Sarah Guo on neo labs, AI trading, storytelling, and what she's watching heading into 2026

What's up, guys?

How you doing?

I'm feeling good. I'm feeling the spirit.

I have to take my ear off so I can get the [laughter]

Yeah, it's a mess. Um,

those look very natural on you.

Yes, I know. They really do. I I I said earlier before the show started, John's more Santa coated. I'm more elf coated. I don't know. It's just [laughter]

Anyways, you're looking like a lot more a lot more polished.

I like the nails. They're like frosted nails.

There you go. We don't have skinare, but we've got nail art.

Okay, cool. Amazing.

Um, how did you process the storytelling uh the the storytelling cycle, news cycle yesterday? We asked pretty much everyone about storytelling. Uh, do do you did you did you I mean, I'm sure you saw the viral announcement that there's a lot of companies that are hiring people, storytellers. Is this just a rebranding? Is this a valuable thing? If a if a founder came to you and said, uh, uh, my my second hire is going to be a storyteller, what would you say?

I'd say that's your job.

Okay. [laughter] Yeah, actually.

Well, so then the question becomes, can like is it is it a skill set that you're kind of born with? You maybe start at an 80% and you can get to 100%. But if you start at like a 10% capability wise, like you should probably just focus on other things, right? You're maybe never going to be great at that. I I was classifying it as like it feels like there's like Joe Rogan CEOs, which are CEOs that could go on a three-hour podcast and deliver an amazing performance. some of which would be about the business, a lot of which would just be about, you know, random stuff. Um, but it feels like have you seen founders in your portfolio go from just okay to like truly excellent?

Uh, I think so. I'm an extreme growth mentality person, right? I'm like if you are smart and you are a high work ethic, like most skills are accessible to you. [snorts]

Not everybody can, you know, be TBPN, right? But you can go from like a D to a B+ and that does a lot for your company. But I don't think you can outsource it.

And so if you [clears throat] take this like storytelling idea and you just kind of disagregate it a little bit, you have you have like channels that have changed, right?

This matters and like traditional media, you know, it's it's having a time of it. It's struggling, right? Um and so if you think about the job like does anybody want to like start off as a junior PR consultant today?

It's a hard job, right? Um, I'm not saying there's not room for that, but the traditional like lowgrade version of that that was I'm going to call a bunch of journalists. I'm going to repeat the pitch from the company. I have no point of view. I don't really have storytelling or taste and I'm a I'm a C. Like, I think there's not a lot of room for that, right? Because if you think about it from the reporter's perspective, even if you think about where the attention actually is, you could be like a midsize AI influencer

and have as much like uh followership as a large media publication that's relevant today, right? And so I think

and weirdly like the hardest job on our team is Nick who's somewhere around here and that is actually just like getting inundated with pitches from from various PR teams, companies, agencies, etc. and he has like functionally he has to say no to 95% of them just because we don't have there's not there's no time in the show to do as many as many as are coming in. So

yeah, I think there are a lot of people now where um they like why do they call it storyteller instead of like content person or PR or whatever

or just advertising specialist

or advertising specialists like uh the meta has changed, right? It's saying like we need people with taste. um we're chasing the moment. There's a lot of noise and we have to like create some signal even if we want the traditional channels.

Um and so some of it is skill set, some of it is the story about like what the job is.

Um but I I do think a lot of it is is hard to outsource and you can get better. We run a grant program um which is like an uncapped note for 10 companies twice a year called embed and the highest rated session for the program is a storytelling session. Interesting.

Right. So you come and you pitch your company and it like it goes from like a D minus to an A+ in terms of where people start.

But like people get a lot better in the span of a week before they do demo day. And so I definitely

it is rel it is relatively like structured, right? It's it's not like if you're making a film, it's not like hey I need to reinvent the hero's journey. Like the structure works and I think there's the same type of approach you can apply to storytelling about your company. I I would have this I would get sent a deck and I'm like your idea is cool and you're great. your deck is terrible. And I would just send a Figma file that is like here's the deck template. Here's like an example for each like six examples for each slide. Just like completely redo it and then it and then it's automatically like a few times better.

But you have amazing intuition for this now. Did you start great at it?

