He made $15K Flipping Pokémon Cards at 13. MarTalks interviews Michael Fanous of nyne.ai

Welcome to Martalks, the number one podcast for e-commerce and marketing applications. Martalks is devoted to covering the latest technology developments that drive the global commerce ecosystem, from advertising to last mile. Enjoy all of our content at rosensteingroup.com/martalks-podcast. Well, hey there, welcome to another episode of Martalks.

I'm your host, Darryl Rosenstein, and joining me today is a founder that's taken a very unorthodox path to his first startup, and he's nodding already in agreement. An individual that's been really at the heart of the LLM movement in Silicon Valley and got some access to some very interesting projects before he even got his first job. Degree. So without further ado, let me introduce you to our guest for today, Michael Fanous. Michael, welcome to the show.

Appreciate it. Yeah, I'll just give a quick gist of my background. I'm one of the co-founders of Nine.AI. Um, very—

Michael, I know that you're actually working on 2 hours of sleep, so, uh, congrats. Aren't you thrilled that you, uh, founded a startup? Isn't it just the bee's knees?

Yeah. Yeah. I mean, I, I definitely am. Glad I founded a startup. You know, it's always a privilege to have, you know, folks bet on you. And when you have customers who need things fast, uh, I think when you run the business and understand sort of the, the outcomes and how it affects their customers, you're obviously incentivized to try to help them out. So, yeah.

Well, it's good that you are on demand, as is your solution, Nine AI. We'll get into that in a second, but, you know, I think we should start off with a little bit about your kind of meteoric career because it's unusual. Tell me a little bit about your initial experience with university.

Yeah. Yeah. So love the question. Very unorthodox educational path. I dropped out of high school when I was 15. So already sort of deviating right there. Worked on an automotive media project for about a year called Speed Suspects. Millions of followers, worked with folks like Jake Paul, Alex Choi, a lot of very famous car folks as well as very large brands. Like, uh, Gintani. Learned a lot there, built some really cool technology and decided I want to learn more. Love maths. Cal was a great option based in California and had a great time.

Well, I'm kind of curious when you were, uh, you know, when, when you dropped out of, uh, a high school, what were you doing from a, a coding perspective at that point? Were you actually in a full-time job as a coder?

Yeah, it's a great question. So I started coding when I was around 12, learned a lot about technology. I love math. I'm actually not a big proponent of software and technology sort of upfront, competed in math competitions. And then one summer when I was 13, I made a very basic scraper that scraped eBay for Pokémon cards and made something like $15,000. And 13-year-old kid over the summer, that's an enormous amount of money. So I think that really drove me towards, Okay, you can apply your math knowledge to technology. So I did have a few years of coding under my belt at that time. And a funny story I always tell is like, I got my laptop taken away a few times in high school because I should be paying attention, but I wasn't. I was writing code in class. And I honestly think that was a big catalyst for wanting to, to move on.

So not quite risky business. You didn't have, you know, your folks' house where you grabbed a bunch of streetwalkers and put them to work, but you were earning money on the web trading Pokémon cards. Were you a Pokémon fan or was just like, this is an opportunity, I'm going to take it?

Yeah, so I was always a collector of things. I probably still am less so now, but for example, I remember when I was, you know, 6 or 7 years old, my dad would take me to coin conventions and he'd make me negotiate with the vendors. And I'm sure I learned a lot of like intelligent things about business through doing that. And that sort of led to ways to utilize that sort of outward worldly knowledge into technology where I can sort of teach robots to behave like me, negotiate like me, look for the best deals the same way I would. But yeah, I guess long story short, was a fan of Pokémon. I don't know if I still am at the same sort of degree, but it was definitely a great sort of jumping pad.

Well, what was it that led you to go back to school? I mean, why did you, why'd you bother? You were already minting CAC.

So I was doing well with Speed Suspects at the time, but like I said, I love math. Like I think math really drives me. I think like reasoning and logic is something I've always just really connected with as an idea. And when I wasn't in school, I kind of felt like, hey, I'm almost stagnant because I'm not learning these really high-level things. And some of my friends were, you know, off in college. They were sort of showing me their problem sets and they intrigued me. Like, as an example, a really cool project before this whole AI craze in one of my first electrical engineering courses was this idea of basically you have this audio, but the audio is ruined. It's a speech. and some guy's yelling on top of the speech, so you can't actually hear who's talking. And what we were sort of tasked with was deploying machine learning models that could erase the guy yelling on top of the speech and make this like a crystal clear speech from the president. And when we were able to sort of do that using like Hertz and looking at noise and all that sort of cool stuff, for me, that was like a really cool aha moment in terms of you can deploy technology to do things that I think the average person would be really surprised is even possible.

Well, I typically find that during somebody's university studies, they start, they get exposed to the ideas that really form the nasus for their first few startups. What was it that you did while you were in school that led you to want to create Nine?

