology, tech- technology exponentiates, then the border and the line between sci-fi and reality just, like, change, changes, like, like, like, immediately. And in that case, like, like, what, what, what is the most optimistic way you can see ASI, AGI? Like, like, I, I truly think it's brain chips. Like, like, if it's not integrated into our brains, then it will not help us in the best way that it, that it can. I am a true AI maximalist to my heart. Um, that's why I say cheat on everything. In reality that means use AI for everything in every instance possible. And if I really believe what I say, then, like, that means, like, fucking brain chips.

    15. AGAakash Gupta35:20

        You've said me versus Elon versus Sam, we're all gonna be competing for titles as conquerors of the universe. Walk me through how we get to you competing with OpenAI and xAI.

    16. RLRoy Lee35:31

        Yeah. I mean, it feels quite clear to me that companies are converging into super companies, and the ways that you can get there are either be a massive company already or just, like, moonshot yourself into massive, massiveness, massive size, and that means flipping a coin and landing heads 20 times in a row. And that mean- you just have to swing the biggest way you possibly can, make the most, most noise you possibly can, and just hope that, that the things that you spike in will just funnel you more and more and more in, in, in, into greatness. I think there are a few short years before the super companies will have emerged and it will have been decided who the winners are. And in that few short years, I have the time to do what I do best, and that is make as much noise as possible and attract the attention of the whole world onto a product that will only get better. Um, and, and, and I, I truly think that if there is a reasonable case for us to be a super company, this is the case. We are capable of commanding more attention than anybody else at our scale with our budget, and we've, we've shown this not once this past few days, but, like, the past few months I've shown that I've been able to command the attention of the entire X timeline. Um, and I think this, this only projects out into ultimate scale. We get more money, we get more funding. With our knowledge of the algorithms, like, we will just dominate the mind, capture entire extreme consumer mind share, and this will project out, moonshot us into hyper relevance and, and super company status.

15. 36:47 – 45:14

    Company culture as a growth engine: “frat house” intensity, comp, interns, and criticism

    1. AGAakash Gupta36:47

       So the critical thing there, I think, is then building that company, right, within Cluely. So let's talk about working at Cluely. You said you're not gonna have any work-life balance if you work at Cluely. Your whole life is Cluely. So what is the type of culture you're trying to build?

    2. RLRoy Lee37:02

       Yeah, I mean, this literally look- is a frat, frat house like exemplified. We have a six-bedroom mansion in San Francisco, three stories, and pretty much everyone lives here or you live a block away. You spend all your time here, you dick around here. We give free Hinge Premium to all our employees, and every single date is sponsored. We spend our time dicking about, d- dicking around about future and, like, our careers, our relationships. It's like, it's a friend group with the sole purpose of living like, like a tribe almost.

    3. AGAakash Gupta37:27

       The frat house mentality, it works pretty well to grow at the beginning. Do you think there are any risks?

    4. RLRoy Lee37:34

       I'm sure that there's things that we're doing that, like, like will end up in lawsuits later, but I mean, you solve problems when they come, not before they come. Maybe the- maybe this will burn me, but I mean, like, right now this, this is what the culture of the com- of, of a company should feel like. If you're clocking out at 7:00 PM, going home, and just like, you know, like hanging out in your second life, then I feel like you're just not invested in the company. Your work, your life, everything about you should be Cluely.

    5. AGAakash Gupta37:59

       That's a really high bar. Usually you have to pay people a lot. What's your approach to compensation, to getting people to buy into this culture?

