YT YouTube2Text

YouTube2Text → Transcripts → $215M AI CEO: How I’d Build a Profitable AI Startup in 30 Days (2026 Playbook)

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

by Silicon Valley Girl · 6,792 words · 985 segments · language en · Watch on YouTube

Open in transcriber Summarize with AI Download SRT/VTT

Full transcript

0:00So let's imagine you have to start a
0:02company again. Walk me through the first
0:0430 days. What will you do
0:06>> if I have to start a company again? I
0:08think the first thing is
0:10>> this is Young, co-founder and CEO of
0:12Opus Clip, an AI tool that turns long
0:15videos into viral short clips. In just 2
0:17and 1/2 years, he built it into one of
0:19the fastest growing AI companies in the
0:21world. 50 million users, $215 million
0:25valuation. LLMs are advancing so fast
0:28and large companies are releasing better
0:30and better products. Do you think there
0:32are any niches or problems that are not
0:34worth solving anymore?
0:35>> If you ask me like 3 years ago, the
0:37answer will be much simpler. I think
0:39every AI founder should be somehow aged,
0:42which means that you can predict what
0:44the financial models can release in the
0:47next few weeks or months.
0:48>> The rules have changed and most founders
0:50don't see it yet. Young breaks down
0:52where the opportunities are, what to
0:54avoid, and what it takes to win in 2026.
0:59This video is sponsored by HubSpot.
1:01Okay, Young, thank you so much for being
1:03here. You've built one of the fastest
1:06growing AI companies, 12 million users
1:09in 12 months, 215 million in valuation.
1:12Can you introduce yourself in 60
1:14seconds? What do you do?
1:16>> Thank you for having me, Marina. Um, I'm
1:17Yan, co-founder and CEO at Opus Clip. We
1:20built the products that turn long- form
1:22dense information content like articles,
1:25blogs, footage, papers into engaging
1:28content like short form videos that can
1:30attract your audiences worldwide. We
1:33have gathered actually more than 15
1:34million users in the past 2 and a half
1:36years.
1:36>> What? This is crazy. Wow.
1:39>> Super crazy.
1:40>> And also like your story is fascinating
1:42cuz you started during co and you
1:44started with all the different features
1:46and only this one stuck. Can you talk to
1:48me about this mindset of like trying
1:51things and seeing them not work and not
1:53giving up?
1:54>> Yeah. Um it's kind of frustrating for
1:57sure. I think in the early days like we
1:58are like new founders or early founders.
2:01Um it's actually the passion that drove
2:04us to continuously try different things.
2:06uh but like probably after half a year
2:08or even one year of trying like let's
2:10say three to four different things if
2:12you don't still get any early signals of
2:15the product market fit you will be
2:17easily fatigued. Um so I think that's
2:19the challenge and we're kind of lucky
2:21that we first built a live streaming
2:24tool. um nobody likes it but there's
2:26only one feature that is like um the
2:29clipping feature in the live stream tool
2:32uh that had somehow early signal of the
2:35product market fit and also thanks to
2:38that time which you know actually in the
2:40same week was launched by open AAI um so
2:43we quickly married that to this you know
2:46standalone feature and pivoted to a new
2:49product which is the open opus plate
2:51right now
2:51>> what should people pay attention to when
2:54they decide to double down on one
2:56particular feature.
2:57>> Good question. So, we were not tracking
3:00any classic OG um a RR kind of metric.
3:06We didn't build a product for OpenClip
3:08in the in the first day. We actually
3:10engineered the result, the final
3:13outcome, the final videos and just
3:15emailed them to all of the potential
3:17customers. Oh, so you're like manually
3:19creat
3:21[laughter]
3:24we we man we we work we work with the AI
3:27to edit all these videos and write and
3:29send out to all the um potential
3:30customers and I think we got like more
3:32than 60% of positive feedback.
3:35>> I love this clip. I want to use it like
3:37I just want to tweak one thing and you
3:39know I can publish it right away. So
3:41those are the very early feedback and
3:43the then like after a few more weeks we
3:45built the product into a discord bot
3:48still no interface right so we save a
3:50lot of time building the you know UIUX
3:52and all that all that kind of stuff and
3:54just focus on delivering the value
3:56validating the outcome so we grabbed you
3:58know hundreds of uh thousands of uh
4:00creators into our discord channel and
4:04they are playing with the bot so what we
4:06were looking at or basically their
4:08retention right their engagement of that
4:10tool
4:10And but but beyond that we also look
4:12into the discussions around the content
4:16like we saw people hey how do you get
4:18that content how do you get that piece
4:20of um the clip. So I think in the early
4:24days quantitative data are important but
4:26qualitative um [snorts]
4:28data or feedback is also super
4:30important. When we start hearing people
4:32uh complaining about the queue,
4:35complaining about the quota of their
4:36daily usage, we realize this is the part
4:40of market fit moment.
