Why Vibe Coding Makes Taste More Valuable Than Syntax

Jared Friedman 0:00 It's like somebody dropped some, like, uh, giant beanstalk seeds at night and we woke up in the morning. (laughs) And they were like, "Whoa." (laughs)

Garry Tan 0:06 Whoa. What's going on? (laughs) I mean, I think our sense right now is this isn't a fad, this isn't going away, this is actually the dominant way to code. And if you're not doing it, like you might just be left behind. (instrumental music plays) Welcome back to another episode of The Light Cone. I'm Gary. This is Jared, Harj, and Diana, and we're partners at Y Combinator. Collectively, we've funded companies worth hundreds of billions of dollars, right when it was just an idea and a few people.

Garry Tan 0:42 So, today, we're talking about vibe coding, which is from a Andrej Karpathy post that went viral recently. "There's a new kind of coding I call vibe coding, where you fully give in to the vibes, embrace exponentials, and forget that the code even exists."

Harj Taggar 0:59 Yeah, so

Harj Taggar 1:00 we, we surveyed the founders in the current YC batch, um, to get their take on vibe coding. And we, uh, we essentially asked them a bunch of questions. We asked them, "What tools are you using?" Um, "What work ... How has your workflows changed?" And sort of generally, "Where do you think the, um, future of software engineering is going? And how will the role of software engineer change as we get into a world of vibe coding?" (laughs) And we got, like, some pretty interesting responses. Anyone have any favorite quotes that jumped out from the founders?

Garry Tan 1:27 I think one of them that I can read verba- verbatim is, "I think the role of software engineer will transition to product engineer. Human taste is now more important than ever, as code gen tools make everyone a 10X engineer." That's from the founder of Outlet.

Jared Friedman 1:43 I got one. Um, Abi from Astra said, "I don't write code much, I just think and review." (laughs)

Harj Taggar 1:50 (laughs)

Jared Friedman 1:50 This is like a super technical founder whose, like, last company was also a dev tools company. He's like extremely able to code, and so it's fascinating to have people like that saying things like this.

Harj Taggar 2:00 There's another quote from a, a different Abi, Abi from Copycat, who said he's, "I am far less attached to my code now, so my decisions on whether we decide to scrap or refactor code are less biased. Um, since I can code three times as fast, it's easy for me to scrap and rewrite if I need to."

Garry Tan 2:15 And then, I guess the really cool thing about this stuff is it actually parallelizes really well. And so Yoav from Kess60, he says, "I write everything with cursor. Sometimes I even have two windows of cursor open in parallel, and I prompt them on two different features."

Jared Friedman 2:30 (laughs)

Garry Tan 2:31 Which makes sense. Why not three, you know?

Jared Friedman 2:33 Exactly.

Garry Tan 2:33 You can do a lot, actually. And I think another one that's great is, uh, from the founder of Trainloop. He mentions how coding has changed six to one month ago. "10X speedup one month ago to now is a 100X speedup, exponential acceleration."

Jared Friedman 2:49 And he says, "I am no longer an engineer, I'm a product person."

Garry Tan 2:52 Yeah, that's super interesting. I, I think, like, that might be something that's happening broadly. You know, um, it really ends up being two different roles you need. I mean, it actually maps to how engineers sort of self-assign today, in that either you're, uh, you know, frontend or backend. And then backend ends up being about actually infrastructure, and then frontend is so much more actually being a PM. You're sort of, um, almost being like an ethnographer going into the obscure, underserved parts of the pie of GDP, and you're trying to extract out, like, "This is what those people in that GDP pie actually want. And then I'm gonna, you know, turn that into code," and then actually evals are the most important part of that.

Harj Taggar 3:37 When I was running Triplebyte, this was actually one of the things we noticed. It was almost as important as the technical assessment of engineers, when you're trying to figure out who's a good match for a specific company, is there's a certain threshold of technical ability you need. But beyond that, it was do you actually want to talk to users or not? Like some engineers are actually just very, a lot more motivated by working on things where they know who the users are and they get to, like, communicate with them, and they get live feedback, and they can iterate. Um, essentially being a product engineer. And other engineers really don't want to do that at all. Like they find it annoying having to deal with users and they want to just like work on like hard technical problems and refactor code. (laughs)

Garry Tan 4:12 (laughs) That's the backend engineer. (laughs)

Harj Taggar 4:13 (laughs) Yeah. That's it. That's what we call a backend engineer, yeah. Sure. And that's certainly a theme that came up in these survey responses, right? This idea of sort of the LLMs are maybe gonna push people to choose, because the actual writing of the code may become less important, and it's about are you really ... Do you have taste and you want to solve product problems? Or are you an architect and you want to solve systems problems?

