EVERY SPOKEN WORD
30 min read · 6,106 words- 0:00 – 0:54
Coming Up
- GTGarry Tan
If O1 is this magical, what does it actually mean for founders and builders? One argument is it's bad for builders because maybe O1 is just so powerful that OpenAI will just capture all the value.
- GTGarry Tan
You mean they're going to capture a light cone of all future value?
- GTGarry Tan
Yeah. (laughs) Yeah. They'll capture a light cone of all present, past and future value. (laughs)
- GTGarry Tan
Oh my god.
- GTGarry Tan
The alternative, more optimistic scenario is we see ourselves how much time the founders spend, especially during the batch, on getting prompts to work correctly, getting the outputs to be accurate. But if it becomes more deterministic and accurate, then they can just spend their time on bread-and-butter software things. The winners will just be whoever builds the best, like user experience, who gets all these like nitty-gritty details correct.
- SPSpeaker
(instrumental music)
- GTGarry Tan
Welcome back to another episode of The Light Cone.
- 0:54 – 5:35
What models get unlocked with the biggest venture round ever?
- GTGarry Tan
We are sort of in this moment where OpenAI has raised the largest venture round ever, $6.6 billion with a B. Here's what Sarah Friar, the CFO of OpenAI said about how they're going to use the money.
- SPSpeaker
It's compute first, and it's not cheap. Uh, it's great talent second. Um, and then, of, uh, course, it's all the normal operating expenses of a more traditional company. But I think there is no denying that you are, we are on a, a scaling law right now where orders of magnitude matter. The next model is going to be an order of magnitude bigger and the next one on and on. And so that does make it very capital intensive.
- GTGarry Tan
So it's really about orders of magnitude. Let's live in the future. There's 10 trillion parameters out there, 10 trillion parameter large language models, two orders of magnitude out from the state-of-the-art today. What happens? Like are people actually going to be throwing queries and actually using these 10 trillion parameter models? Uh, seems like you'd be waiting, you know, 10 minutes per token.
- DHDiana Hu
Yeah. For a bit of context, right now, the frontier models, I mean they're not public exactly how many para- parameters they have, but they're roughly in the five hundreds of billions-ish, like Llama-3, 4.0, 5 billion. Anthropic is speculated to be 500 billion, GPT-4.0 roughly a- around that much. Getting to 10 trillion, there's a two order of magnitude, right? I think the type of level of, uh, potential innovation could be the same leap we saw from GPT-2, which was a- around one billion parameters that was released with the paper of a scaling laws, which was one of the seminal papers that people figured out, okay, this is transformer architecture that we figured out, what if we just throw a bunch of engineering and just do a lot of it? Where does this scale on this logar- logarithmic type of scaling? Then this was proofed out when GPT-3.5 or 3 got released, that was about 170-ish billion parameters. So that's like that two order of magnitude. And we saw what happened with that, that created this new flourishing era of AI companies. And we saw it, we experienced this back in 2023 when we started seeing all these companies building on top of GPT-3.5 that was starting to work and it created this giant wealth. So we could probably expect if this scaling law continues, the feeling will be similar to what we felt from that year of, uh, transition from 2022 to 2023.
- HTHarj Taggar
Yeah. That was the moment when everything changed. So that would be pretty wild if that happens again.
