Are Young Founders Set Up To Fail? | YC's Requests For Startups | "Frontier" Not "Foundation" Model

The irony is we're going back to the future, back to the days where the government was a really big funder and early customer of innovative entrepreneurial technology products. It wasn't always the big primes. In some ways, it feels like we're going back to the future.

It's the week of July 27th. Welcome to the Learning Corner of Precursor Ventures Podcast. Welcome to the Learning Corner by Precursor. I'm Mia Farnam, and I'm joined by our GP and managing partner Charles Hudson, where we will be talking through the articles and good reads that caught caught our team's attention over the past week. We've also got an awesome guest joining us this week, Alina, one of our incredible interns who's been here for almost 6 months. We're so excited for you to join us here.

Thanks for having me. I'm so excited for the discussion.

Awesome. Well, lots to talk about this week. Our first good read this week comes from Ashley Smith at VermilionCliffs Ventures, riffing off a TechCrunch conversation she had on backing founders under 20 years old. She argues today's young founders face a different kind of pressure than the last cycle, since AI-era expectations want a working growth engine within months, not just years. And the patience that used to come with early capital has mostly disappeared. She's also blunt about the rise of predatory incubators that hand out small checks for outsized equity stinks, warning that a young founder can end up with a cap table too damaged to raise a real next round before they've proven anything. Ashley points to Precursor Ventures' trajectory as a case in point, a genuine outlier that too many funds and founders are now treating as the baseline expectation. And her closing argument is that investors owe young founders far more honesty about how normal failure is at the earliest stages, even when that honesty is very uncomfortable to deliver. We've talked about young founders on the podcast many times, and I think we've talked about it from the standpoint of separating being a founder and being willing to take risk with just being young. I think Ashley touches on a lot of different aspects that we haven't blown out in much detail, like the predatory incubator aspect, making sure you're a little bit more diligent on the difference between success and failure early on. I think one great quote from Ashley when she compares cycles of this prior to today was, quote, capital is available. But the patience that used to come with it at the earliest stages isn't. So Charles wanted to pause here first to see if you had any specific reactions to the overall piece and, and not the quote specifically.

I agree with Ashley, like there isn't a lot of patience. I worry a lot more with these younger founders about IQ, but not about IQ, but about EQ and like this managerial journey they're going to go on just like every other first-time founder we work with. I think right now though, it's not just that there's not patience, it's also that the expectations around how quickly good things can compound in this market have never been higher. And I also couple this with the fact that, you know, we've talked about this internally. We go through cycles where at times dropping out is a high status thing to do because as Aoshi mentioned, she's of the generation where finishing college was the thing that you did. So dropping out was like a high risk thing to do. Now I feel like particularly if you're at a top 10 school, It's so difficult to get into those schools that there's a bunch of students who feel that dropping out has become a low-risk activity, which to me means it's low signal. It's funny, I mean, I graduated from college in 2000 and I had a ton of friends in '99 who dropped out of school to do internet startups. Most of those companies weren't successful, but many of those people discovered that entrepreneurship was going to be a big part of their future career. And my expectation is that the results here will be pretty similar. Many of these companies will be unsuccessful, but many people will realize that being a startup founder or being in actual employment is something they want for the rest of their career, and learning that early has value.

I think one of the things that I thought about the most after reading the article was I have no issues with young founders flooding the space, being excited, and bringing all that entrepreneurial energy into the ecosystem. That's Fantastic for us, for our jobs. It's fantastic for the ecosystem. I think it just calls into play the question of, is the infrastructure laid out to help support young founders who maybe didn't have all the mentorship or, I don't know, truthful conversations about what are some predatory things that could happen to you along the way that you might build. I mean, it is similar to first-time founders too. A lot of founders don't know that there are specific predatory terms that exist for some incubators versus others. There are some that have different signals than others do. I don't know if necessarily every first-time founder has that information, and I would say the same is about young founders. So I think it's more of, are the rails ready to help support this flood of new founders, new energy into the system in the way that I hope that they are?

