Chip Motors CEO demos low-speed electric vehicle aimed at replacing golf carts and second cars, with autonomous parking

Speaker 2: Figma. Agents, meet the canvas. Your AI agents can now create and modify your Figma files with design system context.

Speaker 6: And we are very fortunate that we have our next guest, Jamison Dettweiler, here with us in the TVP at Ultradome. Welcome to the show. Thanks for having me. Please introduce yourself and the company. Yeah. I am Jamison Deltweiler. I am the co founder and CEO of Chip. Mhmm. And we are building what we like to think of as the ideal second car today Okay. Because everything becomes autonomous, the ideal primarily owned vehicle of the future. Okay. Can we get

Speaker 1: the chip that I specked out on the big board? Yeah. We will. Special special car, and I think it would would provide some important context.

Speaker 2: So quick facts. It's electric low speed, light utility How do you rate my spec? I

Speaker 6: love the I I love the the spaceship there and everything and The spaceship. Obviously,

Speaker 1: surf package. Surf rack. Yes. And the TV on the front. Yep. This is pretty good. You could get TBPN on there, it would it would kind of complete it. But good start good start for me. Before we get too much into Chip, what were you doing before this? What led you to creating the vehicles of the future?

Speaker 6: Yeah. I before this, I was at TIER Mobility, which was Bird and Lime's largest competitor in Europe backed by Softac and Movedala. They own spin scooters in The US. So got wound up there by selling my HD mapping business, Fantasmo, to them. So basically creating a digital twin of the world and we built a computer powered GPS. And turns out scooters don't know where they are very well because of GPS. And so Bird, Jumpadoober, and TIER were our biggest customers and scaled up. And then also started building kind of a driver assistance systems for scooters as well. Mhmm. So, yeah, I've been in the mobility space for a while. Interesting. I can imagine driver assistance would be

Speaker 1: great for scooters. I don't know how I don't know how far along it's gotten just because driving around LA, some people on scooters do things that that are, you know, maybe not illegal,

Speaker 6: but definitely wrong. Yeah. Yeah. I mean, say driver assistance is like ADAS in cars, but really it was about restricting their behavior. Right? So it's like, you drunk? Are you going are you on a sidewalk? Are you going the wrong way down a bike lane? And so you're putting that on on there and then just, you know, camera and compete at the edge running there. So it's building a lot of that kinda yeah. Why is, autonomy important for this vehicle? This feels like the last place I would want self driving. I I need self driving in the morning commute. I'm tired. I'm going 40 miles or 30 20 miles, and I just wanna, like, zone out and just get to where I'm going. But this looks fun to drive. This looks like you're driving it around a small community.

Speaker 2: It's not that much of a hassle to actually drive. Why is autonomy important?

Speaker 6: Yeah. I mean, for us, like, where we're starting with the autonomy stack is actually kind of like the inverse of what you'd expect when you get from a Tesla anywhere else. Right? So, like, yeah, we like to say chip is fun to drive, but parking never is. So this is not like FSD in a Tesla. In fact, like, when we do bring autonomy eventually to the vehicle, we'll be, you know, unsupervised in in the way that we do it. Got it. But what we're doing today is we can take a chip, you can pull up to, you know, a busy place in West Hollywood and skip the valet line, hop right out. Chip will go off, take care of himself Okay. Take care of the parking, show up when you tell him to come back either from the calendar as you speak to him or text you, let you know when he's on the way back there Yeah. And pick you up exactly where you want to be picked up. Unlike a Waymo, right, which, like, will make you walk a long distance. And the way that we're able to achieve this, by the way, today is with direct remote control of humans because there's actually a data gap there. Okay. Yeah. Yeah. Yeah. Is there also an easier regulatory approval path? Because if the if it's only autonomous or remotely controlled while there's no humans in the car, it's safer. Does that make it easier to get to market faster? Yeah. So it's definitely an easier path. And for us, like, we think a very responsible path. Because, like, when we run our, you know, we're gonna be running our pilot in Miami Mhmm. This winter, a great market. That's where we're based. Yeah. And we're you know, our early customers are gonna have effectively, you know, virtually unlimited access to it and they'll hop out and we'll take that vehicle. We'll drive it. In fact, we'll drive it at lower speeds Mhmm. On less busy roads, etcetera. And so it's just a way for us to validate and bring kind of that safety level up. And then eventually, we'll drive with people Sure. And we'll move from direct remote control to, you know, full autonomy here over time. Yeah. That makes sense. Jordy, sorry.