- GTGarry Tan
I think there's a, one interesting aspect to this, which is clearly the current generation state-of-the-art models that are available, uh, especially given O1 chain of thoughts, um, they sort of basically rival normal intelligence. Like you could make a strong case that AGI is basically already here. The majority of the tasks that, you know, 98% of knowledge workers do day-to-day, uh, it is now possible for a software engineer, uh, probably sitting in front of Cursor, to write something that gets to, you know, 90 to 98% accuracy and actually do what a knowl- a human knowledge worker with 120 IQ would be doing all day. And that's sort of writ large, like there are probably hundreds of companies that each of us have worked with over the past few years that are literally doing that day-to-day right now. You know, the weird interesting question is like at 10 trillion parameters at you know, 200 to 300 IQ like sort of ASI beyond what a normal human being normally could do, you know, what does that unlock? Uh, there's an, a great article in The Atlantic, uh, with Terence Tao, sort of famously this Taiwanese mathematician who is like literally north of 200 IQ and how he uses ChatGPT right now and it's sort of unlocking, uh, new capabilities for him. There are some examples of this happening, you know, quite a few times in human history. Like you could argue that nuclear power was that fission. You had to actually m- model theoretically that something like, uh, nuclear fission was possible before anyone, you know, experimentally tried to do it. Uh, Fourier transforms.
- DHDiana Hu
Yeah. Maybe the thing is, uh, if we think
- 5:35 – 9:53
Some discoveries take a long time to actually be felt by regular people
- DHDiana Hu
a lot of the capabilities right now are here, but is not evenly distributed, if you go walk down the street and you talk to the random Joe, they don't feel the AI. (laughs) They're just living their normal life and stuff is still just normal. It hasn't changed. But I think the counterexample is just sometimes these discovery take time for it to really pan out. This is example we're discussing. Fourier transform was this mathematical representation that Joseph Fourier discovered in the 1800s, it was like a seminal...... thesis that he wrote about representing series of functions that were repeating in periods, and before Fourier transform, they were written as these long sums, and series of sums that are very expensive to add them up and figure out how to really model the equation basically. But he found this very elegant way that instead of just doing sums of series, you could basically collapse all these math function into sines and cosines wave that only need two variables, basically the amplitude and the period, and you could represent every periodic signal and function. I mean, it's not really cool math, which is like how some of this, uh, LLM and use case sounds like, okay, cool, they can do all this coding. But Fourier transform, it took another 150 years until the 1950s when people figured out what to do with this. It turned out that Fourier transform were super good at representing signals, and we need signals basically to represent everything in the analog world to be digital because bits are ones and zeros, and how do you compress that? And one of the big applications as well is, uh, radio waves, and made telecommunication a lot more efficient, image representation, encoding, information theory. It just unlocks so much of the modern world, like the internet and cell towers work because of this theory. But it took 150 years until the average Joe could feel the Fourier transform.
- GTGarry Tan
Interesting. That's a really powerful idea. I mean, that took a while then. I mean, apparently, in the 1950s, that's the moment that color TV happened, so...
- DHDiana Hu
Unlocked by Fourier transforms as well.
- GTGarry Tan
That's right.
- GTGarry Tan
If you apply it to the AI stuff that's happening today though, it's like, one, where do you start the clock ticking from? Like, it's not clear if you start it from the ChatGPT-3 moment two years ago, or from just, like, all of the research that's been going on for decades. Like, we might actually just be, we've talked about this before, but we might actually be, like, decades into this now, and it's starting to hit, like, the inflection moment potentially.
- HTHarj Taggar
Yeah, for sure. I mean, if we run with Diana's example of f- fast Fourier transforms, like all the math that's underpinning all the- this new AI stuff is linear algebra stuff that's like 100 years old.
- GTGarry Tan
Yeah.
- HTHarj Taggar
It just turns out that-
- GTGarry Tan
You just don't know how far you can push it. Yeah. (laughs)
- HTHarj Taggar
(laughs)
- DHDiana Hu
But you have all the GPUs to compute it.
- GTGarry Tan
I guess that's one potential way that, uh, these 10 trillion parameter models actually alter the face of what humans are capable of. Like, they sort of unlock something about the nature of reality and our ability to model it, and then somehow it leads to, uh, either nuclear weapons or the color TV.
