Y Combinator10 People + AI = Billion Dollar Company?
EVERY SPOKEN WORD
35 min read · 7,461 words- 0:00 – 0:51
Coming Up
- GTGarry Tan
What is the state of this, these AI programmers? Like, is it reliable yet? And where are we at?
- HTHarj Taggar
Will we just see software companies have way less employees and converge on a point where you could have unicorns, billion-dollar companies that have, like, 10 people on them?
- DHDiana Hu
If we imagine a world where there could be companies less than 10 employees, maybe you could still be a family, but is that still a good idea?
- JFJared Friedman
I have a controversial argument-
- HTHarj Taggar
All right.
- JFJared Friedman
... against what Jensen said. This one will probably piss some people off.
- HTHarj Taggar
(laughs) Nice. (instrumental music)
- GTGarry Tan
Welcome to another episode of The Light Cone. I'm Gary. This is Jared, Harj, and Diana. And collectively, we funded companies worth hundreds of billions of dollars.
- 0:51 – 1:38
What Jensen Huang said about coding
- GTGarry Tan
And today, we're talking about this one very controversial clip that lit up the internet from Jensen Huang.
- SPSpeaker
I'm going to say something, and, and it's, it's gonna sound completely opposite, um, of what people feel. You probably re- recall, uh, over the course of the last 10 years, 15 years, um, almost everybody who sits on a stage like this would tell you, "It is vital that your children learn computer science. Um, everybody should learn how to program." And in fact, it's almost exactly the opposite. It is our job to create computing technology such that nobody has to program, and that the programming language is human. Everybody in the world is now a programmer.
- 1:38 – 3:16
Now that computers can code, what does this mean for CS?
- GTGarry Tan
So, what do you guys think? Is this true? We're at the dawning of LLMs. We infused the rocks with electricity, and recently, they learned how to talk, and now they can code. What does it mean?
- DHDiana Hu
I guess the question is, are the, are the next generation of founders or young, or anyone who's young looking to figure out what they want to do with their career, should they still study computer science? Is that still a good bet on the long run, do you think?
- GTGarry Tan
Yeah, a lot of us spent a long time telling people over all of these generations, "Yeah, you should learn to code. If you're a non-technical founder, you should learn to code."
- JFJared Friedman
It's like the most important thing to do during college. Like, definitely, no matter what else you do, learn how to code.
- GTGarry Tan
Right.
- DHDiana Hu
So the question is that whether LLMs and AI is just gonna automate all of these jobs. And I think we have different views on it, right? We funded a couple, a number of companies that are actually doing, building coding assistants, that are taking task of developers, and what does the future look like for that?
- GTGarry Tan
I mean, I guess the analogy that you could say, I don't really agree with this, but, uh, you could say that given, um, photography, you didn't have to learn how to, uh, you know, use a paintbrush in order to create representations of real life, and, uh, today, you can prompt using an L- you know, us- using a diffusion model. You can actually, you know, just write out what you want and an image will be developed for you. Will this transition to code? And some of the question that Diana has done a little bit of research on, and I think Jared, you too,
- 3:16 – 11:44
How good are AI programmers right now?
- GTGarry Tan
is, uh, what is the state of this, these AI programmers? Like, is it reliable yet and where are we at?
- JFJared Friedman
Related to, to Jensen's clip is the launch of Devin, which also, like, took the internet by storm and has inspired many founders to go into this area, including a, a lot of the companies that we've, we've funded in the, in the past two batches. It could be interesting to talk about that history and what the state of the art is with AI programmers.
- DHDiana Hu
Yeah, so right now, these are companies that I funded with companies like Sweep. We also worked with Fume. Um, a lot of them are solving a lot of tasks for more junior developers that have to do with, like, fixing the HTML tag here or a bug here and there that's fairly small. But it's impo- a bit more difficult when you wanted to actually build more complex systems, like, "Build me the distributor system over the backend that will scale," that we cannot do today.
- JFJared Friedman
I think it's important to, like, to put context around Jensen's tweet that, like, three months ago, basically, AI could not program usefully at all. It was hitting, like, almost a zero. And what really changed, um, I actually think it goes back to before Devin, I actually think the real unlock for the current surge of interest in AI programmers goes back eight months ago to when the Princeton NLP group released this benchmarking data set called SWE-Bench. And SWE-Bench is a dataset of GitHub issues taken from real programming problems. And so it's a, it's a good representative dataset of real-world programming tasks, the kind of things that programmers actually do. And, um, this dataset finally made it possible for people to really tackle this problem of building an AI programmer and to, like, try an algorithm and benchmark it and see how good it is and to compete with other people on the internet. Diana and I were actually just talking about how if you look back in the history event of machine learning, a lot of the big unlocks came from somebody publishing a, a benchmarking dataset.
