Pod Insights Lightcone Podcast Why The Next AI Breakthroughs Will Be In Reasoning, Not Scaling About this episode There's an ongoing debate about whether AI scaling laws will hold or hit a wall in the near future. However, what's clear now is today's models already have the power to increase productivity in ways that would have been unimaginable even a few years ago.
In this episode of the Lightcone, we dig into the results of a recent o1 hackathon hosted by YC to find out what can be unlocked when founders leverage a SOTA reasoning model.
Transcript Speaker names marked ? are inferred from speaking time and the show’s known hosts — they are not voice-verified.
0:00 Harj Taggar? I remember about a year ago, one of these conversations around, are we going to have AGI? What would that look like? What one of the arguments for it was that, well, like at some point, the AI will get good enough to just like design chips better than like humans can. And then it will just like eliminate one of its bottlenecks for like getting greater intelligence. And so it feels we're on the pathway to that in a way that we just weren't before.
0:20 Garry Tan? The last episode, we were talking about, you know, what are you going to do with these two more orders of magnitude? Since then, Sam has told me that he actually wants to go to four orders of magnitude. It's the worst that these models are ever going to be right now, right this moment. Week to week, you know, there are things that you couldn't do maybe a month ago that you could do really, really well right now. So that sounds like a pretty crazy moment in history.
0:53 Garry Tan? Welcome back to another episode of The Light Cone. I'm Gary, this is Jared, Harge and Diana. And at Y Combinator, we've funded companies worth more than $600 billion. And we fund hundreds of companies every single year. So we're right there on the edge of seeing what is going to work, both in startups and in AI. Recently, Sam Altman wrote this pretty wild essay that predicted that AGI and ASI are coming within thousands of days. Seeing him on Monday, he actually directly estimated, you know, between four and 15 years. Have you guys read this essay yet and what do you think?
1:34 Jared Friedman? Yeah, I read it. And one, one of the interesting places where I think we have a unique perspective is that we were, we had a front row seat to the very beginnings of OpenAI because OpenAI basically spun out of YC. And so what was cool to me, reading this essay, is that it's literally the same ideas that Sam was talking about in 2015, when he started OpenAI. Like, he's been talking about this, like, basically since I've known the guy. And in 2015, when he said these things, he sounded kind of like a crazy person and not that many people took him seriously. And now, 10 years later, it turns out he was right. And actually, we were much closer to AGI than anybody thought in 2015. And now it doesn't sound crazy at all. It sounds like totally plausible.
2:20 Garry Tan? I mean, the essay itself is pretty much the most techno-optimist thing I've read in a really long time. Some of the things that he says are coming are pretty wild. Space colonies, fixing the climate problem, your intelligence on tap, being able to solve abundant energy. Yeah, I think he's basically ushering in this sort of Star Trek future on the back of literally human intelligence, being able to figure out all of physics.
2:49 Jared Friedman? Yeah, Sam was always, like, I remember back when he was starting OpenAI, one of the things that really motivated him to do it was he believed that when we actually had AGI, basically we'd be better at doing science than humans were, and therefore would accelerate the rate of all scientific progress in every scientific field. That was, that was part of the motivation from the very beginning. And I think it's really connected to O1. Even when Sam came and spoke at our, at our, at our batch a year ago, this is long before O1 was publicly released, but it was being worked on in, you know, in, in secrecy by OpenAI. That was the thing that he was most excited to talk about, was giving GPT more advanced reasoning capabilities. And I think this is the reason. It's because,
3:34 Jared Friedman? like, the thing that's missing from its ability to actually do science and, like, accelerate technological progress is it needs, it needs to be able to, like, think through things.
3:43 Harj Taggar? One thing that really stands out about O1 in particular is if you read one of the papers talking about it, so capabilities and potential for the future, it talks about how it does really well in chip design. And I remember about a year ago, one of, like, one of these conversations around, are we going to have AGI? What would that look like? What, one of the arguments for it was that, well, like, at some point, the AI will get good enough to just, like, design chips better than, like, humans can. And then it will just, like, eliminate one of its bottlenecks for, like, getting greater intelligence. And so it feels like that's already kind of like, we're on the pathway to that in a way that we just weren't before.
4:17 Garry Tan? Diana's gonna show a cool demo of her doing exactly that.
4:20 Diana Hu? It's fun, because we run this hackathon with OpenAI, and Sam came over and judged the winners. And one of the participants was actually chip design.