Yeah, it's a great question. So it's a long story. I guess the first iteration of that was I had a friend who I won't name so he doesn't get in trouble, who was doing research at Cal. He had access to one of the foundation models from a very large tech giant. I got access to that. And as you guys can sort of imagine, this is 2022 era. So the model was pretty terrible, but I think I sort of saw a path in terms of where LLMs could potentially go. And my first thought was this idea of AI personalization, AI sales. If I'm speaking to Darryl, why isn't my messaging personalized to Darryl? Why is it not like these sort of just plain templates where it's like, hey, Darryl, I see you work at You know, Rosenstein Group, now come buy this product. It's like, okay, that's useless. Um, and what I very quickly realized was the bottleneck is actually context. You go look at all the other data companies in the space and you sort of see the fields they have. It's usually first name, last name, company, job title, maybe like an email and phone number, and that's really it. Whereas we wanted to take a different vector approach, which is basically we deploy agents across the entire internet, hundreds of millions of websites to learn as much about you as possible. That way, when we actually craft messaging, and sort of insert that context layer, it knows just a drastic amount of information and can actually bridge gaps, which let the agents actually draft good copy.

So I guess the thing that I still am blown away by when I call up a founder is using their mobile number is how did you get my number? Uh, like, well, there's, I don't know, 1,500 different sites that have that data. I picked one. I don't know what else to say. But you actually have developed AI agents that go out looking for non— well, obviously unscripted data, unstructured data, things that people say, things that people do, things that people publish, and then convert that into metadata that can be cataloged, that can be uploaded, that can be used to do a lot more in terms of generating intent.

I think you just made a great sort of understanding of what we do and sort of on top of it, something I would sort of highlight that's nuanced is exactly what you said. 1,500 different sites have your phone number, and I guarantee on almost all of them there's different iterations of like which phone number is right. And by having access to all 1,500 websites, you could essentially create this system of judges that are like— think of a coliseum of agents, and they're all sort of submitting votes in terms of what they think is the right answer. And the source of truth is far more granular because we have so many data points about every little thing versus a company that's just, you know, using one phone number they found somewhere from 10 years ago, let's say.

Well, when you founded Nine, what was the one signal that you couldn't find anywhere else that you were like, I am going to incorporate that signal in a Nine? Yeah.

Um, I think it was less so about one signal, but more so about, I didn't understand why the world was structured a certain way, which is the fact that, for example, right now I'm getting email outbound for like wedding rings, for baby showers, for things that have nothing to do with me. And I've always sort of wondered why is that still the case in 2025? I think Meta and Google have had a pretty good attack vector via ads, but for the 99.99% outside of that, I think we're still living in almost a prehistoric age in terms of how we're targeting people, how we're messaging people, how we're doing outreach, and AI is going to transform that completely.

Well, you know, what was the biggest no-brainer that, like, insight you had when you realized that autonomous AI agents could beat traditional data collection methods?

Yeah, so I think I'll sort of frame this for, for non-technical folks as well. There's this idea of what's called scraping. Like, what that means is you basically would send a sort of robot to a website, it would pull down some information from that website, and you would use that to sort of do something. Could be finding an email, a phone number, a name, um, etc. But the issue with scraping historically is if a button changed from, you know, this point to half an inch above it, or something moved around, all of a sudden your scraper's broken, right? So up until I'd say about 2020, scrapers were not very dynamic. Basically, you'd have to continually fix them over time, every time like a little thing would change in the site versus now post-AI, we can do things like put an LLM/intelligent model into the brain of a scraper. So if the site changes, it can take a screenshot of the page, it can pull down the CSS and HTML and heal itself. And by doing that, you can hit a scale that really very few folks can really hit. And I mean, I think you see good examples of that from Comet Perplexity, for example, the agentic browser. If you sort of see how the agents are able to navigate the web, it's, it's really impressive and it's going to allow productivity gains that no one's really seen before.

Well, as most of my listeners are in the B2B software business, and many of them are sales executives who are always looking for better data and insights on their target clients, most of which are going to be in the Global 2000. What do you think Nine does that would surprise them versus Zoom or Seamless?

Yeah, it's a great question. I guess the first thing I would rattle off is, I mean, no disparagement to those other companies, but I think they really utilize LinkedIn for a lot of heavy lifting. And the issue that folks are beginning to realize now, Nine is working with some health systems for certain projects, is that a lot of blue-collar workers don't actually live on LinkedIn. LinkedIn is heavily over-indexed on technology folks, people in finance, which, which is a good thing to have. Like, blue-collar America is the majority of America, really don't have a good platform that can actually sort of see all the people across all these different platforms and understand them, versus Nine, which we can see, hey, this person's in a forum talking about being a nurse in Texas working at this health system, and they have no LinkedIn account, and we can find out who they are. We know their email, we know their phone number, we know how to contact, we know where they're based. And that's something that's really been a big value add for like large health systems that have been trying to solve this for many, many years.

So there's actually people in the world that aren't on LinkedIn. Is that what you're saying?