- DHDiana Hu
A hard one.
- JFJared Friedman
Going back to the very beginning of deep learning. Do you want to talk about how deep learning actually got started, really?
- DHDiana Hu
Yeah, so this, uh, benchmark with SWE-Bench is very reminiscent of ImageNet, which was a groundbreaking dataset from the lab at Stanford from Fei-Fei Li. And it was a very challenging dataset and one of the biggest one that had a lot of images and lots of classes where the task for a algorithm was to classify and see what the image was.
- JFJared Friedman
'Cause at the time, like, the biggest unsolved problem in machine learning, this is, like, hard to believe, was, like, to look at, to, to get a computer to look at a picture of a cat and be able to tell you, "This is a picture of a cat." That was, like...... totally intractable in 2006.
- DHDiana Hu
For, because a cat can have lots of variations, it's actually a very hard problem, because you have cats that are yellow, that are black. They could be in different positions. They could be, like, sleeping, they could be, like, laying down, and they all look very different. But how do you encode that when you have limited sets on that? So before 2006, the traditional methods in machine learning were more statistical. You would do things, were more discriminant. You would have things like support vector machines. You would use things with feature extraction that were, would hand code it, uh, signal processing feature extractors. And with putting things in, like, the frequency domain or all these sorts of things that people try, or wave lets, whatever. And people tried it, and that's, dataset was really hard. The error rate was like really, really high, like over 30%, 40%. And for a bit of context, human perception on this dataset is about 5% accuracy more or less. And then-
- JFJared Friedman
5% error. Percent error rate.
- DHDiana Hu
Error, error rate, correct. Yes, 5% error rate and then all these standard methods were like 50% or more, or 30 above, so which is really bad. It's like way, way bad. So then came about AlexNet, right, Jared?
- JFJared Friedman
Yep, a group from the University of Toronto, and they had trained a deep learning n- net using GPUs. And it was one of the first cases of people training deep learning networks using GPUs. And AlexNet blew the performance of everybody else out of the water. It was way better than all the other techniques. And I remember the day that that news article dropped. It, like, took the, like, programming internet by storm. I would argue that the AI race that we're in right now was, is, we're literally still riding the wave that was kicked off by AlexNet in 2012. (laughs) Like, it, it, it just kicked off this incredible race.
- DHDiana Hu
Yeah, it was the first time that, at that point, it was getting to that human level perception. Then people found this, this, this phenomenon of stacking neural nets with lots and lots of layers. People didn't exactly knew what was happening in the middle, and people treated it like this black box, was actually starting to work. So the interesting learning from this lesson is that Sweet Bench is that moment in time where we can measure something and then we can get better at it. Because before with ImageNet, there wasn't big enough of a dataset to do that. So we will make progress in terms of programming. But now the question is, are we gonna get to the point that we're gonna get AI algorithms that are just as good as programming with humans? Is coding like an image recognition task?
- HTHarj Taggar
What are the reasons this wouldn't happen? Because so far, like, if you zoom out, you have, uh, programming is one of the most promising early use cases for LMs since they've, like, launched essentially, right? You have, like, the copilot term, which really was the GitHub copilot, specifically, like, a copilot for programmers. Data, compute, everything is scaling. The models keep getting better. Um, we now have, like you said, like, a benchmark and, like, human attention focused on trying to make this better. Like, what, what are the reasons we won't just, this isn't just a straight scaling law?
- JFJared Friedman
Oh, I, I think we will. We're now at, like, 14% on Sweet Bench.
- HTHarj Taggar
(clears throat)
- JFJared Friedman
That's, like, the state of the art performance. And it's still well below human performance. I'm not sure what human performance would be. But certainly a skilled programmer could probably solve most of Sweet Bench given enough time. So, like, I think the Sweet Benchmark is gonna go, like... Is I think we're gonna see rapid improvements for, for the reasons that Diana mentioned. But Sweet Bench is, it's a collection of small bugs in existing repositories, which is quite different from, like, building a new thing from scratch.
- HTHarj Taggar
Yeah.