4:41>> Nice. And was there a metric uh when you
4:44said you were tracking how often people
4:45were coming back to the bot in Discord?
4:47What's a good number? Cuz I feel like
4:49it's not something like Google, right?
4:50When you're like googling several times
4:52a day, but for a feature like yours,
4:54what's what's a good metric? Yeah, the
4:56general use case for content creator,
4:58let's say, they would normally have like
5:00one piece of long form content, uh, long
5:02form footage, and they want to use our
5:06tool to to generate like 5 to 10 short
5:09clips so that they can post to their
5:11channels, one clip a day. Um, so I would
5:14say the average uh frequency of using
5:17our tool is like on a weekly basis. But
5:19when we see people like come to use it
5:22every day or you know multiple times a
5:25week that is a very strong usage. Yeah.
5:27>> Beyond what we can imagine.
5:29>> So based on this story that we just
5:31talked about of you ended up finding
5:34this feature. What would be your advice
5:35to all the startup founders? What would
5:38you say is the key learning from that?
5:39>> The key learning is that you should
5:41build a real business with a product not
5:45a cool demo. Right? Because many people
5:48just start with building a very cool
5:49demo. Um they showed it off to the
5:51people and you know it looks like magic
5:54because the demo really demonstrates
5:56>> you polish the way it looks right works.
5:59Yeah.
5:59>> Yeah. So the demo shows a very strong
6:02capability but um many founders failed
6:05because they didn't find a product
6:08market fit for a real business. What
6:10that means is that they are not solving
6:13probably they're not solving a real
6:14painoint a real painful job to be done
6:17in the real world because for your
6:20product if they if it is a real painful
6:22job to be done there must be a lot of
6:24alternative solutions like you know
6:26humans do the job or they build some
6:29internal tools or they manually just you
6:32know concatenate couple of different
6:34snippets of the or solutions to make it
6:36work right they spend extra time effort
6:39or you know painful hours to get the job
6:43done.
6:43>> So are you really replacing those type
6:46of like tedious or human heavy workflows
6:49that is one part of you know whether it
6:51is a real business. The second thing
6:53would be how your customers uh
6:56understand the values you created,
6:58right? If you can tell your product
7:00value in just like 10 words, maybe or
7:04probably you're close to a real
7:06business. If your product is just cool
7:08and all the feedback was like this is
7:10amazing, this is awesome, but nobody
7:12want to give you uh their credit card
7:14information. Nobody want to ask you what
7:16is the you know what is your pricing
7:18tier. it's also not helping you um get
7:21to the product market fit as well.
7:23>> I really like the way you approached
7:24this. So you figured out the problem
7:26that people are solving manually and
7:28then you created a product manually
7:30without any interface and then you saw
7:33early users feedback. I really like that
7:34approach. If people want to use that
7:36approach, what would you say comes
7:38first? Passion about particular problem
7:41are the or the problem itself?
7:42>> Yeah, I would say both of them are super
7:45critical. Um but my first principle for
7:49passion is actually not necessarily the
7:51passion for solving the problem. I think
7:53passion is something more emotional. Um
7:57I think all the entrepreneurs, all the
7:58founders need to have the passion to be
8:01a problem solver. You need to have the
8:04passion to be a builder, to build
8:06something that people want, to build
8:08something that you know ideally change
8:10the world. I think that's the passion
8:12founders need, right? You don't have to
8:14have a passion for like video clipping.
8:16You don't have to have a passion for
8:19like, you know, running a restaurant. I
8:21think all you need to have for a passion
8:24is to be a problem solver, be a builder,
8:28be a game changer. And rationally
8:30speaking, the real painful job to be
8:33done is something that is inevitable in
8:36your journey. Um you have to be very
8:38rational to figure out this is a real
8:41problems to be solved by my passion
8:43right um so the deep understanding about
8:46the industry about the workflow about
8:49the customer profiles about the use
8:50cases those are all rational um so I
8:54think an emotional passion for like a
8:56bigger picture um just just to be a
8:59builder plus um the rational conviction
9:03for solving a concrete problems that you
9:07know very well. Those are the
9:09necessities for becoming a founder.
9:11>> My guest Young built a tool that
9:14creators love. And if you're interested
9:16in turning content into a career in 2026
9:20or taking your creator career to the
9:22next level, take a look at this new
9:24ultimate content creator toolkit from
9:26HubSpot in partnership with Tik Tok.
