Why AI Infrastructure must evolve for Agent Experience — Akshat Bubna, Modal CTO

0:00 [on-hold music] We're here with Akshat of Modo, CTO- Mm-hmm... of Modo, together with Vibhu. Congrats on your Series C. Thank you. Uh, your party yesterday was amazing. [laughs] Yeah. From all the photos and all the swag. We, we had a bunch of art installations which, uh, is kind of fun seeing, like, our products on pedestals next to, like, Rodin. [laughs] Very nice. Very nice.

0:27 Um, when you started, it was not the GPU inference company. I mean, m-maybe it was in your mind. Take us back to the origin story. Uh, I actually first met Eric, who's the CEO, through an investor. And, um, back then Eric was already thinking about building, uh, a new kind of runtime, and he, he got there thinking through why are workflow orchestration products so hard to use.

0:55 It's because you have to run them on Kubernetes. Kubernetes is hard to manage. It's not built for burstiness and, um, custom images, and- Yeah... has a terrible developer experience. And I'll, I'll interject for- Yeah... pe- for listeners, uh, who are new, uh, Er- we interviewed Eric two years ago, and there's a bit more of the story there from Spotify and all those things. Mm-hmm.

1:14 And I, I actually came across Eric through Data Council because he, he did that talk on the, the sort of serverless container stack that, that you guys did, which was like, that was my, my first like, "Okay, I need to take Modo very seriously" moment. Yeah. But it was still very unclear, like, do I actually need all this for just my data pipelines?

1:31 [laughs] Y- yeah, I mean, initially what we were thinking about was if we build a better runtime, it's a very useful primitive in itself. It's, um, uh... There, there's a lot of things that, um, get solved by serverless functions. Like, you can do, uh, ATL stuff. You can do job queues. You can do all this, like, bursty processing, which it turns out every company had needs for.

1:55 Um, but then we also were thinking about this as like, um, this is a primitive that we can build a whole collection of products on, um, uh, which are very verticalized. So perhaps data engineering would've been the first one, but we were thinking about inference. Back then it was more classical inference, like computer vision stuff and running XgBoost or whatnot.

2:15 But we, we, we added GPUs to the product a year before ChatGPT came out. Nice. We just didn't think it would be that big of a deal. Yeah, just like add A100. Was there any, like, early key problem that really sparked off why you built it? Yeah. Primarily it's just, um, none of the tooling that was out there was built for, um, one, a really great developer experience, and also there's the general trend of, um... A lot of the workloads that we were seeing were very...

2:42 W- this is-- I wish there was a better word for it, but compute heavy. Like, they need, um, um, one, like, need a lot more resources, so you need to burst up and down a lot, versus like Kubernetes designed for, like, slow scaling and, uh, more for, like, web server use cases. And also, there's just a lot more specialization in, like, what kinds of environments these workloads run in.

3:03 Like, we had sometimes they need accelerators, sometimes they need different kinds of images, and this is just like a consistent thing that we saw across a lot of companies. That would be the, the next step. Yeah. Yeah. Be, be nice. I don't know how much this factored into the early story, but I wrote a post when I was at Temporal about infrastructure, software-defined infrastructure or something like that.

3:22 Yeah, the self-provisioning- Self-provisi- Yeah. Yeah, yeah, yeah. I can't even remember my own post. [laughs] And then you put me on the landing page. [laughs] Yeah. We, we really like, uh, the term a-and so we, we stole it. Because you had the insight that everything can just be in decorators next co-located with the code, right? Yeah. Was that a big part of the original- Yes...

3:39 story, or it was just like a DX layer? That, that, that was, uh, really important because we really didn't want people to spend, um, so much time, uh, writing YAML, and it seemed like you could really condense the surface area of what you're doing, uh, put it in code so you can actually operate on it just like you operate on other code and, like, build stuff that's more expressive and dynamic.