- JFJared Friedman
And so even when we get to a thing that can solve, you know, half of Sweet Bench, that's still pretty far from something where you could just give it instructions for an app to build, and it could just go build the whole app.
- HTHarj Taggar
Yep.
- DHDiana Hu
I mean, the way I think about it, um, that was kind of what my question is really Sweet Bench, the kind of tasks that are in Sweet Bench analogous to image recognition. But I think programming falls in a different kind of category of problems that it can solve. It is a bigger set because Sweet Bench is, like a subset. It's still, like, in this idealized world. And maybe to put a bit of context, I think in terms of engineering, there's two categories of problems and how we model the world. There's sort of the design world that is all, like, perfect, where you have all the perfect engineering tolerances, all the simulation data, and all the laws of physics work perfect in that simulated world. And then you have the reality, which is messy. I think the world of AI, LMs and all that, I think do a good job with this design world. But when you encountering real world, a lot of stuff breaks and you end up with, "When I was working and building all these engineering system, hot fixes that come in and it's, like, random magic numbers to make the system work." Or, like, you could imagine all these self-driving car, I'm pretty sure there's a lot of magic numbers because it's just the placement of sensors that are like, "Mm-hmm." Kind of like physics. Physics, you have all these, uh, coefficients of, uh, friction and they're not pretty like the laws of physics, like Newton. They're like beautiful equations in this ideal world. But in the real world when you need to get systems to work, like engineering and systems and... For startups, you solve real problems. You encounter friction and there's all sorts of coefficients of friction depending on all the materials, and that world is infinite. So my argument is that I don't think LMs are going to be able to really encompass and really manage the whole real world. The real world is, like, infinite.
- HTHarj Taggar
I feel
- 11:44 – 14:50
Good ideas come from the building process
- HTHarj Taggar
like going to the Jensen original video, I, th- I, you had, basically you were saying, hey, like, basically the dream situation is you type in, "I want, um, an app that..."... helps me share blah, blah, blah photos-
- JFJared Friedman
Yeah.
- HTHarj Taggar
... and the software just magically figures out how to build it.
- JFJared Friedman
Yeah. And I guess one way, like, to build on that analogy, like if I- I- I think the world that, uh, Jensen was envisioning was a world in which programmers are like product managers today. If you think about a product manager, a product manager basically builds an application by writing English, right? They write a spec and then programmers go and they translate that into, like, working code. And so maybe in the future, that's how apps will be built, is you'll just, like, write English and the, like, the- the AI will take care of the translation.
- HTHarj Taggar
Which I think gets into, like, the heart of a- this debate that has always happened amongst engineers and non-engineers in Silicon Valley, which is, how much of programming is an implementation thing? It's just, hey, like, you have the idea and the implementation are separate, versus actually, like, you only get the ideas in the process of implementing. I know, like, Paul Graham is a huge proponent of the latter, right? Like in multiple ways. Like in programming, it's like the whole reason he's such a proponent of Lisp from the early days is you want a very flexible language because you only get the good ideas once you start building.
- GTGarry Tan
And his philosophy actually, uh, translates over to writing, where writing is literally thinking.
- HTHarj Taggar
Yeah.
- GTGarry Tan
You're- the process of actually writing is thinking. And I remember, um, when I was learning how to do YC interviews, watching him and being in the room with him and asking him, like, "Well, ho- you know, what are you exactly looking for?" And, um, one thing that he disabused me of was sometimes people would come in and I'd look at, you know, what they did in the past and, you know, I generally felt like, well, this looks like someone who's smart and with it, and they did some impressive things in the past. Surely they thought through this and they just didn't say it in the meeting. And, uh, one of the things Paul would always say is like, "Oh, no, no, no. If they don't say it, then they themselves do not know." (laughs) Like the writing is actually thinking.
- HTHarj Taggar
Yeah.
- GTGarry Tan
And, um, I guess to sort of torture this analogy, but I kind of like it, that, um, we ha- we're sort of in this moment where, uh, if we take the analogy of, like, the- the camera, like, made it so that you don't have to paint anymore, the subtlety there is that, like, aesthetics in the world still exist. And I think the artistry of creating software or technology products is actually, um, in that interface between the human and the technology itself. So my argument would be if you're doing backend s- software and you're writing APIs and models, um, that might get a lot of help from these types of, you know, uh, AI programmers, right? Like you can actually strongly type this stuff and then you- you can actually use language to translate that into saying what the product should actually do. But there is still an artistry in that interface of what should it actually even do and how.
