Julius AI hits 2M users with zero VC hype: SEO, HBS adoption, and the case for app-layer AI

product, you need to analyze that data, AI can help you with that. And Julius is the tool for that. Very cool. Uh dude, I'm just going to start. You have 2 million users. How do I How'd that happen? Yeah. What's the secret? Yeah. A lot of it is word of mouth. Um and some of it is SEO.

So if you look up AI data analysis, Julius is the first result on the internet um on all search engines. Basically, turns out like analysis is a fundamental part of many jobs and people just kind of assume that it's a solve problem. Turns out most people don't know how to analyze data.

Um and and Julius helps them with that. So, I think we kind of found that that need um and used AI to solve that problem for people. That's insane. Um where uh where should we even start with this? Uh I mean, I want to talk about just like the culture of uh of this this whole meme of like a rapper.

I think it's really I' I've said this on the show a bunch, but I think it's kind of like a sigh out from VCs to like not like for so long it was like, oh, like you're just going to get steamrolled steamrololled. Like it's not it's not a viable place to build a monopoly business, a trillion dollar company.

and that's all you should be focused on. Go for broke. Now we're seeing, oh, the value actually occurs at the application layer. Oh, maybe there are tools that need to be used. Um, but at the same time, like if somebody said, oh, well, like GPT7 might be able to do the same thing as Julius.

Like I might hear them out on that argument. So, how do you process it? And can you take me through your thesis on just companies that leverage foundation models instead of train them themselves? Absolutely. By the way, shout out to Sam Hoffman for for that steamroll sound. think it really stuck a chord with the VCs.

But um look, turns out like you know every business out there in the world is solving problems for people, right? And as long as you're building something that solves a problem for people, like that's all that matters. Your users aren't paying for a model, what they're paying for is this tool solving a problem for them.

If you can do that faster, easier, cheaper, that's all that matters to your users.

Um now uh in terms of like models and like where does value occurr early stage founders just get too caught up into that meta level thinking instead of just building something like just go build something talk to people um see if you can sell to one user can you sell to 100 users and then worry about these other things.

Um what how the way it has played out is when we started people told us you're a rapper um and we would try to like fight that label um and then over time we just learn to embrace it. Um turns out like these models they keep getting better.

There's multiple model labs AI labs that are producing really good models and what that means is the differentiation is actually in the product layer. Can you create a differentiated product and an experience and a whole infrastructure to power that? Um now um is is that how it's going to play out forever? Depends.

Turns out like your your product needs to have enough of a unique value proposition so that it matters outside of the models themselves. Like for example, we have this infrastructure code execution. Every 24 hours, by the way, every 24 hours, Julius writes and executes over two million lines of code. Mh.

Um, and to serve that to our users. Um, we serve like over a million Jupyter notebooks. Most of our users don't even know what a Jupyter notebook is, right? You're a VP of marketing at a at a big company. You don't know all that technical details. All you care about is analyzing your data.

And then how do we build that infrastructure to power that is really important. Uh, and because we have built that models getting better actually helps us build a better smarter product. Uh, with all the infer and all the UI and the the interface that we have built. Yeah.

Uh how do you uh how do you think about companies in the space doing anything related to AI data analysis? There's companies, you know, coming out uh some of them very vertical where it's like this is just for investment bankers. Yeah. Yeah. Stuff like that.

But but the the thing that comes up for me is like you clearly have chased hype from users versus hype from VCs, right? because some of these companies that have come come out um they maybe have like as the same amount of users as they have raised millions of dollars, right?

So it's like you you know other other companies in the space like with like 300 users maybe at some cool firms or whatever, you know, raising hundreds of millions of dollars.

But I'm I'm curious even understanding like adoption at how your tool gets adopted at at you know big companies that might have some crazy procurement process but then you just have like they're just sign going over to Julius and being like yeah actually like almost positioning the work that you're doing in Julius as kind of their own and they're just like pasting in a screenshot to some slide deck or yeah I'm curious how you think about adoption and and maybe like have you intentionally avoided hype throughout this entire process.