4:33 Diana Hu? This company is called Diode Computer. I think we mentioned them earlier. What they're building is basically AI designer for circuit design. And their previous product, it could handle in the, if you think about PCB design, there's four major steps. The big expensive part that you need a lot of, all of these need a lot of expertise. So the system design, how do you really put together the architecture of it? How do you design all the components, like the resistors I need, the sensors, the specific processing units? Then you need to go do the layout with schematics, then doing the routing. And routing is known to be a NP complete problem because as you have different layers in circuit boards, there's interference.
5:18 Diana Hu? And this is why companies like Nvidia and Intel, Apple have a gazillion electrical engineers because this is a NP complete problem. Up to GPT-4, which is this company have built, it actually put some constraints and was able to automate a lot of the schematic design that you as a human had to design what components it needed to go on the design. And to some extent, the routing, it was simple, which is still pretty cool up to that point. So they were able to automate all that. But the thing that they demonstrated now with O1 was actually able to do the system design and component selection, which is crazy.
6:01 Diana Hu? So it would be able to read all the data sheets and select the right components. So the way the product would work, it could say, I want to build a wearable heart rate monitor with an accelerometer and a microcontroller, very high level. And given this constraint and looking at the database, it would be able to match the specific accelerometer and microcontroller and heart rate monitor sensor and connect it and just output the end result.
6:25 Speaker 04 What we are trying to build today is a wearable heart rate monitor, something like you would see in a Whoop, for example. The O1 is amazing, but one of the downsides is that it's a bit slow. So we actually cached a pre-generated like system diagram that the one was able to generate. It's pretty good. It has a USB-C connector, an IMU like we requested, a heart rate sensor, and like this is a microcontroller. So I'm going to show you how you can go from this and like build a PCB. So we are gonna like build the project. The output of this is code. We actually use AutoPilot, which is an electronics code language. And you can see that it took all the blocks in the block diagram,
7:08 Speaker 04 stitched them together exactly how we want. And the second step is it actually is going to generate a layout for the board. And so now like we can directly open it. And here you go. Here's the board. It's quite, it's quite nice. There's still like a couple of like fine tuning steps required. For example, we could like move like this USB type C connector slightly. We can like change the shape of the board. But, but these are all the components. And then like, thanks to the system that we built, we can like call the auto router on this specific board and actually get a fully working printed circuit board back.
7:48 Diana Hu? So this is actually one of the examples on the O1 paper that it would do EDA, but actually they went a step forward because the example on the paper, they described the EDA as step process is this set of tools for circuit design. It does sort of the design of the schematic, also the simulation and bug verification. It's easier to verify stuff than to select and write it. So this company actually went a step further beyond the paper because the paper did mostly the last stages of verification and simulation.
8:22 Garry Tan? I guess it's an interesting example of using different models for different tasks and in different workflows. So in order to actually pick the correct components off the bat, you know, even before you, you know, place it on a circuit board, you've got to actually have probably RAG on structured, you know, taking unstructured data like PDF documentation and turning into a structured form that then 40 mini sounds like is being used to actually extract the data and then put it into.
8:53 Diana Hu? Into a format for O1. I think this is actually a very common pattern that we're seeing. A lot of the interesting products built with AI, you use different kinds of models. So yes, 4.0 minis for PDF extraction and then 0.1 for the reasoning, because it's actually very hard to select the components for parts. I know, Jared, you also work with a lot of hard tech companies and the whole part of selecting whatever the servos, the motors, the sensors, it's like so, it takes a lot of thinking for human.
9:22 Jared Friedman? Yeah. The other thing I think is interesting about this example is like during the batch, before 0.1 came out, Diode had tried to do this with GPT-4.0 and it just flat out didn't work. And then they basically tried the same thing, the same prompts, but fed it to 0.1 and boom, all of a sudden it worked. And so there really is this sort of like step function capability unlock.
9:45 Diana Hu? They were so excited when I talked to them and they showed me, they were, they had this big smiles like, wow, they themselves are super impressed.
9:51 Jared Friedman? This hackathon that Diana ran incidentally, I think it was a really interesting concept for a hackathon. Like most hackathons are like people who are just sort of like building something that they plan to throw away. And the cool thing about this hackathon is it was all actual YC-funded startups that have real businesses and have funding that have like a real thing with real users. And they were all building actual features for their product that they plan to release to real users. It was, it was really cool, I think, for us to see how 0.1 unlock capabilities for real companies, not just like toy projects.