4:02 A-and so yeah, that was, that was always a very important part. Then the pushback is this is a DSL. Yeah. It's your closed source. Uh, I am locked in to Modo. Yeah. We, we, we never really got pushback for that because the nice thing about Modo is you can bring whatever code you have, and sure, the DSL is at the sort of configuration layer for, uh, what hardware you're using, how you're scaling things up, but you still own the code.

4:26 Mm-hmm. Uh, and that's, that's been an important, um, part of our story, uh, even as we do inference now. Yeah. How much of d-it do you think still stays the same today? Like, if you were to build something today, DevX obviously very important, but I feel like, you know, a lot of this has kind of been changed with just hook it up to an agent, have Cloud Code, have Codex implement a tool.

4:48 Um, there's very agent native primitives that are kind of different than if I'm doing this myself, right? We've actually changed our SDK team to think about agent experience instead of, uh, developer experience and, and we think that the same benefits that apply for DX also actually apply for AX, which is why would you have an agent read through hundreds of Kubernetes files and, like, write YAML that's not even typed when it can basically make a couple changes in a decorator and it gets this sort of self-provisioning runtime of, uh, being able to see its changes live in action?

5:22 Um, yeah, it just seems from, from the customers we talk to, um, they actually find the model is way faster used for agents versus operating on a different substrate. Yeah, because like you, you, again, you co-locate the infrastructure requirements to the code that runs it. Yeah. Well, the, the, the negative thesis now is that nobody's looking at their code anymore, so there's no point. [laughs] Yeah. I mean, people aren't looking at code.

5:46 Uh, one thing we actually still see is, is really important is observability. Yeah. Like, how good is your dashboard? And of course, like, we have, um... We push a lot of it to the CLI so the agents can do their own investigation, but you still need humans to go interpret what's going on and, uh, you know, make judgment calls and whatnot. Uh, and that's, I feel like, um- Maybe more important now than looking at the code itself.

6:11 Yes, because like, you know, it's kinda-- you, you can try to treat the code as a black box and then use, uh, sort of see the observable action that comes out of it and then just prompt a change. [laughs] Yeah. So I, I think actually think it, it takes a bit of restraint to not specialize, to say, "I want to ship a new primitive," and then just, just be general purpose. People ask you: "What are you for?"

6:33 You're like, "Uh, I don't know. We can do this, we can do that." Uh. [laughs] Well, I mean, I'd be curious to see, you know, like, okay, if you, if we were to ask you, like, what is Modal for even at a high level? There's a lot you guys do, sandboxes, GPUs, uh, everything. How, how do you answer? Modal is a cloud platform that's built for, uh, where we've built the primitives from scratch for AI applications.

6:53 Uh, and right now it basically covers, um, inference, training, batch processing, uh, and sandbox workloads. But we're building a lot more and- I noticed you didn't say web server, so there, there is still a role for like the always-on large scale Kubernetes type things. Yeah, absolutely. We're, we're not trying to compete with the renders of, of the world, uh, because yeah, we, we think the differentiator for us is the, um, are the workloads that need specialized compute, need to scale up and down a lot.

7:24 Um, yeah, they're, they're, they're just shaped differently. I think you're building a lot of it alongside the startups, right? They're innovating quite a bit, um, even in your like latest blog posts, like even, even in the series C, the, the customers that you mention here, uh, the cognitions, technical ones, ramps and whatnot, they're, they're innovating with you, right?

7:42 And that's not something AWS is doing directly with. Yeah, absolutely. I, I think, I mean, this is again, classic. We're a small team. We can move really fast. Uh, our engineers are working with our customers and figuring it out. Yeah. So my first week at Cognition, I walked in, there was someone wearing a Modal shirt. I was like, "What are you doing here?"

8:00 They're like, "Yeah, I just, I, I am em-embedded [laughs] inside of Cog." [laughs] Yeah, I think that was Peyton. We, we sent him over- Yeah. [laughs]... because, uh, the latency of communication was too high otherwise. [laughs] Um, yeah, you know, distributed node, you have to put-- you have to place one and collocate. Mm-hmm. Yeah. So actually, so I had a, I had a direct personal experience, right?

