AI Startup Funding: Bubble-Proof Strategy from Top VC - Sophia Zhao, Alumni Ventures

AI bubble is something that's been talked about right now and as venture capital professionals everyone's thinking about it and therefore they are probably a little bit more cautious. We've all seen these headlines in the news of companies raising $200 million on

cautious. We've all seen these headlines in the news of companies raising $200 million on a billion valuation. So it sounds a little too frothy, sounds a little too rich. The bar for startups to build to a point to differentiate themselves is higher and to compete for capital

at a reasonable valuation is is also higher. All of these foundation models they basically exhausted all the public publicly available data out there. Being able to access and train small medium size LLM or train your own model based on proprietary data that will give you defensibility.

Building something for a very niche industry that makes it harder for open AI or claude to touch or anthropic to touch is very important. So for startups, they need to build super fast. They need

to go to market super fast and they need to scream as loud as they can to get your attention because it's all about attention economy. You're watching Today's podcast. I'm Boyuan. If you enjoy deep conversation about building future of technology and startups, please subscribe and stay updated.

Hey, what's up everyone? Welcome to Today's AI. I'm Boyuan. In every tech cycle, there's a moment where ideas stop being experimental and starts defining the next decade. AI feels exactly like that moment right now. Today, I'm joined by Sophia Zhao. She's a partner at Alumni Ventures,

one of the driving forces behind their new AI First Fund. This fund is built entirely around companies where AI isn't just a feature, it's a foundation. Alumni Ventures are one of the most active VCs in the US with over 1,600 portfolio companies and co-investments alongside Sequoia,

a16z, and Y Combinator. Welcome, Sophia. Yeah, thank you very much. Hi everyone. Thank you

Boyuan for hosting and thank you very much for having me. Great to meet you all. Yeah, Sophia,

maybe let's get started with your background first. So, you started your career in crypto and fintech before stepping into AI investment. So, what pulled you into this new venture, this new frontier? Yeah, for sure. Well, let's go way back in time. So, my background is actually in both finance and startups. I work with a lot of different CXOs and entrepreneurs in the past.

in the capacity of corporate banking, capital advisory, BD and operations and oftentimes crossborder. So these experiences really allow me to build a pattern recognition of what it

crossborder. So these experiences really allow me to build a pattern recognition of what it takes to build a successful company as well as forming a global perspective. I joined alumni ventures back in 2021 after spending a few years in blockchain working with you know web3 merchant

banks such as Galaxy Digital. I spent a year at Crypto Exchange, Huobi US in San Francisco, and then Crypto.com. so as an investor, I truly enjoy being plugged into these different hackathons,

then Crypto.com. so as an investor, I truly enjoy being plugged into these different hackathons, accelerators, tech clubs, like tech stars, like Yale's innovation center, Ethereum Global, etc. so I'm educated in both Canada and the US, and I graduated from Yale School of Management in 2017.

so I started in crypto in 2018 because during a course work we were helping a VC diving deep into different verticals including u blockchain and AI. so I did some deep research into blockchain.

I purchased my first crypto in 2017 and later exited that really helped me pay some of my student debt and then I became a believer. So I've been focused on investing in web 3 since 2021 after I joined Alumni Ventures blockchain fund and over the years we made some you know

decent investment. everyone's pretty happy with our performance and we've been entrusted to then

decent investment. everyone's pretty happy with our performance and we've been entrusted to then being staff on the AI fund. So currently I am also looking after our investment in our AI fund specifically for a newly launched fund which is the AI first fund and prioritizing opportunities

at the seed to series A stage. So before we dive into your new AF first maybe let's talk about your experience as an investor working across different regions and cultures. So you work across Asia, Canada and the US. So I'm curious from your view, how do funders and investors approach

AI differently in these places? Yeah, that's super interesting. I've have exposure in Asia, US and Canada and it's quite interesting comparing contrasting the three different regions. maybe

starting with Asia for example for Asian VCs or just the way of doing business in Asia, it's very collective and very relationship driven. So everything is built on trust. You're building

relationship with the long term. You're aligning your interests before any transaction is done.

