Lightcone: Consumer is back, What’s getting funded now, The vibes immaculate

It feels like there's more energy around this batch than there has been for as long as I can remember for any YC batch. Like, what do you think's happening? There's a platform shift, and this is the moment where every single SAS dollar in the world is up for grabs again. The batch 3x ARR in 3 months, which is pretty cool. That's a great growth rate.

Yeah, it's a fun time to build. It's the best time ever! I mean, as a builder, it's like the technology just does such a different thing than what you expected before.

[Music]

Welcome back to another episode of The Light Cone. We are four group partners at Y Combinator, and we’ve funded hundreds of companies, many dozens of which have gone on to become unicorns. This is Jared. I'm Gary. This is Harge, and this is Diana. We just finished the Winter 2024 batch of Y Combinator, and it feels really, really different, doesn’t it?

It does! Let's talk about how this batch is so different from the batches that we funded in the past. Some of it is you actually need to know where we've been and where we are right now in order to actually figure out where we're going to go. A lot of people watching right now are trying to figure it out. Like, how do I go to where the hockey puck is going? How do I get there before everyone else?

The best way to figure that out is what happened in this batch. We're going to connect the dots with actual numbers that I don't think we’ve ever shared before. By stats of batches from four years ago, and contrast them with the numbers for this patch so you can see the actual trends, the way they're playing out here in the center of Silicon Valley.

Yeah, I'm curious. Like, what are some of the trends that we've seen? What's made this last Winter 24 batch different from previous batches?

Well, the strongest trend, the one that everyone is writing about, is AI. That's definitely like the big mega trend in the past year.

Yeah, I was surprised when we were pulling up some of the numbers. It's just under 70% of the ideas are AI, 70% of the batch— that's wild! Yeah, it’s about 170 companies. Yep, versus Winter 20, we only had 8% of the companies.

Maybe one of the notables from Winter 20 that you worked with, Jared, was Replicate. Yeah, those Replicate founders, they were into AI before it was cool, which was really awesome because they got to then ride this wave. But the first three years of Replicate was slow going because there weren’t a lot of people working on AI. They were building tools for people working on AI. There weren't many customers, and we didn’t call it AI as much back then. We called it machine learning.

Totally! Let me say one thing I've noticed that's different about this batch is consumer ideas. They are certainly coming back. Like, I've gone from working, it feels like for many years working with zero consumer startups. Now, just even in the group of companies I'm working with, several of them, I've noticed founders who are pivoting during the batch are pivoting into consumer ideas, whereas previously I think they would have pivoted into like an Enterprise B2B SAS idea.

And I'm not sure what to think about that, actually. Do you think it's bad or good?

I'm not sure. Okay, here's Harge and I have opposing viewpoints on this question. Okay, I’ll make the case for why it may be bad. Like, I think you could argue that pivoting into consumer ideas is sort of lazy because so many of the canonical tarpit ideas are these bad consumer ideas, where it's like travel planning, or splitting the bill at a restaurant, or finding a roommate, or all these kinds of things, right?

And people gravitate to them because it's so easy. You're like, I just want to build. It's the advice to build something you want, which is great advice. But it means that you can often build these like very easy ideas, and it's really hard to get lots of users for them. Whereas when it felt like people didn’t want to work on consumer ideas, oh, okay. I actually have to go out and become like an expert on something. Like, I have to go out and figure out how expense management works, and see if there are any interesting ideas there. It led to lots of really, really good startups being funded.

So I partly worry that people will pivot into tarpit consumer ideas because it's easy. My perspective is I find it so refreshing that consumer is back because when YC got started, when we all did it in the 2005 to 2012 era, there were tons of consumer companies. In fact, the first YC batch was like 80% plus consumer companies. What happened is all the consumer ideas basically got done, and there were no good consumer ideas left, or very few.

And so we went through this whole super cycle where the only non-tarpit ideas were B2B ideas. And the problem with B2B ideas is they’re a little boring. Let’s be honest. Like B2B SAS is like a little boring. If you think about what like Drew and like young technical founders—like the original founders in the early YC batches—they were building consumer apps because they were fun.

Both you and Gary, when you went through the batch, had consumer companies, right?