8:19 So I worked on Small Developer three years ago. Uh, it was inspired by Cloud One. I think you onboarded me at some point, like just before, and I was like, "Oh, like I need some bursty compute. Like I was just gonna try using Modal." And it was a, it was a pretty pleasant experience. Uh, apparently, I showed up in the board meeting, like the analytics. Yeah, you blew up on Hacker News and, and- Yeah...

8:41 we got a big traffic spike. Uh, I actually, I, I think the way you, you used Small Developer was Modal functions for, for running stuff, uh, which, which was... Like that, that was a good use case. Um, but then- Yeah, that was-- So to me, that was proto-cognition. Right. If only I had like stuck to it. Like that, that was like, if you... Just like draw the tech tree out- Absolutely...

9:00 it was like, yeah, like probably this will happen. Yeah, like it was so close. You were just rebuilding upon us and- I just didn't realize. [laughs] Um, but, but the, the funny story there is at the same time, uh, we were talking to a bunch of customers who needed something like sandboxing. Yeah. This is like twenty twenty-three. Yeah, yeah.

9:16 Uh, so we built our- Introduced a new API right after that. Yeah. Yes. Like we built sandboxes in May of twenty twenty-three before anyone was even knew this was gonna be a thing. And the first example we published was, uh, we took Small Developer- The looper... and put it in a loop, uh, so the agent can iterate on itself. Uh- Loops are hot these days. The looper. [laughs] Yeah.

9:35 [laughs] Lo-loops in... When was this, twenty twenty-three? Yeah. A small check. Yeah, yeah. It's like mid-twenty twenty-three. I mean, so the, I mean, the, those sort of obviously for listeners, like the, the problem was the models are not built for any of this, right? Mm-hmm. Like you, you're just trying to like...

9:48 They're not post-training to understand like, you know, looping and like self-correction and, and tool calling was there, but like also not that great. Yeah. [laughs] I, I don't remember if you used tool, tool calling in this one, but yeah, the models would just diverge after like ten iterations and not produce anything meaningful. Yeah. But like then...

10:04 So I mean, okay, like now talking to myself three years ago, the answer is- Of course it will get better... collect all the failures, build benchmark, and then collect all the, you know, examples, build the RL environment- Right... sell it for like ten billion dollars to Meta. [laughs] Uh, and then, and then also train a model and then sell that for sixty billion dollars to, to Elon. And this is [laughs] Yeah, of course. For the funny machine. Like it's like it's actually about the hardware.

10:26 [laughs] I mean, it's hard to have that kind of inherent conviction that the, the stuff will get that much better. But in retrospect, it's so fucking obvious. [laughs] Fair enough. Like what else were we doing back then? I don't know. Um, anyway. Yeah. So, so, so I mean, this, that, that was the start of your sandboxing- Mm-hmm... journey, right? Uh, I feel like it didn't blow up, blow up until like last year. Yeah.

10:50 We- So there was like a couple of years of quietness. Exactly, yeah. We were- I feel like very underrated product value. Like, um, my experience with Modal, Charles, before he had joined Modal, met this guy at a hackathon, and he really insisted we wanted to run some small model, you know, not hosted anywhere, and he's like, "Ah, there's this cool company, Modal.

11:09 They'll like spin up a GPU sandbox, we can throw it on there. They'll take a Hugging Face link." And you know, like there's so much value just right there, right? Like instant hosting, spin it up, spin it down. It'll stay cold, but you know, we run the demo a few days later, it'll come back up and like all this stuff in retrospect, like it's still what we needed like today. Yeah, I mean, it, it's still needed today.

11:29 Uh, obviously workload shapes have changed a lot as, um, we, we run stuff for people with really massive production scale and, uh, there it's, it's not about scaling from zero to one, but it's how do we scale really elastically, uh, from like thousand to fifteen hundred GPUs very quickly in, in a given region.

11:48 It's the same shape problem. Okay. So you look at, say, Cursor Composer, right? Yeah. They gotta... will do RL on a model every couple hours. Um, you guys have a whole version of RL inference gym and whatnot. Mm-hmm. When you look at workloads like that, you're basically doing train runs where you need to scale up, scale down every hour thousands of GPUs, right? Um, that's the example for we do need it, right?