So in Asia you know founders and VCs sort of re rely on their network to help validate hey how do you think about this project? Is this VC someone that you recommend working with? so that's

sort of how the relationship is being prioritized because network builds trust and trust is what anchors relationships. in terms of the type of startups, gaming and entertainment is very popular

anchors relationships. in terms of the type of startups, gaming and entertainment is very popular especially in China, Japan, Korea and Southeast Asia because the industry is quite deeply rooted

in K-pop for example, anime and gaming. and we can also see fintech and super apps especially in Southeast Asia like consumer fintech payment neo banks they tend to dominate early stage funding

and for China even though fintech is now more regulated but legacy giants like and financial tensen they they have a huge influence on the ecosystem. I find capital is less patient versus capital in North America. Yeah. So they expect a faster returns or invested in startups where

signals are a little bit stronger. Yeah. They tend to want to see returns a little faster than other regions. so that's what I would share about what is it like in Asia. I think for Canada the keyword

regions. so that's what I would share about what is it like in Asia. I think for Canada the keyword is a lot more conservative. I think there's a more balanced approach to risk appetite. You can see

more more cautiousness like they analyze more but they're definitely a lot more open than the Asian counterparts. I think the community focuses more on sustainability, inclusion and long-term value

counterparts. I think the community focuses more on sustainability, inclusion and long-term value creation. And then naturally because of that then in terms of the industry you're looking

creation. And then naturally because of that then in terms of the industry you're looking at clean tech, climate tech because both of these are supported by public policy and there's a lot of government co-und funding that can go into supporting those startups as well as health

tech life sciences. There's a lot of strength in biotech genome tech digital digital health etc. And I think a lot of people are not aware that Canada is actually a AI powerhouse, especially in Toronto and Montreal. we've invested in a Canadian LLM called Cohere. It is born in Canada,

but is actually a foundation model company that's serving the world prioritizing enterprise users.

So that's an example of a AI huge AI player that's came out of the CA Canada ecosystem. Yeah, Cohere

is another foundational model company. It's kind of low profile compared to OpenAI or anthropobic.

It's focusing on mostly enterprise customers and it's got a quite strong profitability. Yeah. And

I think it's good that foundation models are not just concentrating in one country. we see Mistral in France. So I I think it over time it's great to see some diversity in terms of these foundation

in France. So I I think it over time it's great to see some diversity in terms of these foundation players. so yeah that's an example of Canada as a AI player u within our AI market. Maybe in terms

players. so yeah that's an example of Canada as a AI player u within our AI market. Maybe in terms of the approach I think people are a lot more okay let's focus on prototyping let's focus on

you know getting some revenue. So I feel like the Canadian companies tend to approach it a little bit more steadier than the US counterpart. and not to mention Shopify is also a Canadian legacy in a way. U B2B company. Yeah, Shopify recently has also rolled out a few AI native commerce features

a way. U B2B company. Yeah, Shopify recently has also rolled out a few AI native commerce features and it's actively partnering with AI and the cryptonative startups. Yeah, absolutely. and then

if you're looking at us, this is the fun part. everything is fast. I think the move fast and break things mentality is still very prevailent especially in the valley and there's a lot more prioritization on hey success is tied to the founders charisma and sto storytelling abilities

and the VCs are a lot more willing to place big bets with you know uncertain outcomes for outsized returns and everything's faster in the valley and I don't know if you listen to the recent not recent but there's 20 VC episode podcast interviewing some of the European founders and

the underlining narrative is the same that they came to the valley to raise capital because things are just done faster and a lot more efficient. So over here I feel like we're just prioritizing bitscaling. We're prioritizing growth over profitability. And with that the exit expectation

bitscaling. We're prioritizing growth over profitability. And with that the exit expectation is that you IPO or you have like over a billion dollar acquisition within the next five to eight years which is really interesting. And I do find US often shape global investment patterns. And

fun fact like CB Insights according to a recent CB Insights quarterly report US is the leader in the total number of deals that's done this year. Seven of the largest rounds were based in the

US including Anthropic, OpenAI, DataBricks and Figure. So it's a vibrant market here.