Speaker 1: Yeah. That use case is cool. I think, like, think about, like, you have driver surfboard off the beach. Yeah. Yeah. That that. But, like, yeah, there there's something of the biggest benefits of, like, having a driver is just being able to be, I need to run to the grocery store, pull up outside, go go straight in even if the parking lot is, like, crazy or chaotic. You're just going in, coming out, you're not thinking about it. To me, that's the most underrated The number of people that have drivers but not they don't order delivery groceries, it's like very small, but yes. No. No. I'm I'm just saying that's been a in in my experience, one of the most underrated parts, and that's something that you expect the the Tesla to be doing soon, but

Speaker 6: can even make sense in this form factor. When originally we saw the launch video and we were like, why would we need why would I want autonomous driving for something that I'm just driving to the beach or whatever? Yeah. Well, I mean, so you can go to the beach for like if you're living in Manhattan Beach, like, where golf carts are really popular Yeah. You know, it's small, it's easy to park, but there's still no parking. So in that context, you can take your friends there, get dropped off, and you just go back to home and come back. And, you know, so like, really what we're trying to do here is bring this like fun robotic, you know, private driver to everybody. It's why it talks. And it can go pick up the groceries. Right? It can go there and it can do curbside pickup and, you know, the frunk opens. And because the confused Target employee comes out, the vehicle can talk and say, hey, put it in there. We're good to go. Oh, yeah. Do that. Yeah. Exactly.

Speaker 1: It's What do you think what do you think like, it feels like there's been an explosion of, like, just new vehicle startups. You have Slate. You have you guys. You have Amble. Yeah. Like, clearly a bunch of smart people at the same time were like, why is there not this super specialized vehicle? Right? Slate was like, he's going after everyday pickup truck. Yep. Ambles, you know, focus on resorts. Right? You guys are focused on everyday consumers and like in these sort of like, my view, like coastal hubs at least like at least that's where I see you guys getting a lot of traction. But like what what is it snowballing now? Should we expect a lot more vehicles? Is there some like is it just the the advancements of batteries and

Speaker 6: declining costs? Like, what do you think is driving it? Yeah. I mean, I think it's really interesting, like, when consumer behavior changes in such a major way, right, especially around cars. And, like, you know, in a lot of ways, I think, like, the American love affair with the car is dying. Everything's too expensive. And increasingly too, like, we're reurbanizing.

Speaker 1: And even in places like Sorry. I was gonna I was gonna hit the boob. There we go. Yeah. No. No. No. I I I think you're I think you're right. And the the the one some even you can see manufacturers now, all the all the cars started looking the same ten years ago. Yeah. Less and less was like a like nobody goes in there like, oh, need a new Daily. I'm gonna get a yellow car. Right? Whereas historically, like somebody would be like, I need a new car. I'm gonna get like a yellow Volkswagen Beetle or something like that. And

Speaker 2: now it's all like converging around like the Model three format. Right? Or even John was complaining the other day because, like, he's in the market for a Tesla and he's like, I wish that they still made the s. Like, that's the one I want. Or a convertible or a full size SUV. It feels like there's a it's almost like a barbell effect. Like, they're they're in the heyday of the love love affair with the American car, it was you got one car that was an expression of everything you needed. So maybe four doors, but also some power and and a supercharged V eight. But now the the standard, like, pod that just takes you to work is so bland and standard that maybe there's room for a second car that is just completely on the other side of the barbell.

Speaker 6: Yeah. And I think for us, I mean, there's a lot of variables here to, like, how we think about this. And a large part of this is, like, how do you you know, almost like bought on a budget. Like, our car like, our cars are like our fashion and they reflect who we are. Yeah. But today, if you wanna buy a new car, that's cool. That's out of reach for the majority of people. Sure. And so that's part of what we're trying to provide here and make it, you know, from design, but also, like, add a lot of function that supports the lifestyles that we live today. Because, like, we are living closer to home, you know, today. And so it's I think that's a large impact as well.

Speaker 2: Do do you think this can slot in it says $15,000 estimated MSRP for four seats. Top speed, 25 miles an hour, 100 miles of range per charge. I'm gonna test the top speed too.

Speaker 1: I'll I'll be the one to decide what the top speed of the chip is.