12:12 Well, so actually I'll, I'll take a step back and, um- Maybe talk about like how, how people use Modal today. Uh, because our, our biggest use case actually is, uh, elastic inference. And the thing we first found product market fit, uh, with was inference for custom models. So we kind of stayed away from the LLM space, uh, and we were serving companies like Suno for audio, Runway for video, robotics, uh, comp bio companies that train their own model elsewhere.

12:39 But Modal is the best black box that for deployment, um, scaling to however many GPUs you need as your traffic pattern changes. And we saw all of them actually, um, w- like, have a very unpredict- predict- predictable, uh, traffic pattern. Uh, it's like diurnal, it's, um... Some days, like, the company will do a launch, and, you know, they'll need like, uh, way more. And it's not just one model that they deploy.

13:05 They-- all these companies deploy, uh, lots of different models in different regions, and so the autoscaling problem becomes even harder, because then you have to scale within a certain region. And those cycles sort of are offset, so different times you need to scale up in different regions. Hmm. Uh, so, so that's like our sort of- And that- Yeah... you know, that, that in and of itself is a huge category.

13:25 There's a bunch of inference providers which, you know, provide this fireworks, does this as a service together, whatnot, base 10. Um, that's kind of carved into its own niche for language models, at least right now. Yeah, I mean, the thing that we, we have actually specialized in is the autoscaling aspect. Yeah.

13:41 Uh, because we, we, we found that i-it's not universally true that everyone else can autoscale, and we've gone deeper into it on the tech side by, uh, we've incorporated GPU snapshotting into the product, so we can actually, uh, take the GPU state, uh, like your Torch compiler model, snapshot it, and the next cold start is way faster.

14:00 And so going back to your question, it's, um, that's why you need a lot of burstiness for inference. But then people also do a lot of on-demand training, like for RL stuff, your rollouts are bursty, as you said. People also do a lot of batch jobs. So we'll see, uh, a lot of companies, uh, before they have a training run, they'll need thousands of GPUs to run encoding or something like that.

14:19 And I think those things are much more bursty than... I agree that agents are, are not that bursty. Uh, sandboxes are, uh, except when you're doing RL. RL is just- Or Ebos... insanely bursty. Yeah. Yeah. Like, uh, when, when you're doing, um, rollouts, uh, you, you sometimes need 100,000 sandboxes [laughs] and new sandboxes. Yeah. I'm curious if you've seen early sparks of continual learning.

14:42 There are some people, like our friends Engram recently announced this- Yeah... they're, they're trying to do training. That also seems like a different workload, right? If you're doing training 24/7 per se, there's a very weird dynamic of how you're using GPUs between people and whatnot, but seems like something you guys would work for.

15:00 As you said, we're, we're fortunate to work with a number of, uh, customers at the frontier, and, and Engram's one of our customers. Uh, and, and they are taking the primitives we have, um, and trying to use them in very interesting ways, like continual learning. It's possible as the stuff gets better, some of that will be part of, uh, you know, our offering as well if, you know, more people need it. Um, but we're, we're just waiting to see how- Yeah... how it shakes out.

15:24 Is there a, uh, primitive that you added after sandboxing that was the next step in the story? I, I, I guess we've been going much deeper into LLM inference- Yeah... because we realized that some of the advantages we have with like autoscaling, again, especially in different regions and whatnot, uh, are, um, not present elsewhere.

15:45 A-and the place where we had a gap was we weren't, uh, working on the model there itself. Like, we were a black box. And, uh, we realized that, uh, we actually can get to frontier-level model performance, um, uh, with, you know, by having great people who, who work on, uh, on this. And, uh, we've actually been open sourcing a lot of our work, uh, in terms of, um... Recently, we, um, shared our work on D-Flash, which is a block-based, uh, uh, speculator, and we've open sourced, uh, all of it.