- GTGarry Tan
The other big thing is just because this is all in software, like, um, comparison like Fourier transforms, that like a lot of the applications we're seeing physical devices, right? Like record players or- or telephones, like you said. And so it takes a while for the technology to get adopted because you have to, like, buy your updated device and all these things. Now, we have like Facebook and Google who have like, you know, pretty decent percentage of just, like, the world using their software already, like as soon as these things start rolling out. And I feel like it's another thing that's starting to be noticed is Meta in particular coming out with their Meta Ray-Bans, the consumer, like, device. Like, I think consumers, once this becomes, um, something that's like visual in your, like, smart glasses plus like a voice app that you can talk to and it, like, is indistinguishable from human being, like that's going to be a real change the world moment for people. They will start feeling the AI once they can, like, talk to it all the time.
- 9:53 – 14:26
Distillation may be how most of us benefit
- GTGarry Tan
I mean, it seems like there's, uh, really a bifurcation in what we might expect when we have this capability. Um, at the extreme end, you're going to have people like Terence Tao pushing the edge and boundary of our understanding of our, you know, modelable world. And then, you know, maybe that's actually worth tens or hundreds of millions of dollars of inference to run these 10 trillion parameter models. And then the more likely way this ends up being useful for the rest of us is actually in distillation. So taking, you know, there's some evidence that, for instance, uh, Meta's 40- uh, 405B was mostly use- useful to make their 70 billion parameter model much, much better. And so, and you actually see this today, there's sort of this moment there where we thought that, um, you know, people might just go to GPT-4 and distill out all the weights, and it seems like there's some evidence that, uh, certain governmental entities are doing that already. But GPT-4 itself and 4- you know, it became 40, OpenAI itself has now enabled distillation internal to its own API. So you can use 01, you can use, uh, even GPT-4 or 4- 40 to distill it down into a much cheaper, uh, model that's internal to them, like GPT-4- uh, 40 mini. And that's sort of their, you know, lock in capability.
- DHDiana Hu
Yeah, I don't think this is talked much about, but it is interesting that you have these giant models like the 400 or 500, whatever, billion-parameter models that are basically the teacher models because they're the mega trained with everything and took forever, and they are the teacher model, master model that teaches a student model which are these smaller ones that are faster and cheaper, because doing inference for a 405 billion-parameter model is very expensive. So we have evidence that all these distillation models are working. Companies in the batch, the- they're building from the latest and greatest, they're not going for the giant model with all of the parameters and give me the biggest thing to do so that it works the best. We have evidence that's not the case. People are not going for the big model, and we actually have stats in the batch. I mean, Harsh, we kind of talked about them.
- GTGarry Tan
Yeah, Jared ran some numbers on this, and it's- it's fascinating, but I think the bigger meta point is even the fact that talking about...... the startups or the founders building this stuff are choosing, like, the smaller models versus bigger models. Highly suspect they just have choice and even, like, a year ago when this, like, entire industry started existing, like, everything was built on top of ChatGPT, right? There was a... It was 100% market share, the ChatGPT wrapper meme. And I feel like we've, especially over the last six months, seen people start talking about the other models like Claude and Sonnet being sort of this word of mouth for almost being better at cogen than ChatGPT, and people just starting to use different models. And so the numbers that Jared ran for the summer '24 batch are fascinating because it seems that that trend has just continued. Like, we have more diversification of LLMs and models that developers are building on top of. And some of the stuff that really stood out is Claude has even... Just in six months, from the winter batch to the summer batch, has gone from, like, 5% developer market share to, like, 25%. (laughs) It's like-
- DHDiana Hu
Of companies in the batch?
- GTGarry Tan
Yes, of companies in the batch, which is huge jump.
- HTHarj Taggar
That's right. I've never seen a jump like that.
- GTGarry Tan
Right. LLaMA's gone from 0% to 8%. (laughs) Like...
- HTHarj Taggar
One thing that we know from running YC for a long time is that whatever the companies in the batch use is a very good predictor of what, like, the best companies in the world are using, and therefore what products will be most successful. A lot of YC's most successful companies, you could have basically predicted which ones they would be based on just looking, basically just running a poll of what the companies in the batch use.
