$215M AI CEO: How I’d Build a Profitable AI Startup in 30 Days (2026 Playbook)

0:00Where is Daredevil?

0:03Marina Mogilko:I'm right here. Don't miss the return of Marvel Television's Daredevil Born Again. So what's next? We're gonna take this city back. In an all-new season, now streaming only on Disney+. They're hunting us. It's time we started hunting them. I can work with them. This should be tons of fun. Marvel Television's Daredevil Born Again, now streaming only on Disney+. This episode is brought to you by State Farm. You know those friends who support your preference for podcasts over music on road trips? That's the energy State Farm brings to insurance. With over 19 ,000 local agents, they help you find the coverage that fits your needs.

0:45Marina Mogilko:So you can spend less time worrying about insurance and more time enjoying the ride. Download the State Farm app or go online at statefarm.com. Like a good neighbor, State Farm is there. So let's imagine you have to start a company again. Walk me through the first 30 days. What will you do? If I have to start a company again, I think the first thing is... This is Yang, co-founder and CEO of OpusClip, an AI tool that turns long videos into viral short clips. In just two and a half years, he built it into one of the fastest growing AI companies in the world. 50 million users,$215 million valuation.

1:23Marina Mogilko:LLMs are advancing so fast and large companies are releasing better and better products. Do you think there are any niches or problems that are not worth solving anymore? If you asked me like three years ago, the answer will be much simpler. I think every AI founder should be somehow AGI-peeled, which means that you can predict what the foundation models can release in the next few weeks or months. The rules have changed and most founders don't see it yet. Young breaks down where the opportunities are, what to avoid, and what it takes to win in 2026.

1:57Marina Mogilko:Okay, Yang, thank you so much for being here. You've built one of the fastest growing AI companies, 12 million users in 12 months, 215 million in valuation. Can you introduce yourself in 60 seconds? What do you do? Thank you for having me, Marina. I'm Yang, co-founder and CEO at OpusClip. We build the products that turn long form, dense information content like articles, blogs, footage, papers into engaging contents like short-form videos that can attract your audiences worldwide. We have gathered actually more than 15 million users in the past two and a half years. What? This is crazy. Wow. Super crazy.

2:36Marina Mogilko:And also like your story is fascinating because you started during COVID and you started with all the different features and only this one stuck. Can you talk to me about this mindset of like trying things and seeing them not work and not giving up? Yeah, it's kind of frustrating for sure. I think in the early days, like we are like new founders, we're early founders. It's actually the passion that drove us to continuously try different things. But like probably after half a year or even one year of trying like, let's say three to four different things, if you don't still get any early signals of the product market fit, you will be easily fatigued.

3:14So I think that's the challenge. And we're kind lucky that we first built a last streaming tool um nobody likes it but there's only one feature that is like the clipping feature in the last mean tool uh that had somehow early signal of the product market fit and also thanks to that time which you know actually in the same week try gpt was launched by open ai um so we quickly married that to this you know standalone feature and pivoted to a new product, which is the Opus Clip right now.

3:47Marina Mogilko:What should people pay attention to when they decide to double down on one particular feature? Good question. So we were not tracking any classic OG, AARRR kind of metric. We didn't build a product for Opus Clip in the first day. We actually engineered the results, the final outcome, the final videos, and just emailed them to all of the potential customers. Oh, so you're like manually created? That's how you start. Yeah. Like we work with the AI to edit all these videos and write and send out to all the potential customers. And I think we got like more than 60 % of positive feedback. I love this clip.

4:32I want to use it. I just want to tweak one thing and I can publish it right away. So those are the very early feedback. And then after a few more weeks, we built the product into a Discord bot. Still no interface, right? So we save a lot of time building the UI, UX and all that kind of stuff and just focus on delivering the value, validating the outcome. So we grabbed hundreds of thousands of creators into our Discord channel and they are playing with the bot. So what we were looking at are basically their retention, right? Their engagement of that tool. But beyond that, we also look into the discussions around the content.

5:12Like we saw people, hey, how do you get that content? How do you get that piece of the clip? So I think in the early days, quantitative data are important, but qualitative data or feedback is also super important. When we start hearing people complaining about the queue, complaining about the quota of their data usage, we realize this is the product market fit moment.

5:37Marina Mogilko:Nice. And was there a metric when you said you were tracking how often people were coming back to the bot in Discord? What's a good number? Because I feel like it's not something like Google, right? When you're Googling several times a day. But for a feature like yours, what's a good metric? Yeah, the general use case for content creator, let's say, they would normally have one piece of long-form content, long-form footage. And they want to use our tool to generate five to ten short clips so that they can post to their channels. one clip a day. So I would say the average frequency of using our tool is like on a weekly basis.