Yeah! Yeah, we both started consumer companies because those weren't tarpit ideas at the time. The problem was like in 2020 starting a US-based consumer idea was ill-advised. If you were doing it, you were probably working on a bad idea.

Well, the other fact too here was just Facebook sucking all the oxygen out of the room, right? Like, it felt to me like that era you guys are talking about, so 2007 through to 2010, maybe there was just lots of optimism around building consumer ideas.

That sounds familiar!

Yeah, right? But then it felt like for a period, it just felt like, hey, anything you build Facebook is going to like clone or crush, and it just seemed not exciting to get crushed by Facebook. You could argue that a foundational model might come along and crush that, but if you're working on a consumer idea, but I'm just kind of skeptical that that's true. There's a lot of white space in there to actually build real revenue.

What would your advice say--if a founder came to you, Gary, and said, hey, like, uh, somebody told me that I shouldn’t work on consumer ideas because Facebook's just going to crush anything I do?

Well, OpenAI! Well, OpenAI now! Open, open, OpenAI is the Facebook of--yeah, GPT-5 is coming out soon. So what should we tell founders? They come out to you, like, okay, should I really work on this?

I guess it's early enough that it hasn't happened yet. I mean, when I see Facebook actually come after Replica in the AI boyfriend-girlfriend space, then I'll sort of believe it. But some of it is the capability expands the ability of computers to sort of operate with human beings in such a broad way that they couldn’t possibly be in all the places.

Another big trend that we've seen is more developer tools. And Diana, you worked with a lot of those. Do you want to talk about why? What's happening with developer tools?

Yeah, in this batch we funded about 30% more dev tools than four years ago. This is like one of the largest dev tool batches we’ve had in recent years, and I think a couple reasons is I think it's attracting a lot of the super technical founders that want to build this future with AI. And before you build it, you need better tools, right?

It's kind of like this technology trend where you have like two phases. I think it comes from this book from the Industrial Revolution, from these economists. I think it predicts a lot how technology cycles happen. The first cycle is sort of where you're laying like the railroads, infrastructure, all the tooling before the installation and proliferation of apps.

There's a lot of this kind of tooling because even right now for building an AI app, there's so much plumbing you need to do and customize it. And right now, there are certain patterns that are emerging, like RAG and doing a lot of querying and indexing and getting results to be more accurate and fine-tuning. Those are not well-known patterns and everyone is building the same stuff to build the actual end application.

So we have a lot of founders that are like really good tool builders that are excited to kind of build the hammer. It would be really cool to see, at some point, all the way from distributor systems to evaluations to even as hardcore as probably at some point custom silicon.

Like, we could probably see the next Nvidia being funded and gone through YC. One thing that's really interesting is I remember in 2010, when I first started working at YC, dev tools were not seen as a good idea to work on because people didn't think they would ever make money.

It was only when Docker started taking off, then MongoDB started taking off; there was sort of this era where it was like, oh, like, dev tools are things you can actually work on well for open source companies in particular because it was very unclear at that time that open source projects could actually be successful companies. Like Red Hat was like the only example at that time.

Why? And so I think this is a great--like, dev tools is just a great example of the build something you want for yourself. And if you're an engineer, you can just be very self-indulgent and build like the tools that you want, and there's actually a business there.

Well, in that respect, dev tools are basically B2B SAS but consumer-style. So you only have 20 million people who you have to actually market to, and you actually have to market to them in the same way you would to for a consumer product. But instead of a billion consumers, you're talking about 20 million developers.

And the cool thing is most of those 20 million at some point are on Hacker News, which is a YC website. It's like the New York Times for hackers. That's right.

There's a lot of parallels because consumer ideas are not judged on how much money they make, typically; consumer social in particular, right? Like it’s all just like growth and daily active users and monthly active users.

And what I've noticed about open source--same. Yeah, Diana, how do investors judge whether an open source startup in the batch is doing well or not?

Some of the early signs are whether this is getting adopted by the tastemakers. And that happens to be in a GitHub project. You have a lot of GitHub star growth, and also if you have actual like hardcore developers that are good using you and early signs of getting into production with companies.

At this point, like consumer early in the early days for infrastructure and dev tools, you don't really make money because things like installing a new database is such a big bet for a company. You need to make sure that it's battle-tested.