9:28It's packed with resources that help you
9:30cut through the noise, what tools you
9:32actually need, AI prompts that save
9:35hours, and most importantly, how to
9:37connect your content to real revenue,
9:39not just views. My favorite section is
9:41the Tik Tok breakout, and I love how
9:43actionable it is. Tik Tok Ads Manager
9:46doesn't just help you launch ads, it
9:48shows you what's already working. I wish
9:50I had this tool 10 years ago. You can
9:52peek at top performing ads, see which
9:55products are blowing up, track trending
9:56sounds and hashtags, and even use Tik
9:58Tok's own AI tools to make your videos
10:01stand out. And it's not just about
10:03views. It's about turning attention into
10:05signups, customers, and recurring
10:08income. Tik Tok users are actually way
10:10more likely to sign up, buy, or take
10:12action compared to other platforms. And
10:14with a HubSpot integration, you can
10:15literally see which videos are
10:17converting into leads and clients inside
10:19your CRM. This toolkit was created by
10:21HubSpot in partnership with Tik Tok and
10:23I'm partnering with them in this video
10:25to share it with you. So, you're
10:27learning directly from the people who
10:28see the data every day. Download the
10:30toolkit for free in the description
10:32below. Honestly, if you're serious about
10:34growing as a creator, this is where you
10:36start. Let's talk about your new agent
10:38opus cuz I was I heard this presentation
10:42where you compared uh where you tool
10:44your tool is doing now to human editors.
10:47And yes, human editors are slightly
10:49better, but when it comes to someone
10:51who's just starting out multiple
10:53channels, agent opus becomes invaluable
10:55because it just takes all the content,
10:57creates, I guess, all those channels.
10:59Can you explain how it works?
11:00>> We currently have two products, right?
11:02The Opus script product is like a very
11:04refined workflow um or agentic workflow
11:08that follows um a rigid um plan and then
11:12execute it. The agent opus on the other
11:15side is more um is more agentic. Um
11:19there is no preset workflow. Um and the
11:21interaction for users is that you can
11:23just drop your um drop your idea, your
11:27story or some links or some you know
11:30assets and instruct the agent to do
11:34whatever you want them to create.
11:36>> So the agent is more is playing the role
11:39like a director, right? It's not a video
11:42editor. It's not a video creator. It's
11:43not a video generator. It's not a
11:45designer. It is a director. And think
11:48about you know what how director works
11:50in those uh movie studio. They are like
11:54a true leader um managing and
11:57collaborating with multiple different
11:59functions like there are designers,
12:01producers, um artist um researchers and
12:06script writers, scene writers, right?
12:08there are so many sub agents in agent
12:11opus all reporting to the central
12:15director agent um so I think that's the
12:18magic part of it um because when well
12:22everyone says I'm using AI I'm building
12:24AI but I think building a team of AI is
12:29another level of challenge but it also
12:32unlocks another level of superpower so
12:36when when we saw that um agent the agent
12:39office was the central director agent um
12:42you know designing a plan and also
12:44orchestrate among the other 8 to nine
12:48agent.
12:49It is a truly end-to-end autonomous
12:52experience
12:54where the input is just some article or
12:56pieces of news while the output consists
12:59like you know well-written script
13:03polished voiceovers
13:06um photorealistic avatars
13:08real world assets sourced from the
13:11entire internet and also YouTube plus AI
13:15generated scenes animations infographics
13:19along the video. So, it's a true multi-
13:23aent, multimodal agent workflow.
13:26>> And so, or I can drop my LinkedIn post,
13:28right, which is a written post, and
13:30you'll be able to create
13:31>> a video out of it.
13:33>> Yeah, exactly.
13:34>> So, can you show the new AI feature? Cuz
13:38th this is something that we're doing a
13:40lot. We're trying to repurpose whatever
13:42is going viral on LinkedIn. This is so
13:44funny. I posted myself sitting in the
13:46garage on a call because my kids were
13:48going crazy upstairs and I just couldn't
13:50take the call upstairs and I'm like in
13:52this messy garage I'm freezing. Um and
13:55it got 120,000 impressions
13:58>> and ideally I want it to become you know
14:00something else I can post. So for sure
14:03>> so I copy link right?
14:04>> Yeah you copy and paste. You can just
14:06say like create a video of this LinkedIn
14:11post
14:13and
14:13>> should I add more prompts like uh make
14:16it a viral re or
14:18>> Yeah.
14:18>> or it should or it's like pre-prompted
14:20or something.