- GTGarry Tan
If we just take o- take OpenAI's latest fundraise off the table and the latest, like the 01 model off the table for a second, it would seem like amongst developers and builders, OpenAI was losing. Like, they went from being the only game in town to just, like, bleeding market share to the other models at a pretty rapid rate. The interesting thing though is maybe they are coming back. (laughs) Like, what was the stat that you pulled there? It seems like over 15% of the batch are already using 01 even though it's not, like, fully available yet.
- HTHarj Taggar
Yeah, 01 is only, like, two weeks old now.
- GTGarry Tan
Yep.
- HTHarj Taggar
Yeah.
- DHDiana Hu
And we're seeing some
- 14:26 – 21:17
o1 making previously impossible things possible
- DHDiana Hu
interesting things, uh, with 01. We're actually hosting right now, in person right now as we speak downstairs, a hackathon to give YC companies early access to 01. And Sam himself was here, did the k- kickoff. There's a bunch of OpenAI researchers and engineers working on it. And it's only been about four hours of hacking, and we already heard of... I already saw actually some demos as I was walking by to see some teams, and they already built things that were not possible before with any other model.
- HTHarj Taggar
Do you, do you have some examples?
- DHDiana Hu
One of the companies I'm working with is, uh, Freestyle. They're building a cloud solution fully built with TypeScript, with, if you're familiar with Durable Objects, with this really cool framework to... That, that makes front end and back end seamless to develop and is really cool to use. What was cool about them is they'd just been working on it for a couple hours and I saw a demo that I was mind blown. They basically got a version of Replit Agent working with the product. All they had to prompt 01 with was all their developer... Well, sure, some of their developer documentation and some of their code. And they could just prompt it, "Build me a web app that writes a to-do list or this..." And it would just, boom, just work. And it was able to reason and inference with the documentation and took a lot longer, but it arrived and built the actual app.
- GTGarry Tan
What's interesting for us to talk about is if 01 is this magical, what does it actually mean for founders and builders? And, uh, one argument is it's bad for builders because maybe 01 is just so powerful that OpenAI will just capture all the value and, um, everything that could be valuable and built on top of this stuff will just be owned by them.
- HTHarj Taggar
You mean they're gonna capture a light cone of all future value?
- GTGarry Tan
Yeah. (laughs) Yeah, they'll capture a light cone of all present, past and future value. (laughs)
- HTHarj Taggar
Oh my god.
- GTGarry Tan
The alternative more optimistic scenario is we see ourselves how much time the founders spend, especially during the batch, on the tooling around getting the prompt, like, getting prompts to work correctly, getting the outputs to be accurate, human in the loop. Like, all of this time spent just getting the core product working is not deterministic. But if it becomes more deterministic and accurate, then they can just spend their time on bread and butter software things, you know? Like better UI, better customer experience, more sales, more relationships. In which case it's like, it may be a better time to start now than ever because you don't even have... Like, maybe all of the knowledge you learn around how to get, like, the prompts accurate and working was just temporary knowledge that's no longer relevant as these things get more powerful.
- DHDiana Hu
Actually, we had this conversation with Jay Keller from Case Text where getting the legal copilot to work 200% was the huge unlock.
- HTHarj Taggar
And it was really hard.
- GTGarry Tan
Yep.
- HTHarj Taggar
He, like, he... We heard this whole talk about all the things he had to do to actually get the thing to be accurate enough.
- GTGarry Tan
Yeah. Imagine if he didn't have to do any of that. If, if they just on day one, he could be guaranteed 100% accuracy as, as though you're just building a web app on top of a database, the barrier to entry to build these things goes way down. There's gonna be more competition than ever and then it will probably just become, look more like a traditional winner takes all software market.
- DHDiana Hu
Derek has an example. So, there's a company, Dry Merge-
- HTHarj Taggar
Yeah.
- DHDiana Hu
... that you work with, and they went from 80% accuracy to pretty much 99, or for intensive purposes, 100%, using 01 and it unlocked a bunch of things. You want to talk about them?