Garry Tan 4:34 Oh, and interestingly,

Garry Tan 4:35 I guess, um, one thing the survey did indicate is that this stuff is terrible at debugging.

Harj Taggar 4:40 Yeah.

Garry Tan 4:41 And so you still, the humans have to do the debugging (laughs) still. They have to figure out, well, what is the code actually doing? Here's a bug. Where's ... You know, spot the bug. Where's the code path that, you know, we have some uns- you know, logic error, you know, just didn't figure this out, right? There doesn't seem to be a way to just tell it, "debug." You were saying that you have to, uh, be very explicit, like as if giving instructions to a f- first time software engineer.

Jared Friedman 5:08 Ye- I have to really spoon feed it the instructions to get it to, to debug stuff. Or, you can kind of embrace the vibes. I'd say Andrej Karpathy style, is just like ignore the bug and just like re-roll. Just like (laughs) just like tell it to try again from scratch. Like, it's, it's wild how your coding style changes when actually writing the code becomes a 1,000X cheaper. Like, as a human, you would never just like blow away something that you'd worked on for a long time and rewrite it from scratch because you had a bug. You, you, like you'd always fix the bug. But like, for the LLM, if it can just like rewrite a thousand lines of code in like six seconds, like-

Garry Tan 5:42 It doesn't cost anything.

Jared Friedman 5:43 Like, well why not? (laughs) You know?

Garry Tan 5:44 That's kind of like, um, writing, you know, taking the approach of, uh, you know, how people use, um, Midjourney or Playground when you're trying to generate images. Like, if there are artifacts or things that I don't like, sometimes I don't even change the prompt, I just click re-roll. And I do that five times, and y- sometimes it just works and it's like, "Oh, I can use that now."

Jared Friedman 6:03 (laughs) Yeah.

Garry Tan 6:04 Which is a very different frame of building systems, because you're not building foundationally step-by-step. You're really doing it from scratch, because fundamentally what's going on is like all these-

Diana Hu 6:16 ... tools today are coming from the world of, uh, generated code, that are in this latent space, hidden somewhere, and you have to do it from scratch to find like a different gradient and not get stuck, and then you wanna like add a bit of randomness, get it to regenerate. But I do think maybe, I don't know, whatever next generation of 0.5, maybe we'll get to the point that actually is able to build upon... I mean as of right now, I think most of it is you need to reroll and rewrite, but it doesn't build upon it yet. But we haven't seen any of the coding tools right now work well with reasoning. I think we have-

Jared Friedman 6:48 Well, well, 0.3 is infinitely better at debugging than 3.5 Sonnet, so like it definitely feels like we're headed in the direction where this may not be true in, you know, (laughs) six months, the next time we do this episode. Diana,

Jared Friedman 6:59 do you want to talk about like the models that people are using and the IDEs that the people are using? There's some really interesting trends there.

Diana Hu 7:04 Yeah, I think as we mentioned a couple episodes ago, we already saw this in, uh, the shift started happening, the vibe started to shift back in summer '24, when Cursor was being used by a big portion of the batch, and now by far is the leader. But the other thing that's happening, this is a very fast and moving environment, Windsurf is a fast follower. It's starting to be a very good product as opposed to Cursor, and I think, Jared, you have some first experience with why Windsurf is like better than Cursor.

Jared Friedman 7:35 Yeah, I think the number one reason that people are switching is that Cursor today largely needs to be told what files to look at in your code base, so if you have a large code base, you can tell it what to do but you have to tell it like where to look in the code base. Windsurf indexes your whole code base and is pretty good at figuring out what files to look at on its own. Th- there's, there's other differences too but I think, I think that's the most important one at the moment.

Diana Hu 7:56 Notable, Devin does get mentioned but the drawback of Devin not really being used for serious features is that it doesn't really understand the codebase. It's being used mostly for small features and b- barely, it's like barely mentioned. The other one people still use, uh, ChatGPT, and the reason they use it is because they want to actually use the reasoning models, so it does get posted. People post some of the debugging questions to, to figure out the, use the more powerful models for reasoning, because right now, Cursor, Windsurf are still in the old world, I mean old world less than six months ago, (laughs) of pre-reasoning models, not in the test time compute, so founders are using that. And, uh, there's some founders, some of them are self-hosted as well, self-hosting models because maybe they have more critical sensitive IP, they do that. And now talking about the shifts in terms of, uh, models, the thing about CodeGen, the big game in town that we saw, uh, six months ago was Claude Sonnet 3.5. It's still actually a big contender. Most are still using it but 0.1, 0.1 Pro and O3 meaning all these, uh, reasoning models are starting to see, uh, it's almost like getting neck to neck now close with, uh, Sonnet 3.5. The other one is 4.0, virtually no use for CodeGen, and the other interesting thing is, uh, DeepSeq R1 is getting mentioned, is, uh, it's been used, it's like a viable contender as well. And Gemini not really mentioned.