I think one thing I think about a lot is I've been fortunate enough to hear a lot of interviews with Dylan, the CEO of Figma. And I just think about all the people, like notably John Lilly, who were super experienced, not just as entrepreneurs and founders, but as people who understood how to navigate the Valley and fundraising ecosystem. And my, my sense is that having those people around the company was really, really helpful in its development. And I just wonder who's advising, supporting, and helping this next generation of really young founders figure all of these things out. Cause as much as things have changed with AI, a lot of things about company building, you gotta find great people, you've gotta be a great leader, you've gotta figure out how to scale yourself. Those, those principles haven't changed.

And I think on that note of truthful conversations, Ashley also argues that investors have this responsibility to be honest with young founders earlier sometimes than might feel comfortable about the failure rates and what the hard parts of being a founder actually look like. I'm curious, how do you be honest about that without getting in the way of that fearlessness that can make young founders so special?

I think one of the challenges for a lot of young first-time founders is like knowing when the answer is to change ideas versus to stick with it. And I think, um, as you mentioned, Alina, a lot of the young founders I meet have what feels like boundless energy. And so like the capacity to work isn't the issue, but not every problem is an effort problem. Some problems are, hey, you're just pointed in the wrong direction. What you really need to do is actually take a step back. And like look at whether the thing that you're building makes sense, whether it's still a good idea, whether it's still worthy of your time. And I think this sort of unbridled, unbounded amount of time that young founders have sometimes means they just keep going when maybe the answer is to actually press pause.

I think Ashley had a really good quote here and she said, for a young founder who hasn't watched peers go through the harder parts yet, that silence creates a distorted map. Not knowing what to be afraid of can be a strength, but so can the myth that if you build something great, it will succeed. Great products fail every day, and she said a different word, but bad products make it also. I mean, it's so true. And I think, again, we talk about— we're talking about this a lot. There's a lot of parallels between first-time founders and young founders of just, you're only privy to the things that you see in media. And the headlines that come through. And so if you truly believe, if I bring a good product to market and it works, then people will come and it will be the one that people grasp. That we've seen so many examples of that not working. Bad products hit the market every day. Like, that's just not something that's new.

You also just made me think about one other thing that's different is I think a lot of young entrepreneurs I met back in the day had very little contact with startup formation or fundraising. And now with the proliferation of on-campus scout programs and VCs mining college campuses, trying to find the next generation of entrepreneurs, I've met many more people who, they're kind of wise beyond their years in the sense that they've like, are familiar with how cap tables work. They know what a SAFE is. They have some sense of who the good invest, like they're just more familiar with the ecosystem and like that part of it. But that's a very different part of the process than actually building a company.

Our next good read is Y Combinator's Fall 2026 Request for Startups, where YC partners and outside voices lay out the problems that they want founders to build for. This batch centered on AI moving into the physical world, spanning categories from multiplayer AI tools to compute at sea, to even rebuilding trust online as deepfakes get cheaper and more convincing. The one that stood out the most to me was the request from the sitting US Secretary of the Army, which seems to be the first time that a government official has contributed to the list. Calling for founders to build low-cost interceptors, next-gen sensors, and hardware that plugs into the Army's new open system architecture. Across the list, YC seems to be betting that the next generation of durable companies gets built at the intersection of AI and the physical world, not purely software. I'm curious, Charles, uh, it seems like the Army Secretary putting out this request for defense solutions via YC and going straight to startups instead of traditional procurement is something different and maybe big. I'm curious, what do you think this implies for the startup ecosystem?

It's funny. I always find the YC requests for startups to be an interesting list for two reasons. One, it usually reflects some flavor of sci-fi. And I was like, when I read about the always-on AI tutor, I was like, that's a Neal Stephenson reference, as they mentioned. But so is the compute at sea. I don't know if anybody's read Cryptonomicon. I was like, seasteaders, like these are, these are ideas. That have been talked about in the past in science fiction. I read the list and I was like, these all seem like near-term, high-probability things that can be solved with tech today. And when I read this, I was like, these are problems that if you're already working on, you should probably be excited that there's an audience for. But if you take this as a list of things you should start, my hunch is you're probably too late. And if I think about many of the great companies that I've met through YC in the whatever, over a decade I've been going to Demo Day. Many of them were pioneering new categories, were taking advantage of things that had just become feasible but weren't yet well understood to be big ideas. The irony is we're going back to the future, back to the days where the government was a really big funder and early customer of innovative entrepreneurial technology products. It wasn't always the big primes. There was a lot of overlap between the government and Silicon Valley, and it brought us a lot of really important innovations. And in some ways it feels like we're going back to the future.