No, I made six years old. No, no, no, no. Yeah, it's definitely uh Yeah. I don't I I think I started I've I've I've been sort of n naturally I've studied advertising since I was a kid. not directly but just like understanding like why do I want this product so much more than this product right and that started with like skateboarding and snowboarding stuff right and so I like fully understood okay this this brand it's like their event athlete strategy the design this collaboration that they did how you know but uh but again like it was very very refined to get to that ultimate structure that I have in Figma that I can just be like this is the structure that that has worked for me

and it's worked for a bunch of other companies It's like tried and true. This structure even just my personal one which I haven't shared broadly has like directly contributed to raising like hundreds of millions of dollars, right? Um just across the years. So um I don't know. I think you can systematize it.

Yeah, there's art and there's alpha in it, but it's certainly something you can practice.

Totally.

Um and I think people have discovered that like you need to be good at this now. Yeah, I think that's

Do you think people

Well, the other the other thing I've I've noticed with like company storytelling is like

if it is truly

and people take you more seriously if you do it in an elf costume.

Yeah. Of course. Of course. Uh no, but the the thing that I've always found is like if it's really difficult to get the story out in a cohesive way, then your whole strategy is just

effed. Like it's just bad, right? And if it comes out very naturally, it means that like

there's alignment, the strategy makes sense. But if you're like if if making the deck takes, you know, 50 hours, it's like you probably just don't have that great of a story in the first place. Like the actual facts of the situation are not that great.

Yeah, this is so true for me. So the version of it is at some point I actually thought I wanted to be a writer. Like thank God I'm not. Um I find writing incredibly painful, but I still have to write every week for my work as an investor. Right.

Like memos, LP updates.

Yeah. Yeah. Mostly memos. And and like the problem is writing is thinking, right? storytelling thinking, right? How do I actually communicate the strategy of the business and do I believe in it?

And the difference between like

I'm procrastinating, I'm avoiding writing this memo for six days in a row because I got to talk to my partners about it is I actually understand the strategy of the company or I don't. Yeah.

And sometimes it's because like I just don't know enough, right? we make our first defense investment and I'm like, well, I I need to call like 45 people to understand what I don't understand about this yet in terms of the procurement process or whatever else. Like,

how important is electronic warfare? Like, how quickly is this cycle going to go? Um, but uh

but some part isn't knowledge, it's just like does it make sense at all? Yeah.

And I have learned to trust that signal of like, well, if I'm not missing information and it's still like not super cohesive, it's not cuz I'm an idiot because I have to go tell this story to recruits. and to the media um and to the next investor. And if it's hard, they're not going to get it either, right? And so I think there's substance in that.

Yeah. I think I think part of the at least backlash or virality around the idea of

still so funny

to me.

We almost did Santa. We almost did Santa for for uh for Evan from Snap on on Monday. And

for it, I not the first time. [laughter] It's the first time you're meeting someone. Like can you really go full Santa mode? Um but uh yeah I mean part of the reason there was like some backlash is I think people are worried about startups leaning into like over rotating to all style no substance all storytelling no plot like you have to forget that like without some plot points you can't tell a story around it if you haven't built anything you haven't raised you know you didn't raise a funding round like a lot of the stories get get told around we built this thing we satisfied this customer we hired these people

you also need you also need tension you need you need some like adversity Right. So, we'll have founders come on and we're like, "Aren't is this not

like the most hyper competitive category?"

And they'll be like, "No, it's not." And then like not give like the followup as to why it is. And it's like it's okay if it's competitive if you're going to win.

Uh and you might be the one to win, but you it's better to just admit like it's competitive and like here's what here's what the market is like missing.

Uh sometimes people don't want to share whatever their alpha is, but

yeah. So, I um I've like learned a little bit more about I'm just going to go ahead and say marketing, advertising, storytelling, but like that category of things that has existed for a little while over time, but I'm still like

I taught marketing at Wharton, but I'm I'm like at my core.