Jared Friedman 9:26 The, the, the one thing I've heard from Gemini is because it has like the longest context window, I've heard from a couple of founders that they do use it and the way that they use it is they put their entire code-

Diana Hu 9:36 (laughs) .

Jared Friedman 9:36 ... base into the Gemini context window and they just like tell it to like f- fix a bug, and it doesn't always work but like sometimes it can just like one shot fix stuff because it's the whole thing in the context window.

Diana Hu 9:47 It will be interesting to see as people get more adoption on the newly released reasoning models with Flashback 2.0. Uh, I don't think people have tried it yet. But the long context window plus reasoning could be a good contender. What is

Diana Hu 10:01 the estimated code that's being written by LLMs in the current batch? (laughs)

Jared Friedman 10:05 I think this is pretty crazy. So we, we explicitly asked this question, "What percent of your codebase do you estimate is AI generated?" The way I interpreted the question is like, (clears throat) like of the actual characters in your codebase, not including any libraries that you imported, like w- what percentage of like the characters were like typed by human hands versus like e- emitted by an LLM? And the crazy thing is one quarter of the founders said that more than 95% of their codebase was AI generated.

Diana Hu 10:31 Wow.

Jared Friedman 10:31 Which is like an insane s-

Diana Hu 10:33 (laughs) .

Jared Friedman 10:33 ... statistic. And it's not like we funded a bunch of non-technical founders, like every one of these people is like highly technical, completely capable of building their own product from scratch. A year ago they would have built their own product from scratch but now 95% of it is built by an AI.

Garry Tan 10:48 Except for, uh, you know, maybe it sounds like we have-

Jared Friedman 10:50 Oh, yeah.

Garry Tan 10:50 ... one or two examples-

Jared Friedman 10:51 Two examples, that's true, yeah.

Garry Tan 10:52 ... of people who, um, they're so young-

Jared Friedman 10:55 Yeah.

Garry Tan 10:55 ... that they learned to code in the last two years, (laughs) so they actually don't know a world where-

Jared Friedman 11:00 Yeah.

Garry Tan 11:01 ... Cursor didn't exist.

Jared Friedman 11:02 Yeah, this is, um, one of my best companies this batch actually i- is exactly this, the founders are extremely technical minds but they're not classically trained in computer science and programming, and they are incredibly productive and able to produce just a ton of like really amazing product, and AI is writing almost the entire thing.

Diana Hu 11:24 It kind of makes me think a lot of, uh, the discourse around sort of Gen Zs are the first digital native that grew up with-

Jared Friedman 11:32 Yeah.

Diana Hu 11:32 ... the internet. This is like the generation that grew up with native AI coding tools, that they skip the classical training of a software engineer and they just do it with, with vibes, but they are actually very technical minded, I mean they have degrees in math and physics.

Jared Friedman 11:48 Math and physics, yeah.

Diana Hu 11:49 So they, they, they have that raw, let's call it more like system thinking type of mind that you still need, maybe we should talk a bit about that, is like what's

Diana Hu 11:58 still the same and what has changed?

Jared Friedman 12:00 I think this vibe coding will enable people who have those kinds of technical minds, who come from other technical disciplines like math and physics, to become highly productive as programmers much faster than it was in the past. Like I remember there were like coding boot camps, like back in the day they would try to like retrain physics people into programmers and then like it didn't work that well because it just takes too long to learn all of the syntax and all of the libraries and all the stuff that you have to know to be really productive. But like now...

Garry Tan 12:27 Now, it's a new world.

Harj Taggar 12:28 The coding boot camps are also very specifically focused on getting you hired at companies. And I think there was, it was during, this was around like 2015 era, where just companies themselves were rethinking how to evaluate software engineers in their hiring processes. And it was moving... There was a real shift away from, like, "We want to hire classically trained computer scientists, um, white board algorithmic problems," towards, "We actually want people who are just really productive and write code quickly." And, uh, some- some of these arguments are, like, evergreen eternal, right? Like, I remember when Rails first came out, there was just, like, a real sense of, oh, like, I don't know, like, ActiveRecord as a way to, like, interact with your database was seen as a great abstraction. But, like, there was still the same flavor of argument, right? Like, oh, no, like if you don't really understand the internals, like, you're just gonna write, like, crappy low performant web software.

Garry Tan 13:18 How do you feel those arguments have aged, if you look back on it now?