Um because like clearly I would say like Yeah. Yeah. Anyways, just just take it from there. Absolutely. So um I think you're spot on. You know that's how users discover Julius. You know they're looking for they're working a job. They need help in the moment to analyze data.

They're looking something up on the internet and they come across Julius. Um and once they start using it slowly what happens is they're able to first of all bypass all the procurement process and just get the value right away but then analysis and insights are inherently sharable.

So they start bringing their teammates you know onto Julius especially if it's a manager or an executive that discovers Julius they bring the whole team with them. Um this is an early shout out but we're we're about to launch real time collaboration in Julius.

So you can collaborate on analysis with your teammates in in the product. Um and and that's going to be huge uh for collaboration in general. How do you how do you think about data analysis?

you you guys are the AI data analysis company but how many of your decisions are based on vibes vibes a lot of the so the for the channels that are established for example SEO growth all that we use Julius pretty heavily and it's also very datadriven for a lot of the exploration and that's where we bring in a lot of qualitative data like talking to five users that are superpower users and observing how they're using the product what would they need um and how do simp create a simple version of what they need so that many more users can actually access it.

I have a couple questions on the actual platform. Uh what models are working the best for you and what languages are most popular? I know that Python's obviously super popular but then a lot of data analysis use R and Julia even which is kind of a funny name.

Uh yeah, are you kind of language independent and then you just service the results and then how much are you swapping models under based on like what's hot on Twitter today or X because you hear about a new model every day but then you know is it actually worth switching? Yeah, great question.

So for the models, you know, it this changes quite often, but you know, if you're watching this a month from now, it might be a different answer, but you know, usually it's just OpenAI and Enthropic doing a lot of the heavy lifting on the model front. Uh both both labs have really good models.

Uh we do work with other model providers as well in different parts of the stack. Uh but usually it's OpenAI and Enthropic doing just a killer job there. Um, in terms of languages, turns out like a lot of users don't know what Python is.

You know, for them it's just like something that the AI is doing to do the analysis for them and they like seeing that, but whether they understand it is is like not that common among our users. Python is pretty popular as a default because it's just so versatile, right?

It's it's like the most popular language on GitHub. It's got the biggest ecosystem of modules and all that. So, it gives you a lot out of out of the box. Um there's there's also um R that we support and we might add support for SQL at some point in time. Sure, makes sense.

Do you think the power of giving anyone uh in the world access to this sort of complex analysis is under just like priced in yet?

Like because to me because to me I think about okay every Shopify store owner can now get the power of like this sort of like worldclass CFO for basically nothing right they have their Julius subscription but it's not super consequential that to me doesn't feel priced in because you can enable millions of businesses to run more efficiently and but I'm curious what you think.

I think the market hasn't priced this in yet. In fact, the cost of getting an insight from the d from your data has gone down 99% in the last two years. And I think most businesses, whether small or big, haven't really like realized that value yet.

You know, if you're a big company, you have to realize that every marketer on your team can now go do really deep marketing analysis, channel analysis within seconds using good product like Julius. Every finance person on your team can get financial insights within seconds or minutes using Julius.

And these businesses when we talk to them um most of them haven't priced this in yet. Um so I think we're going to see like a big reckoning in the next few years.

How do you uh advise your team around hiring and how do you think about hiring when you know how powerful these tools are and uh you know I'm assuming you really push people to say like you know do we need to make this higher or you know I'm I'm uh curious what you think hiring internally at Julius or just at his companies?

No, no, no. Just internally at Julius, like you know, today the Shopify CEO came out and with a leaked memo where he was saying like, "Hey, we're we need to be AI native.

You need to question every single hiring decision uh that that you want to make uh and compare it to just like getting that same service or value from models, but it's different at a high growth startup potentially. " I think you guys had Keith on earlier today, right? Yeah. He has this like memo about barrel and ammo.