Yeah, US is definitely leading and shaping how AI is evolving especially given the sheer size of capital concentrated in the BC and tech ecosystems. So I want to dive into alumni ventures investment thesis. You've had a very strong track record of successful exits and your portfolio

investment thesis. You've had a very strong track record of successful exits and your portfolio spans everything from AI infra to applications to even crypto projects. So what made Alumni Ventures

so unique and also what led you to create an AI First Fund? Yeah, absolutely. I think our model is very unique because number one, we give retail accredited investors access to venture investment that are historically very hard to access unless you're in institutional investor. so we provide

that unique entry point and then secondly is we are connected to all the top universities in IV leagues like Harvard, Yale, MIT, Stanford, UCLA etc. I went to Yale for for grad school. So I'm

plugged into the Yale community. We have this huge network of people including over 1,600 portfolio companies that becomes a network for our founders and our partners because everyone's looking for adoption. They're looking to hire, they're looking to partner. So, we don't bring the capital, we

adoption. They're looking to hire, they're looking to partner. So, we don't bring the capital, we bring the network to help our port host grow. So, that's super unique. Plus, we're stage agnostic and we're sector agnostic. So, that gives us a lot of flexibility to get in on interesting deals that

are co-invested with a16z, Sequoia, Founders Fund, Khosla, etc. So we're very unique in in that in that perspective and now we're launching this AI first fund and and thank you for highlighting that

we did a lot of big infra projects like rock like cohhere like lambda like together AI and now we feel that there's a lot of opportunities that are at the industry level in the science sector that

are also for the knowledge workers or continue new ways of AI tooling for further info improvement etc. So there's a lot of experimentation innovation at the very early stage now that we have a lot of foundational tools that we can build upon. So we're looking at seed to series A stage

of companies that are being an AI first knowledge worker tool being an AI first industry application an AI tool discovery you know whether it's like a AI chip or how how do you optimize and make

very clunky for example healthccare backend a little bit more better interesting opportunities like that. So, in my past episodes, we've been talking a lot about what AI native companies could

like that. So, in my past episodes, we've been talking a lot about what AI native companies could be, and it seems to me that everyone nowadays trying to call themselves AI native or AI first, and the term gets thrown around quite a bit. I'm wondering from your point of view,

what does an AI native company really look like? Yeah. maybe I can use our portfolio company as an example. literally today, 1mind announced their 30 million series A, led by an amazing woman founder,

example. literally today, 1mind announced their 30 million series A, led by an amazing woman founder, Amanda Kahlow. so we we've invested alongside Battery Venture, Wing Venture, Primary Venture,

Amanda Kahlow. so we we've invested alongside Battery Venture, Wing Venture, Primary Venture, and a lot of amazing investors. So essentially, 1mind is building this AI superhuman teammate that

actually does a sales job and not just assisting a sales staff. because when you think about it, like what is a modern B2B go to market? It's a little broken for a lot of people. For example, sellers are stuck choosing between growth and efficiency. Buyers are waiting days for answers that should

take minutes. and a lot of buyers choose whoever respond first because everyone's time is precious

take minutes. and a lot of buyers choose whoever respond first because everyone's time is precious and our attention is very limited. So most companies can't really deliver that not because of lack of effort because as humans we have limitation in terms of time, capacity and

recall. So for 1mind they are building these AI superhumans. They're not building co-pilots and

recall. So for 1mind they are building these AI superhumans. They're not building co-pilots and they're not building a wrapper. They are creating this full digital teammate with a face, voice, personality, and a full go to market brain. So, there literally is the AI work for you. they're

on 24/7. They can join video calls. They can give demos. They can take actions. So,

they're now starting to replace the chat bots and STRs. so essentially this is not an assistant tool but this is the actual AI as your sales teammate. So to us this is very AI first and this is almost like the future. Yeah, I can definitely see there's a massive demand on B2B sales side,

especially for startups. When founders are moving into enterprise or B2B, they need to either hire a sales team or upscale themselves in order to do the B2B sales. And despite there's another layer for enterprise compliance and trust. So having an AI native sales team feels like a real value

ad for early stage companies or B2B companies. Yeah, definitely. So that is more of a almost like a full stack AI innovation. But there's also like another example would be our investment into Toma AI. We invested at the seed stage and a couple months ago a16z came in and took the entire series

