0:00 A lot of businesses can really scale to tens of millions of dollars today profitably with AI. This is Alex, founder of Hexafield, an AI company that hit a $200 million annual recurring revenue in just 9 months, faster than Slack or Zoom. And what he told me about starting a business today completely blew my mind, and you can copy his strategy, too. Focus on bringing like the first dollar by day 30 of product development, and maybe 1 million ARR by day 90. That's a lot. I just think it's the next industrial revolution. It's probably more powerful than the internet. For many, many young people, AI becomes social elevator. >> For someone who still has fear, like this is moving so fast, I don't know how I can start. Can you give them one piece of advice? First, I would start from
0:47 I have an amazing guest today. I am so excited to learn from you. Let's get very practical right away. You built Hexafield and achieved $200 million in revenue in 9 months. Let's imagine every I don't want this to happen, but let's imagine a scenario when you have to start from scratch tomorrow, and you have 90 days to launch a business idea. From what you've learned with your experience at Hexafield, what would you do? I think it all starts with maybe a team of two, someone who is builder, who can go within 24 hours from idea to a product. And now it's all becomes possible. There are so many databases, there are so many payment systems, and so on, which simplify creation of MVP. And then
1:33 someone, as I call it, go-to-market person, who has this natural empathy maybe or understanding of the sort of target distribution, whom they're selling to, and who can come up with interesting kind of new content formats, which can resonate with the target audience on social media. And I think this is a very different skill set from the like marketing roles of the previous decades. And how many times should they be ready to iterate? How many ideas? For example, for us last year, we were iterating every day. So, it was six um days a week and um every day we were putting new product release. As we were trying to find workflows and use cases
2:19 which have high frequency and which matter for our target audience. And then, once the technology gets there, it's important to develop sort of the workflow, which is easy enough but gives enough configuration as well. This dilemma of the perfect interface is still not solved, frankly. So, uh this is another reason why we embrace daily iteration. And on top of all of that, every month the whole industry resets. They completely push the boundaries in terms of the capabilities. There are probably, let's say, five leading research labs. And each lab is pushing massive updates every quarter. So, at some months we have even two major updates. And it typically requires to substantially rebuild the whole product
3:05 around those models. So, it's exciting time today cuz product builders, like ourselves at Hicksfield, we we just try to evolve the product so that it highlights the best possibilities of these models to our customers. But, that's a constant embrace. >> Yeah, it sounds like a very challenging race. And I think you mentioned that last year was one of the hardest for you when a lot of things were not working. Can you talk to me about that one thing that actually worked? In 2024, we really started from the mobile apps. And things were not working well cuz retention for mobile apps is relatively low. Things drastically changed for Hicksfield when we started to constantly iterate with creatives. And we just
3:51 asked them very simple question like, "Did you see this video?" This was a cool AI-generated video and they said, "Oh, no. How is this possible? What's the cost?" And we say, "It actually cost maybe less than $500 to make this video." And then we ask, "Did you actually try these models? What's your experience with AI?" And and obviously everyone tried AI by then. Even by maybe February last year, everyone tried AI, but everyone had some issues with that. Back then, we realized that the core limitation was around camera control. Like a lot of creative directors, they really want to control all the camera effects, camera angle, and so on. Back then, there was no system to achieve that. So, that's the
4:36 initial traction of Hixfil.ai came from these camera controls, which we implemented on engineering side based on the feedback from creatives. >> So, how many interviews did you have to conduct to come up with this feature? >> This is a very good question, but it kind of puts me on a weak spot, frankly, cuz we interviewed eight people. Eight out of eight said the same thing. And we talked from Hollywood level movie directors to like regional producers of commercials. Everyone had the same feedback. >> And how did you select those people? Were they customers already or you just wanted to talk to people in the industry? Actually, this was probably a challenge as we wanted to talk to people who we don't have a very close relationship to just get an unfiltered opinion. But, the feedback was very consistent.