Like you you don't want to hire people that are ammo. You want to hire people that are barrels. Like I think what he's looking what he's basically hinting at is you want to find people that are multipliers. And so we look for people that are multipliers.

You give them resources and they find a way to 10x the value creation out of that out of those resources. And AI is a resource, right? a a really competent barrel uh employee, a really competent multiplier employee can take all the power of AI and just make a lot more happen in a shorter amount of time.

And I think that's what we care about when we recruit at Julius.

So whether it's engineering like what is their learning rate not not how much how much like hard skills they have in a particular area all of that matters a lot too but it's do they have this like massive learning rate where they can just pick up new things really fast and get really good at it and ship things um for marketing similar things like you know you may have experience with one particular thing but the game is changing all the time you have to learn how your users are talking you have to learn where your users are can they learn all that quickly is what we care about and with AI it's just the people that are multipliers are going to get so much more productive and get so much more done.

Um, and that's that that's how our hiring philosophy works today, too. I have one last question, we'll let you go. Um, can you tell me about HBS? What's going on at Harvard Business School? How did that come together? Tell me kind of the whole story and and how Julius is being used at Harvard. Absolutely.

So, Harvard is Harvard Business School is teaching analysis with Julius. Entire MBA class over a thousand MBA students. Wow.

all kinds of backgrounds like investment banking, marketing, product managers, tech exe executives, all of them are learning to do analysis with Julius because they realize companies have data and business leaders need to make decisions but analyzing that data is really hard. Um with Julius just becomes very easy.

How did it come about? Well, the professor um at Harvard Business School who teaches data stuff, Professor Yav Bjanov, he's a Julius user. He's been for over a year before he reached out and said, "Hey guys, we want to teach this whole class with Julius and we said let's do it like we launch.

" Um and um yeah, so they're they're all learning to use Julius now. That's amazing. Are you going to scale that to other universities or just hope that Mometics take hold and everyone copies HBS? Uh we I mean we have a lot of interest from other institutions as well.

So like Rice University is teaching a class called AI financial analysis with Julius. Um, it's just like this new way to do analysis. It's like the new Excel. So, yeah, that's awesome. Is that the prize by the way? Is Excel. You want to just replace it entirely on a long enough time horizon? We won't tell.

We won't tell Satia. Absolutely. Excel, Tableau, all of that. You know, number one business software in the world, Excel. Follow that. I love it. Uh, last take AI 2027. Uh, did you read it? Maybe you didn't have time. Uh, are we all getting paperclipip by a year and a half from now?

What does Julius have to say about a AI timelines? Import import the GDP growth rate and the energy growth rate and tell me when will we be paper clips? Yeah. Well, um, I think I think it's a really exciting timeline.

You know, I think the best I think the way to I think one one easy way to look dumb right now is to make super bold predictions on how the future's going to play out. And so I'm a big believer in like just like I think it's an exciting timeline and we just have to like ride the tiger. Ride the tiger.

I'm going to make a bold prediction. Uh 4 million users by end of year for Julius. Let's go. Um I'll take you over on that. We're going to We're going 10 million users. It's great to have you on. Um we uh we got some data to analyze. So you'll you'll you'll uh see us in there.

And yeah, come come back when there's a new product release or funding announcement. We'd love to have you back on. I'd love to come back. See you guys. See you. Talk to you soon. Love that he wore a suit. Yeah. Fantastic. Understood the assignment. Uh, next up we got Flo from Lindy. Another AI company.

You could call it a rapper, but I've seen him doing really, really cool stuff. Um, I'll let him describe it, but I'm excited to to chat with him. And he is in the studio now. Welcome to the stream. Hello. Yeah, thank you. Uh, how you doing? Uh, uh, thanks so much for joining. Uh, it's been too long.

Uh but uh good to have you here. Could you just give us a kind of a basic overview of who you are and the company for everyone? You are on mute. Sorry, we're lost. We lost audio. We're working on it. We heard him for a