Unsupervised Learning with Jacob Effron
00:32:39 · Erik Bernhardsson, Jacob Effron, Patrick Chase
I I feel like, in a way, like, because of the training data and and the training loops, like, there's more, like, you know, like, incumbency power in a way. Like, there's more staying power in software today. Like, if you get inside the training data and you get the models to recommend, you know, a a certain framework, I... I've heard people say, like, oh, I'm just gonna use React for everything because agents are really good at it. I'm not gonna try these, like, new fancy front end frameworks. Right? Like, I I think there's, like, something similar with, like, a lot of the frameworks today. It's like, that staying power is actually very, very valuable now. So so we obviously wanna make sure that agents recommend modal as much as possible because that creates, a little bit of flywheel. [00:33:19] S0: And one thing I'm I'm struck by is you obviously had this, like, very prescient bet in 2023 on on sandboxes. Right? It sounds like you guys had kind of this instinct of, like, hey. This is gonna be something that's required for coding agents, and then it took a few years for it to really take off, and then it became this really important primitive. I'm curious, is there anything else that you guys are working on today that you feel like you're a little early on, but you're equally as excited. And and how much of your product portfolio is, like, those kind of bets versus, like, okay. We know this is a hair on fire, like, here and now problem. [00:33:48] S2: Jacob, are you are you fishing for the next thing we should be investing in?
00:33:48 · Patrick Chase, Jacob Effron, Erik Bernhardsson
[00:33:48] S2: Jacob, are you are you fishing for the next thing we should be investing in? [00:33:51] S0: A 100%. A 100%. I'm nothing if not shameless. [00:33:54] S1: I I I feel like Sandboxes, they did create a little bit of, like, a high bar for us. Like, everyone's like, what's the next thing? Like, you know, what's the secret sauce you're working on? And, like, what's cooking in the lab? You know? And and I I I wish I had, like, an easy answer. Like, yeah. Like, a 100% confidence, you know, this thing is [00:34:09] S0: Would you have answered sandboxes in '23, or was it just, like, part of a bunch of things you were doing?
00:34:13 · Erik Bernhardsson
[00:34:13] S1: No. Because what sandboxes represented back back then was, a little bit of, like, an asymmetric bet. Like, sandboxes was not, like, a very hard thing for us to build back then. Like, to getting out, like, a prototype of sandboxes took, like, Akshat, like, a few days. And then I spent a little bit of time, you know, like, improving the SDK, some other people, like, started building. And then we started seeing some customers. We rewrote it. And then in 2025, when it started exploding, then we had to rewrite the system, like, five times just to handle, like, more and scale and make it faster and and and more reliable and and things like that. So... But but but the initial investment was actually pretty pretty basic. So in in that sense, like, it was a little bit like a sort of asymmetric bat, like a a kind of a low downside, high upside bat. I don't know if there's, like, anything similar today that that we're working on. I am... I mean, like, just just kind of called out, like, a couple of, like, you know, things that I'm quite bullish on, and and these probably represent, like, way larger investments, like, you know, larger risks and and, you know, in terms of, like, the the bat size, Low latency. Like, I I think, you know, there's gonna be a lot of voice to voice, a lot of real time video type stuff that requires us to rethink a lot of how we manage to compute, like, globally. How do we get down the number of hops, like how do we scale up and down in different regions, like almost like building a CDN for inference. So I'm pretty excited about that. I think training is actually kind of an underrated thing. I think there's a lot of opportunity in in in training, like mid sized models and reinforcement learning. That's been a big bet for the next... For for the last couple of years, like, getting all the RMA InfiniBand type stuff working. What else? I I think a lot of VM, like, stuff is is quite interesting. Like, I think, you know, supporting full blown operating systems like Mac and Windows and Android and iPhone could be kinda cool. I think there's lot of upside there. So I I I have a lot of those ideas in my head, and we we we... We're, like, building a lot of those things. I I don't know if there's anything that represents the same, like, sort of explosive potentials like sandboxes.
00:36:05 · Erik Bernhardsson, Jacob Effron
I mean, maybe, like, if we have this podcast again in, like, three years, like, we'll we'll kinda look back and [00:36:11] S0: We're we're gonna make it recurring at this point, so we'll we'll we'll look we'll look forward to to revisiting that. You know, I I guess obviously sandboxes have become this fundamental primitive for coding agents, and it feels like that then brings you into this world where there's like, I imagine you look at the current state of coding agents, you're like, God, there's so many more tools that would be helpful for these agents to make them more effective for people that use them. How do you think about, like, you know, whether you further build into that world or, like, the the set of the things that are interesting for you there?
00:36:39 · Erik Bernhardsson
[00:36:39] S1: We we do think about it, but we... I think we think about it more from an infrastructure perspective. I I think it's very easy to kinda get lost in, like, you know, whatever is, like, the latest today. When I look at, like, the architecture of, like, agent harnesses and, like, how we're building agents and stuff like that, I I still kinda feel like we're so early in these architectures, like, evolving so fast. And, like, what our job to some extent is, like, trying to figure out is, like, what are, like, the sort of core building blocks that we can put out today that, like, will last, you know, like, people would need them for the next five years. And and that to me is, like, less obvious. Like, we we are interested in, like, sort of agent orchestration as a category. I think there's a lot of interesting stuff there. We're interested in, like, you know, evals and model gateways and all kinds of other stuff around, like, you know, like, the the whole, like, you know, life cycle of of of learning from agents and and innovating on agents. But but I think it's, like, less clear exactly, like, how the architecture. My my view is, like, it will be more cloud centric, and it will be probably more, you know, like like, durable objects, like, kind of workflow engine type driven. And we have some ideas of, like, you know, stuff to build in that space. But but I I I I find it, like, harder to, like, you know, lay out exactly what agentic development will look like in five years. Partly because, like, I also think, like, to some extent, like, all software will, like, basically be agentic. Like, I I kind of hate hate the word agentic in certain ways because I I feel like it's like, okay. So it's using AI to make decisions. Like, well, shouldn't all software do that? So, like, I I almost, like, feel like if you look at it on a five year horizon, like, I I think, like, software... Like, all software will become agentic. And then what does software look like then? Like, I I don't know, but I, you know, I I I hope we can build a lot of the building blocks to support all kinds of aspect of like, know, in general, just like how software uses AI.
00:38:22 · Jacob Effron, Erik Bernhardsson
[00:38:22] S0: Yeah. I mean, it's still like one of the hardest parts of building AI for a company, right, of... You know, across the board for many people, you know, building AI for companies of, what actually persists versus is a pattern is like rag from two years ago or whatever it is. [00:38:35] S1: Rag is a good example, exactly. Like if we had made a bet on like rag like two years ago, it would have been kind of a bad thing to do. But but I I would also... So so it's hard. But I would also say, like, one of the benefits of building an infrastructure is, like, you can take a longer view, and you don't have to, like, listen to all the noise. I mean, you you you have to listen to the noise, but, you kinda have to, like, ignore a lot of it trying and to figure out, okay, what are the patterns here? [00:38:58] S0: You kind of alluded to some of these other workloads that like low latency workloads, like voice to voice and real time video. I feel there's a lot of chatter in the air about multimodal inference, robotics inference, these other categories. I'd love to hear you just riff a little bit on like, to what extent are those problems similar to the set of problems the inference markets had before? To what extent are they like different and new?