The Exchanges

Q from tab to tab and try things out. But you not only have different offerings of LLMs on the platform, but you have different kinds of AI agents for different tasks, which is so fun because I go between research mode and just general search. But I'd love to know, could you just share more on how you.com has evolved and what your vision is for AI powered knowledge work?

A Yeah. Uh, we actually started as a search engine, uh, years before, uh, ChatGPT came out. So the BC era, um, a lot of people kind of call that now that before ChatGPT era and somewhat, uh, forgotten times, but we started with search wanting to revolutionize and change what search is and actually summarize and give you answers. And we've done that for a while. And we realized that a lot of folks are very stuck with, with Google, even though Google has gotten worse and worse, But also there are a lot of normal people out there who have fairly simple informational needs. Uh, if you're have a very simple normal life and you're mostly asked Google, like, what's the score of this game? Uh, what's the weather tomorrow? What's the stock price? When was France founded? Like simple questions, then you're not going to be able to give a 10 backs, 10 X better answer than Google using some really sophisticated AI. You know, there's only so much you can tell people about the score of a game or the weather tomorrow, and we're the first to bring LLMs into the search engine space and have some exciting patents, uh, some already given some pending, uh, on that, uh, but we realized at some point when we charge people for the product, because it was getting expensive, and there are no really good advertisements, people don't want ads in their answers in the chat engine that they are hoping they can…

AI assessment note: “we actually started as a search engine... it turned out that... there's a lot of knowledge workers”

Q Talking to the customer side, who are your key customers? What's their profile and how do you expect them to grow or expand?

A Yeah, there, there's sort of two categories on the API side. Uh, we have, uh, some very large consumer companies. They don't unfortunately always want to share exactly who they are. Um, cause they don't want the world to know how they get their answers, uh, to be more accurate. Um, but on that side, uh, we have. Companies that themselves have hundreds of millions or more users and Infuse our search or LM or both capabilities into their own end facing user products. And then on the enterprise site license, uh, line of business, we have a whole host of different folks, which kind of is both a little bit scary, to be honest, as a startup founder, I wish it was just like, oh, they're all in this one tiny niche and we dominate this one niche. Right. But the truth is we have one of the top 20 largest hedge funds as a customer. We have a bunch of, like, tech unicorns. Mindcast, the cybersecurity company, recently just increased their scope again with us, seeing incredible results, and a couple of other tech unicorns like them. We have biotech firms for, that are doing research with the tool. I have some very exciting announcements coming out where I actually partnered with someone who's been working in this space for a while, and They've been trying out different solutions, and like, it's kind of mind-blowing. I think I have a question here that we're able to answer for them, right? S…

AI assessment note: “we have one of the top 20 largest hedge funds as a customer.”

Q Um, you touched upon this a little bit earlier, but what does the current AI funding market landscape look like? I think we're all aware of the massively inflated valuations of some companies, but Are you seeing this trend continue? Are you seeing it wane? What justifies those valuations? And I just, I just don't understand if that's going to be sustainable for much longer.

A Yeah. As an investor, I have shied away from what I call unicorn seed rounds, uh, largely because they merge seed stage risk with late stage returns. And if you understand expected value, that's just not a good place to be as an investor. And so we at AIX have, have not gone, uh, that, that crazy on some of these massive Uh, very early pre-product market fit, uh, pre-product of any kind, pre-demo, like, kinds of rounds, and so that doesn't mean that none of them will work out, right? There's some parallel, maybe some will actually, uh, be successful, but it's, it's, it's quite hard, um, to do that. I think if you start to show revenue over time, uh, the AI funding situation is still very positive, and you just need to balance So the longer term vision with, you know, some concrete milestones as you, as you scale and, and start raising more and more.

AI assessment note: “I have shied away from what I call unicorn seed rounds”

Q How are you guys able to do that?

A That's a pretty complicated question. Yeah. We've been at it for, for quite some time. Uh, we're able to, uh, distribute the load. We're able to, um, when people just use our standard agents, which we'll get to maybe also in a sec, like we can choose the right LM for the task. And overall, the user doesn't ask five times more questions when they have access to five times more different LMs, right? So you get a similar amount of load and then we just distribute that, um, To, uh, some of the LM providers or hosted ourselves in some cases. Uh, and so, so yeah, that's how, that's how we can do that. And so maybe the last bit that I didn't answer for your previous question was on the agent side. Like, why do we call them agents? We had this genius mode, we used to call it, or genius assistant for awhile. And we realized just like with rag where we had invented rag already, we're doing it at a web scale, but it wasn't yet a term. And then at some point you said, oh, I guess we're doing Retrieve augmented generation. The future is already here. It's just not equally distributed. And in this case, genius mode was basically one of our agents that can decide when to search, when to follow up additional deeper searches on the web for you, when to program, when to actually run that code, which is non-trivial, like for general purpose, Python code, right? You can like just to run that on yo…

AI assessment note: “we're able to, uh, distribute the load... choose the right LM for the task.”

Q So at what point does u.com become five employees with just a ton of AI agents?

A It's, uh, it's unlikely, I think. Just cause we, we do have like, you do need to program all of this and it's, it's non-trivial, uh, to, to get this all like implemented correctly. Right. Um, and, but I do see us already using the technology more and more also. And a lot of people think we're like several hundreds of people. Cause you know, we have a whole search engine stack. We have an LM stack. We incorporate all the different LMs. We have agents. We do all of this usually before the big companies do it and others copy it and we're getting it accurate. Now we're also, you know, selling it more and more, which is great. Uh, revenues, uh, doubling every quarter this year and, um, it's becoming really meaningful now. And so we're really excited about that, but I don't think we'll have in the next few years, like the sort of unicorn with a single, single employee type of situation. I think that will probably take a little bit longer and we'll be non-trivial to get, to get right. And probably there's some luck involved, but also some really exciting ideas. I do think. I want to, I want to kind of balance the excitement of the future with also the hype bubbles that are happening, right? I do think that all boats are rising in a good way because of AI getting better and better. At the same time, there's like hype bubbles on top of it, right? You see this kind of inflection and a bu…

AI assessment note: “It's, uh, it's unlikely, I think. Just cause we, we do have like”

Q Maybe this is more of a technical question, but how do you ensure the accuracy and reliability of the information you're providing, um, especially when they're more complex, require up-to-date information? Do you have certain types of testing you use in-house? How do you, how do you ensure that?

A Yeah, there's, the, the true answer is very, like, it could take hours to, to give the full thing, and then of course we don't want to Fully explain it to all our competition. Um, there's a lot of copycats out there already as is. And so, um, we, uh, just like on some level, we just care about it a lot. And it's, it's sort of my background. I'm German. I'm a former researcher. I don't like overhyping things and then not like delivering, uh, on the technology side. Uh, and so, you know, likewise, my, my CTO and co-founder Brian McKen and I, we, with others to like invented prompt engineering in 2018. And We love sort of pushing the technology further and further and making it, uh, really, really good. Uh, so that's sort of organizationally and, and, um, philosophically kind of where we are. Uh, and then, uh, you know, concretely, there's sort of two stacks that you use when you try to work with LMs and getting accurate answers. There's a search stack, and then there's the LM stack. And LMs are great at Synthesizing, uh, a lot of the facts that they get and then reasoning over them and giving you, uh, answers, but they're not very good at the retrieval bits. You have to have an accurate search engine and a lot of companies that started sort of after or around when ChatGPT started, they kind of skipped that whole search engine bit and they just, uh, work on LMs. And the problem is…

AI assessment note: “concretely, there's sort of two stacks that you use when you try to work”