AI in Venture Capital: Investing, Startup Valuation & the AI Bubble | Nick Davidov, DVC

160 million is no longer interesting. That is, he didn't approve a single deal for us. Getting a rejection from a venture fund is a stroke of luck, because usually they just disappear. A cool employee left Big Tech and is starting some startup.

I am a venture matchmaker.

AI is the new electricity. Is it a bubble? Hello everyone. This is the Hoptks podcast. A podcast about IT and startups from Central Asia. My name is Alisher. I am the host of the third podcast. And today we have a special episode. We are recording it as part of Digital Bridge 2025. And today we have incredible guests — venture capitalists, people who believe in innovative and sometimes shocking ideas among the very first. Today we are joined by Yelzhan and Nick. Nick, Yelzhan, thank you very much for joining. Welcome. Today's topic of our podcast is AI in venture investments and in finance in general — yes, we will try to cover everything that AI and investments and venture investments in general can touch, right? So let's dive straight in. AI, large language models and multimodal language models — they were, well, probably among the first to be used in financial technologies, and there are a huge number of examples; among probably the most respectable is BlackRock Aladdin, which uses Generative AI to handle asset management and accelerate all of that, right.

Please tell us how artificial intelligence helps you in your fields — in venture investments — if it helps at all, right? And then we'll try to develop that topic further. Go ahead.

Well, we actually started using something even before ChatGPT existed. We built a thing just for fun. I hate writing rejections to startups. And founders won't let me lie — getting a rejection from a venture fund is a stroke of luck, because usually they just disappear, like a guy after a date with a girl; in English it's even called ghosting, the same word as after a date. Because writing rejections is unpleasant. You understand that a person always takes it very much to heart. I mean, you can say: "I don't want to buy your product." And then you think: "Well, maybe my product is bad." But when you say: "I don't want to invest in your startup," people go: "My ideas are bad," my company is bad — it's very, very personal. Plus, venture capitalists always try to leave themselves the option to come back and say: "No, let's actually write you a cheque" — if, for example, something has changed.

So the task is a difficult one, and I, as a person who received about 150 rejections in my own startup before we raised a round — I took my 150 rejections, stuffed them into the GPT-3 playground, and we built a thing that writes a rejection based on a startup's blurb.

Mm-hmm. Right. And, well, we built it just for fun and even thought about making a website where you could sort of practise — practise your mental health, in general, before talking to VCs. That is, you pitch your startup, and it rejects you in different ways. And then all these jokes actually turned into what we are doing now, because venture is very heavily concentrated, and mega-funds manage something like 25% of all assets in venture. Well, that is actually much healthier than all other asset classes, because in public markets and in private equity, out of 128 trillion in total assets in the world, 1% of firms — that is, 207 firms — manage 66%. So venture is a somewhat healthier thing, but still. And we decided that venture needs to be democratised, because the real value actually comes from people who were previously founders. The value comes from people who are technical, who have expertise, but those people have no way to build themselves a proper pipeline.

They don't do this full-time. They can't do comprehensive diligence — they can only do it within their own area. And we said: "Let's bring them together, build a community, have a fund where everyone can add some kind of value." And at first it worked great, but then, when we crossed 60 investors and around 50 portfolio companies, it started to break down, because I was the bottleneck.

And we started to automate me.

Basically, my wife Marina spends about half of her working time automating me. Then a product manager joined us, then an engineer, then another product manager, another engineer — and now we have a tech company. And we arrived at a point where all sorts of solutions have been built — for pipeline management, for portfolio management, and various additional perks — we'll be sourcing them and making them available to all angels and venture investors. Right. But we can no longer work without this thing.

Yes. Yes. You mentioned in one of your interviews that your proprietary AI — or rather, the AI that has around 220 variables written into it proprietary — manages to analyse everything before it even lands in your inbox, so that by the time you can read it, it's already done.