That's what I'm saying.

Wow. You wouldn't know that from, from, you know, them. They don't share that very openly. That's a statistic that I was not aware of. So you can actually target people based upon what they're doing, what they're posting, what they're saying online, anywhere, anyhow. I mean, is there any rock that I can still hide under electronically and not be found by Nine?

Yeah, so it's actually a really good question because I'm actually a big privacy advocate myself, and something I would sort of recommend is there's this idea of expectation of privacy that's really a pillar for a lot of the privacy laws in the US, and the premise of that basically is if a user makes an intentional effort to be private, no data aggregator or collector should be collecting their information, right? So like for myself, all my social media accounts besides X and LinkedIn are private accounts, right? So I'm intentionally setting a barrier saying, hey, like, don't collect data about me. But if, if you're sort of publicly facing and intentionally making public accounts, then really any LLM can aggregate data from that, right? So you have no expectation of privacy if you have a public account. So what I would say is if privacy is like a big thing that matters to you, like it is for me, I would be intentional about setting sort of privacy restrictions based on the platform I'm posting on.

So what if you made some embarrassing tap dancing videos when you were in your 20s that you really don't want anyone to see, but Somehow made it into the public domain. Are you screwed?

It matters if you own the account. If you own the account, I'd recommend deleting that video as soon as you can. But if you're, you know, your friend named Alex decided to take a crack at making fun of you, there's really not much you can do besides telling Alex to please take the video down. But I think a lot of platforms are getting better at sort of monitoring, hey, this person has content of me that I really don't want up, and they'll sort of work with you to take it down. Like, I think YouTube's been doing that very, very well. Had a different company we were working with and a sort of quantitative trading firm which bet against that company, tried to publish some information that wasn't actually true to increase their sort of margin based on their position. And YouTube was actually fairly proactive in like sort of looking at the proof points that it wasn't true and seeing that they had a financial incentive to take it down.

Can't believe that a company would try to do something so scurrilous, so wrong. Well, I mean, being able to keep track of that is another thing that those Agentech agents can do, you know, are there, are there any kind of no-go areas for these Agentic AI agents that are gathering this data?

It's a really great question. So something that I like describing to some of our investors is that there's a lot of complexities and barriers for future entrants. What that could look like, for example, is a lot of our sort of compute is based on Linux servers, where we have agents that are deployed from these servers to do certain things. And as a big part of that, there's what's called IAM permissions, which are sort of well-known technology. And basically this idea of IAM permissioning is basically like, hey, like this API has access to these pieces of data. It can only do these things and only these users can sort of engage with this API, right? And the same thing applies to agents too. So for example, I have a server in Amsterdam and I have a server in London and London is UK GDPR and Amsterdam is General GDPR, which are two separate rules, which sounds weird, but that's, that's the truth after the UK left Brexit, uh, or via Brexit, I should say. But yeah, like there's different parameters we're allowed to collect. There's different cadences of what we can collect. There's different ways we're allowed to store data. There's differences in terms of what we can store. So yeah, like I would definitely say that there's definitely no-go areas. There's reductions in depth and granularity and cadence that we can sort of crawl and collect on. Uh, and some regions have basically purely anonymized data and you would have to basically just run ads against who they are.

Well, building a graph-level context across 250 million sites is pretty massive. What's been the most surprising pattern or signal that you've seen so far?

It's a good question. Honestly, less of a pattern, but just a plethora of possible use cases. Like I can give a couple examples, one of which is a like foundation AI company is utilizing this data to find people to create training data for humanoid robots. So like the way I would describe that is like, imagine a person who's wearing a camera on their head a certain number of days per week. And the requirements include they must live at home, they must work remotely, they must do something creative with their hands, such as painting, coloring, woodworking, et cetera. They can't have kids or like under 18s in a video. So it's like with that sort of criteria, how would you find them? You can't go on LinkedIn to sort of find these people. But with Nine, we have all these sort of aggregate signals that could sort of look at property records and see that they're living in a studio. And sort of determine amongst all the criteria, see, hey, he's remote, he runs this sort of creative studio from home, this would be like a perfect fit. It matches all the other pieces of that. But then there's other things like AI social platforms that are using us as context. So for example, right now, the way, you know, social platforms work outside of Meta and really just Meta is like a user has to log in, they have to sort of insert a ton of information about themselves, which is a huge uphill battle for these social platforms. Or they can come to Nine and Nine already knows Michael loves boxing, he loves weightlifting, he's, you know, big into technology and codes and Python and Perl and Go. And Michael knows these people already, so if they're already on the platform, they should probably recommend that Michael knows them. And it's just this huge amount of context that sort of decreases the friction and allows users to get immediate value from signing up to platforms without them having to, for example, prompt back and forth for an hour.

Well, let's say I want to get to know the CEO of Blue Origin. Because I happen to like things that fly and go up in the air and some come down, some don't. What would the advantage of utilizing Nine give me?