Wow. She didn't just study no studied storytelling. I studied a little bit. Um but [laughter] uh but like I'm a product absolutist to your point of I actually had just this like pretty full contact debate with a founder in our portfolio very recently about um how much of uh like the success of a company is its ability to raise money, get to momentum, look like it's winning to customers, look like it's winning to hires and go faster than others. Yeah. And I'm like, well, you know, there's some it it depends on the space, right? And uh if you're in take an example like like late 2010 SAS,

the degree of freedom you have for like what you're going to go build in any given category. Not to denigrate the product work. It's just like you can only do so much without fundamental technology change if you have certain integration points and certain workflows. Totally. And you can have amazing execution there. But I'm like, you know what? you probably want to do invest in people who understand the workflow, have amazing execution, and are total animals about like storytelling and momentum and all of that, right? Um, and that doesn't mean there's no substance. It's just like

Well, there there's like there's not as much creativity.

That doesn't sound like 2010 advice, though. That sounds like advice today.

No, no, no. Okay. So, today I would argue actually like

is it different?

Yes. Okay. Um, so

there's more white I mean 20 late 2010 SAS there was no there it was struggling to find whites space. I think like if you if you look at

the things you build were not that different.

Well, yeah. The 2021 era, you saw this like it was crypto and fintech, sure there was a lot of SAS, but like even with fintech,

there wasn't there w there was some new infrastructure with fintech that made it easier, but yet that you were still competing with companies that started five, six years ago. And so that was that was just really rough. And so I think the difference today is like we have super powerful LLMs, we have machines that can think. And so maybe there's like a little bit more like kind of random white space and categories that you can

uh find and and go digging in.

I I was talking to um an investor friend yesterday about this and he like we were discussing whether or not um you just want to invest in the smartest people you could within some constraints of like they can convince other people to come to their cause and and such. And I'm like I actually think this is a much better strategy now than it was in the late 2010s 2020 period. Yeah. because you're like

well the r the variance on what you can build is really wide right and that's both information and like ability creativity people are trying to you know they're trying to build very different products even in software and it depends on like what you think you can do with models

or [clears throat] robots or hardware whatever it is right and I think it's just like broader and if if you can do more different things with product it matters more so substance matters more than ever

yeah storytelling too

if you believe in the midwit meme then you should invest in the smartest founders, but also the dumbest founders because who knows, they might lock into the same thing. No, I don't think

I think I think that I think I think that works. I think that works in CPG actually. I think it works in CPG. I've seen some people on on both sides do very very well and in the middle it's like innovating on four to four to five different dimen dimensions. It's like way too complicated uh and and tends to uh

anyway uh tell me what is a Neolab? Okay. So, there's this category of company that has been um you know a handful of these things have been funded this year that are quite controversial in the investing community which is essentially like

we're going to raise a solid amount of money let's say you know $50 hundred million plus and uh we are going to invest in largecale AI research and training upfront.

Yeah.

Um

so not a rapper company

uh not a rapper company though we can come back to that. Um, and I think this is like one of the reasons it's so interesting is because you had this dominant narrative for a lot of the last couple years of just like how could you compete with Google and OpenAI.

Yeah.

And maybe people threw anthropic in there.

Yeah. Yeah. And then that was the first and it's notable because you have you have if you're if you're

trying to build like a new LLM or maybe have some new ideas around that type of product and people are already saying like

oh open AI can't compete with Google because Google has all this cash flow and even though they have like almost a billion weekly activives like they're burning so much money and then and then you get a new company a neolab that's like oh yeah we're also going to compete and we don't have any users and we have a fraction of the money and so that's why I think it's been like almost a narrative violation if you're going to like or or at least like contrarian to say like no we actually are going to bet even more into this

point of clarification uh thinking machines SSI anthropic are those neolabs or are the neolabs post those labs

I think the definition would be post those labs but that was the first generation of this because they also fought that narrative um like and opening I did before them totally right um and so I think this is like this big open question in venture investing like it it opens a can of worms of like

well how much money do you need to go do venture investing at conviction I got to say like 14 times like we would argue not that much necessarily if you're very early but um

no it changes the economics completely to like growth fund you need to be able to build a position because it used to be you could take a flyer for two you know get get 10 company for a couple million bucks and if that's just not going to happen on day one it's just the economics just don't works, right?

Yeah.

So, so is that changing your fund or have you figured out a way? How are you thinking?