14:21>> Yeah, it's pre-prompted but you can
14:22always always add
14:24>> um or you can you add some thought on
14:26it. Um up to you. So and for voice you
14:30want to also create your own voice but
14:32let's just like use our default all
14:35voice
14:36>> um everything
14:37>> what is the hook
14:38>> hook is that you we have different um
14:41hook templates like the fast cuts
14:45article highlight right now it's kind of
14:48slow uh so probably takes about like 30
14:50to 60 minutes
14:52>> oh 30 to 60 minutes
14:53>> yeah we're optimizing this the speed
14:55hopefully we can get to like 20 minutes
14:57uh in two months I'm going to show you
14:59the result. [laughter]
15:00>> Work and home lives blend together now
15:02more than ever. A constant zigzag
15:05between professional calls and family
15:07chaos.
15:08>> Okay. So with all of this, what will
15:10happen to creators in 3 years? So like
15:11in general creator economy, how do you
15:13see it evolving with all the agentic
15:15tools?
15:15>> I think the entry barrier to create
15:17something is going to dissolve so that
15:21everyone has the superpower to become a
15:25creator. Um that's a brutal fact.
15:28um and it makes the competition much
15:31more challenging to be honest. But the
15:33good thing is that um everyone now don't
15:36have to think about you know what tool
15:37do I need to use or what techniques what
15:40expertise do I need to learn but now you
15:42can focus on figuring out what is your
15:45uniqueness how should you stand out
15:48right what is your unique narrative your
15:50of your story what is your unique
15:52messaging what is your unique tone um
15:55what sets you apart um so just focus on
15:58yourself focus on your own storytelling
16:01and in you know point of two years, no
16:03more than 3 years, AI can just deliver
16:05the job for you. They're doing actually
16:07the dirty work for you, but you're going
16:09to be the real creative one.
16:11>> But it's crazy. The competition is going
16:12to be insane.
16:14>> It is.
16:14>> Do you think there's less and less time
16:16to build a personal brand? Because what
16:18happens if in 3 years everyone has
16:20access to tools, everyone tells a story,
16:22and then we don't have enough eyes to
16:24consume all that content. Do you ever
16:26worry about it?
16:26>> I don't worry that too much. Um because
16:28think about like 10 years ago, 5 years
16:30ago when you want to build a personal
16:32brand, um I think most of the time
16:34you're you're wasting is to um you know
16:37edit the video scene by scene, right? Is
16:39to design design your colors, design um
16:43everything like page by page. I think
16:45that's how you wasted the time. So the
16:48analogy is that is that if you have to
16:50spend five months um building your
16:53personal personal branding from zero to
16:55one but nowadays it's like you just need
16:57a couple weeks
16:59>> my like thing because I started 10 years
17:01ago and I remember editing all the
17:03videos.
17:04>> Something that set me apart was the
17:06willingness to do that. So my content
17:08was super mediocre but because I took
17:10the chance of like you know spending six
17:12hours to edit a video I got my
17:14followers. It's impossible today. like
17:16if my content is mediocre, it's not
17:17going to be noticed. That's the thing,
17:19right? So, your story either has to be
17:22extraordinary or or that's it. Do you
17:25think we still have this opportunity to
17:28showcase our personal brands or because
17:30there's so many, we're getting
17:31saturated?
17:32>> The good storyteller um or the the real
17:36unique people can still um have the
17:39advantage. U think about um you know,
17:41carriage versus cars, right? 100 years
17:43ago very few people have cars. Uh so it
17:47become a privilege to be you know one of
17:48the fastest people in the world.
17:50Nowadays everyone are able to to drive a
17:53car. Almost everyone are able to drive a
17:55car. Um but there are still formula 1
17:57drivers. There are still you know WDC
17:59drivers who are you know way faster than
18:02the ordinary people. So I think this
18:04probably you know 10 years ago you
18:07showed your willingness to become to
18:09build your personal branding by uh
18:11spending extra time um writing content
18:14um polishing your website editing the
18:17videos. But I think in the future you
18:21people who want to stand out should
18:23still probably spend the same amount of
18:25time on something else. not using the
18:29tools, not you know building by yourself
18:32but the time should be spent on thinking
18:35it through how you should stand out. So
18:39I think time spent probably the same but
18:41the actual word are completely different
18:43which is more human actually.
18:45>> Yeah. Yeah. I like that refle I like the
18:47car analogy actually. Thank you.
18:49>> So let's imagine uh you have to start a
18:53company again. Now walk me through the
18:55first 30 days. what will you do
18:58>> if I have to start a company again? The
19:01first thing is always to figure out what
19:03is what is a real painful job to be
19:05done. Um I want to so I I'm the
19:08prototype of founder who don't start
19:11everything from technology but from um
19:14users from the market. Um I have to
19:17segment my market um to a very clear
19:21very vertical niche that I can clearly
19:25understand their existing workflow,
19:27their existing pain point um their
19:30existing alternative solutions and then
19:33so I think probably I I would have to
19:35spend you know the first couple weeks um
19:37two or three weeks understanding the
19:40real use case um the very specific ICP.