Yeah, it's a great question. So we would know, you know, how to contact him. We would even know his favorite sports team. We would know a lot about him. And the way I sort of describe it is this idea of instantaneous serendipity. When you know what people like and you know how they think and you know what they care about and you know their business objectives, For example, Nine is tracking like 8-K and 10-K filings from public companies. So if a company like Uber has regulatory hurdles that are, let's say, tied to SOC 2 compliance, then a great SOC 2 company like Delve can go and say, hey Uber, I saw, you know, X, Y, and Z. Like there's clearly a strategic initiative versus just doing guesswork and saying, hopefully this matters to him. Maybe it does, maybe it doesn't. So what I would say is it gives you sort of like a weapon to go into battle with in terms of like, you know about him. You know how he thinks, you know what he likes, but then you also know from a business perspective what specific things you should harp on.

So you're, you're collecting data that is unstructured there and, and spread from here to yon about what somebody is doing. How do you, for instance, figure out what their, what their hobbies are?

Yeah, I mean, it's really anything they're doing publicly facing. So for example, if there's a YouTube podcast of them, If you, Daryl, looked up, um, like Dave, who heads Blue Origin as a CEO, like we would basically deploy an agent to sort of source all the content on the web in addition to what we already have in our DB. And then it would, for example, watch an hour-long YouTube video to very deeply understand what he cares about, the jokes he cracked, um, how he thinks. And having all that context is really useful. For example, if the CEO believes that you saw a part of this podcast that is 45 minutes in, he's gonna be more inclined to wanna work with you because he's like, hey, Darrell put in effort. Like he cares about what I think. He did his research, did his homework.

Yeah, I'm not good at homework. It's never been my strong suit. So I think having 9 would probably be a good thing for me so it can kind of expedite that. But you know, the CEO of Blue Origin is Jeff Bezos and yeah, I'd love to get a ride on the rocket on New Shepard whenever he's available. He's got an extra spot. I'm just going to throw that out there for the hell of it. You know, quick, just an aside here, you know, as a founder, would you rather hire another machine learning genius or like a street-smart recruiter for your next 3 hires?

It's a great question. I think right now we're seeing this big shift. Like, I'll be completely transparent with you. I think we're a little bit early in this LLM resolution. I think a lot of folks are thinking, you know, the market's saturated and they're sort of worried. I think the exact opposite. I, genuinely believe these LLMs are not at the capability to be completely autonomous yet. Like, Nine has a lot of guardrails and things we've learned from trial and error to optimize and orchestrate these agents. But I think like a big trend we're seeing, especially in the Bay, is this premise of forward deployed engineers. So how do you take this sort of base set of information, this base product that you have, and you sort of add that last 10 to 15% to map it to the customer and make it actually useful for them?

Yeah, the lack of forward deployed engineers is just a huge hassle for most of the LLMs because without that last 10 to 15%, which is the training on your specific business, on your specific problems, on your specific workflows, it really doesn't add value. It has all of this potential. It's like that employee that just has all this potential, but they never really amount to much. So what, you know, what, what do you think the solution is? How do we, how do we overcome that, that paucity of talent in the forward deployed engineer arena?

It's a great question. Uh, I think the first part is obviously identifying the best talent possible. In San Francisco particularly, like we're very bottlenecked by, by great talent. I think that's like very apparent right now. And in fact, obviously we'd love to, you know, bring on some great forward deployed engineers. Um, that said, I, I think the, way we do this is by aggregating the talent, by getting them excited about the product they're working on, um, and also just showing them the input impact they're actually creating. An important statistic I actually like mentioning to folks who are very bullish on AI, and I am as well, but more from like a longer tail horizon, if you will, is that 95% of AI pilots are failing today. And the reason for that is because startups aren't meeting corporations where they are, right? So like corporations, I think a lot of times even now have legacy workflows, for example. And startups, they're basically just giving them a platform saying, you sign in and figure it out. When in reality, that's not the way to work with these large enterprises. What you should be doing is saying, hey, you know, this Fortune 500 company is already using Salesforce. They have, you know, data about their sort of consumers living in here, but they need this extra information. And if they had it, that'd be great. And instead of 9-forcing them to log into, you know, a brand new platform, figure it out themselves, figure out how to use it. Obviously we have some great engineers where we can say, hey, you know what? You guys already know how to use Salesforce. It's easy for you. We can just click a button and give you sort of the additional information.

I guess the sticky question here is how do you balance, you know, privacy, ethics, and power when you're building systems that map people?