Harj Taggar 13:23 My feeling is that many of the most successful companies, I would say Stripe, Gusto, are just two that really spring to my mind as ones that really heavily leaned into the actually, "We just want people who are really productive with the tools, and we're gonna change our whole hiring process to just select for people who are good at..." Like, the interview shifted from, "Teach us how you think," to, "You've got three hours and a laptop, and you need to build a to-do list app and build it as quickly as you can." And those companies have had a tremendous amount of success. Uh, it does, it does seem like at some point as they grew and they scaled then the bottleneck did actually become having people who are classically trained and systems thinkers to sort of scale up and architect things.

Garry Tan 14:04 It does seem like, uh, how people are hiring engineers is changing, but maybe not changing fast enough yet. The results of the survey are relatively surprising to the four of us here.

Harj Taggar 14:15 (laughs)

Garry Tan 14:15 (laughs) Uh, they're probably-

Harj Taggar 14:16 Yeah.

Garry Tan 14:17 ... pretty shocking out there, even, it's just, like, this thing that popped up in our backyard only in the last six to nine months. My guess would be engineering hiring period has not actually caught up to this. People are still standing at whiteboards and doing that kind of thing, as opposed to, "What can you get done?" And so it sounds like the Stripes of the world, they were ahead of the game, and everyone has to hire engineers this way now.

Harj Taggar 14:41 I mean, I- I wonder if the, actually if even that's going to be sort of the old meta. Like, I mean, something that stood out from the survey responses was this idea of, um, two themes we talked about, right? Like, one is, oh, okay, we're all just product people now. Like, actually the thing that you need is really great taste and understand what to build. And the second was a idea of, um, actually now what's really valuable is to be, like, a- a systems thinker and an architect and to really sort of understand the bigger picture. In which case, actually, like, maybe being a really productive coder, 'cause that's definitely something that always fit my definition of when you're talking about who are great engineers, you know, and like, the, one of the dimensions are they're just really produ- Like, they can write code-

Garry Tan 15:21 Really fast.

Harj Taggar 15:22 ... really fast.

Garry Tan 15:22 Yes.

Harj Taggar 15:22 Like, maybe that's outdated. Like.

Garry Tan 15:25 Yeah.

Harj Taggar 15:25 Maybe that's not i- If the LLMs are actually-

Garry Tan 15:27 That's fascinating.

Harj Taggar 15:27 ... really good at writing code quickly.

Garry Tan 15:29 Yeah. Right.

Harj Taggar 15:29 And to your point, like, now it's actually just cheaper to, like, re-roll and just, like, write everything from scratch-

Garry Tan 15:34 (laughs)

Harj Taggar 15:34 ... and try and debug. Like, the skills might just be completely different.

Garry Tan 15:38 Problem is, there's, like, two different stages. There's zero to one, which, in which ca- speed is the only thing that matters. And then, to your point about ActiveRecord and Rails, that battle was actually fought to a standstill. Because of course using ActiveRecord or Rails allowed you go, to go from zero to one very quickly, but then what happened at Twitter? It became the fail whale, right?

Harj Taggar 15:57 Yeah.

Garry Tan 15:57 Like, basically-

Harj Taggar 15:58 Yeah.

Garry Tan 15:58 ... once you get to one-

Harj Taggar 15:59 (laughs)

Garry Tan 16:00 ... like, one, you know, that architecture-

Harj Taggar 16:02 Yeah.

Garry Tan 16:03 ... will not get you to-

Harj Taggar 16:04 Yeah.

Garry Tan 16:04 ... a billion or 10 billion-

Harj Taggar 16:05 Yeah.

Garry Tan 16:05 ... or $100 billion (laughs) in valuation or users or whatever. Like, it's just not going to actually work. So, I think you're gonna see a, the- the same thing. And then th- there's nuance in what you just said, right? Like, getting zero to one quickly and then being able to scale to a billion users are two totally different sets. And then I think that that might be irreplaceable for now around people. And one of the things I discovered as we were scaling one of the biggest Rails sites, I mean, you scaled one of the Rail- biggest Rails sites too, is that, um, there aren't that many people who have to do it. Getting to one is so rare. (laughs) Yeah, that's true.

Harj Taggar 16:42 That, uh, did you reach this point where how you got to zero to one very quickly was, like, you used lots and lots of open source, and then at some point, like, maybe two years into the st- maybe a year and a half into the startup even, we could not use random gems anymore?

Garry Tan 16:55 Gem- gems anymore 'cause they- they were just never designed for companies-

Harj Taggar 16:58 Right.

Garry Tan 16:58 ... at our scale. And so we had to-

Harj Taggar 16:59 Like, you'd deploy it-

Garry Tan 17:00 ... like build our own stuff.