Aline, I know we were talking about this before you even started recording, but if we have, you know, like a government official making a request for startups, will that change potentially what the procurement cycle looks like for them? And again, hand in hand with the conversation that we just had earlier, the age of YC founders continues to drop every year. So what does that look like when you have younger folks going through, you know, government procurement cycles in a way that they haven't prior. Again, I think part of this is wonderful marketing and making sure that they get really great people to, or notable people to come up and walk through what the requests look like. But also it, it makes you question like, oh, what would this look like if it actually came into play? Um, I mean, we've seen young people start coming into the government. We had Doge happen. So that's not like the first time something like this has occurred, but it does make me wonder, how do you actually put this in play? And I think Alina, you had a good comment earlier of Do you think this will be the beginning of newer cameos from tour leaders of different industries making an entrance into the Request for Startups, YC batch to kind of ask and call for new ideas that can help them push the needle? And then I wonder always if that's going to help them make the decision on, do you think they'll buy or build whatever comes from, do you think they'll use it as inspiration or do you think they really will partner with all the folks who are building in this? YC batch. To my excitement, the YC list also signals that consumer is back. I have not seen that come up in batches in many cycles. Their argument is that every platform shift mints new consumer giants, yet 3 years in, and I think ChatGPT is likely the only big app that's hit most people's iPhones since things have hit the market, even when it's getting cheaper to build in consumer year over year. Precursor Ventures Ventures, like, we remain bullish on consumer even when so much of venture capital was running away from the industry. What, Charles, do you think other investors are seeing now, or maybe signals of that they hadn't seen before?

It's pretty funny. I was talking to one of our longest-tenured LPs. He's been with us since the very beginning, and he said, I think it's really interesting that you have companies in all of these categories that are doing well. And I said, well, it's I don't tend to think in terms of categories. I think in terms of founders and ideas. I don't think we've ever had a year where consumer has been less than 20% of what we do. And we've had years where it's been half. As great as AI harnesses and inference and compute are, there's nothing quite like finding a consumer experience that really touches people's lives. Like, uh, I've made a few investments lately in consumer that will hopefully come out soon. That I think are tackling like really important problems in really novel ways. I do think in some ways consumers always had this problem of differentiation and moat. Most things you could build in consumer are relatively lightweight, simple experiences. The, the real knack is like, how do you find an audience that wants them? And I think about things like Twitter, not a, not a, in the beginning, not a technically complicated product, just happened. To really strike a nerve for people. Instagram, Snapchat, a lot of these products really like tap the nerve. And so I'm glad to see that YC's open for business for consumer startups again. And I will say this, I tend to see certain categories like the AI-powered parent assistant apps. I've probably met 10 of them, all started by really great people that are all really smart, but all kind of have the same idea.

Yeah.

And so I think we will probably see hyperconcentration in some of the obvious categories to consumer, and I hope that that ends up producing some really great experiences for end users.

And going off that line, I think we've had conversations about that, Charles, of what are subcategories where maybe we don't need tech as more, maybe tech isn't the solution to maybe helping folks find a better way. I think we also talked about that a lot with You know, uh, parental controls on tech with kids, like it, not like I, I don't think that there's going to be something that wins there. I just think that the consumer itself has to evolve to get comfortable with, you know, how do they see AI or tech sitting in their day-to-day lives within family.