So, yeah, it's a great question. It's something that, you know, I have to think about every night before I go to bed. You know, I think we've been very intentional about who we partnered with recently. So like our average customer is worth over a billion dollars. They're like very, very large enterprise companies and Typically the use case looks like, hey, we want to target people better, or hey, we do really well with couples in Texas that have over $100,000 of expendable income who really care about staying fit, for example. Right? So an example I sort of like denoting is we're talking to a bank right now that's releasing a new card that have benefits tied to, you know, like fitness and Pilates and saunas and cold plunges. But they have to front a big bill up front to sort of like actually enable those benefits for the users. So the only way that sort of card company is going to get lift is if they're able to target people who care enough about the benefits, one, and two, make enough money to actually like drive enough spend on that card to offset the benefit cost, right? And right now the only way to do that is you advertise, which is essentially a black box, and you hope the conversion rate is high enough to like find the people you need. Versus with Nine, we show you proof. Hey, Adam works at Apple as a PM. He's probably making somewhere in the realms of $250,000 a year. And you know, from it, you know, his social media accounts, we know that he loves weightlifting and running and, He just, you know, made this podcast with his friends where they're talking about how they're all trying to bench 400 pounds first, right? So unifying all that information about him is able to qualify him in a unique way that really no one else can. But it's like you said, we want to be intentional about our partners and how people use us. And that's why, like, the use cases are typically around, like, just buying products, figuring if someone needs something else. And also, like, from a work marketplace standpoint, like I mentioned earlier, all these blue collar workers We're finding them jobs that didn't exist before. So for example, like the state of Texas has a database of unemployed people looking for work. And what Nine's able to do is like, number one, find those people, but then also say, hey, this, you know, oil refinery is really trying to hire oil rig workers. And this person in Texas who has experience as an oil rig worker can't find a job. And maybe they're just not like up to date with technology. Maybe they're not looking online, but we bridge— like us bridging that gap creates a lot of value. And it sort of reduces the lift required from job applicants when we can just find them jobs.

Wow. That would be fabulous. That would be fabulous. Feeding all of the open jobs into Nine and then have Nine spit out, well, here are your people. How do you want to reach them? You know, as I mentioned, most of my clients are startups and they are in the martech and supply chain universe. And they're typically targeting, as I said, Global 2000, upper mid-market. Where they're trying to— it's very important to have a point of view when you're engaging with a company. And as you mentioned earlier, reading those 10-Ks, those 10-Qs, those financial filings really gives you a great deal more insight into what challenges a company is facing right now. Is there any, I would say, comparable methodologies or solutions to gather that data? Other than utilizing Nine to scrape through those financials?

Yeah, it's a great question. I think some good examples might be Perplexity, for example, where it's like this AI search engine, which is really great. The one sort of caveat I would say with Perplexity is it's limited to sites it can see, right? And my understanding of how Perplexity works is it's going to leverage the first 10 or so results from Google and sort of make a conclusion based off of that. But unlike Nine, on most of these sites, they're not actually going to the site. They're just reading the snippet that's available on Google. So by doing that, you get some good information, but for example, like these 10-K filings can be up to, you know, 500 pages, like a giant PDF. And most LLMs aren't sort of optimized or fine-tuned like we talked about to sort of find and aggregate and actually make conclusions based off of large corpuses of information.

What do your clients typically specify in terms of the data that they're looking to gather and how much customization is involved in piping that data back to their, I would imagine CRM system.

Yeah, it can be a CRM. It could even be products. Like I think we've been doing a lot of really cool product partnerships recently that get, you know, my, myself personally sort of excited about it. I, I'd say the two things are this, like we have customers who are very broad. They're just like, hey, I want some nurses in, in Texas, for example. And it's like, okay, we can do that. But then like I mentioned earlier, we have, you know, pretty crazy requests in terms of, you know, training humanoid robots and things, for example, like dating apps to find people within 5-mile radiuses who love biking and work in fashion and all sorts of really crazy requests. So I think it's interesting to see the very large set of people who are all getting value from Nine in just completely different ways because it gets me excited. And I think the future is really going to enable a lot more, like something I just spoke about to, to someone else this morning is that like as AI gets better, AI will become increasingly participatory in human decisions. This can only occur once it gains enough context about the person it's serving, which will then unlock sort of new values. For example, what I like saying is, Daryl, let's say you have like a physical therapy appointment on Thursday. If AI knew that and you got an email from like a physical therapist, sort of like confirming your appointment or saying you have an appointment, it's going to be able to route that and understand that much better than if it knew nothing about you and you just got a random email from a physical therapist. Right? So the ability to have context about a person unlocks a lot of productivity gains and just abilities in general for these models because it actually understands who you are.

Let's talk a little bit about partners and partnerships. What, uh, what companies or product types within the martech and e-commerce universe would be good partners for you?