19:44Um and the second thing is to just
19:46engineer a you know prototype or pro
19:49proof of concept. I can do that very
19:51easily with all the web coding tools
19:52right now. Right.
19:53>> What's your favorite?
19:54>> Uh well I'm a heavy um cursor user. Um
19:59but I I tried many other different tools
20:01but cursor is my goto goto tool. Yeah
20:04>> cuz I came from engineering background
20:05so it's um yeah more familiar with the
20:07IDE. The step two is actually just um
20:11couple days should be good enough u to
20:13have a proof of concept. Uh and then I
20:16go back to share that with the early
20:17users um the ICPS to get you know their
20:20feedback. Uh the feedback should not
20:23only about do you like the product but
20:25also like what kind of problems are am I
20:28solving for you and what is the value
20:30you perceive which means how much money
20:33you are willing to pay for it. Um right
20:36so that that process probably is the
20:38majority of the early 30 days but I
20:40think on the other side I'm also I also
20:44need to think it through about what kind
20:46of propri proprietary data can I possess
20:49along the way and will that data set
20:51grow as the product grow as the user
20:54grow right um we don't have to build
20:58toward a moat um or a very clear plan of
21:02differentability but I think We need to
21:05have that idea
21:07at least well thought in the first early
21:09days. Um because you have you can't
21:13avoid it right when you really launch
21:14the product. So it's better to have
21:17something in mind have some idea in mind
21:19>> like what makes you stand out market and
21:21what makes people come back.
21:23>> Yeah. And if you ask this question,
21:24especially nowadays, especially we are
21:27entering 2026,
21:28>> we're switching between apps like crazy.
21:30Like there's new shiny app, you install
21:32it, you're like, "Oh, I love it now."
21:33But the next day something else comes
21:34out and it's so easy unless you have
21:37history within certain app. Right.
21:39>> Exactly. So the last thing I think we
21:41should consider is the distribution
21:42channel. Um I think the competition is
21:46totally different. Yeah. Two years ago.
21:48two years ago you can easily say you are
21:50the chachibility of ABC right it's so
21:53easy to go viral right at that time
21:55nowadays like everyone knows AI everyone
21:57has used AI um so figure out a
22:00distribution channel will help you
22:02better refine your user experience your
22:04on boarding and also um your target IP
22:08as well
22:08>> so [snorts] basically you start with
22:10expertise it looks like you need to
22:12understand at least one niche to be able
22:14to build something not like oh this
22:16market looks good I'm going to try and
22:18build something for it. No, you need to
22:19be deep in that market and understand
22:20the problem.
22:21>> Exactly. Because you asked me a
22:23question. Uh the assumption is today,
22:25right? AI has been there for like more
22:27than 3 years. The LM has been there for
22:28more than 3 years, right? It's because
22:30if you ask me like 3 years ago, the
22:32answer was will be much simpler. But
22:35nowadays,
22:36>> just build with AI. [laughter]
22:38But now when you know LLM are advancing
22:41so fast and large companies are
22:43releasing better and better products. Do
22:45you think there are any niches or
22:47problems that are not worth solving
22:49anymore just because big companies can
22:51take over in like a week?
22:53>> I think two major types of problems that
22:55a founder should avoid. Um the first one
22:58is that you are just building a feature
23:01for an existing type of ICP ideal
23:04customer profile within an existing
23:08workflow built by the incumbent. So
23:11basically what that means that is that
23:13um the incumbents can easily build a
23:16feature that you know bundles everything
23:19right and probably you won't have your
23:22own distribution channel for example I
23:24think like if you want to build a
23:26noteaker think it through right probably
23:28find another direction because it's
23:30super easy for zoom or you know Google
23:32meet to have that feature in their
23:34existing workflow because you're
23:35targeting basically the same ICP same
23:38market same almost you know same or
23:40adjacent use cases attached to a big
23:44workflow, a big platform. I think every
23:46AI founder should be somehow AGIP, which
23:50means that you can predict or you should
23:53be confident to make some predictions
23:56about what the financial models can
23:59release in the next few weeks or months,
24:02right? if they are already doing you
24:04know some job 80% very well 90% very
24:08well I think it's you know in their
24:10second in their next few release they
24:11can probably do that job like 99% well
24:14or even 100% right so um run that test
24:19in your internal strategic discussions
24:22if you're just becoming a rapper with
24:24some prompts then probably you don't
24:27have to write any prompts in the second
24:29release in the next release of um Gemini
24:31or CH GPT you need to focus on solving
24:35like a vertical business problem um by
24:38like integrating
24:40the workflow end to end right you need
24:42to own the workflow end to end so that
24:45you know AI is part of the workflow but
24:48it's not all right instead like if you
24:51build a rapper with some prompt
24:52engineering AI is almost the endto-end
24:54workflow itself
24:55>> can you also walk me through um the
24:57process of coming up with pricing for an
24:59AI tool how do you decide on cuz I know
25:02you have lower tier but then you also
25:04have like a 20k check for um bigger
25:07companies. How do you come up with the
25:08price?