So a lot of open source companies take a little bit longer to monetize, kind of like consumer, where it's all about user growth, right? And the second thing I would say, um, is ultimately open source companies win when they really have the developer mind share.

It's sort of like Facebook won with the network effects by capturing a lot of the users. Like, I don't know. It's like a third or half of the world uses Facebook. Same thing for dev tools; it's like if anyone thinks of, let's say, building a full-stack application and easy deployment, they could think of Supabase.

I think they've done a really good job, and I think you worked with them, Jared. What was it at the beginning when they were doing the batch, and what does it look like now?

Well, the cool thing about Supabase is literally what got Supabase off the ground is Hacker News. They had built this open-source project that was an open-source competitor to what's it called? Google the Firebase? Firebase. Firebase, yeah, which is a YC company that got acquired by Google and then became a really big product inside of Google.

And they were building an open-source challenger to Firebase, and they built it. And like, how do you get users for something like that? Well, the thing you do is you launch it on Hacker News.

And so they had this blowout Hacker News launch, and to your point, Diana, it was clear, if you literally just read the comments on the Hacker News post, it's clear that really good developers are like, "This is exactly the thing that I wanted! Thank you for building it!" And that was what, like, launched them into the stratosphere.

And there’s a recent stat about the percentage of the batch that's using them. We'll have to pull the number, but it's like a third of the batch—73 companies—out of the 243 are using Supabase batch.

Yeah, that's crazy! Wild! Which is really going up against the big infrastructure clouds with AWS, GCP, right? Yeah, investors really pay attention to like the Hacker News launches of these dev tool companies.

See, that again, Supabase had a phenomenal round that they raised around demo day, and it was really all directly attributed to the Hacker post. Yes, this is a free alpha leak for all of you out there: you could basically take almost any close source dominant dev tool or platform and create an open-source version of it, and you might just kill the close source version of it.

And this is a shot again—you know, a shot across the bow of every dominant sort of dev tool or SAS platform. By the numbers in Winter 20, we only had five companies that were open-source dev tools, and in this current batch, we have 22. So that is like a 5x plus increase—22 open-source companies! That's a big shift.

And we've seen this over and over again. I mean, there’s sort of like the Slack and then Mattermost; there’s sort of the GitHub and then GitLab, which was a YC company.

Um, you know, in sort of analytics we have--and Ploog, exactly. So the other cool thing about this batch at color of getting a lot of founders getting excited to work on AI and dev tools, this is the most technical batch ever, right? It's like 99% of the companies have a technical founder.

In the current batch, versus just 88% during the pandemic. Let's talk about why that is. I mean, we talked about some of the driving factors from ideas, but I think there's a couple more things at play.

I think one thing that feels very different now, versus if we go back to the pandemic era, is I think there was this whole "the software eats the world" idea, which originated with a Marc Andreessen blog post, I can't remember, like maybe 2012, something around that—um, or a decade ago. Great essay!

Yeah, great essay, right? But I think what it boosted was this idea of, hey, not every business is going to be a core software company. It'll be like software reads the world. It'll be software that sort of enables non-software businesses to become software.

What do you mean by that? It’s like kind of more businesses that are operations-heavy. Sort of like, give examples!

Like Flexport, I think would be a great one, right? Like Flexport was, hey, there's this giant trade and freight brokerage business moving things around the world. So much of that is done manually with humans filling out forms.

And Flex was, hey, that could be automated. Why don’t we just have a software team that builds software to help the people who are managing like the freight and the brokerages do this more efficiently? It’s tech-enabled!

And I think like a consequence of that is that, especially with the SaaS era, where I just felt like there was lots of money available to fund ideas, that the profile of founders became a little bit more tilted towards like, can you do, do you have domain expertise in a non-software business? Like, are you like someone who's in the shipping industry who now wants to start like a tech company?

Some spectacular examples would be like WeWork, right? Like, probably the poster child of like tech-enabled, but I just think the profile of founders shifted a little bit away from like geeky engineer. Adam was not technical.

Yeah, right, exactly, right! He does not look like at all. And actually, to that point, this whole trend is like somewhat controversial about these tech-enabled businesses, and there are some that seem to be on the right side of the line where they actually were tech-enabled, like Flexport, which is working.