16:14 So, um, you can get... By using open source D-Flash, you can get the same performance as you would with one of the proprietary providers. And the next thing we're thinking about here- I thought, I thought this was actually- Yeah... an interesting blog post as well, right? Like, uh, I think in here you make a claim that, you know...

16:28 Not a claim, just that how, how effective speculative deco-decoding really just gets you. Yeah. Anything you wanna point out from this around, you know, what people should know? Uh, yeah, absolutely. I mean, the high-level summary is, um, it would help to describe what speculative decoding is at a high level. Yes. I think, like we- Yeah. So we've covered like Eagle and all this- Yeah...

16:48 like Hydra and all those things, but it was like two years ago. Yeah. I think it doesn't hurt, right? The speculative decoding is you have a smaller model, uh, pre-- called a draft model, predict tokens ahead of the bigger model, and then you have the bigger model, uh, verify all of this, all the tokens that are predicted.

17:05 And the reason it's faster is if you're predicting, uh, one token at once, you're kind of, um, bound by memory bandwidth. But if you can batch the verification of, uh, the draft model, uh, then you're much more efficient using compute and it's faster, and as long as your draft model is producing a lot of tokens that can get accepted, which is called the accept length, you can get a speed up that's, uh, multiple times of, um, you know, the original model speed.

17:32 Uh, and w-well, that's what we highlight here. It's, um... Like people talk a lot about we made these kernels faster and whatnot, but improving kernel will only give you like a few percentage points of improvement, and, uh, increasing accept length, uh, literally is a multiplicative decrease in- Like two to four X.

17:48 Yeah, exactly. Without much head-on performance. Yeah. I think it may be... I-I mean, you, you are running a second model, right? So it may be something more expensive in, in the compute, but- I think quality performance... might not be much. But yeah, I mean, I think- So, so there's no drop in quality performance- Yeah...

18:01 because you're always-- you're never accepting a token that the big model- It's strictly better- Yeah... or it's same. Exactly. Right. Yeah. And, and so we've been working a bunch on D-Flash, which is a, a block-based speculator. Uh, so it's, um, uh, instead of predicting, uh, one token at a time, it's predicting a block.

18:20 And we've been open sourcing our work with it. The, the next thing for us here is for, for helping people train speculators and custom models. Uh, it's, it's something that traditionally is very FDE driven, uh- Mm-hmm... support deployed, engineer driven, like you work with customers and help them do that. And our vision for... This is why we launched Auto Endpoints, is we want to make frontier-level performance available to everyone.

18:43 And so, uh, we mentioned this in the announcement, we kind of teased it. The next thing we're, we're launching is, uh, basically as you run an auto endpoint, uh, we, we shadow traffic and- Do, do you wanna explain what auto endpoints are? Yeah. A level, yeah. [chuckles] Yeah. So, um, this is I guess going back to your modal is, is you touch the code, but, uh, sometimes people actually don't wanna touch the code, and they wanna get started with an endpoint that works and has all the great performance and, uh, scalability that modal has.

19:13 So we've made that easier with, uh, basically a way to create an endpoint from our UI, from the CLI, um, that has all of our optimizations that we talk about, uh, like the deflash stuff already baked in, and there's full transparency. So we give you the code, uh, you can go run it yourself, and if you want, you can eject out into the full modal experience, uh, which we see as people get sophisticated, they, they do wanna tweak the models, uh, they wanna, uh, fine-tune stuff.

19:39 You, you can still do all of that. It's, it's not a black box. And yeah, the next thing, as we teased later in the post, is how do we give you value even beyond this in terms of having your draft models evolve as your data distribution evolves, again, without having to talk to a, a person and, um, yeah.

19:59 I guess just to kind of understand it directly, I mean, you know, obviously you, you have the GPUs, you have an endpoint that's compatible, you serve open modal. If someone was to do this themselves, what's the delta that you guys provide? So you do a lot of open source, great work on effective inference.

20:15 Um, how does it compare to, say, I take the same model, GLM 5.2 FP8, take off-the-shelf inference engine, VLLM, SGLang, um, you know, get compute of similar capacity, similar cost. What's the kind of delta that plugging into something this, like this offers outside of the benefit of, you know, scaling?