    6. RLRoy Lee38:06

       Yeah, I mean, we do pay people a lot. Like, like, I think it's, it's ridiculous to me that, that companies are seemingly having a race to the bottom to see what's the lowest amount I can pay my software engineers while still having them, like, like, like dedicated. If you want employees that are dedicated to you, you should be dedicated to your employees, and this means, like, paying them what they're worth. Every software engineer we have is bringing us upwards of, like, a few hundred thousand dollars of revenue per year, like by head count. So by that metric alone, they should be getting paid that much. Um, and I think I'm not hesitant to pay employees massive salaries and give them massive equity because they're vital. Like, if I expect you to be a part of the cult religion of Cluely, then I should be rewarding you as if, if you are a valid contributing member of the cult. And everything we do is, like, extreme. Um, our compensation ranges are extreme. The, the company outings are extreme. The, the, the Hinge Premium is extreme. Like, like, everything we do is extreme, and as a result it, it, it's a super unique culture. And the thing is, like, you don't apply if you're not interested in the culture. Like, like, right now what we're doing, I mean, it's interesting to almost everybody in the world.

    7. AGAakash Gupta39:05

       The parallels it gives me most to are Uber, right? And Uber-

    8. RLRoy Lee39:08

       Yeah

    9. AGAakash Gupta39:08

       ... was able to take that culture. Travis was able to get everybody to invest all in, and I think the lawsuits came down the line from, you know, women were at the workplace finding it too fratty and things like that. How do you kind of protect yourself from those downsides?

    10. RLRoy Lee39:24

        I mean, it's not like we're, like, sexually harassing or assaulting people in here. Like, we're just like, like dicking around and, um, like I, I, I don't... I think this is a pretty friendly work environment for anyone who's, like, of the age of, like, I don't know if I'm allowed to say that, but I mean, it's, if you're like a chill, youthful, fun guy, you will come here and, and you'll... Or a guy or girl, you'll come here and you'll have a good time, and I, I, I don't think that means, like, sexually harassing your employees or anything.

    11. AGAakash Gupta39:47

        Okay. And you said you wanted to launch the best paid internship program. You said you wanted to launch and hire 50 interns. Is that how it's turned out?

    12. RLRoy Lee39:56

        We have more than 50 interns if you count UGC content creators as interns, and I think that there's a very strong case for you too. I mean, like what is a marketing internship? You're working part-time for a company in the hopes of a full-time return offer, and you're providing value for them by getting more eyeballs on the company. That is what a marketing internship is. And right now in today's world, in today's, like, economy, like what that looks like, uh, for us, is you sit in front of a camera and you make videos, and you get paid for the videos that you make. By all means, like, our UGC creators are growth, they're marketing interns. They're better marketing interns than like every com- every, every, every intern that every other company hires. Like, I guarantee you this is more profitable and lucrative than having some 17-year-old college kid, like, try and design a billboard or whatever the fuck the other marketing interns are doing at bigger companies. And, and if you, if you count them as interns, then yeah, we're like over 70 then.

    13. AGAakash Gupta40:41

        Okay. So I really like that. So you're using your internship program to find your UGC creators. These are generating your hundreds of millions of views on Instagram and TikTok.

    14. RLRoy Lee40:50

        Yeah. Yeah.

    15. AGAakash Gupta40:51

        So I told my dad, a 60-something professor of computer science, that I was interviewing you this weekend, and he said, "You have to ask Roy [laughs] are we all gonna get dumber?" Because his students are using these cheating tools, and he feels like they're not-

    16. RLRoy Lee41:09

        I mean, like, like, like what exactly are they learning? Like, I bet the blacks- blacksmithing instructor from 1600s would've thought like, "Man, it's a shame that all these students, these people nowadays, they don't know anything about blacksmithing. Like, like these kids are all getting dumber." But like, like why exactly do we need to know blacksmithing? Like, as technology advances, there's bits of knowledge that we don't need to know and bits of knowledge that become exponentially more relevant, and I think with AI this is going to, uh, massively change the scale of how things work. It's going to make 99% of knowledge in the world redundant right now, and the 1% of knowledge that is left, it will be 100 times more useful to know. So no, I don't think people are getting dumber. Like, maybe people are memorizing facts less. Like, maybe there will be less people who know what the 50 capital states of America are, but I'm like, "Okay."