- HTHarj Taggar
Yeah, yeah. Just by swapping out GPT-4o for, for, for 01. I think there might be an even more bullish version hard, which is that their use cases right now that people...... are not able to use LLMs for because even though they're trying to get the accuracy high enough, they just can't get it accurate enough-
- GTGarry Tan
Mm-hmm.
- HTHarj Taggar
... for, to actually be rolled out in production. Like, especially if you think about, like, really mission critical jobs where the consequences of mistakes are dire. Like, pretty hard to use LLMs for that. As they keep getting more accurate, those applications will start to actually work.
- GTGarry Tan
I guess there is, uh, a lot of evidence inside the YC, uh, greater portfolio. You know, I was meeting a company from 2017. I think I tweeted about them. They were, um, you know, $50 million, uh, annualized revenue at that point, but growing 50% a year. A year or two ago, um, they were not profitable. They knew that they needed to raise more money. But, uh, in the year since, they automated, uh, about 60% of their customer support tickets, and they went from something that needed to raise another round imminently to something that was totally cashflow break-even while still growing 50% year on year. That's sort of, like, the dream scenario for, uh, building enterprise value because you're big enough that, you know, you're a going concern, and then you're literally compounding your growth with, like, no additional capital coming in. So it's companies like that that actually end up becoming, like, half a billion, a billion dollars a year in revenue and, like, driving hundreds of millions of dollars in free cashflow. I mean, that's sort of the dream for founders at some level, and I think that that's one of the more dramatic examples that I've seen thus far, and I think it's sort of not an isolated case. You know how we're sort of talking, it's, you know, 2024 now, and we're, um, still in this overhang moment where, uh, companies sort of on this path raised way too much money at, you know, 30X or 40X, uh, you know, next 12 months multiple revenue. (laughs) Like, seemingly struggling but, you know, also never going to raise another round. Like, this is actually pretty good news for them because they actually can go from, like, not profitable to, you know, break-even to then potentially very profitable. I think that narrative is not out there, and I think it's, uh, really, really good news for founders.
- GTGarry Tan
It'll probably start to catch attention. I... Didn't the Klarna CEO got a lot of attention a few weeks ago for... I mean, uh, it's not unclear how much of it is real or not, but at least they're pitching that they're just r- you know, replacing their internal systems of records for HR and sales with home-built or LLM-created apps. At least was, like, the insinuation.
- GTGarry Tan
Yeah, what is it? They got rid of Workday.
- GTGarry Tan
Yeah, that was it, yeah.
- GTGarry Tan
That's pretty wild, honestly. I mean, so that's good. If you treat OpenAI as the Google of the next 20 years, you want to invest in Google and all the things that Google enabled, like Airbnb. Google could do Airbnb. It probably won't. (laughs)
- GTGarry Tan
Yeah.
- GTGarry Tan
Just from, like, I don't know, Coase's theorem of the firm probably. (laughs) It's just, like, too inefficient and too difficult, requires too much, uh, domain expertise to actually pull that off.
- 21:17 – 23:47
The new Googles
- DHDiana Hu
So what are the, sort of the new Googles that are getting built out? There's these vertical agents. What are some examples that we'll have that we can talk about?
- GTGarry Tan
I loved working with this company called TaxGPT, um, from the last YC batch. They started off actually, uh, really literally a rapper and, like, you know, it's in the name, TaxGPT. (laughs)
- GTGarry Tan
(laughs)
- GTGarry Tan
Uh, but my favorite example about them is, like, you know, it turned out that tax advice, you know, doing basic rag on, um, you know, uh, it's sort of like CaseText actually. It was, uh, you know, being able to do rag on existing case law and existing policy documents from the IRS or internationally. That was just sort of the wedge that got them in front of, you know, tens of thousands of accountant- accountants and accounting firms. And, uh, now what they're doing is building an enterprise business on, uh, document upload. So, you sort of, you know, get them for cheap or free for the thing that people are Googling for. And then once they know about you and trust you, you get, like, this $10,000 or $100,000 a year ACV contract that then takes over real workflow that, you know, actually extinguishes tens to hundreds of hours of work per accountant.