5:22 Everyone was missing these camera controls. So, that's what we delivered March last year. Then in April, we delivered a library of visual effects. And then, I think in June, industry completely changed. We saw the emergence of AI native marketing agencies. So, essentially those agencies, they completely go end-to-end with AI. And very often, they try to bypass incumbent tooling like for example Adobe or something else and go end-to-end with AI. On the one side they are very limited because AI capabilities back in June were a little limited. On the other hand, they drastically improve their margin profile
6:08 and they kind of show their clients that they can build ads within days. A lot of brands actually want to have constant content flow on their socials and they want to embrace AI. And then from June to December last year, this um this new industry of AI native agencies completely exploded. But basically you said the the start of your growth was the multi-angled uh camera view and it came from talking to people in the industry. I love that. And also the number eight is actually very consistent from what I'm getting uh talking to other founders. It's normally like 12 to 20, but it's not too many interviews cuz I feel like a lot of people think they need to talk to thousands of people to figure out the problem, but it's actually like around
6:54 10 interviews. >> Yeah, absolutely. So eight people who actually helped us to shape the product and then I think we hired four of them. Oh, you also >> Yes. So now we have this feedback loop within the team. That's awesome. And that's how we realized that probably the best products in creative AI is going to be built in symbios uh like in collaboration between engineers and between cre- creators. So today roughly half of our maybe 40% of the team are um engineers, maybe 40% are creators. Like you said, two founders, right? One is technical, one knows the consumer. It's basically reflected in your team. Okay, let's let's get back to that tough year because I feel like for a lot of people that's what they're scared of. So when you were building this and nothing was really working and
7:39 then Google releases the new video model. Did you ever think about giving up on that particular market and starting something else because this was getting so crowded? >> I think we were committed to figure this out. There are two reasons why we had a conviction. First, is that prior to that I was at Snapchat, I was running Genie I there, and I saw the uprise of TikTok and CapCut. CapCut became top five apps in the world. And it's unprecedented says it's not a messenger, it's not a social media. And this was a strong signal for me that the needs of social media creators are simply unmet in the markets. And the second reason why we had conviction, we constantly heard that creators feel a
8:27 burnout from sort of feeling pressure to record multiple videos a day for socials with their own face. Mr. Beast, he spoke very openly about that. But that's I think that that has been a primary challenge for the whole industry over last four or five years. And then what was surprising to me is that advertisers have the same problem. Most of the advertisers talk about like maybe mid-market brands, they can be spending hundreds of millions of dollars on marketing and they don't have any production team. So you chose the right part of the market. This is what I'm hearing. So I feel like for every entrepreneur who's watching, uh when you see a big opportunity in the market, you also have to spot who's paying the most. Not just go after each
9:14 user, right? Absolutely. We live in a very interesting um era where all the powerful companies try to give AI to everyone. Literally, Meta, XAI, Google, Microsoft OpenAI Anthropic ByteDance and many, many more. Think each of those companies want to give a yacht to all its customers really. That's why startups have to have more nuanced view of the world and have to have more nuanced set of customers. And I think um startups today are sort of incentivized to stay cash flow positive and build real business from day zero. This is what I see across Higgsfield and
10:01 other top application AI companies. So, I think this also creates a very interesting dynamic that the target audience should be maybe tens of millions of users who are willing to spend hopefully couple thousand dollars a month. So, this is not for everyone, but the delivered value should be so strong so the product should be so good so that it kind of sells itself. >> I love that and I love that you have a concrete number, $2,000. Yeah, $2,000. I'm not sure like with Higgsfield, for example, this is the core metric which we are tracking. Like how much of the value we believe we provide, which is more like through the user interviews, but also how much we charge