AI. We invested at the seed stage and a couple months ago a16z came in and took the entire series A round. So Toma AI is building AI agent for the automobile retail industry. as you can imagine,

A round. So Toma AI is building AI agent for the automobile retail industry. as you can imagine, like there's these car dealer shops in our in our neighborhood and they what's surprising is they actually miss 45% of its inbound calls which is significant because they just don't have a way

of handling customer communication. Like when you call a dealer shop, it's a landline. Then you have an operator and it's like one for sale, two for service, stuff like that. So, Tama is building this new voice AI that doesn't just sound like human, but work like your employees to really

personalize to every dealership, so it's not one-size fit all. They're trained on the data and workflow specific to that particular dealership. they can delight customers and they can save staff a lot of hours so the staff can focus on what really moved the needle versus the mundane

work that could be automated. So this is another example of company that's thinking AI first for an industry application because car dealership is a very sleepy. Yeah, it's interesting that some of the best companies aren't always in the spotlight. I think Toma is a very good

example that it's transforming traditional capital intensive industries behind the scene. So Toma is also backed by YC, right? You guys co-invested it at early stage. That's right. Yeah, we we joined YC in Toma seed-round last year. And maybe just a quick call to action for our portfolio company.

If they're looking for great senior and staff engineers that's working on platform or product, they're also looking for product designers. So if you or if the audience you or you know someone who might be a great candidate for the team, they're looking to grow and they would love amazing human

talent to join the team. So now let's shift the gear a little bit. Alumni Venture is known for co-investing alongside a strong network of top VC firms. So for an angel investor like me, how can I participate in the AI first and capture some of the great deals? Well, thank you very much

for asking that question. we have this AI first syndicate on our website av.bc. So it is free to join. You're not obligated to invest in anything. Just free to join the community. Every months

join. You're not obligated to invest in anything. Just free to join the community. Every months

you'll get maybe one or two deals that we are participating in. So for you it's a great learning experience to see what's new and exciting as well as we'll be sharing our diligence publish our investment memos so you get to see how we evaluate deals is a way for retail investor to have their own framework or build their own framework of how to evaluate these opportunities. So I think it's a

great learning experience for anyone looking to learn more about AI what's happening as well as learn more about how to approach investment. so I encourage you to join our AI first syndicate free to join. So let's take a step back. I want to ask a tough question. According

to an MIT recent report that 95% of generative AI pilots are failing and if you look at the current market a lot of these current momentum feels like a circular momentum. Like for instance, the AI startup valuations are getting higher and higher because the hyperscalers are pouring

billions of dollars into these companies and at the same time these AI startups they spent most their capital to buy back the compute power from hyperscalers. So how do you see the current market and are we close to the AGI or are we close to the AI bubble? Well, that is a very tough question.

Thank you for asking. I I feel like there's a lot of concern about okay well where are we in terms of AI development and where are we in terms of AI adoption. Me being an AI every day in and out I feel like I'm inside a bit of a bubble because I'm a power AI user. I use all the foundation model

tools. I'm experimenting with AI note takers. I am trying out these AI videos, content creators,

tools. I'm experimenting with AI note takers. I am trying out these AI videos, content creators, and to me, and this is so exciting because I can see how it frees up my time so I could use my time on more strategic things that I enjoy that only I can make a move the needle on. So,

as an AI power user, I love it. And I'm also experimenting with like vibe coding with with cursor. But when I speak with people that are not in the AI industry or pay close attention

cursor. But when I speak with people that are not in the AI industry or pay close attention to AI and I do feel there's a bit of a delta in the rate of adoption and the extent of adoption. I I ask people around me whether it's my hairdresser or high school people because I

adoption. I I ask people around me whether it's my hairdresser or high school people because I volunteer to deliver financial literacy courses at high school. Just everyday people when I ask them like how do you think about AI? They're like well I've only heard about chat GBT. you know,