Well, practically, yes — that has actually allowed us to fundamentally rethink the process of how a typical venture fund operates. People hear about a deal, someone makes an introduction, a partner or an associate meets for the first time, they say it's interesting and send it to analysts. The analysts then write an investment memo. The investment memo essentially includes the data that was in the startup's presentation, plus some market data that could be gathered — who the competitors are, how things are going, you look at traffic, you try to understand what the exit scenarios are. And you end up with a document on the basis of which the partners can make a decision: invest or don't invest. Then you wait for the partner meeting, and then after that — the partner meeting said "invest" — after that you do due diligence and invest. Due diligence is also a very clear set of steps, structured according to rules. You don't even request most of these things from the startup — you do it by talking to their customers, talking to some experts.

And at some point you go: "Oh, wait, actually the market is smaller than we thought — we can't invest in something like this." And you pass.

Mm-hmm.

But since now your analyst, instead of the eight hours it used to take to write a memo, writes a memo in about a minute for thirty cents, you can flip the process upside down. You can do due diligence before you've even had a call with the founder, so as not to waste your time or the founder's time. So you can invest twice as much per partner, or three times as much per partner, because the partner isn't making those calls that, three weeks later, you realise were a waste of time.

Right, and yes — so when someone sends us an email, the agent automatically pulls out everything it can and writes a memo, posts it on the platform, and pings me in Telegram that there's a new memo to look at.

Great, great. Yelzhan, please tell us about your experience. As one of the most active venture investors in Central Asia — as far as we know, you have more than 90 projects in your portfolio. Tell us, does artificial intelligence help you in any way? Yes, how did you manage — how did you manage to get to as many as 90 portfolio projects?

Yes, well, 90 — that's my personal portfolio,

which I invested through mostly various syndicates and angel communities around the world. Right. I'll say what we use at MI7 — what are called, well, VCTK — yes, that's what they call these products — there are actually a lot of them on the market right now. There's even a site called vcstack.com. There's pipeline management, CRMs, portfolio management, founder engagement — that is, tracking how they submit reports or finding new founders who are just getting started. Lots of different products that sell as B2B SaaS to VCs and build large businesses out of that. We tested some of them, looked at others — well, we found a good CRM that can integrate with other tools. Then we asked Arman Sulemenov to write us an agent that does the work, well, as Nick said, of an analyst — and it does it much better, probably. And it prepares such memos that, well, probably there are no analysts on our market who could produce such memos,

right?

It was, well, fed all the other accelerators and so on and so forth, and it is, probably, stricter than this investment committee — it hasn't approved a single deal for us at all. Yes. Well, it works the same way — an email comes in, the email integrates with the CRM, it passes through the agent

and, well, makes life much easier, yes.

Yes.

And then there are tools that monitor the portfolio, calculate various things — there are also a lot of those. And, well, the whole industry is saying that soon AI will be doing everything — watching everything, searching everything, finding that some cool employee left Big Tech and is starting some startup, being the first to reach out to them: hey, pay attention to me, what are you doing, tell me about it, and so on, right? But I believe that in this business, what they call «насмотренность» — a trained eye — is very important. And, well, intuition, right? Intuition is

all the experience you've had, just fast decision-making.

So that, probably, hasn't gone away, and it's a very important element. And even this: startups sometimes write cold, right? And those who reply to cold outreach — 10–15% — those are, as a rule, very in-demand funds that have the time to reply and not miss out. But if the AI always replies on its own, you can miss out just as easily as with an incoming message that arrives. So the human element is very important in our industry.

Yes, yes, thank you very much.

From cold emails, by the way, over the entire career, we have invested in one company.

That is, out of hundreds of thousands of cold emails received.

Mm-hmm. And is that because the quality of the startups didn't satisfy you, or for some other reason?

Well, it's a bad signal right at the entry point — that people don't know how to raise money.

Because if a person couldn't find a warm intro,

then they don't understand how signals work.