- 14:50 – 17:52
The evolution of programming languages
- DHDiana Hu
I think that's a very good point, Garry. I think maybe the other thing- way to think about this advent with LMs and programming, if you think about the history of, uh, computer science and programming languages, as we progress, we became more and more in higher language abstractions. So we started with, in the early days, it was just very, very much like coding and assembly.
- JFJared Friedman
Assembly. Yes.
- DHDiana Hu
And it would took, like, so many lines of codes to just do addition, right? Then you went up and did a bit of things like with Fortran and then C++, where you had to like really know about the metal still and manage your own memory. Then you went into things that- with more, uh, dynamically typed languages where you- you didn't have to think about the type, like JavaScript and Python, right? Or duck typing, right? And now this is like a new thing with programming with English-
- JFJared Friedman
Yeah.
- DHDiana Hu
... but you still need the artistry, craftsmanship to come up with the design and the architecture.
- JFJared Friedman
And interestingly, the best programmers today, even if they are programming in Python, they've learned C, uh, and they actually, like, know a lot about how the computers... like, how the steps below the stack work-
- HTHarj Taggar
Yeah.
- JFJared Friedman
... even if they're using the- the higher abstraction.
- HTHarj Taggar
Yeah. I- I was cu- curious to ask, um, everyone here, like, a- another potential counter-example is the natural language to SQL idea that w- has been around for years and years and has never really taken off. And I always wondered how much of that is because it's hard to build and implement, and how much of it is it because it's actually, like, it's not as simple as just, I need someone to, like, translate my thoughts into a SQL query. It's knowing, like, the right questions to ask about the data and, like, having some representation of how the pieces fit together. You have to have some sense of, like, the relational database in your head, at least the concepts to ask the right questions. If it's true that that's- there is some step before of, like, thinking involved, then you can't just extrapolate from like, hey, it's- it's just like we were- we started with like, you know, binary code and we just, like, abstracted all the way eventually to natural language. There's gonna be some, like, gap between, like, the highest level of abstraction you can get and actual natural language.
- DHDiana Hu
I think so. I mean, we- we kind of looked into a lot of these kinds of ideas and funded some companies doing this kinda- this kinda idea. Um, I think AI will get to the point that you could actually do the translation from English to SQL, but I think the hardest part is not that. The problem with all these data modeling, why data engineering orgs are so big, because when I- I had to kind of manage these teams, they're very messy. The reason is because the hardest part is the data modeling, because that's trying to encapsulate the real world and the real world is messy.
- HTHarj Taggar
Yeah.
- DHDiana Hu
We have all these, like, annoying coefficients and frictions that we have to model. It's like, okay, this person talks to who and this workflow works to who? And it's all very, very messy that a perfect model in AI can't really encapsulate, and you kind of need the human to kind of think through it.
- HTHarj Taggar
Yeah.
- DHDiana Hu
And that layer is like, how do you put an LLM to kind of parse through that and translate to the business requirements of the data model? Because if the data model is wrong, then it just causes all sorts of issues and that's where things get hard. What do you think, Jaryd?
- 17:52 – 18:57
The benefits of learning to code, even if computers can do it
- JFJared Friedman
I have a controversial argument-
- HTHarj Taggar
All right.
- JFJared Friedman
... against what Jensen said. This one will probably piss some people off.
- HTHarj Taggar
(laughs) Nice.
- DHDiana Hu
(laughs)
- JFJared Friedman
My argument is that even if everything that Jensen predicts comes true, and in the future, you will be able to build a great app just by writing English, you should still learn how to code because learning how to code will literally make you smarter. We have an interesting piece of evidence for this, which is there's a lot of studies now that show that the way LLMs learn to think logically is by reading all the code in GitHub and basically learning how to code. And I think programmers have long suspected this, that learning how to code made them smarter, but it was kinda hard to prove with humans. And now, we have some actual evidence that this is really true.
- GTGarry Tan
There's definitely some evidence that, um, for some certain class of, uh, problems with LLMs, you're way better off having the, uh, LLM write code f- to solve the problem.
- JFJared Friedman
Problem than to try to solve the problem itself.
- GTGarry Tan
Exactly. (laughs)
- JFJared Friedman
Yeah.
- GTGarry Tan
So tool use is actually, uh, a very weird emergent behavior and property of these systems.
- 18:57 – 23:58
Will we see more unicorns with 10 people (or fewer)?