Yeah, so it's really a wide array. I think the first one that's really interesting is like very large D2C brands, for example. So let's say you wanted to find, you know, pregnant parents as, as just an example. The time delta between someone getting pregnant and having a kid is so short where really they can run a bunch of ads, but like some of our partners have done that and very quickly realized like, hey, we spent a ton of money. We're not really seeing lifts and they can't get, you know, sort of approved spend from the C-suite to just continue lighting my money on fire, hoping for a different result. Right. So with Nine, because we have, you know, access to the web, we can deploy these agents across the internet, figure out who's actually, you know, pregnant, having a baby shower, creating sort of a baby registry in some way, and map to, you know, who they are, where they live, where they work, and all this is really relevant. So then, you know, our partners can actually engage with these users. But what I would say is it enables a new way of targeting and segmentation. For example, another thing is like, hey, we do really well with people like this, and we need to find more of these people. Like lookalike audiences and targeting has been something for going on for a while, but the granularity is honestly fairly bad. And the ability of Nine is to go to such a granular point where it almost becomes a no-brainer once they see the proof of each sort of piece of criteria.

What about technology players, technology categories? I know that CDPs are obviously in the business of collecting and managing the customer profiles of their customers. How would they collaborate with you and have you worked with them in the past?

Yeah, yeah, we have a couple of very large partners in the CDP space as well as like AI sales, AI go-to-market, AI marketing. I think a lot of folks are trying to create intelligent workflows that derive value from context and from data, right? And almost any AI sales, AI marketing company needs some semblance of people data. So they're actively sort of buying and looking around and sort of they realize there's, there's a lot of limits. And it's the same realization I came to, which is partially why I built Nine, is that a lot of the companies that are selling data are just purely selling LinkedIn data, which is really not a good utility for what they're trying to do versus with Nine, it's like you create this serendipity to understand the person, but then you also understand the business. It's sort of this two-pronged approach. You understand who they are, if they're the decision maker, what they care about in their job right now. Like let's say on the website, there's a new product initiative to add more workflows related to compliance through a company, right? Like we would know that, we would know who's in charge of that. We would know that that person is a big fan of the Golden State Warriors, for example. And then if you sort of sold some sort of compliance workflows, you could say, Hey Brian, I'd love to take you out to the Golden State Warriors game, and I know this is top of mind because X, Y, and Z, and I think we can help you. And it makes a very clear value prop versus just guessing and doing, I mean, what we both sort of know is spray and pray.

So in terms of other technical partnerships that you might have evaluated, have you worked with any of the players within the e-commerce platform space or within the personalization space or within content marketing or within mobile market?

Yeah, personalization's obviously been like a clear really big play. A lot of these companies are basically creating UIs against the same set of data, right? So it's really hard to get some semblance of like an alpha or like some differentiating factor. Like something I really enjoy doing whenever I test a new AI sales or go-to-market platform is trying to figure out like what data do they have, how would they potentially get it, and how is it different from all the other platforms? And there are a few that I actually think are doing some really cool things. Like Unify is one of them in terms of like warm signals and leveraging data the right way. But I think also think there are a lot of companies that are leveraging like very static signals. So an example of that is there's a company that's basically sourcing folks who may have immigrated recently to another very large company. And the way they're doing that is basically just looking at like last job title on LinkedIn, seeing the location of that, and then looking at the new job title and say, Daryl's last job was in France and now his newest job's in San Francisco, so he probably needs immigration sort of help. Right, which can be useful, but it's, you're getting a lot of noise in that because you have really nothing to sort of confirm or deny. Like, for example, my roommate from Cal, he's going to do a job in London and he's not going to really need help with immigration. So you'd get a lot of sort of false matches. So I think something that's really important for Nine is we really want to get granular about qualifying every little piece of criteria. So you're not guessing, you're not wasting time talking to people who really have no utility for your product or service.

So when you're doing, when a client engages you to, for your dataset, how many data points does the average adult male have by the time they're 30 in the United States?

Yeah, I would say somewhere in the realm of 50 to 60 data points for someone of that age, which is really, really great. Our biggest competitor probably is People Data Labs. So I won't give them a huge shout out. Like they list their own sort of statistics on their website. They list over 250 fields, but when you actually looked at statistics, on average, it's only 6 to 7 fields. So you, they actually denote the fill rate of each field and you very quickly realize that over 50% of them have a fill rate below 30%. So that's less than 1 in 3 people. So I think a lot of folks get excited about number of fields, but they don't actually realize that these platforms aren't actually actively collecting new information from these fields, which is when you get stale data.

Yeah, I've, I've certainly had my, my fill of stale data from LinkedIn itself and from, you know, my ATS that's supposedly connected to LinkedIn, but it is not, and all sorts of other solutions. So I guess we can segue a little bit here. Let's, let's talk a little bit about your experience as the founder of a startup. How many people do you have with you now at Nine?

Yeah, full-time we have 6 right now. I have an offer out too, so hopefully we get number 7. By the end of this week. We'll sort of see how that goes. Yeah, fingers crossed. But yeah, we should be expanding pretty quick. I should have a really awesome announcement that would sort of lead to a big headcount jump pretty soon that I'm excited about. So look out for that and stay tuned.

Outstanding. Are you in the midst of a raise or is that, are you sitting still for the sitting pretty for the point?

It's a good question. What I would say is we have inbound interest to sort of accelerate the speed of the company's growth. We're not actively fundraising, but we have been approached by some great institutions that are very well capitalized and they're big fans of Nine. Cool.