- GTGarry Tan
And another interesting thing about the o1 model is we were just saying originally ChatGPT was the only thing you could build on top of. OpenAI was the only game in town. Then there were all these models. I think the sort of alpha leak we have here, like, right now in this room is downstairs people are building at the cutting edge of o1 that even the public doesn't have access to. Um, and what we're seeing is that this is a real major step forward. Like, o1 is going to be a big deal for any programmer or engineer who is building an AI application. The interesting thing is, will this cycle repeat where it will give OpenAI a temporary lead, their market share will just, like, go, you know, back up towards 100%, but then within six months, LLaMA will be updated, Claude will come out with its new release, Gemini will keep getting better, and there'll just be, like, you know, four different models that have equivalent levels of reasoning? Or will this be, like, the first time OpenAI has a, a true breakthrough? And I would just define a true breakthrough as something that's actually defensible. Like, if no one else can replicate it, then that puts them in a really powerful position. But we don't know, and I think that's what's interesting is, like, OpenAI seems like it is continually the one pushing the envelope, but they always seem to be the first ones to make major breakthroughs, but they have never been able to maintain the lead so far.
- 23:47 – 25:44
o1 makes the GPU needs even bigger
- DHDiana Hu
I think the other thing that's interesting about o1 is that it makes a lot of the GPU needs even bigger because it's moving a lot of the computation needs a lot higher for inference because it's taking a lot more time to do a lot of the inference. So, I think it's gonna change also a lot of dynamics underneath for a lot of the companies building AI infrastructure as well.... which is something, food for thought.
- GTGarry Tan
Hmm.
- GTGarry Tan
It seems like there are two different types of use cases. I believe they did just enable distillation from 01 into 40. And so it's conceivable that, uh, for relatively rote and repeating use cases, you could just sort of use 01 for the difficult ones, and then you distill it out, and then you pay 40 or 40 many prices from there. And then there are, again, there's this other type of problem that is, like, very specific. I mean, I imagine most co- many code gen situations are a little bit more like that, where you need to pay, like, for the full 01 experience because it's, uh, you know, fairly detailed and specific.
- GTGarry Tan
It depends on who you're building for too, right? If you're a enterprise software and you can pass the cost onto your customer and they can tolerate at a higher latency and don't care as much about it being instant, then you can just use just maybe 01 a lot. If you're building, like, consumer apps, probably not. But talking of consumer apps, I mean, the other thing that was striking about OpenAI's latest release is, like, this real-time voice API, um-
- DHDiana Hu
Super cool.
- GTGarry Tan
It's pretty remarkable, and I think the, the most telling thing to me is that the ongoing usage-based pricing is $9 per hour, and that sort of points to a sort of powerful thing. Like, if I were, um, a macro trader, I would be very, very bearish on countries that have, that rely very heavily on call centers right now, um, because, you know, $9 an hour is sort of right there at, uh, what a call center would cost.
- GTGarry Tan
This is another thing we're definitely seeing within the batch, right? Like, it's
- 25:44 – 27:05
Voice apps are fast growing
- GTGarry Tan
clear that voice is a, almost like a killer app (laughs) , like, arguably. Like, um, there's a company I just worked with in this batch, um, or in my group at least, Domu, who just do, um, sort of AI voice for debt collection, and their traction is just phenomenal. It's working incredibly well.
- DHDiana Hu
A whole bunch of the voice apps in S24 were just, like, some of the fastest growing, like, just, like, explosive companies. It was, it was a clear trend for S24. And, and I remember working with companies in the prior two batches that tried to do voice, and it just, like, wasn't working well enough.
- GTGarry Tan
Yeah.
- DHDiana Hu
Like, the latency-
- GTGarry Tan
Pause.