Speaker 2: I mean, that does feel like it would be if you are truly a budget conscious buyer who's looking for a commuter Yep. 15,000, you'd go with, like, used Honda Civic instead of this. But is there something where the math actually does work out with more charging places where this does become a commuter? Because 25 miles an hour is just, like, too slow for anyone who's hopping on the freeway for a daily commute. Right? Sure. But we're in LA. Right? You know, we've driven this thing. You know, like, we've driven That's so sad. But that is, like, the future. Like, the traffic's so bad. Yeah. You really do go 25 miles an hour long. Mean, it's totally

Speaker 1: think he's just saying, like, you can have like, very easily in LA, you can just

Speaker 6: stay off of freeways. Sure. And it's not even necessarily slower Yeah. Often. No. No. That's right. Well, yeah. And that's the thing. It's like, if you're, you know, it's 05:30 on a Wednesday Right. And you're coming from downtown back to Venice, you know, like, you can be in a barley to zero to 60 in two point five seconds sitting on the 10, but you're still gonna get there in the same amount of time. It's gonna be a lot more Are

Speaker 1: you being on the you're like betting that there's gonna be wall to wall traffic and then you get your chip out on the freeway and you're like, woah, like. Yeah. Lot less traffic than I was expecting. Forgot it was the World Cup and there's no one on the road.

Speaker 6: Yeah. But, yeah, for us, like, it's like the ICP that we're really focused on is like I I built this for myself. Like, I like to say that, like, golf carts, like, escaped golf communities during the pandemic and the growth area has been, like, young families in Sunbelt regions. And so for me, I live in Miami. I live in Edgewater, basically, downtown area. I'm not gonna put my kids into a golf cart. It's not safe. But I like the idea. Sure. This is built to to facilitate that. And so that's kind of where we're starting in terms of the function and price point. Yeah. And the the golf cart companies have been so asleep at the wheel.

Speaker 1: Like they like when it comes to customization, it's like, oh, you wanna customize your cart? You wanna order this cart? And it's like, you have one thing that you can decide and it's the color and there's like six options, right? And you can't even you can't really change much of anything. And went down, I was excited to see you guys come up and Amble and some of these other companies because I was like, I've just talked to all these legacy manufacturers and they clearly don't care to try to innovate or try to allow this specialization, customization. And for a car like this, which is this weekend, you know, this weekend vehicle Mhmm. You want it to be really expressive and fun because, you know, you're not thinking, I'm gonna just be driving to the office and parking and coming back. It's like it is the place that you wanna express yourself.

Speaker 2: Paul Graham had this back and forth with the creator of Friend, Avi Schiffman, who's working on an AI device. And the question was, how do you counter position against Apple? If you come up with a new electronic device, a consumer electronic device, Apple's gonna do what they did to Pebble with the Apple Watch where they Sure. Launched something with the Apple design language. And and Paul Graham's advice was do something so bold with your design and your aesthetic that Apple can't copy you. And this feels like a little bit of this. This feels like very antithetical to what other car companies are doing such that you could not be fast followed even if the technology is on the same supply chain loosely. This would be such a brand change for most car companies, they couldn't jump in. Is that conscious? Is that accidental? What's the thought between behind all of the customization? I don't know. I think Ferrari might have knocked the

Speaker 1: with the luche a little bit. They might have gotten a little inspired with the luche. Been like, hey, Chip looks pretty good. We should make it no, it's it's it's

Speaker 2: You're gonna chop the top on that thing. Anyway

Speaker 6: Yeah. I mean, I look, I think Apple tried.

Speaker 2: They missed. Oh, yeah. The car. Right? They missed? No. They're coming back. I like Aren't they announcing a They're playing absent in the '40. Yeah. And then and then they're also launching a delivery for, like, a hyper car or, like, a Laman Yeah. It doesn't count. Yeah. Yeah. And you had that in, like, you know,

Speaker 6: Ford had a product in this category, like, twenty five years ago. Really? Ford Think. There's one that, like, actually, like, is around Venice all the time still. And, like Interesting. But, you know, these guys are like, you look at all of these companies and, like, they're all tied to products, I think, that are huge money makers, but, like, create problems for innovation there, right, for them. And so this is an area where it'll be, like, very hard for them to come in kinda come in and innovate and think about it. And you have to approach engineering differently. You have to approach sales differently and marketing. Like, we're like, frankly, like, a lot of ways, I think of us more like a sneaker and culture company than I do about, like, a car company in the way that we are selling, marketing, engaging our community, etcetera. And I think that's gonna be really hard for anybody of the and certainly any of the incumbents to do today. For sure. Yeah.