10:25 Diana Hu? There's another one that was similar in here in terms of reasoning for 0.1. I think Harsh, you work with Camper.
10:31 Harj Taggar? Yep.
10:31 Diana Hu? So, wanna tell us what Camper does?
10:33 Harj Taggar? Uh, it's, I mean, the tagline is Devon for CAD, but basically they let you create CAD designs with just natural language. You just type in like, you know, something that you want to design and it just like spits out like a CAD design for you.
10:49 Garry Tan? So can you design me five airfoils optimized for 50 miles per hour with a minimum drag to lift of 15 at a five-degree angle of attack? That's very specific.
11:00 Harj Taggar? Yep.
11:01 Diana Hu? Normally this would require actually a mechanical engineer to be running all these simulations and solving the equations. And what you're seeing why it's like flashing is like running all the multiple simulations, four of them at the same time.
11:15 Garry Tan? So it's actually kind of like a co-pilot to SolidWorks.
11:18 Harj Taggar? Yeah, they actually built their like initially, they were going to build this as a plug-in to SolidWorks, but they went for like the even harder technical approach, which was like now this is just like a executable that runs on your desktop and it essentially opens up SolidWorks for you and.
11:32 Jared Friedman? And then just starts like clicking around in the UI and pretending to be a person.
11:36 Harj Taggar? Yeah.
11:36 Jared Friedman? Nice.
11:37 Diana Hu? And you saw there what was really cool earlier, they flashed that the math rays. So O1 was actually able to write all of these equations, all these partial differential equations and solve basically Naive Stokes questions to actually solve airfoil.
11:52 Garry Tan? That is really cool. The last episode, we were talking about, you know, what are you going to do with these two more orders of magnitude? Since then, Sam has told me that he actually wants to go to four orders of magnitude to get to a trillion dollars in, you know, sort of spend. I mean, pretty wild, but on the other hand, like you could see where that might go. You know, you could imagine an airfoil is still, it's, you know, very impressive and complex, but sort of what we're capable of doing today in 2024. You could imagine abstracting that to like, understanding the nature of physics, I suppose. Like it would be sort of hard to see that maybe in the current version of O1.
12:35 Garry Tan? But if the scaling laws hold, it seems entirely plausible that, you know, far more difficult engineering challenges such as, you know, room temperature fusion. Like, these are all sort of ultimately engineering.
12:51 Diana Hu? Fluid mechanics, there's weather prediction, there's all these complex physical phenomenas that are very hard to solve and you need basically PhDs. And to Sam's essay, this is a glimpse into what AI and where O1 is heading with this chain of thought and reasoning.
13:12 Harj Taggar? Especially like, like Sam's essay, the vibes are sort of training intelligence and this new age of intelligence. And then the O1 paper just.
13:22 Harj Taggar? The I think this whole idea of now you can actually give like feedback, not just on like the outputs and whether you got the correct answer, but like on all of the steps to get there. And like you're basically teaching a model how to think like the Camper guys are mentioning it to you. The reasoning traces and they can probably go back and like fine tune the various steps for like every output to make sure that the model is thinking how they want it to think. That one that just is again, very like the AGI conversations, I feel like a year ago, we're all sort of in this direction of like, what happens once you can actually start teaching the models to think better versus just like spitting out the correct answers. And then the scaling laws, this is like even more surface area for like throwing compute at the problem.
14:03 Jared Friedman? Yeah.
14:03 Harj Taggar? Right? Like now you can just basically put compute at the inference step and.
14:09 Garry Tan? And iteratively have something come out that you know, you can actually spend more money and more time and have a result that iteratively gets better, similar to what you might expect from a human scientific organization.
14:22 Jared Friedman? Yep.
14:23 Garry Tan? Maybe more consistently even.
14:26 Jared Friedman? Dana, do you want to talk about the architecture and how they actually created O1?
14:30 Diana Hu? I think a lot of it is inspired from what they've been working for many years since the beginning of OpenAI. I think one of the inspirations is a lot of the work they did with Dota.
14:44 Jared Friedman? Yeah. Does everyone remember when like before OpenAI was famous for GPT? The one thing that he was like kind of famous for, that at least people in the tech industry knew was Dota was like winning video game competitions. That was their first big breakthrough.