- DHDiana Hu
... was too high, exactly. Like, it got confused if, like, with interruptions and things like that, and it's, like, just turned the corner where it's, like, finally working. So another company I work with, uh, Happy Robot, that landed on this idea that I was doing a voice agent for coordinating all these phone calls for logistics. Think of, uh, a truck driver that needs to go from point A to point B. These are all, like, people just calling to check where you are. There's no, like, (laughs) there's no, like, Find My Friends for it. And they started getting a lot of, uh, usage on this and I think we talked a bit about this before, that at this point, AI has passed Turing tests and is solving all of these very menial problems over the phone.
- GTGarry Tan
That's pretty wild. I guess one thing that's, uh, is maybe under-discussed is to what degree,
- 27:05 – 31:52
Incumbents aren’t taking these innovations seriously
- GTGarry Tan
um, the engineering teams that are in, in these sort of incumbent industries, it feels like it's pretty binary. Either, you know, the vast majority of companies and organizations, especially the ones that were maybe founded f- four or more years ago, they actually don't take the, any of this seriously. Like, they have literally no initiatives on this stuff, and I sort of wonder how generational it is. Like, I'm realizing that, uh, eng managers and VP of eng, like, they're probably my age now. I'm 43 now, and, um, you know, if I wasn't here seeing exactly what was happening, I would be sort of tempted to say, like, "This is just the same old thing, AI, yeah, yeah, yeah." But-
- GTGarry Tan
I, I think it's the rate of improvement that people don't get if they're not as close to it as we are. I just think your average corporate enterprise person is certainly used to technology disrupting things, but over pretty long timeframes. And if anything, they become cynical, because they're like, "Oh, the cloud." Like, cloud was such a buzzword for a long time. It totally did change how enterprise software is built and delivered, but it took, like, a decade or so. And so I, I suspect everyone's feeling that way about AI, is just your natural default mode is to be cynical, "Oh yeah, like, it's not going to be ready for a while," and then probably if you looked at this stuff even six months ago, like we were just talking about, if you looked at an AI voice app six months ago-
- DHDiana Hu
Yeah.
- GTGarry Tan
... you're like, "Oh, this is, f- this is years away from being anything that we need to take seriously."
- DHDiana Hu
(laughs)
- GTGarry Tan
And it's, like, actually, like, three to four months later-
- DHDiana Hu
(laughs)
- GTGarry Tan
... like, it's like, it's hit some real major inflection point, and I think that's what takes even people within tech, it's surprising all of us how quickly this stuff is moving.
- DHDiana Hu
It's, it's the fastest any tech has ever improved-
- GTGarry Tan
Yeah.
- DHDiana Hu
... I think.
- GTGarry Tan
Yeah.
- DHDiana Hu
Certainly faster than processors, certainly faster than the cloud. And it's kind of fun to actually watch. It's been remarkable to actually see another example of this in the batch. So a lot of the technical founders, sometimes I sit with them and s- I just watch how they code. The before and the after, before all of this wave of AI, just standard, you have your IDE and things on the terminal. People ship fine, but the demos and products that we're seeing founders build during the batch is like a next level of polish. And when you see them and see them code, it's like they're living in the future. They're really not just, like, at this point, GitHub Co-pilot is already kind of old news. They're using the latest, greatest coding assistant. A lot of them perhaps using something like Continue or Cursor, right?
- GTGarry Tan
Well, this is something Jared pulled out as well when we asked the founders
- GTGarry Tan
Oh, the IDEs, right?
- GTGarry Tan
Yeah.
- DHDiana Hu
Ye- yeah. We surveyed the summer '24 founders, and half the batch is using Cursor, compared to only 12% that's using GitHub Co-pilot. That was surprising to me. They're not even using the fully agentic coding agents like, uh, Replit. These are still sort of, like, Co-pilot phase stuff, but even just going from, like, GitHub Co-pilot to Cursor, which is, like, the next step up in terms of, like, how much the actual AI does, is this, like, incredible breakthrough. They ship very quickly. I mean, there's evidence today in the hackathon, I was impressed with what they built, and I was just looking at their editors, like, Cursor. It's like, cool.