Harj Taggar 17:01 ... and it would fall over.

Garry Tan 17:01 Yeah, yeah, yeah.

Harj Taggar 17:01 It would just fall over, right?

Garry Tan 17:02 Yeah. (laughs)

Diana Hu 17:02 This very good example is what you're saying, Garry. I think, maybe to summarize a bit more, zero to one will be great for vibe coding where founders can ship features very quickly. But once they hit product market fit, they're still going to have a lot of really hardcore systems engineering, where you need to get from the one to N and you need to hire very different kinds of people. And to that, I think there's a very good ex- historical example, like Facebook as well. I mean, they got-

Garry Tan 17:29 PHP. (laughs)

Diana Hu 17:29 Yeah, they got away with PHP, which personally is a terrible language.

Garry Tan 17:32 Yeah.

Diana Hu 17:32 Maybe fling, but I think it's a bad language.

Garry Tan 17:34 I'm with you. I d- never liked it. (laughs)

Diana Hu 17:37 Uh-

Garry Tan 17:37 Sorry, guys.

Diana Hu 17:38 It was-

Garry Tan 17:38 Anvil products.

Diana Hu 17:38 ... very bad, but you could ship things very quickly. But at some point, it became such a big bottleneck for them to ship features that they had to hire the hardcore system people to build a custom compiler, HipHop-

Garry Tan 17:50 Compile it, yeah. Yeah.

Diana Hu 17:51 ... th- that it would run fast on bare metal because just too expensive to replace all the code.

Garry Tan 17:55 Yeah.

Diana Hu 17:55 And that kind of people that did that are not the vibe code people, are sort of this hardcore systems people that, based on our survey, current tools are not good at that low level systems engineering.

Jared Friedman 18:07 Harj,

Jared Friedman 18:08 I'm not sure everybody who's listening knows what Triplebyte is, but it's actually very relevant. Do you- do you mind just describing for everybody what-

Harj Taggar 18:15 Yeah. Triplebyte is a company I started-

Jared Friedman 18:17 ... Triplebyte was?

Harj Taggar 18:17 ... in 2015, and, uh, we were essentially building a technical assessment for engineers. Like, our goal was, how can you use software to automate evaluating software engineers? And the way we did it was pre all these code gen models, which is, we built all of our own custom software to interview engineers, have humans interview engineers, and then essentially just, like, label the data, like, in every-

Jared Friedman 18:38 And, and interview them by, like, asking them to write code. They-

Harj Taggar 18:42 Yeah.

Jared Friedman 18:42 ... they... Oh, they just wrote high-

Harj Taggar 18:42 It was a, it was a com-

Jared Friedman 18:42 ... highly technical interviews.

Harj Taggar 18:43 Yeah, it was a high tech, yes. Like, it was ask them to write code. Um, we did actually include algorithmic problems.

Jared Friedman 18:49 And is it true that you and your co-founders have done more technical interviews than any other people on the planet?

Harj Taggar 18:53 I think so. Like, in terms of just, like, pure hours. (laughs)

Jared Friedman 18:55 Yeah. (laughs)

Harj Taggar 18:56 'Cause it was, that was just, like, the early days of it were, like, all day, every day, just, like-

Jared Friedman 18:59 Like, thousands-

Harj Taggar 19:00 ... interviewing people. Yes.

Jared Friedman 19:01 Thousands and thousands-

Harj Taggar 19:01 Yes.

Jared Friedman 19:02 ... of them, right?

Harj Taggar 19:02 Um, and then we scaled up and we had, like, a team of about, like, uh, a hundred engineers contracted just, like, that we would pay per interview completed.

Jared Friedman 19:11 And so you're exactly the right person to ask this question, 'cause you've literally spent more time thinking about this than anyone else on the planet. Um, if you were starting Triplebyte again today and you had to design, like, new technical assessments for engineers, what would you have them do?

Harj Taggar 19:26 The- the big takeaway I had with Triplebyte and the screening that, in particular, is just ev-... People want different things. And so you kind of need to know upfront, like, what exactly it is that you're evaluating for, and then design your technical screen, um, around that. It's kind of what I'm getting at with, uh, Stripe and Duster and these companies just knew that they didn't care if someone had fundamental CS knowledge, so it didn't make sense to screen them on that. Like, uh, they, they wanted to screen for the thing that they were actually gonna do in their job. And then our product was more trying to screen for everything comp-... Trying to get a taste of everything companies might want and then figure out what someone's max skill was, um, and then send the people with the max skill to the companies that would value that max skill. And in today's world, I think I would actually have a screen that at least accounted for just how well people knew how to use these tools.

Jared Friedman 20:21 Mm-hmm.