And if they don't understand how signals work, how will they reach customers, how will they hire top people — this art of finding the right path to the person you need. Because essentially, when someone introduces you, they are vouching for the person they are introducing,

right?

And that is a very strong signal. That is, if a founder you trust greatly — you invested in their company, you consider them an expert — brings another founder and says: "Take a look, pay attention." You will naturally look and pay attention. And if that person came to you cold, you don't have that signal.

Yes, thank you very much. We've already started talking about such a thing as intuition, and about the art of finding a path. So we are already moving from the purely analytical part to something more abstract, right? Let's talk about the following. Venture investments are the highest-risk. And such an aspect as belief in a person at a super-early stage — that is actually a super-working scheme for practically all venture capitalists who specialise in pre-seed, right? Please tell us: when the AI — your tech — suggested no, or suggested yes, and intuition said the opposite — has that happened, and what did you do in such a case? Well, I wouldn't track it quite so directly as: what does the machine say, and what do the agent and the people say. The machine is simply a normal process of receiving information and quickly absorbing it — it quickly prepares a one-page memo and you very quickly get acquainted with the startup, what it does, and you decide whether you want to meet with them or not.

As a rule, well, the first meeting is half an hour, an hour online, right? If that meeting generates even more interest, you meet again for a deep dive, well, for two hours, where you look at the whole project from start to finish. So it's not the case that if the machine said no, you don't invite them, right? So it's not connected to that. And, well, you touched on, uh, pitching earlier and believing in the person. Yes, well, that is, so to speak, critically important, very important — that the person is an expert and knows their industry best of all, the one in which they're building the startup. Yes. And, uh, that there is the maximum probability that they can actually accomplish what they, well, are pitching,

yes.

So, well, the person's profile is very important.

And, as a rule, it is hard to convey a profile in a pitch deck, right. As a rule, these machines go to their LinkedIn, some open source, they find something like that. But when you meet a person, well, not everyone has an online trail and their experience is there.

Yes, yes,

that is a super important topic. And our AI doesn't give any recommendations at all. Because, well, in reality, if it did, then it would, most likely, for the large part of the companies we look at, and, seriously, the ones we get on calls with, it would say: "Yes". Uh, but the human perception of the founder, the founder's energy, why they're doing this, the company, how they're doing it — that is super important. Uh, the deck and everything else — those are just signals.

And, well, we look at 200-plus signals, uh, we constantly add new ones — for example, right now we're adding who from the team was let go. Oh,

meaning, like, we try — well, if someone was on LinkedIn as an employee, for example, and then stopped being an employee, listed as a former employee — we try to understand that person's profile: were they not pulling their weight on skills, or did the communication within the team just not work out. That is, cultural problems are, uh, the number-one killer of startups, especially if there is tension between co-founders, things like that. And, as a rule, people who lost one co-founder along the way will lose another co-founder. And then the question becomes whether that one person or those two people will carry this thing through or not. That cannot be automated. There are several funds that, for example, TRXON, uh, or Quantum Light by Nick Staronsky, that invest entirely algorithmically,

like quant funds.

And Trxon made a small cheat code — they invest only in Y Combinator,

so they, uh, well, roughly at the median of Y Combinator everything still exists. Whereas Quantum Light invests at later stages, when the numbers already become much more important. But we looked at, in fact, a large number of different algorithms, who uses what, what weights each one assigns to what. And it's quite funny, of course. Just today I even wrote a jokey tweet with a chart of weights, on the topic of: you look at, like, net revenue retention, speed of traction, founder profiles, market size, all of that — it has some weights. And then there's this massive weight on Sequoia having led the signal.

Well yes, yes, yes. Thank you very much. And, well, when we talk about investing at early stages, right, first of all I want to congratulate each of you, because one way or another Nick, Yelzhan, you are directly or indirectly connected to Hiixfel, right? And I want to congratulate you on the fact that they, yes, uh, have preliminarily announced that their valuation has reached one billion dollars. Yes. And Nick, as far as I know, you were the person who introduced the founders.