And there are some that were on the other side of the line where they claim to be tech-enabled, but really weren't, like WeWork, and those ones didn't go so well. But I think AI is a force in the complete opposite direction, right? Where it feels like if you want to work on a good AI idea, you need to be at the cutting edge of actual AI technology and tooling, which a table stake.

Yeah, right, because all this stuff is cutting edge. Which I think it gives a bit of, um, edge to a lot of founders that don't have baggage because everything is so new. All of the latest progress in AI is just like one couple of years old.

And this is one of the batches that also has shifted the median average age of the founders is also a bit younger. Here's another version of the story that I've heard told, which is that in the 2020 era, there hadn't been a technology platform shift in a long time. Venture capital funds had billions of dollars to deploy. They had to deploy it someplace. The best place to deploy it was these tech-enabled businesses that were going after industries and companies that didn’t really look like the traditional tech business that venture capital was set up to fund.

But there weren't a lot of great new tech opportunities because there wasn't a new platform shift. Now, there is. And so it's a much higher ROI use of those venture capital dollars to fund stuff like AI companies and stuff like WeWork.

Yeah, I think one of the interesting subtleties is in my head it's a little bit less about whether it's tech-enabled or not—that is certainly one frame, and a lot of VCs actually really stick to that. I mean, there are some really famous firms that famously only want to fund pure software businesses that are monthly or annual recurring revenue, and that's a whole strategy.

They're a bunch of those firms that we've all heard of that are our friends, like that's all they do, and then you know there are just as many who actually look at it and say, oh, actually I’m willing to do tech-enabled, but there are a lot fewer of those people. And then I think the real subtlety—I’m sort of a little bit more in the latter camp—because what really matters is actually the gross margin.

So if you look at a paler, for instance, you can have a 90% gross margin or 80% gross margin type of tech-enabled, you know, quote-unquote almost consulting business. But if your gross margins are extremely high then people are actually, you know, willing to give you good multiples, and you're actually able to raise money at a reasonable valuation.

I think what's funny is when we give the t-shirts at Y Combinator, like when you come to YC, you get a t-shirt that says, "Something people want." Yeah, there you go, there you go, because it's the end of the batch. What's funny is I noticed none of this mentions anything about whether you're tech-enabled or not, whether you're a software business or not, or even gross margin—it's just purely a function of if you make something people actually want, people are going to pay for it, and then the rest is just sort of details.

I mean, you can actually look at today, the biggest YC companies by, let’s say, they've gone public—Airbnb, DoorDash—it's not clear that like on the surface at least, like the technology is what sets those apart. So much. As it was for Airbnb like, you know, it's a website, but the core thing is building like a network and a reputation system.

Most powerful network effects ever!

Yeah, right? DoorDash and Instacart are arguably more logistics companies than true tech companies. They're the best example actually of maybe this—the tech-enabled label is a tricky one because those are actually probably technology companies. But you could one lens you could put on them and say they're tech-enabled.

But they're also two of like the biggest companies we've ever funded, but they come from a different era when a lot of the rest of the world was still coming up online, right? That's what it feels like to me. If we talk about trends, it just seems to me to your point, Jared, that there was a period around 2006, 2007. It was just pure software businesses. Then it was like, hey, software is going to be bigger than just pure software. And we got kind of DoorDash and Instacart and these interesting businesses, and then it really maybe it pushed too far where there was like WeWork.

Yeah, WeWork! Where it's like, hey, there’s no software! There’s no software here at all really, right?

And that’s what it feels like has been a big reset, which AI has sort of—it's almost like AI has taken us back to that start where it’s like, okay, actually, we just want to fund things that—what’s interesting is like there’s good software opportunities again. Like we ran out of them; that’s why the venture capital dollars shifted to the WeWork stuff, and now they're back.

It's a platform reset!

We’re so back, guys!

So, what have we funded less of this batch? What have we seen move in the opposite direction? Well, we funded a lot less stuff going after local markets. So in the 2015 to 2020 era, YC and just like the world in general funded a ton of companies that were basically the second wave of all these online to offline things.

The first wave was like DoorDash and Instacart in the US, and then the second wave was like, well, what about DoorDash if you're in Brazil? They want their food delivered too. What about Instacart in India? And so there was a whole wave of taking these models and copying them in international markets—fintech too, right?