Harj Taggar 20:21 Like, and I'm... So again, it's very contradictory to what I was saying earlier, but it might be the case of maybe how produc-... How quickly you can code and you can build product is actually something to explicitly screen on and just set the bar much higher.

Jared Friedman 20:34 You probably have to ask different questions, 'cause I'll bet if you go back to the original Triplebyte assessment, I bet a lot of those questions, you could literally just copy and paste the question into ChatGPT and it would spit out the perfect answer.

Harj Taggar 20:45 Yeah. Well, yes.

Jared Friedman 20:45 In which case, you're not really proving that much-

Harj Taggar 20:47 Yeah. Well, this is-

Jared Friedman 20:48 ... competence if you're just copy... Like, the, the questions probably have to be, like, a hundred times harder.

Harj Taggar 20:51 Well, it depends. I mean, this gets you the deeper stuff, right? And no- not necessarily, 'cause if you have someone else moni-... It, it depends on, like, what conditions you're going to put on the screen, which I think is interesting. So, I don't know, a classic question was, like, build Tic-Tac-Toe. Um, yes, of course, if you do that unsupervised and you just let someone, like, come back with their Tic-Tac-Toe solution, that's going to take, like, two seconds, right?

Jared Friedman 21:09 (laughs)

Harj Taggar 21:09 Um-

Jared Friedman 21:10 (laughs)

Harj Taggar 21:10 ... um, if you want to watch them code it and force them to not use an LLM, it won't ask that question.

Jared Friedman 21:16 Well, I- I- I- I guess that's the question. Do you force them to code it without an LLM-

Harj Taggar 21:20 That, I think-

Jared Friedman 21:20 ... with, like, the old questions? Or do you let them use an LLM and now you need new questions-

Harj Taggar 21:25 Yeah.

Jared Friedman 21:25 ... 'cause the other, the old ones became trivial?

Harj Taggar 21:28 That, I think is what everyone hiring software engineers right now should be thinking about and trying to figure out. Yeah, I'm not sure I know what the correct answer is to that.

Jared Friedman 21:36 Ah. Interesting.

Harj Taggar 21:37 Yeah.

Diana Hu 21:37 I think there's gonna be... Probably you're gonna test for different things, because it also did a lot of engineering hiring. I think one key skill that's gonna, I think, remain that's constant, I do think skills are reading code and debugging are maxim. Is like, you have to have the taste and enough training to know that the LLM is spitting bad stuff or good stuff. So like, bad code or good. And I think you can see it clearly sometimes if a candidate is using the tools and there's actually a reasonable solution that the LLM outputs, and then the candidate's like, "Oh, this is actually bad." That is a sign. So I think knowing kind of more the high level thinking to know what is good versus bad. In order to do good by coding, you still need to have the taste and you still need that kind of classical tra-... Maybe not necessarily classical trained, but enough knowledge to judge what's good versus bad. And you only become good with enough practice. I think that will be one that will be constant. That would be my opinion.

Harj Taggar 22:34 Yeah, that's interesting. Just coder-... So like, more like code review as the interview versus, like, actually, like, producing code.

Diana Hu 22:40 Yeah, I- I mean, you could have a- some form of system design. You want to know how good they can put a product out there. So it's re-... Again, it's testing for taste. So we're gonna test for debugging and then taste. But then how do you get to... I guess this is a good question, going to these kids that you have that we call it AI coding natives.

Jared Friedman 23:00 Yeah.

Diana Hu 23:01 How do you develop taste when you don't come from a classically trained world? Which would be interesting for next generation founders.

Garry Tan 23:09 Well, you have to, because if you don't, the startup dies.

Diana Hu 23:12 (laughs)

Garry Tan 23:12 Right?

Jared Friedman 23:12 (laughs)

Garry Tan 23:13 So let's say this founder, they go off, they have 95% written by AI. The proof is in a year out, two years out, um, they, you know, have a hundred million users on that thing. You know, does it fall over or not? And then one of the things that's pretty clear is these systems, uh, you know, in the first realm, the first versions of reasoning models, they're not that good at debugging. So you actually would need to descend down into the depths of what's actually happening.

Jared Friedman 23:43 Yeah.

Garry Tan 23:43 And if you can't, then you got, you... I mean, let's hope that they can go-

Jared Friedman 23:46 You're like, stuck, right?

Garry Tan 23:46 ... find another architect who-

Jared Friedman 23:48 Yeah.

Garry Tan 23:48 They're gonna have to hire someone who can.