Yes. Uh, Yerzat and Alex.

I'm a venture matchmaker.

Yes. Yes. And, Ilan, you essentially invested at the early stages, right, as part of a syndicate, and then later you had follow-on investments, right, in Hiixfel. And so, uh, speaking about investments at early stages, right, in projects like this, where it is essentially very difficult to make a decision — whether this project really is, well, when it essentially has no traction at all — how do you determine, uh, whether this is really a project that solves an existing problem in an existing market, or whether it is simply a pretty AI wrapper?

Well, a pretty AI wrapper doesn't mean it's bad. Uh, well, Hiixfel specifically — that is a no-brainer investment. We have quite a few of those, because we have 172 investors in the fund, and they are all, uh, either technical people or founders. And since they are investors in the fund, they are fairly successful — uh, that is, they can afford to invest in a venture fund. Now, we, again, because of the restrictions of, uh, the American regulator, cannot take investors with less than one million dollars. And then, uh, once we bumped up against the 100-investor limit, we can now no longer take people who have less than 5 million dollars. So when one of our investors launches a company, we naturally don't scrutinize it that closely. Alex sold his previous startup to Snap for, I believe, 166 million. And we had a lot of conversations with him. He brought us several great deals for the fund. I really liked the way he thinks. He made his product incredibly successful at Snap. Got tired of playing corporate politics, said: "I'm leaving."

And think, if you know some good technical co-founder.

I said: "But you yourself are a CTO."

He said: "No, more technical, because," he said, "I want to do something really big this time. 160 million is no longer interesting."

And so that was a no-brainer. Same thing. We recently invested in a startup by Zhenya Kuida, who made Replika. And we invested in a new startup by Tim Ustargen, who made Pandora Music.

Also a multibagger. We invested in Zencoder by Andrey Filev, who built and sold it for 2 billion dollars. That is, in this regard these are no-brainer investments. That is, if this is a super successful founder, then at the very least they know how to build companies, they know how to raise money, hire great people, sell, and so on. That is, all our initial questions — the answer is yes. And then the only question remaining is: in what mental and emotional state, from what mindset is the founder now building this new company? Are they doing it because the technology is interesting and their hands are itching, or are they, so to speak, well, wanting to prove something, or do they want to build something big. That's what — what drives them and how long-term that motivation is.

Yes, yes, thank you.

Yes, well, the founder of 7, my partner Murad Abdurahmanov, he is the first investor in Hiixfel, in Yerzat first and foremost. After that Yerzat met Alex, and they joined forces. Hiixfel altogether is, I think, Yerzat's nickname on GitHub or somewhere. Yes.

And Yerzat, he

is very talented and a top expert in his industry. That is, he — right now somewhere online there is a video from 2009 or 2012, I think, where he was already saying at that time what the future would look like, how LLMs would develop, and so on. Yes. And he is, well, a very well-known researcher online. Everyone knows him, everyone invited him to work for them — OpenAI and so on. So when Murad met him, he simply understood that he is, well, in the top 0.1%, 1% of understanding in that field, so also, well, simply believed in him, there was nothing yet, and so to speak,

well, also very important — that he is a person of high moral character,

because the average founder, uh, given that, uh, as I understand it, Murad invested in principle in whatever he was going to do, right, but not specifically in Hiixfel, not in the video network, in what it became,

yes, that is one of their pivots. And this is, so to speak, well, and any other founder could very easily have said, like, well, sorry, it didn't work out, here I've started a new company, do you want to invest in that one too.

Right. And so that is also important — integrity, well, a great person,

yes. Well, at a very early stage, so to speak, uh, the right human qualities — right, decency, some degree of humility, friendly, kind. Well, so to speak, by default that should be a given. And he is also a top technical talent with experience and so on.