20:34 It's kind of interesting because we, we've taken the approach of open sourcing our contributions and upstreaming them. Uh, we work closely with the SGLang team. We, we actually want the improvements that our team, uh, comes up with to be, uh, there in open source for others to use, even outside of modal.

20:51 The benefit to us is we have a team that has significant expertise in terms of if you do have something that is not there, our team can help you get that performance, uh, first. Um, the, the other thing is with these endpoints, we are way more elastic, as you said, uh, than, uh, anyone else, and you have true scaling to zero.

21:11 Uh, you have true, uh, burst CNS, and in practice, that matters a lot more to people than just finding, uh, the GPU and, uh, running modal code on something. Yeah. And I, I will say it's actually not that straightforward to just... Uh, like what I said, is easier said than done, right? Yeah. Um, it's, I think still for the average person, still hard to just gut check using different... There's, there's quite a bit of combinations you can make there. Um, the trade-offs aren't really known at face value. Yeah.

21:41 I mean, it's, it's not just that. I think it's, it's that running production-grade inference is a hard infer problem. Uh- Yeah... even if you subtract out the autoscaling- Yeah... is p- uh, controlling things like tail latency and, um, making sure every, uh, request is delivered at least once and whatnot. There's a lot of innovation that you can do here.

22:01 I think, uh, it's very interesting that you're starting to encroach on, like as you become a full cloud, you're starting to encroach on other people's turf. Mm-hmm. [laughs] What will you not do? Well, we wanna follow our users and, um, make sure they, uh, they get like a, a platform that has everything that works well together.

22:23 Uh, so right now we're kind of focused on the modal lifecycle and the agent, uh, lifecycle. Uh, so both like going from data prep to training to inference and then also if I want to deploy a background agent, let's say, you know, sandbox, super persistent storage, a whole bunch of other stuff- We talked to Cole, uh, who did, uh, Open Inspect.

22:42 Yeah. Yeah. And obviously Real Inspect also is on modal. [laughs] Yeah. So Ramp Inspect was, was a, [laughs] a great example of a background agent that was really successful because they, uh, were able to use some of the primitives like snapshotting and fast scaling to just have something that feels really reactive and, uh, and works well.

23:02 Yeah. That's the new CTO of, uh, Ramp right there. Yeah, Rahul. [laughs] It was really, really fun. Um, yeah, I mean, okay, you know, I, I think, uh, all very bullish. Like, you know, one of my reflections was also I did not originally... So obviously when I met you guys- Mm-hmm... you weren't that much in the GPU game, and now you're all about, uh, inference. And one of the points that I hinged on for Jensen's keynote at GTC this year was, what we're calling it, like the inference inflection, right?

23:33 That let's say in AI workloads or machine learning workloads, it used to be like, let's call it eight to one GPU to CPU, and now it's more like one to one, which is like a interesting... Like the-- because of how much agents basically are blocked or call out to, to CPU heavy stuff- Mm-hmm... the actual, like limiting factor, like swings back and forth from GPU to CPU a lot more than it used to be all GPU and then occasional CPU. Yeah. GPU, CPU.

24:03 And now it's like just constantly, and you just have to co-locate everything. Yeah. And, and that's one of the things that, um, actually, like again, we see as, uh, something appealing about modal, which is we've built this capacity pool that spans, uh, 17 cloud providers. So we're, we're very good at Running on various kinds of cloud capacity across the world You don't have your own data centers? We don't have our own data centers. We, we just run across a lot of neo clouds and- Yeah...

24:30 metal providers. Yeah. Question mark. I- Yeah, yeah. You're, you're running the math, and you're like, w-what's the cutover point where you're like... Yeah, it's a good question. I mean, part of it is we, we see our differentiator in, in the software layer, and, um, being capital light and focusing on the software helps us move really fast.

24:46 Uh, so far it's worked out well because there are so many other people building data centers that we're able to work effectively with them and again, focus on what makes us, uh, special. Yeah. But- 17 gets you into, like, the local providers sometimes, like- The, uh [laughs] Which was the most interesting one?