There was like—in the first wave primarily in the US—like Chime for example. But a lot of fintech businesses are actually local because regulations are so different in each country. And so then there’s a second wave of like the international copies of all the US-based things.

Yep, definitely during that period, like the 2020 to 2022 period, we were funding a lot of international teams that were like Robinhood for LATAM or like a local crypto exchange or yeah, like DoorDash for X market—lots of these kinds of things.

Yeah, and a lot of those were really good. YC funded some amazing epic companies, like Monzo, which banks some ridiculous percentage of the people in the UK; Grow in India, which is Robinhood for India, is doing phenomenally well.

Zepto, which is the fastest-growing YC company of all time, which does 10-minute grocery delivery in India. I think that's a really interesting one, actually, because it's like, it’s very easy to say Zepto is like Instacart for India, but it’s not quite right because their actual model is different.

It is, yeah.

By the numbers specifically around the 2020 era, Winter 20 batch, only about 45% of the batch was international, and now it’s only 20! Yeah, this is the most US-centric batch we’ve had in a long time. Most of the teams when they applied are in the Bay Area, like so about 29%, which interestingly means like San Francisco is definitely back.

So we looked at the numbers, and pre-COVID around 29% or so of the companies were in the Bay Area when they applied to YC, and it was half of that during pandemic, right? Went down to like 14%, something like that, and now we’re back up to where we were before. So, you know, even higher actually!

Yeah, even higher than we were pre-COVID. I think because there’s so much about Silicon Valley, all the AI progress happening here. I think the fundamental reason—it’s not to be clear that we woke up one morning and we were like, like, we got to fund more US founders! That was not what happened.

What actually happened, I think, is basically the best founders chase the best opportunities, and YC funds the best founders. And so like, what are the best opportunities? Well, in 2020 there were amazing opportunities to take models working in the US and launch them in other countries.

And so amazing founders, like Aude from Zepto, that's what they worked on, and now the best ideas are like that trend has sort of run its course now and like most of those opportunities have been done. And so the best founders have had to move on to other opportunities.

I'll give you stats for that specifically. Like Winter 24, we have four times less marketplace ideas than Winter 20. If we're seeing more consumer and dev tool ideas, it also makes sense because those aren't local at all.

They're exact opposite of your point. Like actually if you want to build the best AI dev tool, yeah, there’s no such thing as an AI dev tool for Brazil. It’s just like use the same ones.

Yeah, right?

Another thing that we're finding a lot less of now is crypto. And here’s something that the audience might not know about the two of you, which is that Gary and Harge are two of the most successful crypto investors of all time. Like literally, you two were the first investors in Coinbase, and you made literally billions of dollars investing in crypto—billions of dollars, right?

Some of it is going back to being around Y Combinator, reading Hacker News, and finding out about this thing called Bitcoin—reading the Satoshi Nakamoto white paper and just saying like, well, what is this? You know, Mt. Gox—like Magic: The Gathering Online Exchange website in Japan—having to, you know, do some weird wire to, you know, some sketchy country on Western Union in order to get money into this weird website that would sell Bitcoin.

And having that experience be so bad, like I remember doing that and thinking, well, this is a very interesting idea. And then, again, if very smart technical people on Hacker News are doing this and believe this might happen, well this might just be a thing.

So as two of the top crypto investors, what happened? Why were there no crypto companies in Winter 24?

I mean, I was looking for them. So I think what’s really interesting about this—zooming out—is if you talk about when Coinbase applied to YC and we funded them, there was— it was a very counter-culture idea. Like, you had to be like, Gary, like really into this stuff. Um, there was no hype around crypto. It was not seen as a very fundable thing.

And then what happened is crypto will go on these bull runs, and when there's like a—which is basically really the price of Bitcoin—so when the price of Bitcoin tends to go on these sort of like euphoric pumps, anytime that happens it brings lots of people into crypto. And I think at YC, what we would see is like it would bring in lots of people applying with crypto ideas, right?

And then that’s clearly what was happening during sort of the COVID era. Crypto had this huge run; Coinbase went public, and we sort of surged in crypto applications. What I think is really interesting about this current moment is we’re going through—we’re in the middle of another crypto bull run, like Bitcoin just hit another all-time high recently, but we have not seen a surge in crypto applications.

Fascinating!