25:09>> Pricing is a is a real science.
25:12[laughter] So couple of factors we uh
25:14considered in the first place. For
25:16example, first what is your value
25:18creation? How to measure your value
25:21creation? Right? Um before using your
25:24product, what would your users do to
25:26achieve the goal uh to get the job to be
25:29done? uh they probably you know spend
25:31their own time which there is a price
25:34for their own time right they probably
25:35outsource it to some vendors or um yeah
25:39they pay humans to get the job done or
25:41they may like use some like OG tools uh
25:45by some professional guys like in our
25:48use case professional editors to to get
25:50the job done. So in our industry like
25:54editing a clip, editing a like a
25:56[snorts] viral ready high quality clip,
25:59one minute clip often takes about 1 hour
26:02or at least 30 minutes and the price the
26:06market price is about 25 to $50.
26:10>> So that's the value creation we're
26:12benchmarking against. The second thing
26:14is about your unit economy. um because
26:17you have inference cost um and that can
26:20even that can go down in the future but
26:23um it's still very pricey in the early
26:25days and also for videos another
26:29elephant in the room for the cost is the
26:32storage. The storage can be just 5% in
26:36the early days but it can become like
26:3850% of your total cogs after 3 or 5
26:42years.
26:42>> Yeah. So the unit economy you want to
26:44make sure it's healthy, it's um
26:47sustainable. The third one I was I would
26:49say just do as many experiments as
26:51possible. We actually send out like
26:53thousands of surveys in the first few
26:55months and we tweaked our pricing logic
26:58again and again well in the early days
27:00right you don't want to change
27:02drastically in the in the late stage but
27:04in the early stage um we we did a lot of
27:06survey customer interview to understand
27:08how much they are willing to pay. Um
27:10because even you have a clear value
27:13creation and a well- calculated um unit
27:16economy, can users still perceive that
27:18right? Is your metric clear enough for
27:21user to understand the value creation,
27:23you know, to buy in your value creation
27:26that is still relying on your market
27:29messaging and also your um metric setup
27:32like most of the products are using a
27:35metric of usage instead of seat. I think
27:38that makes sense especially for solo
27:40printers. Uh but the experiment really
27:42tell tells us what price to go because
27:45you don't have to uh set up a price that
27:47everyone are happy with. You probably
27:50have a assumption or hypothesis of what
27:53kind of customers you want to target
27:55too, right? All you need to do is to
27:56make sure those ICPS are okay, are happy
28:00with the price. you probably have to say
28:03no to like 70% of the early stage users
28:07but you should be to only focus on your
28:09ICAP in in the early days and then maybe
28:12you can develop other you know pressing
28:14or even product um capabilities for the
28:17rest of the users.
28:18>> Yeah, the the importance of saying no
28:20right all the time. When you said you do
28:22customer interviews, how many interviews
28:24do you need to do in order to land on
28:26some kind of a decision?
28:27>> I think the number of customer
28:28interviews are not the key. Uh but how
28:31you do it is more important. Um I think
28:34we did about 20 to 30 um for on average
28:38for one single um critical product
28:40decisions. But we set the distribution
28:44of the those 20 to 30 um customers in a
28:47in a very um mey way where like okay
28:51four marketers you know five creators
28:55and also they're in different
28:56industries. They're they come with
28:58different um you know purchasing empower
28:59come from different geographic locations
29:02backgrounds. You want to make sure they
29:05are representative enough to help you
29:07make a decision.
29:08>> You don't need a lot but you need to
29:10have a clear concise set of um testers.
29:15>> Yeah, you want to make sure you get like
29:16a full picture of the market.
29:18>> That's really helpful for the research
29:20and validation phase. Now once you're in
29:22building mode, what about the AI skills?
29:24What is the number one AI skill everyone
29:27should be working on right now?