I just look at like Tin Can. I look at all the people who've been like trying to get kids off of screens and trying to get, and I was like, if you think of the screen, not just as an entertainment experience for a young person, but also as a communication experience, it becomes very clear that like, if you're gonna take away the entertainment experience, you have to like replace it with something that allows them to communicate and talk to each other. And it turns out that a landline phone is awfully good at that, particularly if the people in question mostly are at home. And I think like not everything has to be some LLM-powered, agentic solution. Sometimes it's just like looking at existing problems and thinking about more clever ways to solve them., and I hope that this next wave of consumer companies also includes some of that energy too.

I think especially since people have such a desire for IRL, and we see that with all these like connection economy type of apps that are coming out, it'll be interesting to see how there's that competing desire for in-person connection and then also the height of what consumers can get from AI. Um, you know, where those things overlap and where they diverge.

And I think the one question that I ask consumer founders who are building in those specific spaces are, how do you keep the magic coming up every single time? I think there are so many different apps and new things that come up that the first time you use it, it's the most magical, fun experience. You blast it off to 15 of your friends. You're like, you have to use this with me. And then by use 4, 5, 6, the magic wears off and your friends have certainly abandoned it. But if you can instill something that continues that thereafter, that's what, that's what creates the long-term durability. And I think that's a really difficult thing to start. So to your point, Charles, it's not that it's easy, but it's lightweight to build something that has like a, a simple, um, a solve to a personal consumer-facing problem. Very difficult task to make that defensible over years, which is why companies like X, Instagram, just, they are truly giants in that industry because they keep people coming back.

And Charles, I'm curious if you were writing some sort of Precursor request for startups, what would be on that? Is there anything that you wish more founders were working on that you're not seeing enough of right now?

It's so funny. I feel like all of my co-investor friends are always like, what are you looking for? And like, I honestly don't know. I only know what's gonna pique my interest when I see it. It's a terrible answer, Alina. To tell people when they're like, what are you looking for? You're like, I don't know. But that's like the most honest answer. And every time I've tried to write down, these are the things I think I want to see, it usually ends up being like, oh, I've seen a bunch of bad versions of thing X. If somebody else— I ended up funding a team that's doing something in the screen time limiting usage space. And I've seen some hardware solutions. I've seen some software solutions and none of them felt quite right. And I thought the team that we ended up backing is certainly doing something different. I don't know if it will work, but they have like a different conceptualization of like what the problem is. And I'm, I'm like, like genuinely curious to see how it works. And I kind of feel the same way about a lot of categories where I'm like, oh, I've seen 8 or 10 versions of why this doesn't work. If someone could show me a version that does, I'm all ears. You made an investment in this company, Bluecore Energy. And a bunch of my friends are like, oh, you guys are doing energy now? I'm like, not really. I made this investment at this company because I like this founder's specific idea. It's not that I now have a general thesis on energy. And I think so many people are, people want you to have like some overarching global thesis of how the world works. And I'm just like, I just like follow the people. I try to find the best people I can who are working on the most interesting problems and like work with them on it. And then I learn and Then you learn about these other adjacent problems. And so, I don't know, guess that's why I've never had one.

Cool. Well, our last Goodread piece is by Chris Peck called Frontier, and it explores why AI labs stopped calling their models foundation models and started calling them frontier models. He says that a foundation implies something durable that you can build on top of and collect rent from forever, while a frontier is something that you can only hold temporarily and have to re-earn with every single new release at rising cost. The piece argues that labs are essentially betting that demand for AI capability is unbounded, so the frontier never closes and a good enough competitor never catches up to squeeze their margins. It draws this comparison to AOL, which was once inseparable from the internet itself until competition caught up with it and distribution advantage disappeared. The author points to agentic coding as a current battleground where falling prices and strong open weight models are already chipping away at the profit pool that labs are counting on. Um, Charles, if the frontier never closes the way that labs are betting that it will, then does that make them a better or a worse venture bet than a company that is selling something with a durable moat sooner?