25:02 There are actually a lot more neo clouds than you expect, and they all have various degrees of, um, various levels of reliability. And, uh, s- that's why it's something we've in- invested a lot of time in, is actually, uh, building our own reliability layer on top. Uh, so if the GPU falls off the bus or something happens, we w- uh, user workloads are not affected, and that, that actually lets us use a lot more capacity than, um- Yeah... you know, you as a user would be able to.

25:32 It's a useful thing to have because like now everyone knows, like, what layer you are and, like, you, you sort of optimize for being the super cloud- Mm-hmm... of all clouds. Yeah. That's, that's, uh, that's the idea. Uh, a-and so I, I guess when you mentioned colocation, that's, that's another interesting thing where, um, one thing we've seen is people come to us when they want, uh, very specifically located, uh, CPUs or GPUs, uh, like they want- Oh, they pin it in like- Yeah...

25:59 EU? Ex-exactly. Or EU, US. Right. Data resiliency- Australia... locality thing or performance or what? It's either data locality or latency, yeah. Yeah. Like, you want your... They're running sandboxes and model. They want them to be right next to each other- Yeah, it's easy to, to- Yeah... you know. That is important in, in, in all those things.

26:15 A-and so, like, you've kind of accidentally, I don't know if it's accident, but, like, you've built the perfect primitive for agents to, to express themselves. And then, uh, you know, like, it's almost very funny how every extra development just involves more file system, just involves more CPU. Yeah. Uh, just like the things that you already have.

26:33 I don't know much about, uh... If there's any, like, networking usages that are interesting, but you've also done some good work on networking. Yeah, I mean, that's exactly right. Like, we-we're sort of just taking compute storage and networking and building stuff on that layer, uh, for, again, the stuff people need. Um- Yeah... we, we see a few interesting networking things coming up.

26:52 Um, one is people actually want networked sandboxes. Uh, so we have a- For like a Docker cluster type thing. Sorry. Yeah. Uh, Docker Swarm. Uh, fuck. What is, what is it called? Compose. Compose type thing. Yeah. So actually if you want Docker Compose, uh, our sandboxes now support, uh, this thing called sidecars. So you can... A sandbox is actually a pod of containers, and you can run multiple containers in, uh, a sandbox.

27:16 Uh, also useful because, uh, going back to networking, people want a lot of control over, um, outbound networking from a sandbox. Yeah. They might wanna run man-in-the-middle proxy for, like, maybe logging stuff for RL or, um, controlling how egress can happen to a domain, injecting credentials. Uh, and yeah.

27:35 So we've, we've kind of had to build a lot of that stuff ourselves. Yeah. Uh, but then also sometimes people actually want, um, sandboxes spanning multiple nodes to talk to each other, uh, which is an emerging thing we're seeing. We have support for that for, for a different reason, and yeah, we'll see if that becomes stable. Like just an open socket. It's a... This is directly like mTLS.

27:56 We do support that, which is you can, uh, expose a tunnel inside a sandbox. Yeah. And then you can either expose it to public internet or it can be, um, you can add like a HTTP, uh, auth layer above it. But we have this thing called I6PM, which we haven't talked about, uh, which is this like overlay network using IPv6 addresses.

28:17 Uh, so if model containers, uh, within the same workspace, uh, when this enabled can actually address each other using this private IPv6 address, uh, and no one else can. Mm-hmm. Uh, so it's like sort of private networking, uh, for containers. We actually built it because we needed it as a primitive for our disputed training product.

28:38 Uh, so we have this other feature, which is you can add a decorator to a function, and you get a cluster of GPUs, uh, and they have RDMA networking. Uh, so you can run a disputed training job, uh, that's truly serverless, uh, and we did the overlay network for that. But then we've seen that people are using it for other reasons and, and I'm kind of intrigued to see- [laughs]... yeah, what would people do with it. Build primitives and let people figure it out, right? Yeah, exactly.