- HTHarj Taggar
Summing up, it, it's like, okay, let's say that one thing is probably uncontroversial is there is absolutely going to be some subset of programming work that will just be subsumed by LLMs. Maybe it's gonna be junior engineering work, like glue code, a whole bunch of certain type of programming work we can all admit does not involve high creativity, high human reasoning.
- DHDiana Hu
I should worry more about all the dev shops where all the stuff, it gets, like, outsourced, that type of stuff that gets outsourced to dev shops.
- HTHarj Taggar
Or even, like, frankly, like, FAANG companies that have-
- DHDiana Hu
(laughs)
- HTHarj Taggar
... like, armies of junior employees. And so one potential consequence of that is if we're not that far away from the junior AI software engineer is, will we just see software companies have way less employees and converge on a point where you could have unicorns, billion-dollar companies that have, like, 10 people on them?
- JFJared Friedman
Sam Altman had a recent comment about this that also went kind of viral on the internet, the idea that in the future, unicorns could have 10 employees or few- or fewer, which has only hap- well, it's never quite happened. I think WhatsApp and Instagram are probably the closest to that ever happening.
- HTHarj Taggar
Yeah, it feels like we've always had... This has been a, a thought for the last decade plus at Silicon Valley, and we've always had flashes of, oh, like, Instagram gets bought for a billion dollars with, like, 20 employees. WhatsApp gets bought for $13 billion with 15 employees or whatever the numbers are. But we've never seen, like, a sustained trend that we can point to. It's always, like, these flashes. But maybe now we're at the point where we will just see a sustained trend.
- JFJared Friedman
It's interesting, I feel like people who are new to Silicon Valley and new to being founders, they want to have more employees-
- HTHarj Taggar
(laughs)
- JFJared Friedman
... because employees are, like, correlated with status-
- HTHarj Taggar
Yeah.
- JFJared Friedman
... essentially.
- HTHarj Taggar
Yeah.
- JFJared Friedman
And w- we know the, like, more experienced founders who've been doing this for a while, and they are obsessed with this idea of having fewer employees, having as few as possible because after, once you, like, manage a large company with lots of employees, you realize how much it sucks. And that's why everyone (laughs) ... That, that's why this meme has been around in Silicon Valley for a long time.
- HTHarj Taggar
Yeah. It, it feels like there's often two types of people who really push for and are motivated for this smaller employee idea or smaller teams idea. It's that profile, and then it's also just engineers who are naturally more inclined towards, like, computers versus people, don't, are not excited about the idea of, like, managing lots of people and hiring lots of people.
- JFJared Friedman
Which was totally the Paul Graham thing. Like, he was into this in 2005, long before-
- HTHarj Taggar
Yes.
- JFJared Friedman
... it was, like, a trend in S- in Silicon Valley.
- HTHarj Taggar
Yeah. And it had to be a combination of foresight and per- personal preference, right?
- JFJared Friedman
(laughs)
- HTHarj Taggar
Like, just not wanting to be, like, in an office with hundreds of people.
- GTGarry Tan
I met up with, um, Mark Pincus from Zynga here at YC recently, and the most interesting thing he told me was, "I think, at some point, a company gets to about 1,000 people, and even the most forceful, the most sort of with it CEO, uh, you sort of lose the capability to ha- really impose your will on the company right around 1,000 people." And if I reflect on some of the founders that we interact with sort of regularly who have thousands of employees, like, that's actually, uh, sort of what their daily lived experience is. Like, there are these things that, you know, you know are sort of extremely true, the company must go in this direction, and then even then, you're, like, a little bit boxed in, and you're, like, unable to enforce that.
- HTHarj Taggar
I have to say, I feel like of founders I work with, especially sort of the younger, hardcore technical engineers, I think they actually grow into the leading bigger teams and just viewing people as a resource that should be used well. The example I can have, like, uh, Patrick Collison of Stripe. I worked with him on our first startup together when he was, like, 19, and he was definitely the sort of archetype of incredibly intense engineer who wanted to be working on hard engineering problems all the time and viewed sort of too many people around as, like, a distraction from, like, the core work, did not want to be hiring people, did not want to be doing any of this stuff. At some point, I think once he started Stripe, like, something changed where he realized that the way to achieve, like, his ambitions was to just take an engineering mind, like, view the company as, like, another product that needs to be, like, engineered and built. And people are a core component of that, and I think he just embraced the, "I need to be a very effective leader, hirer, manager of people." And so I'm not saying s- in this new AI world, Stripe wouldn't have less employees if it were started today, but it... I don't think he would have this internal motivation to be like, "I need to just not hire anyone so much anymore." It'd just be, like, more of, like, a expected value calculation of what, is it better for me to automate, or is it better for me to, like, rally people and use them as a resource? What do you all think?