Speaker 2: Figured. You gotta check it out? Yeah. We gotta go check it out. I well, Tyler, what was that car that I was showing you earlier today? The 1991? We gotta pull that up at some point. But I want your reaction to that car because it's a it's a fantastic one. But let's go take a look at the the actual vehicle. You brought it here today. Yeah. Let's go see. Let's do it. Do we have everything set up? Are we good to go? You got the camera? Okay. We're gonna we're gonna switch over so we can go take a a tour of a real chipmotor. And you got me on the lav. Thank you for coming along with us, Jordy. You want to hop up? Take us through it. Where are we going? Come right out here. Let's see. Outside the Ultradome, second car review we've done. Here we are. Woah, here we go. Wait, it's driving itself. There we go. So Hi. I'm Chip. Look at this thing. Hey, Joey. Hey, John.

Speaker 6: This is amazing. See it. Wow. It's a lot, man. Taller than I thought it would be. Wow. Yeah. It's like and golf carts are actually really tall. So, like Yeah. The typical golf cart that people have is actually, like, taller than every single trim of f one fifty. Sure. And so what happens here is there's a couple things. So you actually want the deck here and your safety structure to be tall because you need to interact with trucks on the road. Yeah. But, also, while this looks big, it's a very efficiently packaged size. It's four feet 11 foot four feet 11 wide. It's about 30 and a half feet long. Yeah. And we've designed this. You'll notice, like, you're sitting like a dining room table, basically. Right? Yeah. You're not gonna be in this thing typically for more than forty five minutes. Sure. So we wanna create something that's really comfortable and that's gonna actually have a ton of space. And so, actually, in production here, we're moving the second row back a bit, and everything is gonna enable you know, for me, I got I still have two kids in rear facing car seats that fit in here well. And so the design exercise, like, started from, like, inside out. There's a boxiness to it that's necessary because for us, like, function does, like, drive form. Yeah. And we also, like, leaned into it because I don't want people to look at this and think that it's a car that can go 60 miles an hour. Right? And so the funkiness of something that we've embraced

Speaker 1: Yeah. It feels to me out of the Pixar universe. Yeah. Yeah. Is that that was intentional?

Speaker 6: Yeah. For me, I mean, I grew up going to Disney World, and, like, I really, like I grew up at, like, the world of tomorrow, and I love kind of the future that the Jetsons created. And we've leaned into, like, a little bit of this kind of retro kind of futuristic aesthetic because I wanna see this positive vision of the future. Yeah. If I'm true,

Speaker 2: It feels increasingly like we're maybe headed toward a Blade Runner future and Yeah. That's not what I wanna see. Yeah. No. Yeah. I like that. I like that. What was the process for actually building this one? Because obviously, this is sort of hand built at this point. This is hand built. Yep. What what how much of the supply chain are you building on something as a demo unit? Are you using others others something else as a platform?

Speaker 6: Well, like, I imagine that, like, you know, you don't have your own leather tannery. That's about Yes. What what how how did this come together? How long did it take to build? Who built it? I'm curious about this process. This this, you know, prototype has been built, oh, literal like, over the last eighteen months and tinker with and optimized a lot. And so we actually imported a small, like, Chinese EV Okay. That we wanted to learn from and Tear it down. Top it? Tore it down. And so I ended up rebuilding. Distilled it. Yeah. Distilled it. Exactly. I tore it down to the chassis, ended up replacing, like, the entire drivetrain, the battery, the whole thing. And so, like, it's kind of over time, most of that, like, disappeared. Yeah. And they allowed us to kinda learn a lot from I mean, frankly, they've been doing these small EVs for a long time. Yeah. And that was really, really good for us. And so, yeah, that that kinda exploration, it's gone for eighteen months, and now we should have production in 10 units in hand in the next, like, four months or so. There are manufacturers for, like, these types of bench seats that you can go to at this point and pull those together? So, actually, our VP of vehicle engineering came to us from Canoe. He's the line executive there, but he thankfully started his career in seating and restraints. And so he's like and it by the way Yeah. In a lot of ways, seating is one of the harder kind of industrial design problems for this. Like, there's so much nuance to doing it well, and there's a lot that gets improved from here. And so we're having actually, like, optional heating and cooling in here too. There's also, like, optional HVAC. And so that's a huge a huge part of our focus. Yeah. But Wait. So you'll have a covered version of it as well. Yep. Yeah. We have a soft top that it will come with. We have hard doors. We have an enclosure here. We have HVAC as well as an accessory. But, like, it's you know, even if you're gonna be operating this in a cold place, you know, like, we took this through Aspen and people absolutely loved it or if you're in Seattle, it's raining. Like, if it's a nice day, you should be able to top off. So our designs, we like to say Driving this home from dinner in a winter storm would be electric.