Yes, yes. All just like Warren Buffett, right. Integrity, energy, intelligence, yes. Thank you very much. And here we get a sense of the fact that this was a no-brainer, it was a very easy decision, right, when you have a founder like that in front of you. And some people get the impression, I think, that it feels as if this happens very often right now in projects that operate with the help of or use AI, right, or that build their own LLMs and other, uh, yes, neural networks. Tell us, uh, about this very popular question right now, a very hot question. Uh, is this a bubble, right? And if so, where are we in it right now? If it bursts — or doesn't burst — right, then how high is it flying, or how painfully will it burst and yes, hit all of us?

Yesterday I gave an hour-long talk on this topic,

and in an hour I didn't manage to get through all the slides, so I posted them — well, the forum organizers are supposed to post them alongside the broadcast. But briefly: uh, the easiest comparison is with the Dot-com Bubble of 2001.

Uh, what's similar? Well, what's similar is that companies shoot up in valuation very quickly. And the most expensive company, one of the most expensive companies in the world right now is NVIDIA, a company that provides the infrastructure for AI.

The same was true in 2001 — the most expensive company in the world was Cisco, which provided 80% of the internet's infrastructure. So, what's the difference? First, uh, the laying of infrastructure for the internet was happening in anticipation of future demand, and at peak the amount of infrastructure exceeded the amount of demand by 100,000 times.

By the end of 2002, 94% of the fiber optic cable laid in America was not being used by users.

Users were not connected, Last Mile had not been done. That is, this infrastructure is just lying there and nobody is using it. Now, infrastructure is being built for existing demand. And not just for existing demand, like for research, for some future thing. This isn't Open AI buying it for new types of networks. 80% of deployed GPUs are used for inference, that is, for applying artificial intelligence. At Facebook, 92% of GPUs from their enormous data centers are not even occupied with LLMs — they are computing advertising.

Yes.

Right. So, the second big difference is that, uh, the internet required a very significant change in user behavior. That is, a person used to walk into a store on foot — they need to shift to e-commerce; a person, well, used to consume content in a newspaper — they now need to consume it on some website on a small screen. Everything was different. Now we are doing all the same things we did before. We're just doing them faster, together with AI. Mm-hmm.

That is, if before I needed to, in order to book a hotel, go to various aggregators, Google things, look at reviews, go back and forth, and then book a hotel on Booking — well, now Perplexity simply does that research for me, and I still book the hotel on Booking, or I press something in Perplexity and it books it on Booking for me. That is, my user behavior hasn't changed, business models haven't changed. The internet brought new business models that nobody understood. AI operates on those same business models, and only in the future do we see that it will probably change them. That is, the advertising business model is slowly starting to go stale, uh, while transactional business models, subscriptions, are starting to grow very strongly. And probably the last point, the biggest one, is that there were no large internet companies at the application layer. That is, there were infrastructure companies and there were some, conditionally, foundational platform companies, things like AltaVista, stuff like that.

Whereas in AI we have the Magnificent 7, and seven of the ten most expensive companies are already AI companies — companies that are already in the internet, they have distribution, they have hundreds of billions of dollars in AI revenue — this is not a new thing. It is simply growing at an incredible pace for them right now, because users like you and me are simply paying. Right. And, uh, we see that adopters pay a lot and that power users actually use it more. And we see how all of this is growing. For example, Chat GPT — recently their usage statistics leaked. 4% of Chat GPT users use it — only 4% — to write program code. Uh, but coding accounts for 10 to 15% of all of OpenAI's tokens. That is, these people use Chat GPT three times more than the average user on average. And they pay more. That is, our, well, our company — we are early adopters, that is, we need to be ahead of this technological penetration, so to speak.

We pay for tokens — probably even more than the average salary at our company, every month. And we understand that this will continue to spread across the market in a similar way.

Yes. Yes. Thank you. Thank you, Zhan. Tell us about your vision specifically in the context of Central Asia, right? And your, yes, probably the most well-informed expertise here — what do you think in the context of the AI money question?