I use it sometimes, but I don't use it that much. So, there's a huge disconnect of where I think we are versus maybe people outside of the AI circle are actually are. But, nonetheless, I'm I'm very excited about this because right now we've done so much research. There's a a very mature foundation

architecture like transformers like RLHF, they're largely established. So now, you know, we're we're building on top of this. There's a lot there's a lot more interest in exploring AI, investing in AI. but in terms of the AGI specifically, I don't think we're close to it in in a general meaningful

AI. but in terms of the AGI specifically, I don't think we're close to it in in a general meaningful sense, the current systems are very powerful. but they don't have that genuine reasoning, planning, and world modeling capabilities. For example, if we think about autonomous driving zero to five. So

zero is like fully, you know, driver has a full control at of the vehicle at all time. That's

like zero automation. Then there's the automation level one driver assistant and there's a partial driving automation. So Waymo is there. It's full automating driving experience. But I think in

driving automation. So Waymo is there. It's full automating driving experience. But I think in terms of AI, we're probably between one and two right now. We're not fully at the full autonomous stage, but that stage is coming for sure. And then maybe on the tough part of your question of AI

bubble, I will have to think about it because I definitely have been reading articles of, you know, discussing AI bubble. I talked about it with our team. but there's also real revenue from from a lot of the companies that we back. So we're still formulating how we think about this. Yeah,

exactly. I think you have to be a power user to understand where we actually are and spot which AI products are going to have the same powers versus the ones that just a brief fats. And if you ask a general public what AI is, you might get a 10,000 answers, right? Like AI as a chatboard or AI is

just a image generation tool or it's a automation tool. I don't think AI is going to stop at chat box which is quite a prevalent UIUX at the moment. I think it's just a starting point. So I want to know from your perspective as a power user what are signals when you're using those AI products

what features or UIUX tell you that this might be a great product and or even this might be a good investment opportunity. Personally, I feel like maybe from a usage versus like investment perspective. From a usage perspective, let's just use like the image editing app. If you go to the

perspective. From a usage perspective, let's just use like the image editing app. If you go to the app store, there's a ton of beauty cam, camera for, I don't know, taking putting your baby in

like a cute pumpkin, stuff like that. Well, so as a user, if you're if your whole purpose is to just have a a cute Halloween costume for your baby, that simple tool fulfills your need and you're okay with that. as an investor, I think when we're looking at AI tools, I think the top three things

that we look for is I think for example, one, the data mode because all of these foundation models, they basically exhausted all the public publicly available data out there and I think being able to

access and train small medium size LLM or train your own model based on proprietary data that will give give you defensibility against these big LMS. so having some sort of data mode and having that continuous access to that specific type of data will reinforce your mode over time. So

that's definitely number one. number two related to number one is some sort of niche industry that you're in. Like for example, we've seen Harvey AI and Lora AI. These are AI tools for law firms.

you're in. Like for example, we've seen Harvey AI and Lora AI. These are AI tools for law firms. if you look into what they're doing is they are basically building their defensibility around an industry that is highly sensitive to confidentiality IP. so you know building

something for a very niche industry that makes it harder for open AI or claude to touch or anthropic to touch is very important. as well as like partnership you have if you have some sort of proprietary partnership with Nvidia or partnership with Intel partnership with whatever industry that

you're going going after and the key players of that industry that give you additional mode over other players. If I'm using an AI tool, I hate to download a new app and open it every time I

other players. If I'm using an AI tool, I hate to download a new app and open it every time I need to use it. I would love for it to be already integrated as part of my workflow. For example,

HubSpot have this AI agent or even Gmail has these little these little AI capabilities. I'm like,

okay, this is perfect. I don't need to form a new habit. I don't need to open an extra something.

So, you know, I think these four points are the ones that I feel will build the defensibility and that's something we're really excited about as a VC. And I think the metric will prove that you have a product market fit anyway. If you build something people want to use that solves a person's painoint, people will be coming over be like, "Please take my money. I want to use your

tool." So when we see that natural pull versus push from the from the startup, I think there's

tool." So when we see that natural pull versus push from the from the startup, I think there's something there. Yeah, I can totally resonate with that. If you can build an imperfect MVP and people

something there. Yeah, I can totally resonate with that. If you can build an imperfect MVP and people are still begging to pay for it, I think that's definitely early strong signal of a product market fit. And there has been this idea floating around that the speed is the startup's only mode. Do you