29:29>> The number one AI skill is should
29:31actually go for first principle. It
29:34doesn't matter like if you're using AI
29:36for um you know design your poster um
29:40you v code your prototype. I think
29:43everyone should treat AI as your
29:45thinking partner or even thought
29:47partner. Which means that when you are
29:49having a problem of understanding your
29:51users, when you are having problem of
29:54managing your teams, when you are having
29:56problems of um figuring out your pricing
29:59or you know all the critical decisions
30:01in your lifetime uh when you are a
30:04founder, I think traditionally you will
30:06just go reach out to your coach or some
30:08more senior people uh or people with
30:11relevant experience, right, to ask for
30:13their advice. But I think in this era of
30:15AI, you should run through it with your
30:18with you know Gemini or Chad GBT. They
30:21are actually a very very senior very
30:24important thinking partner. Um just you
30:28know instead of like asking one line of
30:30questions um throw as many context as
30:33possible um and also you know do like
30:35more than 20 rounds of back and forth um
30:38communications.
30:39you will be mindblowingly enlightened
30:42through these conversations. So nowadays
30:45that's how I practice myself.
30:46>> Do is there a certain daily practice
30:49that you have? Uh I talked to Mustafa
30:51Sleman on this podcast. Uh he's a CEO of
30:54Microsoft AI and he shared that every
30:56single day he would talk to Copilot and
30:59just tell Copilot about what the day has
31:02been like and the decision that he's
31:04made and how those decisions made him
31:07feel. So that when in 3 months he has
31:10something like a similar problem, he
31:11still talks to the same threat and
31:13copilot and copilot will be like, "Oh,
31:15when you made that decision 3 months
31:16ago, you actually regret it. So this
31:18time let's do this." Is there something
31:20similar that you have?
31:21>> Yeah. Exactly. Exactly. Because I think
31:23the current uh chatbot uh memory is so
31:26powerful that uh so what I um like
31:28monthly one of my monthly ritual is that
31:31I ask chat to PT about um you know what
31:35are my major decisions in the past month
31:39um you know give me some comment um
31:41feedback on
31:42>> but that means you shared every single
31:44>> I shared everything. Yeah.
31:45>> How does it like every night you just
31:47sit and like these are the decisions I
31:49made or
31:50>> Yeah. Sometimes I you know talk to the
31:52AI in that way and sometimes I would
31:54just um
31:56>> um forward my decision just um like I
32:00captured like a screenshots of our group
32:02discussion. Um I dropped a link of my
32:06the document of of a PRD of a spec uh
32:09that kind of thing too. I yeah I I I
32:11kind of forced myself to
32:13>> to document.
32:14>> Yeah. To document it.
32:16>> So you said Gemini is the best for it or
32:18>> Yeah. I'm um I'm I'm heavily using both
32:20JA and ChBT. Okay. Yeah.
32:23>> But I think recently I realized they
32:26really memorized almost everything I
32:28talked to them. So now I can start a new
32:31question like what is the biggest
32:34mistake I made in the past 6 months? Uh
32:36what is one thing you you want to
32:39suggest to me? Uh if you could three
32:41months ago. I think this becomes super
32:43powerful right now.
32:44>> I love these questions. Let everyone go
32:48go to your favorite app and ask those
32:49questions. I'm going to do this right
32:50after the interview.
32:52>> It's a good like end of the year
32:53practice [laughter] to see what's what's
32:55going on. Love it. Absolutely.
32:57>> Do you have any top three AI business
32:59ideas you would start
33:00>> if you hadn't had Opus Club?
33:02>> It's hard to name like three ideas um
33:06because I'm not an investor. I you know
33:07I I don't meet as many founders as
33:10possible. Um but I think the general
33:12first principle uh for now right the
33:17question is to start a business like in
33:192026.
33:20>> Yeah
33:20>> as we discussed the market is becoming
33:22super competitive and there are like for
33:26most of the markets or industries uh
33:28there are like more than 20 50 AI tools
33:31out there already. Um so the first
33:33principle I would say is to find a
33:36really niche business. Often times many
33:39people don't really understand what is a
33:41niche. For example, is restaurant
33:43business a niche? No. Is Chinese
33:45restaurant
33:46a niche? No. Is um canonist Chinese
33:50restaurant business a niche? Still no.
33:54Um you have to drill down. Uh there are,
33:56you know, so many different types of
33:57canonist food. Um and even for one
34:00specific vertical um are you going to
34:03target to be like the um the Chanel of
34:06Canada's restaurant or the Zara of or
34:09Uniqlo of Canada's restaurant right so
34:11the different pricing will match to
34:13different um markets different ICPS as
34:16well so pick a really really really
34:18small niche that you cannot further
34:20segment it
34:21>> to start with um
34:23>> Wow that's that's a good one yeah
34:25>> yeah you have to ruthlessly segment your
34:28niche
34:28The second thing is ideally pick
34:30something that's boring because the non-
34:33boring the cool ones are definitely 10x
34:36or even a more 100x more competitive.