Yeah, right? And it’s like it’s clearly because all of those minds want to work on something else, and we know what the something else is. It’s like it’s AI. Like I think AI has just dominated the mind share of engineers whereas previously Bitcoin hitting a new all-time high—Don, don’t you have a story about when you went to MIT last year that you could still see some of the remnants of that like crypto mind share at the college level?

Yeah, so MIT has some of the smartest kids in engineering, and what was really interesting to me were a bunch of kids that dropped out. And UND dropped out, and they dropped out to start crypto companies. They raised like millions of dollars, and they thought they were like on top of the universe, like high flyers.

And then what happened? Then around that time, things crashed, right? And things stopped working. So they were at the top of the mountain—they were like hot shots! We dropped out of MIT; we just raised $5 million! We’re going to be the next Mark Zuckerberg for crypto!

And then I talked to them, and they were back in school like normal kids. But there was a bit of kind of like shame on their shoulders; it was embarrassing actually to come back to school.

It was not seen as a badge of honor to drop out of school. And then when I asked them what they wanted to do next, it's like, oh, I just want to finish school, and then once I finish, I’ll figure out another startup.

The failure of the crypto startups gave all startups a bad name rather than just like crypto scams, which is the actual problem because they didn't want to work at a startup anymore afterward. Some of them were like, there were like, oh, I don’t know yet.

I think there’s like another undercurrent behind all of this because it can be very jarring. You have this kind of very bipolar experience from going to the top, you’re like getting investors to throw money at you. You have this—an on Twitter account that has like hundreds of thousands of followers doing—you bought your board ape. Yeah, you got all the board apes, a collection of them, and you're running this like giant exchange, and then things just crash.

You get sued by the government—that could be another case! And then what's your plan B? And then you come back down, hit ground floor so deep, and it can be very demotivating.

The government stuff you mentioned is actually for on a very crypto-specific topic is a huge thing where like the US has chosen this regulatory regulation by enforcement approach, which is just incredibly scary if you want to do anything interesting in crypto.

Is this casting a chilling effect because people are worried that if they’re successful they could literally go to jail?

Yeah, exactly that. I mean, Diana and I have a company that we just worked with in this batch where the founders had previously—they were young, smart technical founders who had previously started a crypto exchange and were sued by the government. And they are clearly still traumatized by that experience, right?

So I think it's a real shame because in a way, this is like a great time actually to work on a crypto idea because at a high level, it’s like, hey, programmable money! And now we’ve got like AI agents that can like do lots of things autonomously.

And the tourists are gone!

Yeah, right! Like this is actually a great time! But I think in the US, at least, until it feels safer to build these companies, it’s going to be hard for crypto to recapture that imagination. But like, I still think by far away the big reason is just AI is like the exciting thing to work on, which actually I do see—I know it’s a bit of a meme of former crypto founders going into AI.

My hope is actually that for a lot of these crypto founders that went through this ride, that they kind of get back up on their feet and get back to building because it's actually fun. I think the sad thing is some of them really got defeated, and my hope is that they get that optimism back again because that's the thing that, as a founder, if you lose it, it's like game over.

But they're definitely back! We were talking about this, right? Where it just feels like we come across more applic—we’re talking about this on our part! We funded a record number of MIT grads in the last batch, the most YC has ever funded.

So it cast a chilling effect for a year, but it’s—and in particular now when we talk to young founders. And I think this is why the median age of the batch went down slightly, right? The median age was around 30 four years ago, and now it’s like 26.

And I think I’m just seeing more people being willing to drop out of college. Often what we say to them is, hey, like, there’s no rush! Like, you should just like graduate college! Why do you have to start a startup right now?

And the response is, well, like, this might be like a once-in-a-lifetime opportunity. And I think for the first time, I'm like, you might be right! You’re right! The reason this time is different as it relates to this sort of AI—

Oh no, you said the magic word!

Yeah, I said it! I think it actually is different. I think, like, crypto has always suffered from a couple of problems. One is that it's always been very hard to explain the products, right? Like, they tend to be very complicated and not user-friendly.

And so it's just hard to explain like, what even does the crypto thing do? Because it tends to be some sort of complicated lending thing that only other crypto people understand.