What's been your favorite aspect of being the CEO of a startup?

Yeah, I mean, this is kind of like a brutal fact, but it's something I really enjoy, which is I sort of control my destiny, right? Sort of for better or for worse, I'm the driver of Nine. I have to be the one to live with the decisions I make and as a result of that, I think that's how you really get some of your best work. When you know that, you know, like I talked about last night and getting, you know, a couple hours of sleep, that if the client doesn't get this done, there's some downstream effects and they're going to look bad, then incentivize to, you know, do your best work, get it done, get it done quickly. So what I would say is it's definitely instilled this level of agency that I haven't seen for myself prior. And it's very exciting. It's like for all the wins, you feel so much more, all the losses, you feel a lot more, but What I would say is it brings out the best in you.

Yeah, it's not a job where you can kind of sit back, kick it back, kick back and watch things happen. Well, if there was anything you could change about your career path, as bizarre and twisting as it has been, what would it have been? What would you, would you have given any advice to a younger Michael Fanous?

Yeah, I guess advice to a younger Michael. I'm someone who is constantly wanting the best outcome at every single path. And I think I sort of realized that honestly, some of the most growth and the sort of best endpoints to my trajectory now is when you have to navigate uncertainty. And I think something as a founder that I would tell younger me is like, I've always been someone who wants to see the full picture before making a decision, right? And as a founder, you very quickly realize that's never the case. So I just explained this to actually a high schooler yesterday, which is this: to be a founder, you really don't need much. You just need a thesis about a situation. So an example I like giving is if you believe a helicopter had 5 propellers, it would be 50% faster. Like we both know that's probably not true, but you'd be able to sort of test this example, collect feedback as an experiment, and very quickly realize if you had a like sort of correct thesis or not. And that's really all that's required. What I would say is if you're hardworking, you could be wrong 50 times on your thesis. And if you're working in the same problem space, eventually you're going to find some unique alpha or something new that someone else hasn't thought of, and you can very quickly scale that up and monetize it in ways that you really couldn't do in any other sort of life path or occupation.

Well, there's an endorsement for being a startup CEO if I ever heard one.

Yeah.

You know, you can screw up with other people's money and eventually figure out something useful. Yep.

Yep. And they're happy to do it too. Like, I think something that's really cool is like, find me another place in the world. I guess this is more about San Francisco in general, but You know, you get some 22-year-old, you decide to bet millions of dollars on him, and the people betting these millions of dollars are people who are, you know, very successful. They have like a great filter for identifying potentially who is next or who's going to be able to sort of replicate some of what they did. And it's very inspiring to know that, hey, you know, this person's a billionaire and they believe in me to do something similar to what they did, you know, and they were willing to even put their money up to sort of suggest that they feel that way, right? So I think that is a really sort of interesting take and we're in a unique time and I would really say if it interests you, And you're willing to just, you know, give it your all. I mean, big endorsement. That's really all I can say.

Big endorsement. What's your, what's your favorite part about being in San Francisco?

Yeah, it's an interesting question. This will be an unpopular opinion. I grew up in Los Angeles. I love Los Angeles. I know some folks haven't liked it as much recently, but I actually like LA a lot more than San Francisco. Just being completely transparent. I'm in San Francisco because I think it's the best place to build a company today. Right. And that's what I'm mainly optimizing for. Day to day, maybe I guess some more real feedback about being a founder is you should only do it if you're willing to like go all out, right? So, um, my day to day right now is like 15, 16-hour days of pure grind, right? So I guess that is a more cautionary tale in terms of the commitment that you're going to have to put out to sort of build a business from the ground up. But at the same time, I think San Francisco has created this great ecosystem where you have very smart folks building technology. I'd say the most talent-dense place on the planet. You have, you know, the financial infrastructure and the institutions that are willing to, you know, bet on these 20-something-year-olds and sometimes even teenagers to, you know, build something tremendous. And we're already seeing results. I think folks forget, like, this idea of ChatGPT was like nonexistent 3 years ago. And to sort of move from there to today, like, we've been taking quantum leaps in terms of the deviations and the abilities of AI.

Well, if you look at the state of the industry today, you have the early leader of ChatGPT, which is now being called essentially the MySpace, and you have the 800-pound gorilla is alive and awake with Gemini, and Anthropic is certainly making some real headway with developers. Where do you think those, those LLMs are, and who do you think is going to win, and does it matter?