10:49 users annually. So, this is one key metric and we are not chasing just monthly active users, for example. Cuz the monthly active users number can be inflated through some viral effect. Monthly active user doesn't really speak to the frequency of the usage and the value delivered. So, that's why for us really daily active users and average ACV, average contract value, those two metrics are the most important ones. One of the reasons I was so excited to sit down with Alex is that our team has actually been using Higgsfield for a while now. My producers and editors absolutely love that all the top AI models are in one place. Man and bananas, seed ants, clean veal, Sora. And I love that they don't need 10 different subscriptions. And it's kind of incredible. Alex just told
11:35 us that last year they were shipping a new product release 6 days a week. And now everyone knows them, but they haven't slowed down. They keep launching new features. And one of the latest ones is Sora 2.0. Sora 2.0 is Hex Fit's own image model, and it's not like anything else out there. Most AI image generations, you type a prompt, you get something that looks fine. Sora was built specifically for creative work, fashion, editorial, content. It actually understands aesthetics. You give it a reference photo, and it doesn't just copy it. It reads the lighting, the grain, the mood, the era, like a creative director would. There are three models. Sora, that's the core model. You prompt, you get a beautiful image with real aesthetic awareness. Sora reference, you upload a reference and it
12:21 generates new images that match the vibe, not just the composition. Same visual DNA, different shots. And Sora ID, you upload 10 or more photos of yourself, it learns your face structure, skin tone, expressions, then it generates you in any style. Y2K, editorial, film photography, Polaroid, 20 different presets at launch. And here is what actually got me. You can specify the camera medium. Say, shot on Kodak Portra, or disposable camera, and it changes the grain, the color science, everything. It doesn't slap on a filter, it actually shifts the entire feel of the image. And the same thing works with color. They just added hex colors, so you can pull the exact palette from any photo and apply it straight to your generations. If you want to try Sora
13:07 2.0, I'll leave the link in the description. And now, let's get back to Alex. Can you give advice to people who are watching who haven't started yet, but they haven't started because they think a large company is going to take over? How should they think about their defensibility? I think each company can really keep their focus on maybe two or three top priorities. Interestingly enough, people like to say about Anthropic that their major success from MCP and Claude Cods actually was not like a top-down, but really bottom-up. And it like Claude Cod's success was not sort of planned. And I think it's true that like these large companies, they definitely can benefit a lot from applying top-down approach from two to three initiatives
13:54 and really consolidating all the resources to make the most progress there. I just think it's the next industrial revolution. It's probably more powerful than the internet. So, the number of products and ideas to be built, I think just outweighs uh uh number of ideas which OpenAI or Anthropic can push internally. That's why I would encourage builders to build. Especially today when like quite small team of maybe 10 people can build like high-scale products. Do you think there is that we have a certain gap of in time when we can build? So, for example, I've been hearing a lot in social media that we only have 2 years to build something new because then we're going to have some companies that reach AGI and it's going to be impossible to find a gap in the market because those companies are going to be
14:40 filling those gaps. This is a good question. We definitely make um tremendous progress as an industry to automating uh work in digital worlds. And for sure, maybe next decade, there are going to be a lot of applications of AI in physical worlds. It's difficult to forecast how much progress we all are going to make there, but this decade is definitely the era of digital economy completely changing with with AI. That's for sure. I really don't want to believe in the future when there are going to be maybe three labs who have the best models and these models controlling all the worlds. This could happen. That would be very difficult future. So, I
15:25 would I try to stay optimistic and encourage everyone to stay optimistic and really focus on building and delivering value. Like, for example, um today I just want to give you a very concrete example. So, I recently spoke to a very large, very, very large like property management company, which actually uses Higgsfield to to sort of advertise their buildings, their uh apartments, and so on. Like, no one is building for this industry. Like, in this industry, there is a very specific um workflow of a customer. Like, customer needs to learn about about the property, then they need to go to the website and get all the details, then they need to call, then they need
16:11 to show up, then they leave deposit, and the whole customer journey. No one is actually building solution specifically for this industry to like cover this journey end-to-end with agents. And the reason why agents are going to deliver a lot of value in this specific business is that uh customers who want to, maybe, let's say, rent apartment, especially in certain price points, they want to make a decision rather quickly. So, every day of delay, every day of just moving from one stage to another, is just is just lost revenue. So, and this business owner just said to me that no one is building for their industry. And And I'm confident there are many more examples like that. Yeah, so riches are in the niches, as as they say. From your