I think we're in sort of an interesting moment in the cycle. I think for a while everyone was like, oh, the frontier labs are going to win everything. We're all gonna be paying Anthropic and OpenAI and maybe Google Gemini forever. We're gonna be like constantly paying this tax and rent to them because they have this interesting bundle of like LLM capabilities and inference and training and like we're gonna be stuck on their infrastructure. It's gonna be the next AWS. Now I think we're in a moment where people are saying, oh, that's like the old, like, that's like driving a Ferrari to the grocery store analogy. Why would I want to use some high-end model to convert a PowerPoint to a spreadsheet? Shouldn't I use an open-source model? I think the answer is like, people are gonna get smarter about this stuff now that cost is a part of the equation. And now that cost is a part of the equation, people are gonna say, well, do I need to use the fanciest, most sophisticated models? The right answer isn't actually an open-source model. It's a cheaper version or an older version of one of the hosted models, cuz they work fine. Unless you're an open source purist. And I, I really do think that if I were at one of the big frontier model companies, my belief would have to be that there will always be new hard to solve problems in AI. Will there be an audience where only we have the resources to tackle this problem? It could be biology, it could be physics, it could be math. I would say coding, we're approaching the point where like maybe you don't need the most sophisticated model in every case. But there are lots of domains where that hasn't happened. And so if I were trying to make the bull case for the frontier models, I would just say there will always be a new category, vertical, or area of exploration where a frontier model gives you the best answer. And if I were an open source zealot, I'd be like, over time, that window of things where you need the most fancy frontier model is going to shrink. And it's not going to shrink to nothing. It's just gonna shrink to the point where people will use cheaper or open source models to do more things. And I'd argue that like history has shown me, at least if you think about AWS, you would've said, well, eventually like people will start hosting their own infrastructure again. A few companies have, but in many cases you're like, oh, like the world that AWS enabled became its own ecosystem, which became even more complex. And so at some point you're like, well, it's not just like the hosting, it's the hosting and the routing and the bandwidth. And the load balancing and like to replicate all of that stuff, the problem becomes bigger and bigger over time. I suspect that AI is gonna have the same trajectory where you're like, well yeah, I could have, I could run a local version of this on my computer or I could tap this cheaper model, but what about inference, training, memory, all the other pieces that go around it? And I'm not sure that we're quite ready to unbundle. If you think about what OpenAI and Anthropic offer, really it's a bundle, that old Jim Barksdale, there's only two ways to make money, bundling and unbundling. I think the unbundling phase feels premature.

Yeah, I agree. I, I do think that for frontier models to continue to keep their speak, I think two things need to be true. There needs to be continuous, like, unbounded demand, and they need to continuously deliver, like, a higher tier of capability. And the people who use our product need to care enough to what you're saying, Charles. Like, if the average person— I'm only talking about D2C, but B2B is a completely different conversation. If I'm only talking about D2C, basically, do you need that higher tier capability to do the task that you set it out to do? And if the answer is no, then like, do you need to continue to pay, I don't know, $20 to $200 a month for it? Or could you seek capabilities elsewhere?

And look, we, we're probably in the like $5 Uber era of open. And I think it's interesting, like in a lot of ways everyone's like, oh, well, like Uber is so expensive. And like, but it's become an ingrained habit. And people use it and I use it in all kinds of use cases where I would've never used it before. Maybe not as happily as I did when it was cheap. And I have a feeling that the, the reason I think it's probably too early to talk about unbundling is outside of power users, most people are, have barely started to scratch the surface of what this technology could do. And the only people who I think have are the ones who are pushing it to the limit. And those are the people who are experiencing both the pain and cost of being on the frontier. And I think it's natural for those people to say, well, now that I've used these a lot, it's funny. I started running a bunch of my own evals on my own work and I have downgraded a bunch of the models I used to use cuz I'm like, I can get almost similar performance at a fraction of the price using a different model. But that's like mostly out of my own curiosity and desire to keep our AI expenditures at a level that I'm willing to pay for. But if I weren't using it as intensely, it wouldn't be an issue. It really wouldn't.

Well, we are all outta time for today. Thank you, Alina, for jumping in this week, and thank you, Charles, for joining in as well. Please like and subscribe, share with friends. We will talk with you all next week. Thank you.