Speaker 1: Well, Well, if you're gonna do if you're gonna make a mountain version, you gotta get chains, I guess. Yep. Well, it's all wheel drive.

Speaker 6: We use four in wheel motors as well, so it has a lot of torque. So, like, it has no issue in the snow. Crazy. Yeah. Interesting. Very cool.

Speaker 1: Interesting. Very cool. Very cool. I

Speaker 6: want one. Yeah. I can make that happen. What is this track back here for? This is L track. So this is an option here. This is the standard kind of accessory you'll see in truck beds. Okay. This is actually also the same that they use to, like, hold down airplane seats there. So this is an accessory. You can do tie downs and things like that. We have so these seats actually fold down into a truck bed. Sure. So you can use it and put plywood, and so we designed it that way. So, like, the thing is people, like, end up wanting to buy this kind of vehicle because it's, like, the ideal vehicle to go to the beach. But what ends up happening and you've seen his behavior is, like, they're like, oh, this is functional. They'll go pick up the groceries. They'll go to Costco. They'll take the kids to school. And we wanna take that kind of, like, feeling of a vehicle that really empowers you and feels, like, very light and turn up to 11. And so that's a Swiss army knife that you can do, and you can pick and choose what you want. You know? If you're a surfer, you can have your surfboard rack here. We'll have, like, a hand shower for, like, getting sand off and just go heavily into accessory ecosystem.

Speaker 2: Yeah. Yeah. That's cool. It's a great way to differentiate. What what does it take to keep this thing street legal? It feels like right on the fringe, right on the edge. You have mirrors, but no doors. Like, what what else is, like, the bare minimum to make sure that you're always street legal? Yeah. So the category that we're currently operating in is called a low speed vehicle. That's what a street legal golf cart is. It's FMVSS 500. Okay.

Speaker 6: You have to And that that's because of the top you have to cap the top speed. Top speed. Yep. And you can ride drive on roads 35 miles per hour and under. So you can ride on, you know, basically, every surface road here in Los Angeles, you can go on. Wow. And then, yeah, you have to have seat belts, you have to follow lighting requirements, you have to have a backup camera Yeah. And things like that. But frankly, like, we actually think that the category is, like, severely under regulated. Like, golf carts are death traps, there's a lot of, like, really, really bad stories that you hear. Yeah. Of course. And for us, like, we're doing a lot to not only, you know, exceed but push the boundaries here. And so for us, we are doing designing this for a crash at, like, the appropriate speed that we're operating at. We're putting in features that we're gonna be deploying for active safety, leveraging the cameras, and everything else that we have in there too. So Underregulated.

Speaker 1: Yeah. Gonna be open up the Chip Washington office soon and pushing for

Speaker 2: pushing for Total and complete lockdown. So this is basically a one on one. Yep. Is it street legal? Did you have to tow it here? Can you get something like this approved provisionally for a demo drive

Speaker 6: so that you can test it on streets, or is this something that, you have to go through a whole process and they all have to be street legal before Yeah. No. This is street legal. We've registered. Yeah. There's, a custom build there that we've been able to do. Yeah. And we we typically tow it when we come from, like, West LA to here. Just we can drive it. Yeah. Yeah. But Makes sense. I don't like to think about how much this vehicle is worth because we can't get insured for the value I think we have. Yeah. It's a start up with a with a prototype. Yeah. Yeah. You will now find an underwriter. So, yeah, we we take good care of it. I love it. I love it. Thank you so much for time. Impressive how solid it feels as a prototype. Yeah. Like, it feels like ready to ready to ship. Well, we'd love to get you guys, you know, later. We can take it for a drive. Can't wait. Can't wait. Thanks for coming by. Anything we can see in here? Is this open? Or right now. Here is the parting in the front. Okay. But we have a little bit of of wiring. We've got clean up it right now. But yeah. But, yeah, that's where you'll see the TV in the front here. Yeah. Yeah. And then it has gonna have a huge space that you're gonna fit. Do have miniaturized stuff? Yeah. Exactly. And we have you know, my goal is I get I need to have a double stroller that will fit in there that I can keep in here at all times. Yeah. Very cool. No. I love it. I love it. Well, thank you so much. Thank you.