16:58 experience, when you pitched VCs with another AI idea, what makes a pitch stand out these days? >> I think today, um there is definitely fear that OpenAI, Anthropic, and other labs are going to just uh be uh very acquisitive and uh just trying to expand their product offering to multiple different verticals. Pretty much every week there is cloud for X launched and stocks go down. That definitely happens and that creates a lot of fear. And I would just say that the core inside for me personally was that most of these hot and the hyped AI companies they are actually cash flow positive. Like maybe it is a wrong mindset to go and raise
17:46 venture capital today. I think there is plenty of pre-seed capital available today. But I'm not sure every everyone needs like to raise series A, B, C, D and so on. >> How much revenue did you have when you raised your first round? >> For me with Kickstart AI it was probably easier cuz I had previous exits and we basically raised $16 million without having any revenue just with maybe having like million users for our mobile app. But this is not exactly the way how I would recommend to build today. I would recommend to focus on bringing like the first dollar by day 30 of product development and maybe 1 million ARR by day 90 and
18:32 then decide if someone needs VC funding or doesn't. I mean a lot of businesses can really scale to tens of millions of dollars today profitably with AI. And for such businesses there is no need to attract VC funding. I can give you very simple example. So there are so many websites which just allowed to make professional photo shoot like for basically for for passport. Many of them make tens of millions of dollars. None of them are going to be Yeah, none of them are going to be venture capital backed business. Well, and you said something $1 million by day 90. >> ARR meaning like 80K a month. 80 80K a month in 3 months that's a lot. And so the the playbook to achieve that is basically generate ads and launch them and test whether whether they're landing with your target audience.
19:18 >> Paid ads are very difficult today, I think. A lot of distribution come through organic social media and creator integrations. And just make sure that by day 30, uh there is monetization in place, and then there is a way to constantly grow to revenue to, let's say, 1 uh million ARR by day 90. Many successful uh companies scale very quickly today. Just different verticals have different capacity. Some In some verticals, the ceiling could be just 50 million. In other vertical, 1 billion. In other verticals, like 100 billion. >> Any tips on landing first customers in the first 30 days? >> Initially, especially last year, Twitter has been the social media
20:04 where the distribution starts from. It starts from like small communities, then it goes to AI news pages on X, then from AI news pages on X, it goes to Instagram news pages, then from Instagram news pages to creators, then it goes to Telegram and like other social media. That's what we have seen with Higgsfield and with many other products as well, that they went through the same sort of journey of popularity and news through various social media. But it all originated on X. I think now it's being kind of changed. Today, a lot of hype is like a lot of companies, they sort of try to use X to boost their product. Uh sort of signal-to-noise ratio just drops. Twitter becomes less relevant,
20:50 but still it's it's the main place for new AI products launch. That's awesome. I'm still trying to crack the the X strategy. I feel like if you add a word breaking or just in to whatever you're posting in the beginning, it should be in caps. Like then then it will to perform. >> then it's some something should be like claud, cooked. Yeah, yeah, yeah. It's like >> RIP, like something like that. >> Yeah, just wiped this out of the market. Yeah, it has to be very sensational next. But you're so right. I've heard so many and I know a lot of creators who built their whole like email newsletter, 1 million subscribers just off viral X uh posts. >> Yeah. That's amazing. I love this life hack. Thank you so much. Although this has been the primary life hack of the 25 and I do believe that's the media evolves itself as well. So 26 could be
21:37 different. >> Well, I would see LinkedIn on the rise, though. >> [laughter] >> Maybe LinkedIn. Maybe LinkedIn. >> You had a previous company. You sold it for 166 million. >> Were there any key learnings from that business or mistakes that you made that you will never repeat in this one? It it's it's like never say never, but one of the key learnings for me and the key takeaways was to embrace meritocracy, sort of. Like I'm [clears throat] 30 and I feel sometimes that I am quite old for this new era of AI. Like a lot of new ideas at in Hicks will today come from these kind of fresh grads, maybe 23, 25, who sort of maybe never worked in a large company, who are doing like