14:58 Diana Hu? And the funny thing, I think back then Dota wasn't something that took the world for a storm. I mean, maybe only the research community kind of knew about it, but it wasn't anything practical. But what was impressive, it was beating a lot of the best Dota players. So Dota is this complex game of resources and planning, right? And they implemented a lot of kind of reinforcement learning type of techniques in there, which I think were also inspired early days from AlphaGo and AlphaZero as well on how it solved Go. It wasn't just brute forcing through it, but actually having a reward function and. And trying to solve towards it. And even this is why there's just so much talk about Q learning, because that's sort of the fundamental algorithm behind,
15:44 Diana Hu? the family of algorithms behind RL.
15:46 Jared Friedman? So yeah, so like because of Dota, they got really good at doing reinforcement learning. That's how they got it to work. They just had it play against itself like a million games. And then how does that connect to O1?
15:59 Diana Hu? So I think this is where there's a bit of a big stepping function, because how do you then incorporate that into the family of GPT type of models? GPT is all generative, based on predicting the next token and patterns. And then getting those results to check that they're correct. So I think a lot of it is you had to have a lot of data that was factually correct and fed into probably the model and the training and having a reward function that get it to reason a bit more about the output and make sure that is correct. So they probably done a lot of interesting techniques with that. And there's probably a lot of secret sauce on the type of data source that they done.
16:42 Diana Hu? Maybe one of the speculations we could do is a lot of very factually correct information.
16:48 Jared Friedman? Like math problems and science problems and things like that. Yeah.
16:51 Diana Hu? And that's why I outperformed so much in those.
16:53 Jared Friedman? Yes. Yeah. One of the things I think is interesting, Gary, to your point about the scaling laws is a lot of people are really focused on the next like scale up of the model, like the GPT five series of models, which are being trained now and people are working on them and they are going to come out. But I think people may be underappreciating how big an unlock this other direction is because there's there. The two research directions being explored in parallel, right? Like one is the straightforward scale up of the underlying LM. And then this O1 direction is like a totally orthogonal research direction in which you unhobble the model by having it do reinforcement learning, while actually trying to do things in the real world and getting better at them.
17:36 Jared Friedman? The version that's come out so far, it's still only O1 mini. And if you look at the actual
17:40 Diana Hu? O1 preview.
17:41 Jared Friedman? Sorry, O1, O1 preview. And like, if you look at the performance as they released, like the full O1 model, which is coming out any day now, is a huge step function above even O1 preview, which is what enabled all these incredible results at the hackathon. Sam was just telling us that like O2 and O3 are not far behind. And so like, I think people may be underappreciating just how big an unlock we're going to get.
18:02 Garry Tan? Yeah, and O1 also is really opaque still. I mean, from a sort of business perspective, this is a new method. I think at great cost to themselves, they actually did create a new data set to train the chain of thoughts. You know, it's essentially a giant data set of, you know, given task X, can you break it down and into, you know, break it down into parts? And, you know, what's funny is this sort of rhymes with what Jake Heller figured out for CaseText, that if a given task that you give an LLM is hallucinating or, you know, not consistently giving the output you want, you're trying to make that particular prompt do too many things, you need to break it down into steps.
18:45 Garry Tan? And so what's funny is Jake's prescription is really two parts. You know, one is break it down into steps, and then the other part is evals. And it sounds like basically with O1, the chain of thoughts will replace the workflow. So you might not need to break it down into steps yourself, but the evals are still really important. Even like in the aftermath of that episode with Jake Heller, it sounds like some YC alums are reaching out and saying, that episode helped us figure out and unlock something really big. Like, a lot of people really were just raw dogging their prompts.
19:23 Jared Friedman? Yeah.
19:26 Diana Hu? They got to, you have the example of the company you work with at Jared, they got to 100%.
19:31 Jared Friedman? Yeah, just by doing exactly what Jake recommended, which is like having a really big eval set and being very careful about testing every step of your reasoning pipeline.