It's it's not for us. Um and you know, we have supported some new research lab efforts, but um I I think that's also a tactical thing of how early you are. Um and um but but I think the the broader thing that is really interesting is this narrative violation or question of uh is scale the only thing that matters? And um and like if you're going to spend the GDP of a medium-sized country on training, can anybody keep up with you? Um or

even even if so, Greg had a video that came out yesterday or this morning where uh at OpenAI and he was just saying uh we've tried everything but scale and scale is the thing that works, right? And so again, it's it's very contrarian to be like, well, no, we actually can come up with you're basically if you're raising $100 million to compete

Yeah. with these other labs. You're saying like we don't need scale just because you're never going to be able to compete on scale when these anthropic and open AI are running away in the capital markets.

You know, the I think one of the more nuanced arguments about these things has been uh let's say a bunch of people can go raise hundreds of millions to billions of dollars to you know hundred billion dollar plus which is a wild claim to begin with. Um, even so, like where you spend that money matters, not just size, right? Um, and so like one of the things that's happened with open- source is that there are uh large pre-trained models, some Chinese, some European, now hopefully more and more American ones, um, that people work with and are really powerful. Like the big joke that Silicon Valley runs on Qu

and they're like distilling them. um they're distilling them and and they're also I think if you're a research lab you tend to be post-training or just trying entirely different architectures and and like so people are the argument I think goes something along the lines of you know more is better eventually right like I do need the GDP of Japan [laughter]

in compute but on day one um Ilia Sutsker basically said this like in the age of research we need some amount of money to just validate our ideas

but it's not a trillion dollars

yeah I I And his interview I took away. He's like kind of like we're going to do couple million here, couple million here, couple million here. Maybe

it's a little bigger than that. But

yeah. Okay. Okay. But but the idea is like he's not just like we're oneshotting we're going to try to oneshot this with 60% of the dollars that we've raised. It's much more again experimental research driven.

Yeah. And and then I think the second part of the argument of like you know we're we're doing actually new things versus fighting symmetric warfare with Ant or OpenAI which is like seem seemingly a bad idea in Google you know money machine

um is that uh if you have these pre-trained models like we can just focus if you have the GDP of Japan and you have the GDP of Germany and you know I have some other equivalent country California right

yeah [laughter]

let's give it up for California

let's go California we love you

total failure of governance, an accident of history, but it's the best place in the [laughter] world.

We're still cooking.

Still cooking. Still cooking. Um, thank you, uh, semiconductors. Um but uh but if everybody has some huge amount of money and I put it all toward post-training or self-improvement or diffusion or SMS or some other bet and you're like serving inference and also doing pre-training and whatever else. Yeah. Then like my effort on the thing that matters might be bigger than your effort even if you have more money overall. That's the story.

Let me pitch you on a fund thesis and I want you to push back on it potentially or agree with it. Um, so I I have a

Wait, tell me now and then let me do it. Tell

the world. I don't know. Uh, you know, let's say I have like, you know, a billion dollar fund. I'm going to put a hund00 million in 10 of these NEO labs. One, you know, $100 million each. They're all going to go and try different ideas. And my thinking is, you know, 10% chance one of them goes really, really big. Um, but but I but I'm also underwriting these on talent acquisitions. Yeah. And I say that, hey, for, you know, I'm not going to lose money on basically any of these, even though I'm paying crazy prices at seed for just a couple people with barely an idea. Um, I'm going to get a lot of money out because the because the big hyperscalers are going to continue to acquire. Yeah. Are you telling me I think that party's over or do you think no, there's going to be an endless stream of billion-dollar acquisitions coming out of the hyperscalers?

Yeah. It's like shoot for the moon. Even if you miss, you'll land in in Satcha's arms. [laughter] Yes. Is that true?

Um I do think that is part of the calculus that's happening here where the distribution of outcomes is like

how you know without naming names but how many people want to hire Ilia or Mir Maratti or Barrett or any of these folks that have started new labs like a lot of people that could have

um that think that winning an AI race is very valuable to them.

Yeah. So, so, so you think that the the hyperscalers definitely aren't uh like exhausted of these like lab tuckins, lab tuckins, lab.

Well, and I think my point of view is like as long as you have a bunch of people spending something in the range of a hundred billion dollars a year and they believe that by spending a billion dollars to acquire a team that they can make that

money go further, there's a good chance that they will. It's always been

if there's if there's a again if there's like a pullback and new information then you start to run the calculus of like

well we're actually cutting spend and we've figured out what we need to figure out and just adding five new smart people to the team is not going to give us that much of an edge in this market. Mhm.