22:24 freelancing with some web coding tools, who are doing web coding before the term web coding, basically. And uh they they just think differently and I think that's sort of maybe the right mindset. So traditionally in any like corporation, those people would be just simply no one would just simply listen to them. So and the same applies to the creative role. So there is definitely some resistance from people who especially build like, let's say, like large Hollywood projects, that's all AI is dangerous, it's not authentic. And it's in the same time it's really exciting that for many, many people, young people AI becomes social elevator. And I sort of strongly relate to that personally cuz for me I had to do a lot
23:10 of competitive programming, you know, like who solves more problem within like 5 hours, you know, like and everyone in the world competes. So, this was social elevator in 2010 maybe. And this applies both to uh creative and to software engineering as well. I love how you said that AI is a social elevator cuz I feel like social media was the social elevator for me and now is the era of AI. I think you mentioned that video models could be a path to AGI and that you're also building a world model. Can you talk to me about that? And for everyone who's watching, just wanted to explain I was just in Davos and everyone was talking about LLMs having a ceiling because basically just describing our world with words is something that we're used to and there's a lot of information on the internet,
23:55 but understanding the physics is the next level. And once we understand the physics, then we're going to have robots walking around our house and doing chores if I'm explaining this correctly. >> Absolutely. I think Demis from Google and Elon from xAI, they started this narrative. And definitely they are top influencers in the space, that's why now the narrative goes to message to everyone. It's still unclear if that's sort of the path to AGI, although that's definitely a path to advanced robotic systems. Like I think Elon proved to the whole world that self-driving cars can work really well through just cameras. And I think the same logic is going to apply to more advanced robots as well.
24:40 That's why developing perception and visual understanding is critical for next wave of robotics. And it's a top priority for a lot of research labs as it's the next frontier. And there is no other way to improve video generation without visual understanding. Roughly it takes around I think people like to say that in 1 minute we can read 200 words. 1 minute of the video e can be described with maybe 10 60,000 words. There is just so much going on. If you and I were having coffee after this episode, I'd certainly pull out my phone and say, "Look what I tried this week. It changed how my tea works." This is what I do all the time. To share these kinds of things that don't fit into the podcast, I started my
25:27 own newsletter. Every week I write about AI tools, strategies, and experiments I'm running in my own business with real numbers, real results, templates that you can use, and also honest mistakes. If you want to be in the loop, the link is waiting for you in the description. So, where do you see yourself in 2 years? Are you still working on videos or because I cuz I saw Menlo Ventures announce their investment in you and they they said that Hicks' field is building the next world model. Do you feel like your focus is going to shift to that or you still going to stick to the marketing with video? I think that what differentiates us from many other is we are from day zero we are focused specifically on short-form content. And I do think that world model is going to change the way how social media content is made. So, for example, next decade is going to
26:14 be the era of interactive media. Uh basically when games and videos are sort of blurred in like one experience where like it's kind of choose your own adventure. So, a lot of marketing is going to go this way, especially kind of premium marketing and customer loyalty programs. I think this decade is all about supercharging creators and marketing with those models. You said you're seeing customers with marketing budget over a hundred million dollars, 90% of their ads are AI generated. As a creator whose 90% of my revenue is from large corporations through from their marketing budgets, should I be afraid? This is a good question and um this is definitely where we should make sure that video AI
26:59 can help personalities. Like what's I would love to see is that creators like you, Mr. Beast and so on, who are sort of AI native, start to make like more channels and just expand their media presence and be like the whole media empire. So like the next media empire is going to be built maybe with like 300 people, 500 people and be worth like $10 billion. So that's on the one side. On the other side, what I see personally is that essentially this high-scale AI content creation comes after maybe platforms like TikTok marketplace, frankly. So in the past those brands they could just hire sort of even programmatically