Diana Hu 23:50 I think there's gonna be a generation of software engineers that are like, good enough, because it's so easy to retool there with, with all these code gen tools. Like, the barrier to entry is so low. You're gonna be good enough engineers. There's gonna be tons of those. But to be exceptional, like the top 1%, I think you're gonna need to get into deliberate practice. I mean, the analogy we're talking about is, uh, Mar- Malcolm Gladwell popularized this concept of 10,000...... hours of practice to become an expert, which came from this research from, uh, what was his name? The-

Harj Taggar 24:22 Uh, Anders Ericsson.

Diana Hu 24:23 Anders Ericsson, right? Which it wasn't just the research was very specific. It was about, uh, how do you find world-class violinists? And it wasn't about just putting the time, but deliberate practice. It's, like, hours that are actually planned and thought and is hard work. You could become an expert with less hours. So, I think what's happening with now cogen tools is that it's very cheap to put in the hours because the output is just so quickly. You can get to good enough, but to become the best in the world and the best founder, you're gonna need that deliberate practice to go into the details, and you're gonna have to peel the onion and understand the systems and get to, again, to some extent, being classically trained. I mean, a good example is maybe we go back to history is, like, Picasso, one of the greatest painters, he was amazing at drawing lifelike pictures.

Jared Friedman 25:10 Which is not what he's famous for. Of course, when w- when you imagine a Picasso, you imagine the opposite of that.

Diana Hu 25:14 Yeah. There's this famous sequence of drawings on how he got to a abstract bowl. It starts from being lifelike to iterations until he gets to the essence to kind of the abstract art that he's very well known for. But he could only get to be the best in the world because he was actually a very good painter and classically trained and could draw super well. But that's not what he's known for. So, I do think we'll see these two classes of engineers, you'll still have, like, a very fat class of, like, good enough. You need engineers for those. But the best in the world, the founders that become outliers, are going to need to put in the deliberate practice.

Garry Tan 25:53 Yes and no. I mean, I think, uh, there are lots of really amazing examples of, um, great systems level eng- world-class engineers who ended up being CEO and CEO of the biggest public companies in the world. I think of Max Levchin, I think of Tobi Lutke from Shopify. I mean, these are people who just, like, actually that great.

Diana Hu 26:14 Mm-hmm.

Garry Tan 26:15 Um, and the thing is, there are lots of other people who are not that great, but also still CEO or co-founders of companies. And then it kind of goes back to, to link up what we were saying earlier, it goes back to hiring. (laughs)

Harj Taggar 26:26 I mean, I- I- I keep thinking about the Twitter analogy that you brought up, because I think it's a really interesting one. Like, if you think, if you compare Facebook and Twitter, in both cases, they went very quickly from zero to one into a scrappy, move fast, break things way. Um, Facebook was able to solve the scaling technical challenges in a pretty impressive way. I think most people would, would agree.

Garry Tan 26:47 I mean, Mark Zuckerberg was by far way more technical-

Harj Taggar 26:50 Well, then-

Garry Tan 26:50 ... and way more in the weeds probably.

Harj Taggar 26:52 Maybe, but I, I don't know, like-

Jared Friedman 26:54 I think Twitter's scalability challenges were also harder based on the usage patterns. Like, the thing about the usage of Facebook is that it grows, it- it's, like, it's pretty smooth throughout the day, like, people just use it all the time. The problem with Twitter is that the usage is incredibly spiky. You get, like, a Super Bowl or, like, a, you know, like, like, a world event, and all of a sudden you have, like, 10 times as much us- usage.

Harj Taggar 27:13 (laughs)

Jared Friedman 27:13 The way the fan out of the feed works is, I think, like, fundamentally a very difficult computer science problem.

Harj Taggar 27:17 Okay, that's fair.

Garry Tan 27:18 Though, I also think that they were, like, really hamstrung by their tools.

Jared Friedman 27:22 100%. 100% yeah.

Garry Tan 27:22 Do you remember using this terrible queue system called Starling?

Jared Friedman 27:26 Absolutely.

Garry Tan 27:27 Like, I used it because I thought, "Oh, Twitter's so much bigger than us. They're so smart."

Jared Friedman 27:30 (laughs)

Garry Tan 27:30 "They wouldn't use something that's crap." No, they totally used crap-

Jared Friedman 27:33 (laughs)

Garry Tan 27:33 ... and then I used crap, and I couldn't make it work.

Jared Friedman 27:35 (laughs)

Garry Tan 27:36 It was like, it was dropping jobs on the floor, like, it's just like all these crazy bugs happened, and then I was like, finally I was like, "I'm not using that anymore. I have to switch to RabbitMQ," or whatever the heck the, the, the actually correct thing to use was.

Jared Friedman 27:49 Yeah, and, and, like, Ruby is an incredibly slow language, even, like, 10X slower than PHP, which was already too slow.