19:42 Garry Tan? So one of the theories that I have now is ultimately like if you superimpose that on what is a mote? I mean, that's one of the key questions that everyone's sort of asking themselves right now, you know? Okay, like GP5's coming, two more orders, maybe four more orders of magnitude are going to come in terms of a trillion dollars spent on more training. That's pretty wild. You know, if I'm a wrapper company, or I'm trying to do vertical SaaS, or I'm trying to, you know, build my own business, what do I do? My thing is, if I'm trying to do vertical SaaS, or I'm trying to, you know, build my Theory would be it's the evals. It's, you know, write the 10,000 test cases. And the only way you get access to the test cases that are proprietary data that are not like
20:22 Garry Tan? commonly available is that you literally, you know, that's what a bunch of our companies in this current YC batch are doing. They're doing the hard work of doing enterprise sales. They're getting embedded and sort of going quote unquote undercover into these, uh, you know, sometimes really boring jobs, sometimes really complex or arcane jobs, you know, it's everything from, you know, I think accounts receivable all the way over to how do you do like financial accounting or forensic accounting? Like it's just all kinds of things that are really, uh, not readily available. Um, you, you can almost argue that anything that is consumer and publicly available on the internet, that's going to be in the base model.
21:05 Garry Tan? So then your moat ultimately is for all of the other things that are not already online, whether it's, you know, for case techs being a lawyer or maybe over here on science or, you know, in terms of building airfoils, like what you're trying to find is the, the data that is proprietary in some use case, some vertical that allows you to build that 10,000 test case, um, eval. And then that's actually the value. I mean, this is just a crazy theory, but it's something that might, might be what happens.
21:35 Jared Friedman? Totally.
21:35 Harj Taggar? An interesting implication for startups based on everything you just said, um, is it may be worth thinking about who, like of your customers, picking the ones that will pay you a lot for that final, like 10% like accuracy and perfection. I think like Camper are actually a good example of it where there's lots of interest in this sort of text to CAD design amongst like hobbyists or people who want to prototype things and get something up and running very, very quickly. But there's also like a segment of the market where it's people who are literally designing like, you know, airplane parts where there is no room or like margin for error. And O1 makes it quite easy or easier now, right? To get to like the prototype, like, you know, 80% of the way there.
22:19 Harj Taggar? But I think the strongest technical teams. Have the option to go all the way and go after the segment of customers who want like a hundred percent accuracy and will pay a lot for it.
22:29 Garry Tan? Hard to always go all the way.
22:30 Harj Taggar? Yeah.
22:31 Garry Tan? Have to go all the way.
22:33 Harj Taggar? But I think it's interesting because one of the things that gets pushed is does O1 or does AI in general actually make it commoditize a lot of the tech and make it less important to be a strong technical team? And it just seems unlikely to me. It seems like actually the last.
22:49 Jared Friedman? It seems it's the opposite.
22:51 Harj Taggar? Yeah.
22:51 Jared Friedman? Yes.
22:51 Harj Taggar? Like all of the value is probably going to be captured by like the strongest technical teams who can build on top of whatever the base level of tech is and get the final 10%.
23:02 Jared Friedman? Hey Gary, I think it's the prompts, it's the evals and it's also like the UI layer and the integrations that go around it. Because like just the prompts themselves are not a product for a company to actually adopt. Camphor, like it needs to actually integrate into their existing tools. It needs to have a well thought through UI and workflow and all the tooling to sort of make the prompts useful.
23:23 Harj Taggar? Yeah.
23:24 Garry Tan? Well, and then it's distribution, right? Like, how do you actually get in front of people? How do you establish your brand? And then a perfectly good mote is difficulty switching, actually. And once you have all your data and it's working and you're paying $10,000 or $100,000 ACV, sometimes a million to 10 million ACV. You know, man, it's going to be hard to switch. So all the classic modes still apply. You know, this is still software, but you know, you can unlock this capability and you know, this is a moment, you know?
23:56 Diana Hu? Another point to double down on the importance of evals is that that still applies in the world of O1. As founders are wondering, how am I gonna still build the best product on top of O1? Does it change? And everything we discussed in the episode with Jake Keller applies because GigaML is this company that Harshy work with.
24:17 Harj Taggar? Yep, and Gary too.
24:19 Diana Hu? Gary, right?
24:20 Garry Tan? I adopted them.
24:25 Diana Hu? We tell us a bit about what they, what they do.
24:27 Harj Taggar? The full backstory is we funded them for a completely different idea, something like they're an Indian founding team and the original idea was something about helping Indian Indian high school students apply to US colleges, a very niche idea.
24:40 Garry Tan? But they're super crack IIT AI engineers, researchers.
24:44 Harj Taggar? Yeah, and it just happened, like, we were like, this is not a great idea. AI is, you know, changing the world and you literally, like, your research, like, that you've been doing at university or college is all aligned with, like, in particular, like, fine tuning LLM models.