34:39You probably don't want to work in those
34:41areas. That's why I mentioned earlier
34:43that the passion for building the
34:45passion for problem solving is more
34:48important because your passion for a
34:50specific business is most likely not a
34:53viable business. Right? So you need to
34:56have a passion to build something for a
34:58boring industry but you solve a really
35:00big problem and furthermore ideally
35:04there are a lot of existing services. I
35:07think we are actually you know changing
35:09from we're redefining the term SAS and
35:12previously it was like software as a
35:14service nowadays it's becoming service
35:16as a software. So in that specific
35:19market niche, boring market niche, are
35:22there existing services
35:24um done by like agencies um by
35:27freelancers um or some internal tools or
35:30some hacky solutions, imperfect hacky
35:33solutions. Um that is your opportunity
35:36to tackle. Um so go through some boring
35:39test, go through some niche test and go
35:42through some service test. Um, so these
35:44are the my first principles to find a
35:46new business in 2026.
35:48>> That's great. That's great. That's
35:49amazing. Okay. And for young viewers who
35:52watching, what is one principle that you
35:56wish you understood when you were 20
35:59that would saved you years?
36:00>> Damn, [laughter]
36:01>> it's like me talking to your internal
36:04Gemini tragedy. [gasps]
36:06>> I really wish I can travel back to my
36:0820s. But um I think the first principle
36:11is to be as disciplined as possible. Um
36:16I've been following um Cristiano Ronaldo
36:19and LeBron James for two decades.
36:23They are at their 40s and they're still
36:25one of the best players in their fields.
36:28The reasons are they are super
36:30disciplined in their early 20s. Uh you
36:33can't change your behavior when you are
36:35over 30 or 40. it's it's just
36:38impossible. Um but if you're like in
36:40your early 20s or even younger, um make
36:44sure that you use time effectively. Make
36:47sure you can plan things accordingly.
36:50Make sure you push yourself to your
36:52boundary. Make sure you are okay with
36:56you know suffers. Um you can recover
36:59fast. Uh make sure you have clear
37:02mission and you are well aligned with
37:04your mission. and you work toward
37:05automation every day. I think you know I
37:08can list a few different behaviors but
37:10the key principle is to be super
37:13disciplined if you want to be a
37:15successful founder uh later on and also
37:18be disciplined to your own health right
37:19like how many hours you should log to
37:21sleep um what you should eat or not um
37:25what kind of exercise you should do you
37:27should do every day like I I saw many
37:30people started to develop these um
37:32patterns these behaviors in their 30s
37:35but not in their early 20s.
37:37>> It's hard in your 30s. I love that. And
37:40also that brings you to say no to
37:43different things and it's also a great
37:44skill.
37:45>> Yeah. Yeah.
37:46>> Yeah. This is something I've been
37:47hearing a lot when I'm talking to people
37:50who are building something amazing.
37:52>> Yeah. Discipline. I think Priscilla Chan
37:54just shared that she only has work and
37:57family. Parties don't exist in her life.
38:00>> Not exist anymore.
38:01>> Yeah. And this is what you hear a lot
38:03from successful people here in the US.
38:05They work like crazy.
38:07>> I'm trying to learn. Like I think I'm
38:09disciplined, but like that level of
38:11discipline. Oh my god.
38:12>> Yeah. Exactly. Because with a super high
38:14level of discipline, you can just like
38:17master said, right? You can segment your
38:19your time into, you know, buckets of
38:23five minutes. You can easily you can
38:25easily get super focused and also switch
38:27context to focus on something else. That
38:30is the ultimate result of discipline. I
38:34saw many people like find really hard to
38:36switch context to focus for a longer
38:38time. That is because you are lacking
38:41the discipline from day one.
38:42>> This is something I'll be teaching my
38:44kids. [gasps and laughter]
38:45>> Thank you so much. It was amazing.
38:47>> Thank you. Oh, thank you ever having

About this transcript

This page contains the full transcript of $215M AI CEO: How I’d Build a Profitable AI Startup in 30 Days (2026 Playbook) by Silicon Valley Girl, generated from the public captions YouTube serves with the video. The transcript has 6,792 words across 985 segments, with the original timestamps preserved so you can click any line to jump to that moment in the embedded player.

What you can do with it

Use the transcript to take notes, quote the speaker, build a study guide, generate a summary with ChatGPT or Claude via the YouTube Summary tool, or export it as a timed subtitle file with YouTube to SRT. You can also re-open it in the transcriber to translate the transcript into 100+ languages.

Free YouTube transcript tool

YouTube2Text is a free YouTube transcript generator — no signup, no daily limit. Paste any YouTube link and get the full transcript instantly, with timestamps, click-to-jump, translation to 100+ languages, AI prompts for ChatGPT, Claude, and Gemini, and exports to TXT, SRT, VTT, or Markdown.