Speaker 2: Jordan, you're welcome. No. This is amazing. Yeah. Awesome progress. You so much for showing it to us. Appreciate it. Alright. Yep. Thanks for coming by and Have showing the a great weekend. Cheers, guys. Very cool. See you. Bye. Bye, Spencer. Goodbye.

Speaker 1: Friendly little guy. Friendly little guy.

Speaker 2: Back to the show. We have a couple more guests coming by. Tyler, did you find the car that I was pulling I was talking about yesterday? Can we can we send that? Also looks It's the Nissan Figaro? The Nissan Right. Figaro Figaro. Right? Figaro. Can we pull up a picture of that? I like the way this one looks. And this is just like for sale on cars and bids right now, this old thing. Also looks a little bit like this old a cool vehicle. Yeah. It's so I'm so glad that consumers are getting options Yeah. Because like wildly different design decisions than

Speaker 1: Amble, which obviously launched right at the same time. Yeah. Amble's going for, like, the very low tech route, like, the most simple possible vehicle. Yep. But, like, they're not even really selling, I think, a screen in it. You like that? I like that. I love on Figaro? I love it. Whereas Chip is going for really creating, like, a consumer electronic

Speaker 2: product Yeah. In in many ways. Tyler, do you wanna send a link to the car that you were looking at picking up earlier? I think it was a little bit of an older car. I wanna get Jordy's review of whether or not you should pull the trigger on this car. Have you ever old have you ever owned an old car?

Speaker 1: Not since college.

Speaker 2: Because yeah. What what do they call it? Vintage or retro? I don't know. You go further back. Classic.

Speaker 4: Yeah. It's really classic cars. Yeah. So we if we can pull this up, this is from early twentieth century, 1922 Chevrolet.

Speaker 2: Oh, Oh, it's prewar? Yeah. I thought you were going postwar. There's a few options. Tyler's thinking about getting back. Yeah. So

Speaker 1: Woah. Woah. Don't you think this is I mean, but why not just go for something horse drawn? Like, if you're gonna go this far back, why not go

Speaker 2: twenty years prior and go horse drawn? I would have thought this car would be put out of business by horses. Because, yeah, if you're gonna go a horse is less maintenance than this thing, it's over for the what what is this? Model a? Model t? Something like that? Well, it's not Ford. So Oh, it's not Ford? Okay. Chevrolet. And April

Speaker 4: Touring

Speaker 1: is what it's called. Touring. I mean, they kinda nailed the design. It's beautiful. They made the perfect car and then ever since then, we've been trying to get back to this. Yeah.

Speaker 2: I think it's wildly illegal to design a car like this. I don't think you could even if you modernized everything. Yeah. We don't even know But we do know how to tell you about MongoDB. What's the only thing faster than the AI market? Your business on MongoDB. Don't just build AI. Own the data platform that powers it. Adam Hunt, a researcher at Cambridge University, has a PhD in evolutionary psychiatry. He said he's recently flipped from being bullish to bearish about AI. He says he's updating his bearishness to be more solidly bearish. Early thoughts and his reasoning laid out here. The whole it turns out if you keep training and scaling the models more, they develop broad new capabilities and lots of domains. Thesis was wrong. Sorry, Demis. The recent batch of models haven't gotten more general. They've gotten less general. This is the most obvious fact that their language outputs have gotten much worse in comparison to e g o three. If they were gaining generalist cape capacities, we would expect them to be describing their work in ever more graceful and comprehensive prose. And, I don't know if that's that accurate. I mean, you hear about, like, claudish and, like, the little languages that emerge from the models to be more token considerate. And so, there is an evolution of language there, but I sort of take his points that it feels very much like the spikes are getting spikier. And, if you scroll down, you can pull up this chart. So, Thomas Puero had put out this mainstream AGI thesis, this idea that there's spiky intelligence. If you zoom in on the top left there, you'll see the circle where there's tasks that AI can do and a human is better at every single task. Then the AI is helping me in some tasks means that there's these little spikes where the AI is maybe 80% as good as you, so it's somewhat helpful, but there's other areas where it's useless. Then, all of a sudden, you scroll over and the AI has a jagged frontier. Sometimes it's amazing. Sometimes it's dumb. And then the final stage is, wow, the AI is unbelievably intelligent, but for some reason it fails at x. And then AGI is when it's still spiky, but every possible point on that blurry blob is beyond the human capabilities. And so it is generally intelligent even while remaining spiky. Now if you scroll down, this individual has a different story that he is playing out. AI is a fun toy that is helping me do some tasks, has a jagged frontier. But then if you go over, he's saying you're going to get crazy, crazy spikes while the actual the actual frontier broadly sort of stays within that human circle. And then, at the final stage, have this extreme strength in a few domains like coding and math, but it's weaker in many other areas. And he claims that through the process of optimizing for coding and cybersecurity in these specific business use cases, you might actually shrink the broad capabilities. But, Tyler over there thinks he knows better than the PhD in evolutionary psychiatry from Cambridge University. No. We were debating this earlier. But give me your pushback because you you think that this is maybe