fit. And there has been this idea floating around that the speed is the startup's only mode. Do you

think that's true that in order to win this AI race that startups need to fail fast and build fast? Yeah, I definitely support like being the first mover in a way because let's just use us as

fast? Yeah, I definitely support like being the first mover in a way because let's just use us as an example. as a as a busy minded professional, we do not have the time to evaluate 10 different

an example. as a as a busy minded professional, we do not have the time to evaluate 10 different tools and and decide on the final one or two to try. So for startups, they need to build super fast, they need to go to market super fast, and they need to scream as loud as they can to get

your attention because it's all about attention economy. Like which ones get onto your radar?

you're probably going to think about that one when you are looking for a product or service to try.

So being the first to to market definitely has a huge advantage. Then it's important that you don't screw up and then you retain your customers otherwise you will just be accumulating lessons for for the second mover in in the category. And to me, I feel that the current market

really prioritizes and give those who can tell a good story the advantage. For example, a16z, they just launched this new media fellowship program where they're designing this 8week program to give founders, builders, operators an opportunity to really hone in on their narrative skills because

that's super important in this day and age to to help capture your users's attention to to build your own community, to build your brand, to create legitimacy. because all of those are important for people to give you a chance to try out your product. So for sure company need to ship fast but

also build their narrative, build that stickiness to get people's attention. Yeah, totally.

Attention is everything. Whoever wins attention often wins a community. So I want to know from your perspective should startups focusing on building the product first or it's actually better to build a community first then create a product for that community. I think maybe just

picking on the community part. I think community is a lot more important in web3 versus non-web3.

I think for startups, you just need to experiment with different just experiment and test out your product ideas with a small subset of friends and family. do some initial go to market on X or Blue Sky, which by the way, Blue Sky is our portfolio company. you know, get some eyeballs,

get some attention, get a weight list going. so I think founders should definitely do that. For

web 3 founders, find a community that you want to plug into. For example, if you're building games, probably want to choose an ecosystem that is very friendly and mature for gaming. So, Salana is a very good layer one for people building games. So, yeah, I think I think both both are important for for different types of founders. Yeah, in terms of community building, I don't think you need a

100K followers to build a community. I really love the idea that you only need 100 true fans to give you the initial feedback and the momentum so that you can continue to iterate from there. So

if you look at the current economy, do you think it's actually tougher for startups to raise fund or do you think it's much easier because now we have all these AI tools and resources. So the

other day I was interviewing another YC company. Both founders, they were first-time founders. They

had no idea how to raise fund. So they created a little tool to simulate a dozen VCs to help them sharpen their pictures and finally they closed a seed-round. So do you think these AI tools actually made startup fundraising much easier nowadays? I think it's both easy and hard. I

think the easy part what I mean by that is there's a lot of interest in investing in AI startups.

So there's a lot of capital available that want to deploy for the right product. So it's definitely easier than before cuz everyone's ready and a lot of people are capitalized. I think it's harder for startups to raise capital now because the bar is a lot higher. because it's easier given

the maturity and tools and resources available to build something really fast. But then if everyone can move very fast, then you're competing on who can move the faster or the fastest and who's building the product the best or better than the other ones and who's getting that early product

market fit signal. So I think the bar just moves up a lot higher. And not to mention maybe just to go back to your earlier question on AI bubble. AI bubble is something that's been talked about right now and as venture capital professionals everyone's thinking about it and therefore they

are probably a little bit more cautious. So we've all seen these headlines in the news of companies raising $200 million on a billion valuation. So it sounds a little too frothy, sounds a little too rich. So a lot of VCs are trying to think about okay well we want to we still want to make the

rich. So a lot of VCs are trying to think about okay well we want to we still want to make the returns for LPS and therefore maybe we want to consider startups that are not raising on a high narrative on a super expensive valuation. So I think the bar for startups to build to a point to

differentiate themselves is higher and to compete for capital at a reasonable valuation is is also higher. So you know it's a it's a bit of a mix. Yeah, initial capital definitely helps startups

higher. So you know it's a it's a bit of a mix. Yeah, initial capital definitely helps startups accelerate both product building and the go to market. I want to know your take on web 3.

