The Frontier with Brendan Foody
00:24:26 · Andy Fang, Brendan Foody
[00:24:26] S1: I mean, for us, it's like super exploratory right now, to be honest. We launched our DoorDash Tasks app a couple months ago, and you know, that's really like allowing our Dasher uh fleet to basically, you know, pick up jobs on the tasks app and see like, you know, if it's whether it's recording yourself doing something or other sorts of tasks, um, like just seeing like, you know, what traction we can get there. Learning a lot there, but I don't know. It's like so early. Like we're still trying to figure out, to be honest, like what the there there is. [00:24:54] S0: So well, and I also think the labs still have a lot of ambiguity of what exact data shapes that they're trying to figure out. And so it feels like it's more of a research problem of what data shapes actually improve the physical models right now, and less of a problem where you can just pour in human hours in the way that we have in the digital world. [00:25:13] S1: Yeah, exactly. Like I don't think there's like a like, hey, this is I mean, I think even the labs don't even know what they want. Yeah, we're still figuring out if there's a there there here, you know. [00:25:21] S0: Makes sense. What other things do you spend your time on at DoorDash that you're excited about? [00:25:25] S1: I mean, I think there's something you guys probably don't have to deal with as much just because you're smaller in terms of number of people, but just AI enablement in general, I think, is something that is top of mind for us. [00:25:33] S0: Like how to get everyone using Claude and be more productive.
00:25:33 · Brendan Foody, Andy Fang
[00:25:33] S0: Like how to get everyone using Claude and be more productive. [00:25:36] S1: Yeah, exactly. And I think on the engineering side, I think that's been pretty organic. And I think on the non-engineering side, we've seen a lot of adoption and we've rolled out Claude co-work pretty widely within the organization. And I would say based on like surveys and stuff like that, people really like it. But I think it's like, how do we actually get that like step function change has been a little elusive, I'd say. Like, I think there's areas where we've seen like maybe 20, 30% gains, which is you know, we'll take it. But like in terms of like, you know, people are like, hey, I like I was just talking to someone earlier today about this. I think it was like everyone talks about how hey, AI is gonna do my taxes, but I'm like, I don't know a single person who's like literally let AI do their taxes without like touching it, you know. It's like for us, it's like, you know, for various functions, like what are those like step function change type of automations that AI can provide? And we're still in the early innings of discovering what those are. [00:26:27] S0: How much of that is people using the Claude Cowork interface themselves versus a custom harness that you guys have set up that they're able to use so that you can swap between models?
00:26:27 · Brendan Foody, Andy Fang
[00:26:27] S0: How much of that is people using the Claude Cowork interface themselves versus a custom harness that you guys have set up that they're able to use so that you can swap between models? [00:26:37] S1: I would say on the non-edge side, a lot of it right now is through cowork. Like we want to build a paradigm where um everybody can build their own agents. And so we want cowork to kind of be that playground to start with. If we find things that we're really excited about, then that's kind of when we'll do the harness. [00:26:54] S0: That's definitely something I've been interested in how it develops because I think that most large companies are just using cowork directly or ChatGPT directly. Yeah. Whereas a lot of startups are starting to build their own harnesses. Like I know Sierra has posted a lot about this. We have our own harness. We still use Claude a lot of the time and that harness, but it at least gives a little bit of optionality with respect to how we change the model that we're using over time and build more customizations on top of it. So I think it'll be interesting to see how that development unfolds.
00:26:54 · Brendan Foody, Andy Fang
[00:26:54] S0: That's definitely something I've been interested in how it develops because I think that most large companies are just using cowork directly or ChatGPT directly. Yeah. Whereas a lot of startups are starting to build their own harnesses. Like I know Sierra has posted a lot about this. We have our own harness. We still use Claude a lot of the time and that harness, but it at least gives a little bit of optionality with respect to how we change the model that we're using over time and build more customizations on top of it. So I think it'll be interesting to see how that development unfolds. [00:27:26] S1: Yeah. I mean, we have our own harness as well that like only like the power user non-technical people are using right now. But like I think for me, what I'm optimizing for right now is like try to identify the use cases. Yeah. Cause I don't think we've seen enough use cases where we're like, oh my gosh, like like we gotta like kind of like really like supercharge this right now. It's more like we're seeing a lot of like 20, 30% type gains in various areas that people are able to kind of like self-sufficiently do. [00:27:53] S0: And when it's people building their own agents, would that mainly be them building a skills file and setting up like a cloud scheduled action so that they can run that agent on whatever cadence? [00:28:05] S1: Pretty much. And then, but we also have our own like harness where like we can also you can spin up like a agent on a cloud instance and kind of do stuff. But like I think that one requires a little bit more wrangling.
00:28:05 · Andy Fang, Brendan Foody
[00:28:05] S1: Pretty much. And then, but we also have our own like harness where like we can also you can spin up like a agent on a cloud instance and kind of do stuff. But like I think that one requires a little bit more wrangling. [00:28:18] S0: How do you think about incentivizing AI adoption? Because there's obviously one school of thought of token maxing and like how do you just get people making as many queries using as many tokens as possible? Yeah. Which I think there is some value in, but obviously there is often a disconnect between usage versus actual outcomes. And then the other school of thought is how do you just look at the outcomes that people are accomplishing? But it's hard to know exactly if you set higher expectations for someone. You know, it's hard it's hard to like calibrate those expectations to what they should be with and without AI. Yeah. So how do you how do you think about those two schools of thought of token maxing versus focusing on outcomes and how you guys incentivize AI adoption through DoorDash?
00:28:18 · Brendan Foody, Andy Fang
[00:28:18] S0: How do you think about incentivizing AI adoption? Because there's obviously one school of thought of token maxing and like how do you just get people making as many queries using as many tokens as possible? Yeah. Which I think there is some value in, but obviously there is often a disconnect between usage versus actual outcomes. And then the other school of thought is how do you just look at the outcomes that people are accomplishing? But it's hard to know exactly if you set higher expectations for someone. You know, it's hard it's hard to like calibrate those expectations to what they should be with and without AI. Yeah. So how do you how do you think about those two schools of thought of token maxing versus focusing on outcomes and how you guys incentivize AI adoption through DoorDash? [00:29:02] S1: I would say we definitely focus much more on outcomes. So, like, for example, like we will try to solicit like concrete anecdotes of where people have used AI to make their job a lot better. Whether it's, hey, I got this project done like three times faster, or it's like someone on the recruiting team like kind of vibe coded their own internal like pipeline uh website tool that like saved them a lot of time, or kind of like what you mentioned, like a schedule action type workflow. And so that's what we're really looking for is like, hey, let's try to like and what we also try to do is celebrate those. So like if we identify that, then we share it in a forum and like that way people can kind of learn from example. And so that's kind of where we're at right now.