Yeah. I think this is exactly right where um one of the craziest reframings of this to me made by somebody at a large lab with a a lot of money to spend is like just imagine Jordy as a researcher and like his GPU budget for experimentation is $100 million personally. Yeah.

Right.

Um

and I have to give that to him all the time. Right. So he works four years $400 million.

Yeah. Yeah. Yeah. Otherwise, you start looking around. You say

else.

Well, and and the question is like, well, then how much am I willing to pay for Jordy or the talent that is

1% better, 10% more likely to like make the thing work? Actually, a lot of money, right? Because it like the way the um I'm not saying that you should start your growth fund with this risk calculus. It takes a, you know, certain um

uh strength of gut, right? Um, but it's it's not irrational to be like these players have already committed to spend x billion dollars of compute over the next years and they will look at the people as leverage on that compute.

Totally.

And so like I don't I mean lots of things can happen in the macro and people can change their minds and some of these like contracts are um on performance, right? Like they're not like fully solid but the way it looks now I'd be like I think people are going to keep spending money here. So, you're probably safe with your fund.

Amazing. No financial advice.

If the fund pans out, I'm going big on Christmas. What should I give to all of the founders in my portfolio who have made me so much money? What how do you think about gifting? Is there a top gift recommendation for this Christmas season? Is there anything that you can share about top gifts?

Yes.

Uh is there anything you can think about?

Look at Yeah. Look at the Christmas theme on this, too. We got Santa Claus. You know, this is a gift to advertising. I learned recently that Coke made the uh modern

We fact checked this. It's not

They solidified it. No, they solidified it. They solidified it. They like they basically like the whole stick.

They didn't create the meta, but they they owned it.

They owned it. They story told around it.

Yeah,

they story told around it over substance.

So, so gift guide or maybe like a book recommendation that you could put under the tree if you I I always like giving books. Uh what what do you have? What do you think?

No, I'm a I'm a big uh books person. Um, I want people to give me time and books and nobody seems to be able to give me time. So, um,

that's deep.

Whoa, that's crazy.

Yeah, my my husband hates this. He's like, "What do you want for Christmas? Do you want like uh stuff? Do you want jewelry?" I'm like, "No, I want time." And he's like, "You a terrible answer.

Maybe you should get you a luxury watch on getbasel.com, our sponsor. What's better? The only way you can get time is if you get a watch on getbazzle.com." Of course. What are you what are you uh how are you processing

the ad

how are you processing the current moment and like what are you what are you looking out for in the next 12 months?

Okay.

It feels like it's hard to think that we could go things could get crazier.

I disagree.

Can I give one book recommendation and then we'll like talk about 26? Um okay only one. Um no I'm going to give you three. So I think uh I think Dan Wang's book Breakneck on China and the US was really good. It's worth reading. It's like a little stylized but I think like really substantive actually just good storytelling I suppose. Um there's a book on immua like I think a lot of people are suddenly like interested in biotech.

Okay.

I'm not saying that's bad. I think biotech is amazing and it will get a lot more impactful and like more efficient and but uh the book is called um I think it's called for blood and money

but it's about uh

great title. Yeah. I'm interested.

Yeah. No, it's it's sexy. I think

it doesn't sound like a textbook.

Yeah. as hot as and interesting as a book can be about a non-technical founder of like a breakout leukemia drug, like this is worth reading.

Okay. Very cool. Very cool.

Um and then um if anybody Okay, I mean this is a lot of tech nerds in this audience. So if anybody hasn't read Masters of Doom.

Okay.

Um this this is a story of like the like the

Carmarmac, right?

Best era of the video game industry. Carmarmac worth reading.

All right. These are great.

Um

Masters of Doom is great.

Okay. Actually, I have one more niche one. There's also the Prince of Persia book by Stripe Press. That's a good one in that Masters of Doom.

No, I haven't read it. No, I'm excited.

That's sort of like similar similar theme.