Do you think web 3 can be a meaningful channel for AI startups to build community or even raise capital because I've seen there are quite a few AI startups recently pivot into web 3 and AI projects. Yeah, that's very interesting. I think it depends on what you're building and does it

projects. Yeah, that's very interesting. I think it depends on what you're building and does it make sense to tap into web 3. I think when we make our whole web 3 and AI investments whether it's Sahara AI or zero gravity or hyperbolic we're looking at like does it make sense to have a web

3 angle and if they have a token like what what's the utility of the token in in this type of model we don't you don't want to say or explore a web through your token for the sake of potentially broadening your maybe marketing outreach or building community. I think it has to make

actual business sense. so like how I see AI interact with blockchain is for example decentralized infrastructure. we invest in a company called Hyperbolic. it's like a

decentralized infrastructure. we invest in a company called Hyperbolic. it's like a decentralized GPU marketplace. allowing people to tap into the networks of GPU providers that are scatter scattered in different places and and so it's a little bit more agile use case for founders

that may not be able to pay AWS or or other GPU providers a fixed upfront fee but leveraging this decentralized resource in a in a smart and agile way that works for everybody. As an example,

we mentioned earlier about the the importance of data and being able to collect data that are a little bit more unique. So, how do you incentivize people to give you that data? you could introduce like a tokconomics mechanism using token as a way to solve the cold start problem. You find a

way to pull in suppliers of data. people want to sell their data in exchange for a token and then you can collect these data and sell it to vendors who want to purchase these data for their model training for example and they're willing to pay the token. So these are two examples of you know why why it makes sense to have a decentralized infrastructure or having a token play because

it has a utility in the business model. There are definitely a lot of web 3 projects trying to combine those two hottest narratives like web 3 and AI but it's actually crucial to distinguish which ones actually deliver the real utility and value and I agree that for startups you shouldn't

raise capital just for the sake of raising capital because nowadays it's much easier to ship products unless you are motor company you don't need a massive GPUs to build a working MVP.

Yeah, I think people can definitely use leverage the AI resources to now tinkering with different cool ideas. I think the the maybe going back to like the mo and defensibility, if you just have

cool ideas. I think the the maybe going back to like the mo and defensibility, if you just have a feature, I worry that might be easily easily killed at the next model update of any of these big foundation players. But I think it's wonderful that we have the opportunity and the tool set now

to test out these thoughts and and and tinker with products. I think for vertical SaaS companies, there's a huge advantage where you can move fast by building on top of existing foundation models.

But you should also think a few steps ahead like what will the foundation models look like in the next six months. So that you are not building a product just for now. you're building a futurep proof product because these model companies they continuously integrate those features you need to make sure when the next generation models release your product is not going to instantly become

obsolete. Yeah. like you know how OpenAI hired over 100 iBankers to train or to to work on their

obsolete. Yeah. like you know how OpenAI hired over 100 iBankers to train or to to work on their finance product. I I feel like we really need to anticipate maybe the next moves of these big model

finance product. I I feel like we really need to anticipate maybe the next moves of these big model players. So for founders who want to be a part of the AI first fund, what's the best way to get

players. So for founders who want to be a part of the AI first fund, what's the best way to get involved and which stage of AI startups are you most interested in supporting? Yeah, absolutely.

we're focused on seed to series A stage at the moment and I would love to get in touch with any founders that are building in AI whether you're building AI infrastructure you're building AI tool sets you're building AI for lawyers doctors accountants you know AI for knowledge workers

if you're building at this exciting time I would love to get in touch you're very welcome to find me on LinkedIn yeah we've covered from the mindset of great founders to the edge of the new AI frontier. It's been an incredible deep dive today. Thank you so much Sophia. Yeah,

thank you Boyuan for having me. It's been a lot of fun chatting with you on a lot of thing that's been on my mind and appreciate you sharing the sharing the thoughts and and sharing ideas with everyone. I hope you enjoyed Today's episode. Please like, subscribe, and share with your friends. Every follow will help us make better podcasts. I see you in the next episode.

your friends. Every follow will help us make better podcasts. I see you in the next episode.