- GTGarry Tan
I mean, these are hard things for a young founder to sort of approach. And actually, these are sort of some of the reasons why my startup didn't go as far as I wanted it to. Uh, I think the maybe
- 23:58 – 27:23
A startup should be like a sports team, not a family
- GTGarry Tan
most toxic or, you know, difficult thing that I struggled with was this idea that, like, somehow your startup is your family. And, you know, there's actually a clip online of, um...I think Brian Chesky of Airbnb in a prior era actually like, you know, saying that relatively emphatically. And then today if you ask him, he would say, "Oh, no, no, no. This is definitely not a family." (laughs) Uh, a family has all these old weird traumas. Like imagine, you know, uh, bringing home, uh, you know, a boyfriend or girlfriend and they're like sitting with your family and, you know, they go back and they're like, "Well, what happened there?" Like, why, you know, "Why is that like that?" And it's like, oh, you don't wanna a- you know, like let's, let's not ask about that, right? Like, you don't want to... Like, that, uh, having a family be your model of a company is actually kind of a bad thing. (laughs) Uh, and the much more f- functional version of it is actually a sports team, like here's actually what we're trying to do and, you know, basically we need to win. I think wanting to win, uh, is sort of the ideal analogy, whereas, you know, f- for family, there's these weird things like, "Oh, we just want love." And I was like, "Oh, no, no. That's not what a company is for." (laughs) That's not what a startup is for. We're here to solve problems and win. I guess I really wish that I, uh, had someone tell me that when I was, uh, you know, sort of 27 going through my first, uh, stint at YC.
- DHDiana Hu
I think that's a hard transition. I personally went through that because we were... We went from very small engineering team to very large one once we went through Niantic with Pokémon GO and all of that hyper success with Pokémon GO. It's very jarring when you go from that small intimate team and go into like, uh, engineering org of, like, 50, 100 people. It, it really... That, that, that concept of going from this is your tribe and people and family where, where you really know each other and everyone to getting the best, the p- performance out of everyone is very different. And that's hard. And what could be interesting with this era where if we imagine a world where there could be companies less than 10 employees, maybe you can still be a family. But is that still a good idea?
- HTHarj Taggar
I don't actually believe this truth is worth talking about, is, Jared, to your point of, like, programming sort of makes you smarter, um, th- there's certainly some kind of learning founders go through when they hire people, build teams, deal with conflict, fire people, learn how to get the most out of them. Um, that probably just makes them more effective overall. Like, maybe smart is not the word, but, like, certainly makes you more effective figuring out how to work well with people and get the best out of them.
- JFJared Friedman
Yes. You, you learn a lot about people in the process of having to build a company and a team.
- HTHarj Taggar
Yeah.
- JFJared Friedman
And I, I was thinking about what you said, Harj, about Patrick Collison and how he went from being a programmer to, like, learning how to run a company. And I was realizing, like, that's, that's not just Patrick Collison. That's actually, like, all of our best founders are, like, exactly like-
- HTHarj Taggar
Yeah.
- JFJared Friedman
... that. And sometimes people wonder how we can fund, like, you know, 18-year-olds with no prior management experience and expect them to build a big company someday, and it's exactly that. It's because they treat it like an engineering problem.
- HTHarj Taggar
Yeah. Actually, and that's where you cir- you get back to the sort of programmers are smarter basically. It's like, can you-
- JFJared Friedman
Yeah.
- HTHarj Taggar
... actually just treat everything as a programming problem?
- JFJared Friedman
Programming problem, yeah.
- GTGarry Tan
It all just starts with video games and then learning to code. So (laughs) that's sort of the path.
- 27:23 – 28:55
Applying engineering problem solving to non-engineering issues
- HTHarj Taggar
This is something I took away from... I read the Larry Ellison Oracle biography and it's like a bunch of nuggets from there, but, like, one really interesting one is, there's a period in time where he completely ignored just, like, the finance function at the company because he thought it was the most boring thing in the world. And then Oracle went through a near-death experience where they weren't on top of their budgets and expenses and just almost ran out of money, and he, like, forced himself to have to get on top of it so they would not die from running out of money again.