6:15But when we see people like come to use it every day or, you know, multiple times a week, that is a very strong usage beyond what we can imagine.

6:25Marina Mogilko:So based on this story that we just talked about of you ended up finding this feature, what would be your advice to all the startup founders? What would you say is the key learning from that? The key learning is that you should build a real business with the product, not a cool demo, right? Because many people just start with building a very cool demo. They showed it off to the people and, you know, it looks like magic because the demo really demonstrates the way it looks, right? Yeah, exactly. Yeah. So the demo shows a very strong capability, but many founders failed because they didn't find a product market fit for a real business.

7:05What that means is that they are not solving, probably they're not solving a real pain point, a real painful job to be done in the real world. Because before your product, if it is a real painful job to be done, there must be a lot of alternative solutions like humans do the job, or they build some internal tools, or they manually just concatenate a couple of different snippets or solutions to make it work, right? They spend extra time, effort, or painful hours to get the job done. So are you really replacing those type of tedious or human-heavy workflows? That is one part of whether it is a real business.

7:48The second thing would be how your customers understand the values you created, right? If you can tell your product value in just like 10 words, maybe or probably you're close to a real business. If your product is just cool and all the feedback was like, this is amazing, this is awesome, but nobody want to give you their credit card information. Nobody want to ask you what is the, what is your pricing tier? It's also not helping you get to the product market fit as well.

8:19Marina Mogilko:I really like the way you approach this. So you figured out the problem that people are solving manually and then you created a product manually without any interface and then you saw early users' feedback. I really like that approach. If people want to use that approach, what would you say comes first? Passion about particular problem or the problem itself? Yeah, I would say both of them are super critical. But my first principle for passion is actually not necessarily the passion for solving the problem. I think passion is something more emotional. I think all the entrepreneurs, all the founders need to have the passion to be a problem solver.

8:58You need to have the passion to be a builder, to build something that people want, to build something that ideally changed the world. I think that's the passion founders need. You don't have to have a passion for video clipping. You don't have to have a passion for running a restaurant. I think all you need to have for passion is to be a problem solver, be a builder, be a game changer. And rationally speaking, the real painful job to be done is something that is inevitable in your journey. You have to be very rational to figure out this is a real problem to be solved by my passion. So the deep understanding about the industry, about the workflow, about the customer profiles, about the use cases, those are all rational.

9:48So I think an emotional passion for a bigger picture, just to be a builder, plus the rational conviction for solving concrete problems that you know very well, those are the necessities for becoming a founder.

10:18college deal. Everything you need to study and play with select Windows 11 PCs. Eligible students get a year of Microsoft 365 Premium and a year of Xbox Game Pass Ultimate with a custom color Xbox wireless controller. Learn more at windows.com slash student offer. While supplies last, ends June 30th. Terms at aka.ms slash college PC.

10:38Marina Mogilko:When you need to build up your team to handle the growing chaos at work, use Indeed Sponsored Jobs. It gives your job post the boost it needs to be seen and helps reach people with the right skills, certifications, and more. Spend less time searching and more time actually interviewing candidates who check all your boxes. Listeners of this show will get a$75 sponsored job credit at Indeed.com slash podcast. That's Indeed.com slash podcast. Terms and conditions apply. Need a hiring hero? This is a job for Indeed sponsored jobs. Let's talk about your new agent opus. Because I heard this presentation where you compared where your tool is doing now to human editors.

11:18Marina Mogilko:And yes, human editors are slightly better, but when it comes to someone who's just starting out multiple channels, Agent Opus becomes invaluable because it just takes all the content, creates, I guess, all those channels. Can you explain how it works? We don't have two products, right? The Opus Clip product is like a very refined workflow or agentic workflow that follows a rigid plan and then execute it. The agent opus on the other side is more agentic. There is no preset workflow. And the interaction for users is that you can just drop your idea, your story, or some links, or some assets, and instruct the agent to do whatever you want them to create.

12:06So the agent is playing the role like a director. right it's not a video editor it's not a video creator it's not a video generator it's not a designer it is a director i think about you know what how director works in those movie studio they are like a true leader managing and collaborating with multiple different functions like there are designers producers artists researchers and script writers scene writers There are so many sub-agents in Agent Opus all reporting to the central director agent. So I think that's the magic part of it. Because when everyone says I'm using AI, I'm building AI, but I think building a team of AI is another level of challenge.

13:01But it also unlocks another level of superpower. So when we saw that agent, the central director agent, designing a plan and also orchestrate among the other eight to nine agent, it is a truly end-to-end autonomous experience where the input is just some article or pieces of news. while the output consists like, you know, well-written script, polished voiceovers, photorealistic avatars, real-world assets sourced from the entire internet and also YouTube, plus AI-generated scenes, animations, infographics along the video. So it's a true multi-agent, multi-modal agent workflow.