Speaker 4: maybe too soon to call this or not not exactly. I I I'm not saying that it's, like, not correct, but, like, I think if we want to, like, make models really good at at language Mhmm. Common sense or social reasoning, like, it's it seems pretty clear to me that, like, right now we could do that if we wanted. Yeah. Like, all of the the recent advancements are because, like, we, like, want a really good coding model, so we can make the model, like, really good at coding. Yes. And it's not clear that, like, it's impossible to make yeah. Maybe it's true that the the methods we're using right now to make models better don't apply, like, equally to all kind of types of capabilities. Mhmm. But it it, like, there's a reason that, like, the models aren't really good at social reasoning right now. It's because, like, it seems like social reasoning is not that valuable for the models.

Speaker 1: Just because just Well, even even even writing. Yeah. It's like, look at some of the most iconic exchanges in business history. You're never even like, never like, oh, well when Steve emailed the Adobe CEO, his prose wasn't absolutely electric. Yeah. It's just like, what did it actually say? And so I think it's possible that, like, incredible writing just isn't valuable enough to

Speaker 2: Yeah. I mean, so so I agree with you that that there's a world where, like, every task needs to be hill climbed. There's some debate on the timeline today about chess bench. So Gary Marcus has taken a little bit of a victory lap saying, AGI, smarter than the smartest humans? My ass. Says Kasparov, peak rating, twenty eight fifty one could probably probably could have crushed the best commercial language models in chess when he was seven years old. And so this graph of ChessBench shows that GPT 5.5, Gemini 3.1 Pro, Preview are all around just north of a thousand Elo, which is still impressive. It's crazy, but it's not by any means superhuman. And

Speaker 4: but but, like, every time someone points out something where it's like, the models are bad at this, it does seem like there's a very quick cycle of, like, let's make them good at that. Yeah. Also, like, I think it was Amjad that was talking about this. He basically took an LM and he said, like, help me make a chess bot. And then it was, like, became superhuman. Mhmm. So, like, if the model can just do the thing, like, yeah, maybe it's is it the model doing it or is it the model writing code that then does the thing? Mhmm. It's kind of the same thing. Yeah.

Speaker 2: Yeah. I mean, at Like, certain it's using tools to Well, at a certain point, a really good model would just be, like, I'm going to go download Stockfish and just run the actual chess engine. Like, I don't need Yeah. Wait. I mean, that seems fine to me. I I don't I just think that when people say the model is generally intelligent, they're hoping that it they're just like a natural

Speaker 4: Like, they're implying that chess is, like, indicative of some other cognitive ability. Right? Because clearly, the chess is not the actual thing that matters because the chess can just download Stockfish. Yeah. Yeah. Yeah. I don't know.

Speaker 2: Anyway, fun stuff. Let me tell you about Codex. Codex is a powerful workspace for getting work done with AI agents. Whether you're writing code, analyzing data, creating content, or automating business workflows, Codex helps you move projects forward from start to finish. Faster. No. I really like the dog one. It's great. Anyway, we've been keeping Stanley Tang from DoorDash waiting for far too long. He's a cofounder and chief product officer. Stanley, welcome back to the show. How are you doing?