- GTGarry Tan
(laughs)
- HTHarj Taggar
And, like, the only way he could do it was to be like, "Okay, this is just like... I'm gonna treat this like a programming problem, like it's just numbers, it's a process. Like, I'm just gonna optimize this as though I was, like, coding." And then he got really into it-
- JFJared Friedman
Mm-hmm.
- HTHarj Taggar
... and just actually started really enjoying the whole process of process optimization, which then fed back into Oracle in a weird way because-
- JFJared Friedman
Mm-hmm.
- HTHarj Taggar
... Oracle's business was a lot of, like, going to companies, figuring out which of their processes were messy, and trying to sell them software to, like, solve it.
- JFJared Friedman
He experienced the problem himself and then he built the solution that he wanted and then he was able to sell that solution to everybody-
- HTHarj Taggar
Yeah.
- JFJared Friedman
... else 'cause everyone else had the same problem.
- HTHarj Taggar
Yeah, basically. But, again, it all came from, like, an engineer who wanted to avoid a messy people process problem just taking it on and treating it like a programming problem and actually becoming more effective at it than, like, the team that was built to work on it.
- JFJared Friedman
I see this a lot with our technical program- with our technical founders who are doing B2B companies where they treat their sales org this way. They definitely treat sales like a programming optimization problem.
- HTHarj Taggar
Yeah.
- JFJared Friedman
It's, like, stereotypical actually.
- 28:55 – 36:58
What will happen if AI takes on more programming roles?
- HTHarj Taggar
Yeah. So what do we think the net effect of this is going to be overall? If AI, you know, makes us all more productive, if AI can start taking away some of the junior programming work, do we see a lot more unicorns? Does it make it possible for one company to become worth, like, a trillion dollars or do we see, like, a long tail of lots of, like, unicorns started by much smaller teams?
- JFJared Friedman
And do we think the teams will even shrink? 'Cause, um, if we go back to predictions in the early 2000s, there were a lot of people who were predicting that as programming got more efficient, companies would be smaller. Because in the, in the '90s, to build an internet startup, you had to build everything yourself. You had to build... You had to have people who knew how to rack servers. You had to hire people who knew how to optimize databases. You had to hire, like, people to run payroll. And then all of that stuff got, like, turned into, like, SaaS services or infrastructure, open source. And so, like, you could focus on just your core competency. And there were a lot of people who were predicting that this meant that companies would have fewer employees 'cause they wouldn't need all those people that you needed in the past.
- GTGarry Tan
I remember racking servers, but I bet a lot of people watching this have never even stepped foot-
- JFJared Friedman
Don't, don't even know-
- GTGarry Tan
... in a data center.
- JFJared Friedman
... what that phrase means. (laughs)
- GTGarry Tan
Yeah. What is a, you know, what's a rack? Like, how does that even work?
- JFJared Friedman
(laughs)
- GTGarry Tan
You just go and, you know, click a button on a website and, like, boom, I have a server, right? Like, that's how it works, right?
- JFJared Friedman
Yeah. And the four of us were looking at some data earlier and what we discovered is that it's... I- it didn't happen, actually. Like, companies didn't get smaller. And Har- Harj discovered the reason why.
- HTHarj Taggar
There's this concept in economics called the Jevons paradox, which is essentially once you make any, um, service more efficient, like, you make it cheaper to deliver, you increase demand for it. And so you actually just get more consumption. And, like-Examples would be Excel spreadsheets, making it easier to do financial analysis, did not decrease the number of financial analysts. It actually just, like, increased them. I think typewriters being replaced by word processors come another example of where, yes, the strict role of being a typist and a typewriter went away, but the demand for people with word processing skills went way up.
- DHDiana Hu
So software became cheaper to make but at the same-
- GTGarry Tan
To make and programmers became more efficient, but it did not reduce the demand for programmers. It actually increased the demand for programmers.
- DHDiana Hu
Which I think we actually see it in the number of, uh, companies applied to YC. There was this essay from PG just 10 years ago that he w- he couldn't imagine the world where we'd have more than 10,000 applications per year, and at this point, we're getting over 50,000 applications per year, more than that. It is becoming easier to start companies more than ever because there's so much infra- infrastructure built, but at the same time, the requirements to be good at it and be a good founder are higher. I think it requires having even better taste and more craftsmanship to become the best founder now, right?
- HTHarj Taggar
Yeah.
- DHDiana Hu
Sometimes we joke that if we went through YC now in our younger self, would we have gotten in? (laughs)
- GTGarry Tan
Mm-hmm.