Garry Tan 27:55 So, but I don't know. I mean, basically you should be so lucky to get to one. (laughs)

Harj Taggar 27:59 Yeah. Is there an advantage for a technical founder to be classically trained and be a really deep systems thinker?

Garry Tan 28:05 Definitely. Well, I mean, you just, I mean, a Tobi or a Max Levchin is not going to get bullshitted by people.

Harj Taggar 28:11 Pat- Patrick from Stripe is-

Garry Tan 28:12 I'll tell you-

Harj Taggar 28:12 ... the same.

Garry Tan 28:13 I mean, I'll tell you a crazy story. Uh, when I was, uh, at Palantir, you know, I sort of burnt out there after a couple years, uh, after I designed the logo. And then I actually, between that and going to start my YC startup, I spent six months as a interaction designer.

Jared Friedman 28:31 Hmm.

Garry Tan 28:31 And I was at, it was this, like, terrible venture-backed company that ended up going in the ground, and it was, like, credit card software, it was the worst. I spent six months building, uh, like basically just interaction designs, which was really fun. That's what allowed me to work on my startup in my spare time 'cause I had a lot of spare time. But I remember designing, um, you know, this faceted search thing for, like, rental cars or something like that. And then I go into my meeting with my dev manager and engineers who are gonna implement it, and they were like, "Oh yeah, can't be done. We can't do it that way." Oh, like, and I was like, "What are you talking about? Just make the indexes like this." And they were like, "Whoa!"

Jared Friedman 29:10 (laughs)

Garry Tan 29:11 "What do you mean?" Like, and then they looked up my resume.

Jared Friedman 29:13 Didn't, didn't expect to hear, hear that going back from the interaction designer. (laughs)

Garry Tan 29:15 Yeah, basically. They're like, "How did you know that?" And I'm like, "You fucking lied to me."

Harj Taggar 29:19 (laughs)

Jared Friedman 29:20 (laughs)

Garry Tan 29:20 And that's the thing, like, what founders and, like, you know, when y- when you're hiring people wh- like, tha- that was, like, the wildest thing to me, but it was, like, sort of the ultimate lesson where it's like when you're in the workplace, you sort of assume that the people you bring on, like, they're not gonna lie. They're not lazy. Like, we're all for the goal and the mission, right? And it's like, no, like, people who hire y- you hire, like, they totally will lie to you if you cannot tell that they're lying.

Jared Friedman 29:45 Yeah.

Garry Tan 29:46 And then the worst part is, like, you kind of have to call them on it. Like, you know, sometimes you have workplace cultures that are so polite that people are like, "Oh," like, "I'm gonna let that pass and then I'm gonna talk shit about them behind their back."

Jared Friedman 29:57 The AI-

Garry Tan 29:57 And it's like, "No, you should fire them."

Jared Friedman 29:59 The AI agents, incidentally, will do exactly the same thing.

Garry Tan 30:02 Yeah.

Jared Friedman 30:02 They will absolutely, like, the AI agents will bullshit you just like a human employee will if you don't, i- like, if you're not tactical enough-

Garry Tan 30:08 No way. (laughs)

Jared Friedman 30:08 ... to, like, call them out on their bullshit and be like... (laughs) Like, "No," like, "You didn't make the change that I..." (laughs)

Harj Taggar 30:14 (laughs)

Garry Tan 30:15 What's going on?

Jared Friedman 30:15 It goes back to your, to your point about, like, why being classically trained is still helpful. Like, you, you have to be able to call out all the people working for you, whether they're human or not. Being tactical enough to be able to do that is, is a superpower.

Diana Hu 30:27 So, just to wrap up, basically what's going on with all these tools giving superpowers to the best engineers and making the bad engineers also worse is, this is a quote from the founder of, uh, Trainloop, "How coding has changed six to one month ago, 10X speedup. Now one month ago, 100X speedup. Now exponential acceleration."

Garry Tan 30:48 It sort of crept up on us, actually.

Jared Friedman 30:49 Yeah, it's like somebody dropped some, like, uh, giant beanstalk seeds at night and we woke up in the morning and there it is, like, "Whoa!" (laughs)

Garry Tan 30:56 I know. What's going on?

Harj Taggar 30:58 (laughs)

Garry Tan 30:58 So-

Jared Friedman 30:59 (laughs)

Garry Tan 30:59 I mean, I think our sense right now is this isn't a fad, this isn't going away. This is actually the dominant way to code. And if you're not doing it, like, you might just be left behind. (music)

Jared Friedman 31:11 Right.

Garry Tan 31:11 This is just here to stay and, you know, Vibe Coding is not a fad. It's time to accelerate. So, with that, we'll see you guys for the next Light Cone.