27:46 thousands of creators through TikTok marketplace just to do kind of these templateized videos. Like I use this product, it's so good, go buy that. So like this entry-level marketing definitely gets democratized. Although in my opinion, as we are seeing a lot of AI slop on social media and so on the genuine connection and and understanding of audience now matter more than ever. So there are going to be just a lot of average and above average contents on the socials. And I think this is happening one way or another and just deep understanding of the audience and authentic approach matter
28:33 more than ever. Okay, so you think social media a ladder, social ladder is not dead. Definitely not. I just think that authenticity is going to matter a lot. And and the way how I think about that is if let's say top creators now can make their own shows, their own movies with AI. And creators can drive a lot of traffic. That's going to affect like streaming business a lot, I think. So, I think the revolution is going to come at every level. Although, I think clearly people who already understand their audience, who have authentic approach, those people are definitely going to be the winners. Okay. My last question. For someone who's watching this
29:19 still has fear. Like this is moving so fast. I don't know how I can start. I started something today, it's outdated by tomorrow. Can you give them one piece of advice so they can start? First, I think I would start from a position that large companies, Amazon, Microsoft, Google, OpenAI, Anthropic, think they're all relatively well positioned to be winners in AI era as they simply control data centers, GPUs, and so on. But I'm not sure there is much insurance policy for everyone else regardless, right? >> Buy their stocks, yeah? I don't know. >> [laughter] >> Yeah. You just buy their stocks. Yeah, yeah. Buying their stocks is a good idea. But that's I think I'm not sure we can share that. >> advice. It's just personal chat. Yeah, I
30:07 do think that there is a a lot of value in these companies. But I mean for for lots of others, right? I mean we see the sell-off across SaaS. We see the sell-off across cybersecurity. And then there is an ultimate question. If I want to like basically depend on someone else to figure out AI strategy or I want to embrace AI myself and benefit the most, right? I think that's a personal question for everyone. And as we all know that these technology revolutions, they're both fair and and unfair. I think in long term every technology revolution is fair and GDP per capita and quality of life goes up on the horizon of 10 years. On the horizon of like 3 years, I think it's extremely unfair cuz market just loves to pour
30:54 money into companies which are winning. And companies which are like not winning, they immediately go down. But in the short term, I think that it's going to be fair cuz people um both on the creative side, marketing side, engineering side, those who are embracing AI, they can propel their careers so quickly. In the end of the day, net it's going to be net positive, although I think um individually we all just need to think how I can personally benefit the most from AI, how I can become more efficient with the AI, how the value of my skill set can actually grow with the AI. And um this always require using those um agents and various AI models
31:42 several hours a day to to build this intuition. Let's give them a home task. What should they do right after watching this video? Which tool should they start using and how? >> For me personally, I'm an immigrant, so you know, it just takes quite a bit of effort to come with a logical linear storyline. All three mini became my coach, really. This was the first aha moment for me. The second aha moment was around Gemini 3 Pro model. So, I felt that my economic productivity really depends on like how much I use Gemini 3 model. These capabilities of the model which can process voice, which can make image, but which has also deep reasoning capabilities and deep research, this was mind-blowing to me.
32:29 So, that's why I feel that my just economic throughputs really depend on how much I use Gemini 3 today. Do you use it to make decisions about your company, like strategic operations? >> I really feel that communication with humans becomes way more important skill set for me. Cuz a lot of other decisions, I'm sure Gemini 3 and Claude are going to be better than myself. AI is not is not yet good in communication. So, that's where I think I try to put a lot of my personal emphasis. Mhm. Think other than that, um a lot of process can actually be built with AI. So, um just human-to-human communication and sort of
33:14 conflicts resolution and maybe goal set goal setting. Yeah. And like number-driven goal setting really is where I put a lot of my time, my personal time. For everything else, I'm trying to evolve Gemini as much as possible. But also use Claude for Excel, Claude for X, like for example, Claude for cybersecurity, as it really increases the productivity. Alex, thank you so much. It was amazing. I'm looking forward to reading your comments. What was your key takeaway and what you're going to do right after this video? Thank you so much. >> Thank you.