- DHDiana Hu
It's actually very competitive now because the baseline is just so much higher.
- HTHarj Taggar
Yeah.
- DHDiana Hu
So there's this thing that at the end you still need a computer science degree and engineering degree to really build that taste and craftsmanship to really have, know what to build and build it well. You need to whisper to the AI and LLM, but how do you even whisper to it if you don't know how all the stuff works?
- GTGarry Tan
There's this amazing Rick and Morty, uh, meme where there's a little robot on the table passing butter and he goes up to, uh, Rick, the master and he's like, "What is my purpose?" And it says, "You pass butter." And then he goes, "Oh my God."
- HTHarj Taggar
(laughs)
- GTGarry Tan
And the funniest thing about that is like, you know, there's so many people in the world who basically have that job and they're not, like, robots, they're human beings, you know? Like, their nine to five is something that is incredibly rote and not that invigorating or exciting to them, uh, and yet that's like sort of their entire lives. And how could we not celebrate the fact that now we have more software, more tooling, potentially robotics coming around the way, like, that might free that person from having to pass butter (laughs) and they can go off and do something else, something more creative. Like ideally maybe they learn to code, maybe they learn to actually create things way off on the side in areas that, uh, OpenAI or, you know, sort of Microsoft or like whoever the tech giants are, like those companies can't do everything. They probably shouldn't do everything. Not only that, it's not clear to me that Lina Khan will allow that. (laughs)
- DHDiana Hu
(laughs)
- GTGarry Tan
So, you know, given that, actually maybe that's the opportunity, like rather than just a few companies worth a trillion dollars, my, you know, my genuine hope, and I think that we're trying to manifest this world is actually thousands of companies worth a billion dollars or more. And, you know, some of those might have 1,000 employees, some of them might only have 10, some of them might even be just one founder sitting there doing that thing. But at the end of the day ultimately making it better for a real customer, a real problem, a real thing in society that frees someone from being a butter-passing robot that's a human.
- DHDiana Hu
I think that's such a good point, Gary, and I 100% agree with that. I think part of it is we're in this world of post-abundance of sorts where it's easier to... It, it's easier to build things, it is easier to get the infrastructure up and running if you get the right opportunity. Uh, there's a lot of capital too, if you know where to tap, but the bottlenecks is can you enable this equation of human capital to flourish and match that opportunity and get the smart people that can do it and have a lot of the ambition in front of this capital? And this is why right now our job is one of the coolest. We get to do that and enable this flourishment of a lot of people that maybe could have been passed in different situations and give them a chance to build these companies that will go against the trillion dollar ones, right? Just 1,000 billion-dollar companies.
- HTHarj Taggar
We- we have all definitely lived through and hugely benefited from this trend of the more powerful technology becomes, the easier it is to get a company off the ground. Clearly, like just open source software... I mean, I just think back to even when Jared and I first moved here, like Rails was first taking off and-
- GTGarry Tan
That was a huge innovation.
- HTHarj Taggar
Yeah, right? (laughs)
- GTGarry Tan
Oh, that made me feel so powerful-
- 36:58 – 38:07
The verdict - learn to code!
- GTGarry Tan
Well, so it sounds like the verdict is in, learn to code.
- DHDiana Hu
(laughs)
- HTHarj Taggar
(laughs)
- GTGarry Tan
Yes, you should learn to code. Sorry, Jensen is brilliant, but he is not right every single time. (laughs)
- HTHarj Taggar
I- I think one- one thing that is uncontroversial is that over the last 10 years there have been more unicorns started each year (laughs) , right? Like and that's been because technology has made it more possible for people to get their ideas off the ground. I think I- AI only accelerates that trend, right? I think we should just expect to see more unicorns started per year than ever because it is easier to go from getting your idea to like a prototype to your first users than it ever has been.
- DHDiana Hu
And at the same time, it's still table stakes to be able to program and code because so much of the foundation knowledge, you have to have good taste to build something great and you only get the good taste by going and studying engineering or computer science.
- GTGarry Tan
The most important thing to me that I really want to manifest in the world that I think we get to do all the time at YC is that there are people here who are craftspeople or who could be craftspeople, and those are the people who are going to go on to build the future.
- 38:07 – 38:23
Outro
- GTGarry Tan
So with that, we'll see you next time.
Episode duration: 38:23
Install uListen for AI-powered chat & search across the full episode — Get Full Transcript
Transcript of episode CKvo_kQbakU
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