25:00 Jared Friedman? Originally, it wasn't even the AI version of helping Indian high school students apply to.
25:05 Harj Taggar? No, it had nothing to do with AI, actually. So it's like a classic YC story where it was like, these are two clearly brilliant engineers. We don't like the idea at all, but we should just fund them anyway and hope something works out. And the idea they actually pivoted to initially, which they raised a seed round for, was helping companies fine tune open source models so that they could get to like equivalent performance. As, you know, at the time, it was really only OpenAI, but well, I think in general, what we found is that those have not proven to be great businesses because just like the cost of the models has gone down and also like, you know, the performance of the open source models has gone up.
25:45 Garry Tan? You just haven't had to fine tune as much as people thought you would need to.
25:48 Diana Hu? Because the models are just keep getting better. It's kind of betting on the different, on the opposite direction on AGI that let's just trust that these models are going to keep getting better and better, which ergo doesn't require as much fine tuning.
26:01 Harj Taggar? Yeah. And so they pivoted again into, well, like, let's just find a, like, this is fine, like, we're really good at AI now, we're like world experts in like fine tuning, squeezing performance out of these models, so let's just like find a vertical application for that. And they went into AI customer support, which is like competitive, but again, I just think if you're an intensely technical team, you can still find ways to squeeze out, like.
26:30 Harj Taggar? A comparative edge against other teams in the space. And I think that's what they've done.
26:35 Diana Hu? The problem with customer support is you're dealing with very kind of squishy problems. There's just so many edge cases. It's just the space of things that could go wrong as a customer rep is enormous.
26:47 Jared Friedman? Well, it seems competitive, but the thing is hardly any adoption has actually happened. Like, it's not like the world has replaced all the customer support agents with AI yet. We can all see that it's going to happen, but it hasn't happened yet. And so from that standpoint, like, it's wide open.
27:05 Harj Taggar? What I, what I found, at least when I spoke to the Giga ML team last, is that part of the reason for the lack of adoption is that rules-based systems work fairly well for most like of the simple cases. And there's just not trust or belief that you can like build AI that's good enough to solve like the real messy stuff. And so most companies that were pitched on like an AI customer support agent were like, well, you can't actually go all the way and solve like the like hardest problems that take up most of the time. And the rules-based system works like totally fine for everything else. And so I remember when they were first pitching this idea, people would just be like, this is just overkill. We don't need it. A rules-based system works totally fine, but like seems like that is no longer the case.
27:47 Jared Friedman? Yeah, because they now have some really legit customers. Who's?
27:50 Harj Taggar? Zepto just signed up.
27:52 Jared Friedman? Okay.
27:52 Garry Tan? So last time I did office hours with them, they said that they automated 30,000 tickets per day. And so they, you know, I think Zepto had more than a thousand people working on those 30,000 tickets per day, which is, you know, 30 tickets a day. And then the interesting thing was, you know, on the one hand, you know, this is probably one of the things that frankly, everyone, when they think about AI, they're a little bit worried, like, you know, are these jobs going to go away? And the interesting thing about the Zepto customer support job is that it's so not a fun job that I think the turnover rate was something like a few months, like, you know, most customer support agents only wanted to work there
28:35 Garry Tan? for, you know, six months or less. So, you know, this actually is an interesting case of when, you know, when something is incredibly rote, it's literally replacing butter passing. Like, these are sometimes not really actually good jobs. And, you know, hopefully those people can go and do something way more awesome with their time and, you know, beautiful brains than. You know, these rote jobs that-
29:00 Jared Friedman? Then apologizing for Zephto orders that got misplaced.
29:03 Garry Tan? Exactly, right.
29:05 Diana Hu? But the crazy thing they figure out with O1 is that, to your point, Harsh, their previous implementation before O1 was GPT plus rules and all that, and it would not be able to handle most of the cases. It would have about a 70% error rate. Now, what they did is doing the technique like Jake Keller described with really going hardcore on the evals plus O1. During the hackathon, they got to only 5% error, which is, that's an order magnitude improvement.
29:37 Harj Taggar? The other row in this is incredible too, right? This is what I'm saying, like the complex, like the things that are like very complicated that take up lots of time and are expensive to solve, were like essentially like they could not do them.
29:47 Diana Hu? They were shrunk.