Mm-hmm. Yes. So is there a local AI bubble in Kazakhstan, Uzbekistan? I wouldn't say that our sphere differs from the global agenda — it's generally moving along, because the agenda develops the same way for everyone. It's not like

it develops there and reaches us a couple of years later — we all live in the same world, we're all building the same startups, and everyone has access to the same cutting-edge solutions, these LLM models and so on. What I simply see is what Nick said — there's the infrastructure layer, which is commodity, right? AI is the new electricity, the new internet, and tokens and LLMs — all of that will be very heavily commoditized, and all the value will be captured at the application layer. That's why the creators of these LLM models — for example, OpenAI — they want to control the entire path from infrastructure to the end user, which is why they're buying that Jony Ive company, to have some kind of end device that delivers their electricity to the application layer and monetizes there too. And we see that a large number of AI companies are doing AI at the application layer — which is good. In many people's perception, if you're not doing foundational AI or at least some post-training or other heavy-duty stuff, then you're not an AI company.

Hmm,

but in the right application at the application layer, if you can capture value and deliver your domain expertise in your industry, that's probably the hottest battleground right now. Because the battle in infrastructure and deep technology — well, entering that space takes hundreds of millions or even billions of dollars now, it's impossible to get in there. That game has already been won and lost, long ago. And all startups are going here, competing in every domain. Right now it's becoming a very red ocean — the market is forming, I'd say. And what we're seeing now is that large companies are not reporting any hugely dramatic effect. I mean large in the sense of companies that have existed for decades. And the fact that they've cut costs by 30% with AI — they're not reporting that in their public reports,

but everyone is investing heavily in it, because they don't want to fall behind in the future.

One of the coolest cases from big companies — Klarna was talking about how they could save something like 20-something percent by applying AI, but that's a purely internet-technology company, right? Whereas traditional businesses haven't strongly felt the benefit yet. And we all say this is inevitably going to happen. And if it doesn't happen for some time still, we can say that, well, adoption is just moving more slowly,

right.

It's not that there's a bubble that's been inflating and it's burst and that's it, we're done with this. It's just happening more slowly. There are several problems. A recent MIT study came out saying that 95% of enterprise AI pilots have not delivered financial benefits. There's a very simple explanation for this: AI is deceptively simple.

Meaning people can build a prototype overnight that handles some business process and shows off the magic.

And so very often inside large companies the IT department says: 'Why would we buy anything? We'll just build it ourselves from open source.' Hmm,

and I was speaking here with several large businesses, and it's exactly the same story, one to one. There are even more machine learning engineers in Kazakhstan. You come to Kazakhstan and I met with Kadat Branbaev here. He has an ML department for restaurants, calculating which menu items people like more, and you think, like, wow,

and so people here internally say, like, we'll implement it ourselves, look, we've already built a prototype. Then it turns out it only works in 70–80% of cases. That's not enough to replace people, because humans have something like 96–97% accuracy. Plus, AI is not forgiven for mistakes the way people are forgiven.

And they're waiting for 99%, five nines. And getting it from 80% to 99% takes at least a year.

And the market is still very young. The market is actually 2 years old, really. Meaning a large company most likely spent a year on selection and a year on implementation. So their first result — they still haven't gotten anything to work yet. They say: 'We probably need to change our business processes first.' Building a new process from scratch with AI in a new company is much easier than changing something in a legacy company with AI. Mm-hmm. But it's still possible — and related to this, there's now a big trend called AI rollups. These are companies that have figured out their own AI workflow in some niche — say, in HR or in accounting,

and they go and acquire other traditional firms in that space that are still working on paper the old-fashioned way, buy them out, give them their tool, train them, integrate them — and in doing so they buy their client base. They buy the company for three or four times earnings, add all their revenue together, give them a startup-like multiple, raise more money both in equity and in debt to buy out more companies. It's a trend that's very popular in the US right now. In other parts of the world I haven't heard of startups like that yet. We invested in a company called Dwelly. They're very cool, they've just raised a very large round from General Catalyst. They buy property management agencies — not selling, but managing rental residential property. They started by literally just building their own property agency without people. They bought an agency, fired everyone there, put it on their platform. It works on that, they raised money, bought more, removed the people, raised the margin, took on loans, and now they're buying more.