And the second is that it's always been hard to understand like where the money comes from. It can be this sort of Byzantine complex thing trying to figure out like—like it feels like Monopoly money!

Yeah, basically, right? That's always been the criticism. It’s always felt like these things weren't real often. But like this time around, I think working in this batch with so many AI companies, it's felt very real. The products were very tangible.

There was a cool experiment we ran with Harge in this batch; we ran product day where we had all of our companies come and do a demo run-through of the product on stage.

Other product, yeah!

Running! And a lot of the products were beautiful! I was very impressed with the progress they made from the time they applied to what they had because one of the stats is, in this batch, over 80% of the batch had no revenue. Basically, the product was unlaunched before they came in versus in Winter 20, only about 62%.

So we funded even earlier, and in the span of just a month, these products were pretty impressive, right?

Yeah, I mean, I remember that when we did product day, just some of the moments—it felt like constant wow moment after wow moment! Like someone would demo, like one of the companies, Fume demoed like their AI software engineer, and their demo was literally like, hey, like you can basically tell this AI software engineer to implement dark mode on this website.

And they showed like a website, just regular layout, and like Fume engineer goes and implements like in CSS and everything like the dark mode, and then you go back to the website and there’s a toggle to turn dark mode on, and it’s just like, wow!

Like you could see it writing the code and doing all of this stuff, and it was like you could tell this is something that’s very real!

And there’s another thing that we did that like returned YC to focus on products. Do you want to talk about the Bookface launch live events that you started this batch?

Yeah, during the batch I really felt like, well, these demos are so cool, I’m totally going to steal your idea for this next batch where—I definitely—I think that we should do this type of demos across the whole batch.

And then I also did Friday—every other Friday— we would pick the people who had the most impressive launches on our internal social network, and we’d have them actually demo exactly what they built in front of a live audience, yeah!

And then I would actually ask very detailed implementation questions because literally a lot of these things, it's the first time you’ve ever been able to do like that demo, for instance. And being able to understand, well, what did you do?

Like how did you use retrieval-augmented generation to do that? Like what were the prompts? What was the workflow? You know, how do you test that kind of stuff?

Like going back to the dev tool argument, like we're literally trying to figure out how these things work. And then there’s going to be a whole new reset in even a matter of months with GPT-5 and the next generation.

Like this is a very homebrew computer club type stuff! Like we just suddenly had this thing that could happen, and next week some other crazy thing is going to happen!

I love that because it really felt to me like a return to the YC that I did in 2006, 2007! The focus of YC was really about the products because we were inventing new things that you could do with software.

And then in the decade afterward, because there were so many software-enabled businesses and the technology became commoditized, there was less focus on the product and more on growth and sales.

I hadn’t thought that, you know, just sprung into my mind as you’re saying that is 2007, when I first moved to San Francisco from London. In my YC batch in Winter 2007 were Weebly and Zenta, um, and they were pushing the limits of what you could do with JavaScript!

Weebly was a website builder, and Zenta, which would get acquired by Google, was like a web PowerPoint that was cutting-edge stuff! Yeah, seriously! It really felt like every week, I remember you would come to YC dinner, and you would go and check out what the Weebly and Zenta teams had done the previous week.

And you’d be like, wow! Like you can like create like a slide! You can create a slide with an animation in the browser! Like it’s crazy, and even works in Internet Explorer! That’s insane!

Yeah, that was the thing—that you can do this across all browsers. And it felt like yeah, I hadn’t thought about it until you said, Jared! But yeah, that was so much of the energy!

You felt like you were around people really inventing stuff! It actually, that team, those two founders basically kickstarted all of the Google Doc Suite! Like Doc, Doc, Sheets!

And then when you think about it, back then, in 2007, it wasn't clear that you could replace Microsoft Office with like a bunch of applications in the web. That would have sounded insane!

I think the other thing I remember doing the Bookface live demo, I think that feeling was there. I remember seeing the demo of retail AI. So they were building these voice AI agents, and during the demo, they did a call to the AI agent, and they pretty much passed the Turing test!

That’s pretty amazing! You would have conversations with it, and that’s the moment where I think it was just so fun to be alive now and working on this.

And it’s turning into like real businesses again! The crypto analogy, a lot of the criticisms were a lot of the products were promising like high-yield, risk-free, high-yield products, and we would take like a spread on the yield—again, very complicated to understand where the money is zero-sum, basically!