13:56Marina Mogilko:And so I can drop my LinkedIn post, right? Which is a written post and you'll be able to create a video out of it. Yeah, exactly. So can you show the new AI feature? Because this is something that we're doing a lot. We're trying to repurpose whatever is going viral on LinkedIn. This is so funny. I posted myself sitting in the garage on a call because my kids were going crazy upstairs. They just couldn't take the call upstairs. And I'm like in this messy garage and freezing. and it got 120 ,000 impressions. And ideally, I want it to become, you know, something else I can post. So copy link, right?

14:35You copy and paste. You can just say like, create a video of this LinkedIn post. Shall I add more prompts?

14:45Marina Mogilko:Like, make it a viral reel or it should, or it's like pre-prompted. Yeah, it's pre-prompted, but you can always add some thought on it. Up to you. So for voice, you want to also create your own voice, but let's just use our default voice. What is the hook? The hook is that we'll have different hook templates, like the fast cuts, article highlight. Right now it's kind of slow, So it probably takes about like 30 to 60 minutes. Oh, 30 to 60 minutes. Yeah, we're optimizing the speed. Hopefully we'll get to like 20 minutes in two months. I'm going to show you the result. Work and home lives blend together now more than ever.

15:34Marina Mogilko:A constant zigzag between professional calls and family chaos. Okay, so with all of this, what will happen to creators in three years? So like in general, creator economy, how do you see it evolving with all the agentic tools? I think the entry barrier to create something is going to dissolve so that everyone has the superpower to become a creator. That's a brutal fact. And it makes the competition much more challenging, to be honest. But the good thing is that everyone now don't have to think about, you know, what tool do I need to use or what techniques, what expertise do I need to learn? but now you can focus on figuring out what is your uniqueness how should you stand out right what is your unique narrative of your story what is your unique messaging what is your unique tone what sets you apart so just focus on yourself focus on your own storytelling and in one of two years no more than three years ai can just deliver the job for you they're doing actually the dirty work for you but you're going to be the real creative one but it's crazy the competition is going to be insane it is do you think there's less and less time to build a personal brand because what happens if in three years everyone has access to tools everyone tells the story and then we don't have enough eyes to consume all the content do you ever worry about it i don't worry that too much um because think about like 10 years ago five years ago when you want to build a personal brand um i think most of the time you're you're wasting is to um you know edit the video scene by scene right is to design design your colors design um everything like page by page i think that's how you waste the time so the analogy is that is that if you have to spend five months um building your personal personal branding from zero to one but nowadays it's like you just need a couple weeks yeah my like thing because i started 10 years ago and i remember editing all the videos yeah something that set me apart was the willingness to do that so my content was super mediocre but because i took the chance of like you know spending six hours started a video i got my followers it's impossible today like if my content is mediocre it's not going to be noticed that's the thing right so your story either has to be extraordinary or or that's it do you think we still have this opportunity to showcase our personal brands or because there are so many we're getting saturated the good storyteller um or the the real unique people can still have the advantage think about um you know carriage versus cars right 100 years ago very few people have cars uh so it become a privilege to be you know one of the fastest people in the world nowadays everyone are able to to drive a car almost everyone are able to drive a car.

18:26But there are still Formula One drivers. There are still WDC drivers who are way faster than ordinary people. So I think probably 10 years ago, you showed your willingness to build your personal branding by spending extra time writing content, polishing your website, editing the videos. But I think in the future, people who want to stand out should still probably spend the same amount of time on something else. Not using the tools, not, you know, building by yourself, but the time should be spent on thinking it through how you should stand out. So I think time spent probably the same, but the actual work are completely different, which is more human, actually.

19:15Marina Mogilko:Yeah, yeah. I like that reply. I like that car analogy, actually. Thank you. So let's imagine you have to start a company again. Now, walk me through the first 30 days. What will you do? If I have to start a company again, the first thing is always to figure out what is a real painful job to be done. I want to so I'm the prototype of founder who don't start everything from technology, but from users from the market. I have to segment my market to a very clear, very vertical niche that I can clearly understand their existing workflow, their existing pinpoint, their existing alternative solutions.

20:03And then I think probably I would have to spend the first couple of weeks, two or three weeks understanding the real use case, the very specific ICP. and the second thing is to just engineer a you know prototype or proof of concept i can do that

20:21Marina Mogilko:very easily with all the vibe coding tools right now right what's your favorite uh well i'm a heavy um cursor user um but i tried many other different tools but cursor is my go-to go-to tool yeah because i came from engineering background so it's um yeah yeah i'm more familiar with the ide the step two is actually just a couple of days should be good enough to have a proof of concept and then i go back to share that with the early users the icps to get you know their feedback the feedback should not only about do you like the product but also like what kind of problems am i solving for you and what is the value you perceive which means how much money you are willing to pay for it.