Yes. Yes. I'd like to share a very short story about this. The first thought that comes to mind — when we talk about adoption, we talk about early adopters, early believers, but these conservative people take a very long time to come on board — and the conservative person who turned out to use it more than some cool coders turned out to be my mother. Yes. Because when she was selling her apartment, she wrote a great text and sent it to me as an ad. I read it and she says: 'Have a look.' I say: 'What a great text. Where did you get this? How do you know there's a park 900 meters from our house?' She says: 'I asked Luch GPT and it told me — and the most interesting thing is it didn't even hallucinate. It was actually true.' So it's very, very fascinating. Before we wrap up this episode, I know you're very, very busy today, you have a huge number of panel sessions and presentations ahead of you.

Tell us your checklist — who you won't take into your funds, who you don't invest in — a kind of playbook, a guide for founders who are watching right now about what not to do. Well, first of all, don't break the law, don't act badly or unethically. That's a massive red flag. We always — whenever we ask — for almost every deal we do, we run some kind of background check on the founder. And if we're told that a founder is arrogant, or that a founder argues with competitors, or that they're conflict-prone — but if we're told that a founder deceived someone,

that's a no,

it's an instant pass, even if the company looks absolutely incredible — pass, because a venture investment is a partnership, and if it's unsuccessful, meaning the best case for a bad partnership is a partnership lasting a year. Hmm,

and in the worst case, if it's successful, you'll be stuck with them for 7 to 10 years. The second topic — what we don't invest in: we have our ethical code, and accordingly, we believe — well, why do I do venture? I mean, in general I could probably earn more money somewhere else, or work less. It's quite hard work, a hundred-hour work week. And until the money. Well, things are good now, but yes — until the first carry I waited 11 years.

And now, in the sixteenth year, you can already afford to live normally. But for me the most important thing is that venture is an industry that allows you to change the world with maximum impact. It's like

what Archimedes said — give me a lever and a place to stand and I will move the world. This is a great lever with which you can act on the right people at the right point, and move technology in the right direction, move society, and make some changes — talking to these people who have a very strong impact on the world. And so for us it's very important to invest only in things that we believe make the world better. We will never invest in gambling,

we will never invest in things that are addictive,

we will never invest in things that can kill people.

And even if a technology is more or less peaceful — like it's for defense — if it can kill people, we say: 'No.' AI can be dangerous, it can be safe. Many people talk about this. The only way I know to guarantee making AI dangerous is to give it weapons and teach it to kill people.

That's what many drone companies are doing. So we simply don't want to participate in that and we direct our efforts somewhere we can make people's lives better, reduce the amount of suffering, and make society more just.

Thank you.

I'll probably give some practical advice to young founders. Especially when we're talking about pre-seed, where you look more at the founder and the team than at the seed stage where some initial traction appears. Everything is clear — you're a nice person, decent and so on and so on — but the concept of founder-market fit is very important. Meaning the founder must be an expert in the field where they're building their startup.

And how does that show up? At the first meeting you must know more about your industry than your investor.

When the investor hears that you're deeply immersed, that you know the competitors, know the numbers, know the market, know the benchmarks, know who to look up to, and know something more than the investor in that area — then they'll be interested and will understand that you're an expert there. If you're an expert, it means you can see what the shortcomings are, what the problems are, and how they can be solved.

Mm-hmm.

Yes, probably that's our advice,

yes. Thank you very much. Thank you very much for joining us. We know you're incredibly busy today, and you made time for us. We're very grateful for that. This was an episode of the HubKC podcast. We wish you a wonderful day. Thank you very much for being with us. Subscribe to the AnaHub channel and until we meet again.