29:47 Harj Taggar? Yeah, so basically just like 0%. And that's what I'm, that's what they were encountering when they were selling this is a lot of people like, well, actually, all of the, the stuff that we want to automate are these like complicated edge cases that waste lots of time. And like they just, they couldn't actually do any of that. But like now they're at like 15% and that's with like O1 preview alone. I was like, that's, is that 15% of just error? So now they're at like 85%.
30:11 Diana Hu? Yeah, so they went from 0%.
30:13 Harj Taggar? 85%.
30:14 Jared Friedman? It went from 0% accuracy to 85% accuracy, yeah.
30:17 Garry Tan? So the interesting thing here is that O1, it's not even O1 yet, it's O1 preview. And then it's such a new technique that I think they're trying to protect their advantage right now. So, you know, if you use O1 in ChatGPT, it looks like it will tell you what's really going on. But apparently they have a fake model that just spits out things to give you the impression that it's breaking it up into steps. And they've actually, you know, hidden it because they don't want other people to have access to that data yet. But the next step seems like it needs to be some interpretability, some directability. And then for that to happen, you know, I'd be curious if O2 ends up having that.
31:01 Garry Tan? Like you want to be able to see, okay, well, show me the work, show me the steps and oh, like that step, the third step, you know, can we rerun this, but I want this to branch in this way?
31:11 Diana Hu? Or edit it. I think this is one of the things that would be the next unlock is right now it has the plan that it comes out, the chain of thought, but you cannot edit it. So imagine now, right now, today, O1 just outputs whatever 15 steps to the problem that you need to solve. And imagine now being able to edit each of the steps. Then you get into the super, super fine to next level of Jack Keller.
31:32 Garry Tan? So this is the, it's the worst that these models are ever going to be right now, right this moment. And, you know, literally week to week, you know, there are things that you couldn't do maybe a month ago that you could do really, really well right now. So that sounds like a pretty crazy moment in history.
31:50 Diana Hu? So we've been talking a lot about the kinds of companies and ideas that get this wave of uplift from this model improvement for O1. What are the kinds of ideas that are the opposite? That are not getting benefited as much from O1 and perhaps maybe people even should pivot? Because they're getting, they might just get deprecated from the improvements of O1, O2, O3.
32:18 Harj Taggar? I wouldn't go all the way and suggest they should pivot, but I do think companies that are building AI coding agents or AI program engineers are potentially have stuff to think about here because it seems like O1 in particular is like outperforming on just, you know, solving programming problems essentially. And I certainly know some of the teams I've worked in the past, like a lot of what they've invested in is like the chain of thought infrastructure behind this stuff, which is now just, O1 doesn't, is not actually like any leap forward for them. They've already like invested in that already and so.
32:56 Garry Tan? I think that might be a function of basically the opaque nature of what the chain of thought is. And once you get it to be directable, that's actually, I mean, frankly, that's what users and code gen are struggling with even right now. Like, once it starts going down a certain path, you can't really alter things. Like you, you want it to ask you, hey, do you want me to do it like this or that? And, you know, all of the systems are a little bit struggling with that right now.
33:23 Jared Friedman? I was gonna ask the, the inverse question, Tiana, which is like.
33:29 Jared Friedman? Each new model capability unlocks a new set of startup ideas. Like a year ago, doing startup ideas where like the AI agent would talk on the phone just like didn't work. We had a bunch of companies that tried and all the companies didn't, didn't work. And over the summer, it really started working. And under the trends from the, the, the past two batches, like anything around like phone calling is like blowing up right now because the models finally work. So like with this new O one series of models, what are the startup ideas that like just became possible?
34:00 Diana Hu? To connect to Sam's essay is a lot of things that are gonna make the atom world, physical world better because it's really good at math and physics. So any startup that's working around mechanical engineering, electrical engineering, chemical engineering, bioengineering, all of these things that really will make our lives better. I think really will are getting an unlock as we've seen from the demos we highlighted.
34:26 Garry Tan? That's exciting. I mean, it can't just be helping people click a little bit faster. It's gotta be things that actually create real world abundance for everyone. And it might just be a little bit of a race. Like I think there's sort of the fear of AI out there in society right now. And then it's sort of up to the technologists to try to usher in this age of abundance sooner rather than later. And if we can do that, then abundance will win out over fear. So with that, I think we're out of time for this week of the Light Cone. We'll see you guys next time.
Next Vertical AI Agents Could Be 10X Bigger Than SaaS Page updated November 6, 2026 · transcript and summary are machine-generated