21:06So that process probably is the majority of the early 30 days. But I think on the other side, I also need to think it through about what kind of proprietary data can I possess along the way? And will that data set grow as the product grow, as the user grow? right um we don't have to build toward a moat or a very clear plan of defensibility but i think we need to have that idea at least well thought in the first early days um because you have you can't avoid it right when you really launch the product so it's better to have something in mind

21:48Marina Mogilko:have some idea in mind like what makes you stand out yeah exactly and what makes people come back Yeah. And if you ask this question, especially nowadays, especially we are entering 2026. We're switching between apps like crazy. Like there's new shiny app, you install it, you're like, oh, I love it now. But the next day something else comes out and it's so easy unless you have history within a certain app. Yeah, exactly. So the last thing I think we should consider is the distribution channel. I think the competition is totally different. yeah two years ago two years ago you can easily say you are the chat bt of abc right it's so easy to go by right that time nowadays like everyone knows a everyone has used ai um so figuring out a distribution channel will help you better refine your user experience your onboarding and also um your target icp as well so basically you start with expertise it looks like you need to understand at least one niche to be able to build something.

22:46Marina Mogilko:Not like, oh, this market looks good. I'm going to try and build something for it. No, you need to be deep in that market and understand the problem. Exactly. Because you asked me a question, the assumption is today, right? AI has been there for like more than three years. The LLM has been there for more than three years, right? Because if you asked me like three years ago, the answer would be much simpler. But now when, you know, LLMs are advancing so fast and large companies are releasing better and better products, do you think there are any niches or problems that are not worth solving anymore just because big companies can take over in like a week?

23:23I think two major types of problems that a founder should avoid. The first one is that you are just building a feature for an existing type of ICP, ideal customer profile, within an existing workflow built by the incumbents. so basically what that means is that the incumbents can easily build a feature that you know bundles everything right and probably you won't have your own distribution channel for example i think like if you want to build a note taker think it through right probably find another direction because it's super easy for zoom or you know google meet to have that feature in their existing workflow because you're targeting basically the same icp same market same almost same or adjacent use cases attached to a big workflow, a big platform.

24:16I think every AI founder should be somehow AGI-peeled, which means that you can predict or you should be confident to make some predictions about what the foundation models can release in the next few weeks or months. If they are already doing some job 80 % very well, 90 % very well, I think in their next few releases, they can probably do that job 99 % well or even 100%. So run that test in your internal strategic discussions. If you're just becoming a rapper with some prompts, then probably you don't have to write any prompts in the second release, in the next release of Gemini or ChatGPT. chat GPT.

25:03You need to focus on solving a vertical business problem by integrating the workflow end-to-end. You need to own the workflow end-to-end so that AI is part of the workflow, but it's not all. Instead, if you build a wrapper with some prompt engineering, AI is almost the end-to-end workflow itself.

25:26Marina Mogilko:Can you also walk me through the process of coming up with pricing for an AI tool? How do you decide on, because I know you have lower tier, but then you also have like a 20K check for bigger companies. How do you come up with the price? Pricing is a real science. So a couple of factors we considered in the first place. For example, first, what is your value creation? How to measure your value creation, right? Before using your product, what would your users do to achieve the goal? to get the job to be done. They probably, you know, spend their own time, which there is a price for their own time, right?

26:06They probably outsource it to some vendors or, yeah, they pay humans to get the job done. Or they may, like, use some, like, OG tools by some professional guys, like, you know, in our use case, professional editors to get the job done. So in our industry, like, editing a clip, editing a viral-ready, high-quality clip, one-minute clip, often takes about one hour or at least 30 minutes. And the market price is about$25 to$50. So that's the value creation we're benchmarking against. The second thing is about your unit economy because you have inference cost. And even that can go down in the future, but it's still very pricey in the early days.

26:57And also for videos, another elephant in the room for the cost is the storage. The storage can be just 5 % in the early days, but it can become like 50 % of your total costs after three or five years. So the UN economy, you want to make sure it's healthy, it's sustainable. Third one, I would say just do as many experiments as possible. we actually sent out like thousands of surveys in the first few months and we tweaked our pricing logic again and again well in the early days right you don't want to change drastically in the in the late stage but in the early stage we did a lot of survey customer interview to understand how much they are willing to pay because even you have a clear value creation and a well calculated the unit economy, can users still perceive that?