Okay, I'm going to screw up this title because it's

Speaking of storytelling, I mean the the story of a individual video game is particularly great because it's a tech company, but it's always hard to tell the story of a full tech company that just keeps going and going and going. And the video game companies, they do, but you can also tell just the story of that one game.

Yeah. Like multi-year project.

Exactly. No one's like, "Oh, yeah, we needed to talk about like, you know, the story of Figma in 2018 specifically." Like it's like this whole it's such a broader arc.

Yeah.

Anyway, other book recommendations or we go in 26 predictions.

Okay. Maybe last niche one.

Yeah.

But we could do the rest of the episode this way. Um the uh there's a book I I'm really interested in like what environments make for good ideas,

right? And like consistently good ideas because

part of like extreme growth mentality is you think environment matters. like ah

how do we like help people create the environment where they're doing their best work?

Um there is a book called Apprentice to Genius.

Um and it's about uh a series of researchers at NIH and John's Hopkins that had like just huge contributions to biology, chemistry, science for a while, like way more than you should have in just a small lab over like decades. Yeah.

And um

was it the funue of the building or something? What about the environment? Were it obviously like the academic environment?

Yeah, I think it's a I think

even a couple key people that were kind of like around during that period.

It is definitely about um a couple key people, but the question is like how come their students also are like Nobel Prize winners, right? Because it's hard to imbue the next person with like magical creativity.

Let's go win a Nobel Prize. Here's the playbook.

Yeah, but like three times in a row, man. Um and so I think

just look at this deck. Study this deck. [laughter] Yeah.

Study this Figma file for an hour and you'll be

It's actually an online course now.

Yeah.

Nobel Prize for $1,000

for a grand

with with uh cloner payments attached. So,

how did I get this Lambo? I won a Nobel Prize [laughter] and then they sold a course on how to win a Nobel Prize.

The course is $9.99.

Yeah. Yeah.

Yeah. How are you what are you looking forward to this next year? What are you What are you kind of wary of? Um, I think all the other investors should like go away and leave AI alone.

Okay.

Yeah. And then it's over everyone. It's over. Sorry. The demand.

Look at the Oracle stock price. That's That's what's happening here.

Clean.

No. If I can't move the markets that way. Uh, [laughter] no. I uh

Okay. Here. I I have really high conviction that like,

you know, 26 is going to be a year of a bunch of new applications of AI. And that doesn't mean that there won't be like market panic or um some bumps. Of course, people like claim it's winter, but uh I think there's now so much proof that the models are powerful in a bunch of different domains.

Um and it starts with ones that maybe are not like so obviously commercial. Um you know, reneech aside, people are like, what can you do with math? Right?

[laughter]

I feel like, oh, here's my prediction for 26 that you can like come back to me on is um uh I think somebody out there is going to make a lot of money, like hundreds of millions of dollars on AI trading,

like it's not happening at that scale yet, but people are certainly going to try and I I see no reason that would

But but but do you think that would happen on Wall Street? Will there just be like a pod that is like training models or is that like a startup in Silicon Valley because they're very different cultures?

I I would love to see it be a startup. It would be fun if it was our team, but I think it's gonna be who knows.

I think people are already trying it in lots of different environments.

I have talked to some I've talked to some high frequency traders who are just like, "Yeah, I worked at I worked at Meta and like Meta's better at machine learning than we are." I don't know if that's like holds across the entire industry, but I do think that like the the the adtech industry might be better even though the high frequency trading industry is like a little sexier and a little bit like more discreet.

Yeah. One reason you might see a startup do it is just because the labs have been and just like the west coast has been recruiting from some of these

trading firms on the east coast

and so they already have the experience and then they like hey we made something that's pretty smart maybe maybe

and we're seeing like I I think NASDAQ announced like 24-hour trading and there is a lot of volatility

just in general. So

yeah,

last question for me. Do you uh we we've been wrestling with this debate over whether uh some of these AI tools become like video games. We talked to the founder of Sunno and it's like and Midjourney is similar where there's a lot of people that are using Midjourney as a proumer tool. There's other people that are using Midjourney just as like a game like they have an idea for a funny image they make it and then they're satisfied when they get that image that they created. I think there's a lot of people that are creating songs on Sununo and then just enjoying them themselves and they're not actually sharing them anywhere. Yeah. Uh do you think that like the how do you how are you think about the nature of like Gen AI interfacing with like proumer maybe new consumer flows this like gamification? Is any of that resonating from you or any of the portfolio companies?