- GTGarry Tan
It's another sign why, like, the founders have the advantage, right? Like, it feels to me-
- GTGarry Tan
Yeah.
- GTGarry Tan
... again, like when GitHub Copilot first came on the scene, it seemed, "It's GitHub plus Microsoft. It has all the developers, plus it has all the capital, plus has all the access to, like, the inside track on OpenAI."
- GTGarry Tan
Right. Right.
- GTGarry Tan
"How could any coding IDE compete with them? It'll just get subsumed." And like, Cursor has come out of nowhere and is like, you know-
- GTGarry Tan
Just built a better product.
- GTGarry Tan
... according to our numbers, like five times the size (laughs) of, like, GitHub Copilot within the batch. Which, again, like you were saying earlier, is like, the- the startup founders are actually usually the tastemakers on this kind of thing. Um, I think there's certain types of businesses where it doesn't make sense to maybe go after startup founders as your early customers, but for developer tools, it definitely does. Like Stripe, AWS both wanted to own YC batches in particular, um, and that worked out really well for them. So, it's probably a really good sign for Cursor, honestly, that they have, like, such good presentation- penetration within the YC batch.
- GTGarry Tan
Yeah. I- I would definitely say Cursor is pretty awesome, but AltaVista was awesome too.
- GTGarry Tan
Yeah.
- GTGarry Tan
I remember using that as a search engine, and there was another version, and the next version was ten times better. And so, this is the way it's going to go. I mean, which the only people who win are actually developers because of all this competition.
- GTGarry Tan
So, I think, again, it takes us to, like, the optimistic view of all of this stuff, which is as the models get more powerful, the winners will just be whoever builds the best, like, user experience and gets all these, like, f- nitty-gritty details correct. And so that's why Cursor can beat GitHub Copilot that has all of the advantages. AltaVista's a great example. Like, they're still- like, Google still came along and crushed them, right? So, there's still room for someone to keep doing to Cursor what Cursor has done to GitHub Copilot.
- 31:52 – 33:15
Ten trillion parameters
- GTGarry Tan
So, let's get back to 10 trillion parameters. What world do you think we will live in with this made real, with ASI or something approaching it? You know, what will humans actually do and how much more awesome will it be?
- HTHarj Taggar
Well, I'll give a steel man for a really bullish case, which is that the thing that is holding back the rate of scientific and technological progress is arguably the number of smart people who can actually analyze all the information, um, that we already know about the world. There's m- millions of scientific papers already out there, an incredible amount of data, but like, try reading all of it. It's far beyond the scale of any human's comprehension. And if we make the models smart enough that they can actually do original thinking and deep analysis with correct logic, and you could let loose an infinite- a near infinite amount of intelligence on the near infinite amount of data and knowledge that we have about the world, you can just imagine it just coming out with just crazy scientific discoveries. Room temperature fusion, room temperature superconductors, time travel, flying cars. All the stuff that humans haven't been able to invent yet, like, with enough intelligence, maybe we'll finally invent it all.
- GTGarry Tan
Sign me up for that future.
- GTGarry Tan
(laughs) Sounds great.
- 33:15 – 33:44
Outro
- GTGarry Tan
I totally agree with you. I think, you know, what this might be is not merely a bicycle for the mind. It might actually be a self-driving car or, even crazier, uh, maybe a rocket to Mars. So, with that, we'll see you guys next time. (instrumental music)
Episode duration: 33:44
Install uListen for AI-powered chat & search across the full episode — Get Full Transcript
Transcript of episode lbJilIQhHko
Get more out of YouTube videos.
High quality summaries for YouTube videos. Accurate transcripts to search & find moments. Powered by ChatGPT & Claude AI.
Add to Chrome