Great questions. Yeah, so I'll give you some unique insights because I have friends at some of these companies, and this isn't proprietary information. This is all public, so I can definitely say it. So Google is in this interesting, just sort of space that people don't realize, which is they have their own TPUs, right? So unlike OpenAI, that's reliant on folks like NVIDIA, Google's cost to actually train models is far cheaper. And because they actually own the manufacturing of TPUs, they can sort of scale it up themselves and not be reliant on a third party where supply is limited and constrained. You have all these other companies that want it. Um, so I think Gemini very quickly just had this huge leap in its ability with Gemini 3. And, um, I think it'll make headway because honestly, Google, from my understanding, actually has pretty decent margins because their own TPUs are far cheaper than renting directly from NVIDIA. Um, and they can eventually choose to pass that along to consumers, and if they they do, it would really hurt Anthropic and OpenAI where their cost basis is just so much higher because they're reliant on this third party and they're competing with each other to get the finite supply. So that's the Google side of things. I am a big proponent. I think Google will be able to do a lot of cool stuff because they own so much of the infrastructure related to the internet and so much data. On the other side of things, I think OpenAI will make a comeback. Code red was declared earlier this week at OpenAI, and the rumor is they're releasing a new model this week. That's known as Onion or Garlic, also known as 5.2. So I think it'll be interesting to see when that releases. From Anthropic's side of things, they're taking a different approach that I think folks don't realize, which is they want to be the sort of safe, compliant enterprise/government iteration of AI where there's a lot more guardrails. It's much more guarded, and the intention is to create an experience where it's more family-friendly, it's safer, it's going to be less likely to behave in a way that can be deemed like, I don't want to even say malicious, but maybe edgy is a better word. Right? So I think they all three have their sort of own angle they're aiming for and differentiation. And like you mentioned, Quad Code in and of itself, it got to, I think, a billion ARR in something like 4 or 5 months, which is ludicrous in and of itself.

What do you think of Groq?

It's a good question. I think there's going to be some great utility and use cases for Groq. So an example I would say is that Groq not being constrained on training data has pros and cons. But the pro of that is it's going to be willing to sort of do stuff and analyze stuff in an interesting way. Like, as an example, with Nine, we'll, you know, pass in some context, some of these models, and unless it's like very fine-tuned for our use case, it'll basically barf it out and say, hey, like, I don't like touching like data in some capacity, right? Whereas with Groq, you're definitely a lot less limited in terms of your abilities, which can unlock new novel use cases. And not all of them are going to be deemed bad. I think obviously you can sort of enable some stuff that is not ideal, but the interesting thing with Groq is that you're not constrained to whatever the foundation model company has decided to do.

Well, I think we've covered a lot of ground today, Michael. A lot of things beyond just your background and founding Nine and, uh, you're making investments yourself now as, uh, uh, are you an angel or are you a VC or what do, what are you? You're a multi-talented dude. What do you, what do you want to be called?

Yeah, I mean, I think honestly just the co-founder of Nine is an honor of itself. Like, that's, I think, what's going to be my life's work and what I hope a lot of people sort of look back on and say, that was a really cool thing that Michael built. But yeah, in addition to that, I do scout for a couple venture firms. I'm very connected to the space in San Francisco, so folks are welcome to reach out. I'm happy to sort of make connections, introductions. And the way I would sort of describe it is like, Darrell, what you just sort of do with staffing, my goal is to sort of just connect people, connect the right people to each other. I mean, you obviously do a great job with that. We sort of got introduced through, you know, a great mutual friend who I really value a lot and obviously just instantly gave you credibility. And it's the same thing with founders. It's like there's a lot of great folks who may not have the most traditional background or have all the industry connections. And I think it's a privilege and obligation that I'm able to, you know, make that introduction on their behalf.

I'm going to support that 1000%. I think that too often in the world, you know, tech The tech community is still pretty ageist. It's unfortunate because there's plenty of folks that did great things that are capable of doing them again. But I think more importantly, the startup world is rapid fire. And for every, you know, 9 that makes it, there's 15 companies that bite the dust and people that weren't— learn great things, but their resumes begin to look a little crappy after a couple of these startups where you have people that just want to reach for that brass ring. They just want that autonomy. They just want to be in that novel environment. And it's people making connections that look beyond that one-dimensional piece of paper or LinkedIn profile that, you know, breathe new life into these people's careers. So I'd certainly encourage you to continue to do that and get to know people and be a connector.

I dig it.

It's been great. So with that, I think we've had a wonderful conversation today, Michael. Thank you so much for your time. And for joining us on MarTalks. Any closing thoughts before we, we end things today?

Yeah, thanks, Daryl, for hosting. I think we had a great conversation. I hope the viewers can learn something. And yeah, feel free to check out Nine. Feel free to reach out to me. Happy to answer questions. And how do we reach you? Yeah, how do we, uh, just look up Michael Fanous and should be the first person that, that pops up.

Very cool. We're gonna do that. All right. Thanks for joining us at MarTalks, people. Hope you had yourself a great And a great listen.

Thanks for listening to Martalks, the number one podcast for e-commerce and marketing applications. Be sure to subscribe wherever you listen to podcasts, and while you're at it, leave a rating and review. To find out more about how the Rosenstein Group can help you find the right leaders for your client development teams in martech and e-commerce, Please visit our website at Rosensteingroup.com.