    17. AGAakash Gupta41:50

        So it kinda goes back to that LeetCode example then, right? Of Interview Coder, where it's like the stuff we're not learning, like how to memorize the answer to 600 algorithms, riddles-

    18. RLRoy Lee42:00

        Exactly

    19. AGAakash Gupta42:01

        ... isn't maybe that important.

    20. RLRoy Lee42:03

        In the future, work, labor, the human experience will look so different. There's not one historian that is going to look back and think, "Man, it would be better if we knew this useless information." In fact, has that ever happened in history? Like, I would argue that it hasn't, where technology has evolved and made some X form of knowledge useless, and people are saying like, "Man, the world would be so much better if that technology never existed and people would know more about this, this Y random subject." That has never happened in the history of humanity ever. When the calculator emerged, people were... People definitely are worse at computing fast right now than they were when before the calculator arrived. Not a single soul says that the calculator was like a, like an evil invention. Like, everyone agrees that it was just better for the progress of humanity, despite it resulting in people being worse at computing things, like mentally. And, um, that's just a skill that is irrelevant. And right now, like literally these models are the worst they will ever be for the rest of your life. If a model can do it right now, why would you think that it could not do it in the future? And if it can't do it right now, then whatever task that is is obsolete.

    21. AGAakash Gupta42:56

        On the betting market Polymarket, there's a bet that you're gonna go to jail this year. Why would you go to jail?

    22. RLRoy Lee43:03

        I don't think I will go to jail. I don't think I'm doing anything illegal here. I think, uh, the company is growing, it's flourishing, the employees are happy, I'm happy, and, uh, there's no reason for me to go to jail. I, I think people take what I'm doing, which is applying the markets, applying the principles of, like, engagement bait and rage bait from Instagram, TikTok, applying it to X, LinkedIn, and seeing massive success with it, and they're extrapolating it to a character or persona or traits that don't actually exist in real life. All I'm doing is very good marketing because I'm able to apply a skill set from, uh, uh, uh, a, a wealth of experiences that most peop- most tech bros just never had.

    23. AGAakash Gupta43:37

        Insanely good viral sense, I think.

    24. RLRoy Lee43:40

        Thank you.

    25. AGAakash Gupta43:40

        What's something people get wrong, most people get wrong about Cluely?

    26. RLRoy Lee43:45

        I think they think that there is no product, but I think one thing that people forget often is that the company is just... The first line of code was written 10 weeks ago. This is more new than the recent batch of YC startups, but it is just so dominantly on your timeline. People think this company's been around forever. When are they gonna ship a product? And you just can't get it off, it's so on top of your mind. But in reality, like, we are extremely new. Um, and people say, "This is like, I could have built this in a weekend as, like, a hackathon project." Like, we did. It has been built over essentially a weekend. It is essentially a hackathon project that is getting iterated on, like, extremely quickly. I think that's the one thing that people get wrong. They forget, because of how dominant we've been on the timeline, they forget how new the company really is.

    27. AGAakash Gupta44:24

        Well, I'm looking forward to the ongoing future versions of Cluely. I'm hoping that next year's Cluely looks nothing like this one. Roy, thanks for being on the podcast.

    28. RLRoy Lee44:34

        Yeah, thanks for having me, man. Appreciate you, brother.

    29. AGAakash Gupta44:36

        So if you want to learn more about how to shift to this way of working, check out our full conversation on Apple or Spotify podcasts. And if you want the actual documents that we showed, the tools and frameworks and public links, be sure to check out my newsletter post with all of the details. Finally, thank you so much for watching. It would really mean a lot if you could make sure you are subscribed on YouTube, following on Apple or Spotify podcasts, and leave us a review on those platforms. That really helps grow the podcast and support our work so that we can do bigger and better productions. I'll see you in the next one.

Episode duration: 45:23