But like, or it was like it was the promise of future usage. It was like someday when everybody switched to like a decentralized Airbnb, it’ll be really big!

Yeah, but no one’s actually using it to rent apartments!

Yeah! But in this batch, we're seeing companies add like this AI boom—companies are adding like real recurring revenue by selling software to legit businesses, right?

I mean, I think we looked up some of the DAU, and we actually measured it, right?

Yeah, so I pulled the numbers on this. 80% of the batch came in with no revenue, and the majority had not launched any product at all. They didn’t have any users. If you look at the start of the batch in January, the total—batch—all the batch companies together, if you add up their total revenue, they were making $6 million ARR across almost 300 companies.

So like the ARR of the batch, which is kind of like a funny concept. I don't think we’d ever really thought about the ARR of a batch before. Metric! But like the ARR of the batch was like $6 million in January, and by demo day in April, it was $20 million.

So the batch 3x ARR in three months, which is pretty cool! That's a great growth rate, yeah!

You think of all the economic growth that got produced. It’s kind of think of these companies as they keep growing and compounding and accelerating a lot of this growth, which is new is definitely not zero-sum. It's this whole world of creating a bigger and bigger pie, or like a new matrix or new maze, right?

And I think what's getting people really excited about the future of AI in particular is this is not just taking money away from existing software budgets. So much of this work is replacing labor, and so you open up to like labor budgets.

And so I think like all of that has just fed into this general—if we just like go off the vibes—this batch has felt like it’s like being like the best one yet!

Like there more specifically, it feels like there's more energy around this batch than there has been for as long as I can remember for any YC batch. What do you think's happening?

And one time that I felt that really viscerally was at the in-person investor reception, which Gary created for this batch. Do you want to talk about it, Gary?

Well, so demo day is still perfectly online, and it works great! But one of the things we really wanted to do was thank some of the best people who have funded YC companies all these years—and it was right here in our San Francisco HQ!

It was three whole floors of some of the smartest investors in the world with our batch companies just hanging out, and the vibes were immaculate! It felt to me like, at the reception, talking to investors, that there was a real reset on preconceived ideas of what’s a good idea and a bad idea.

People were just renewed sense of optimism, and it felt like everything's up for grabs. I worked with a company called Octane, which got lots of investor interest. What they're doing is AI sales for—it's Salesforce rebuilt in the AI world—and I just think investors did not want to fund Salesforce competitors for a long time because they just felt like, like, how are you going to compete with Salesforce?

But now with AI, it's like, oh, like we totally funded Salesforce competitors because it seems possible that you could actually win against Salesforce now. There's a platform shift, and this is the moment where every single SAS dollar in the world is up for grabs again!

It’s an exciting time! Founders are more excited to build than ever; investors are more excited to invest than ever! And we're just like right at the center of it all here at YC!

It's awesome! It’s a fun time to build! It's the best time ever! I mean, as a builder, it’s like the technology just does such a different thing than what you expected before!

I think this is why you have the earnest founders that love building coming back and doing this. So do we think that this batch was peak? What's in store for the next batch?

I mean, it feels like because we talked so much about all the progress on the current batch, it seems like it's like done and all the good ideas are done. I actually think the opposite because this batch was the one where we had the most pivots!

30% of the batch pivoted and landed in good ideas versus in four years ago only about 10% of the batch pivoted. So that's one very fast defined ideas! There’s still tons of them that are good.

And if we go back to the analogy with how history kind of remixes and repeats a bit, I think of this time more like Facebook is still getting created in the dorm rooms. So if you all want to be the next big AI company, this is the time!

If we talk about what we were saying earlier, Jared, like if now is sort of the equivalent of 2007 when it felt like web technologies were being pushed forward for the first time, it was actually still like three years until Airbnb was started, five years until DoorDash was started, six years until Coinbase was started. These trends always play out much longer than people think they're going to!

So in summary, we are just getting started! That's all the time we have for today! But Y Combinator is actually accepting applications for this summer! So if you're thinking about it, those questions will help you shake out, is this the time? Do I have the co-founder? Do I have the idea?

And at the end of the day, now is the moment to start! So we hope we’ll see you this summer!

And we'll see you next time!

[Music]