Yeah. Okay. So one thing I have as I think it's inspiring as an idea is that um most people are much more creative than they get to express in daily life right if you are uh a creative of any type like people have more ideas and points of view than they have tools and skills to go execute

and so if AI allows people in images and music and video I don't know if it's a good thing in writing with slot but or or code even right like if you can make things at a much higher level.

Yeah.

Then you're like you're going to enjoy making things and making more things, right? Like I you know what's the like saying inside every adult is an artist like because who doesn't draw?

I feel like some people just don't know what the process

like people that get viewed as like creative people are often people that just take two very different ideas and put them into one idea or take something that's working and make it different. And I think once people I've seen people in their head kind of like unlock that concept and then they realize like oh I can be creative because like here's a product that I like. What if I made it healthy

here, you know?

But did they make Santa or did they clear what if we made a soda and then marketed it with by creating Santa Claus?

And we're back to storytelling.

I guess uh one question uh maybe fun prediction. How much do you how many dollars do you think will be spent on ads within LLMs? Mhm.

Next year,

do you think next year?

I think it's a little early.

Little early.

That's that's my feeling. Like I could see it ramping in the back half of the year and like actually because I think the ads are going to be extremely in effective. People are saying like, "Oh, a lot of these queries are not that high intent. It's just research."

But I think that a lot more of them I I think a meaningful amount are product research and once you're driving these sort of like high intent clicks,

I think it will scale pretty quickly. But I think it will take I think it will I think this will be like a second half of the year 2026 story as it starts to scale. But we've had founders on the show that have said like when somebody lands on my site from Chad GBT they can convert at 7% and my average conversion rate is like 3%. So it's like

I think people don't quite um I think when it happens it's going to happen really fast. And the question is just like how quickly can the companies with a bunch of traffic build any kind of product because the inventory is going to be so good. And I think it's useful to look at um open evidence as a like precursor here. So it has been reported that the company went from like 2 to 150 million of ad run rate already.

Oh yeah it's fully ad supported. This is LLM for doctor ad business in AI today.

Right. I forgot about that.

And like what happened? They made a product that was free that was so useful to doctors because you're literally answering the question of like, well, I don't know what's happening in this area of like personalized medicine for this domain of oncology for this type of patient, right? I'm from Wisconsin, so we'll use Wisconsin as an example. Like,

uh, I am a doctor in rural Wisconsin that's never seen this type of leukemia, right? Um, so I'm going to ask Open Evidence for help. Guess what? Like, that's a really high intent

Yeah. query to go serve ads against here's a treatment. Yeah. Um and and so I think when you compare that to the ad inventory that exists in the world which is just like um look at a bunch of random stuff. My average query is two words or whatever like I I think that

and there's some there's some amount of there's there's a level of trust that I think someone will have with open evidence that they maybe don't have with Google search.

Totally. Yeah. They understand what the sources are. Yeah.

It's it's you know research trials etc. It's generated data from other doctors. So I think um I think that if you can make the right products here it's going to be very big. The my only question is like how how quickly they make those ad products.

Last last question are you long or short forward deployed engineer like the model?

Um I think it is a uh symptom of how quickly AI is happening. And so like if the long is people are gonna pay for a lot of FTE next year, huge bull, right? Um

so maybe more like 2027, 2028.

2028 I think like in the long run like you should deliver that as a product and people are going to,

right? I don't I I don't think um uh I think FD happens when people don't have a transition happens too quickly or they don't have the talent or the ability organizationally to do like change management the corporate term but it's real right it's really hard to like

completely transform uh an asset manager or a big public company and so like that's what you're paying for in 26 and 27 and maybe 28 but I think eventually it's going to be product.

Yep. Well, thank you so much for coming and fantastic. Merry Christmas.

Merry Christmas.

Hit the gong. Hit the gong for conviction.

For conviction, all the progress you've made. And we will bring in Doug Olaflin.

The president of semi analysis is coming on TVPN. Uh while we bring him in from the reream waiting room, let me tell you about graphite.dev code review for the age of AI. Graphite helps teams on GitHub ship higher quality software