29:02 You put, you put out a pretty interesting- They're like, they read the docs we think. Let me use that for something- [laughs] Yeah... they never intended to [laughs] This is literally not even in our docs page. People somehow found it and they're using it. Huh. Uh, s- I mean, the way you portrayed it with like RDMA versus TCP, like very well laid out, but just the transfer speed change at scale for RL like- Mm-hmm...

29:25 yeah, you have it, you have it built in. I'm sure someone found it. It's found it to be a lot more efficient before you actually made a thing out of it, right? Yeah. And, and not to split hairs, I, I guess the, the overlay network actually is the TCP overlay network. Mm-hmm. Uh, the reason we have that is you need that to do the key exchange for RDMA before you set up the RDMA network on top of that. Uh, but then people found the, the TCP part.

29:48 Can I tell you, this is like a big aha moment for me because- Yeah... uh, so I, you know, I review 2,200 submissions for, for the World's Fair. Yeah. And then I got this from John Osterhout- Uh-huh... who I don't know if... Do you know John Osterhout by name? The name sounds familiar. He published a...

30:03 He's a well-known professor, published a lot of interesting software design books, and this is the talk he chose to submit is on t- [laughs] is on RDMA at Inference. And I'm like, you wouldn't think that this guy who is like kind of operating systems guy would care about RDMA. [laughs] I, I, I mean, it, it makes sense to me because I, I- This is the cloud, right? Yeah.

30:25 Like, like the way you move around your KV cache and how efficiently you can do it, how efficiently you move, um, your weights from your training GPUs to your inference GPUs and RL is... There's a lot of degrees of freedom, and it is basically a systems problem of Yeah... moving memory around- Yeah, yeah... scheduling This shows you how primitive my understanding of networking stuff is.

30:46 [laughs] Is this like the domain of WireGuard as well? Mm, not quite. So- It's adjacent? Um, so- Explain everything. [laughs] Sure, sure. Uh- How do we move memory around GPUs? [laughs] Well, so, so sorry, yeah, that is memory. Uh, sorry, I, I was talking more, and maybe I was talking like five minutes back, about the, the private IPV6, uh, addressing that you've set up. Yeah.

31:09 Is it like it's basically a VPN? Yeah, it, it's sort of like a VPN, and yeah, w- WireGuard is, is, uh, yeah, you, you're right. It, it is, it is, um- Right, right. Yeah, you already moved on to new topics... a similar- Okay... in the same space, WireGuard is, uh, encrypted, and, and th- this is, uh- And you don't need encryption. Yeah. Yeah. This is not encrypted. Uh, that's the main difference.

31:26 This is, is TCP, and we have eBPF programs that will reject or allow the, um, uh, the TCP connection based on whether you're allowed to do it. It used to involve a full sidecar, but now you have eBPF in the Linux kernel. Yeah. Yeah. I don't know if this is a natural follow-on to the topic of like my skepticism on distributed training- Mm-hmm... is that while, like, people spend a lot of money on, like, cables [laughs] to hook up GPUs, and even that is not, like, fast enough, and that's the bottleneck, is your networking fast enough?

31:59 Yeah, so I, I guess you're talking about sort of fully distributed training like, uh, DiaLoco or something, which is like cross data center- That would be, yes. That's the extreme. Yeah. You're kind of in the middle, and then other people would have like the Mellanox cables up in, in, like, their actual data center.

32:14 When you run multi-node training on, on model, uh, w- uh, RDMA, uh, I, I, I think Mellanox, uh, is, or InfiniBand is like a, uh, is, also uses RDMA. Uh, but basically it's a way to bypass the TCP networking stack and, um, transfer, uh, stuff much faster, uh, between one node, uh, to the other. And, and we, we have, uh, I think like three terabit per second, uh, inter-node networking- Okay... which is the standard that's needed. Okay.

32:42 So I misunderstood what- 50- What part of stack you were- 50 gigs over TCP- Yeah, yeah... if you point to it or any of them. Yeah. Okay. [gentle music] Yeah, I, I mean, very impressive work. [laughs] So e- effectively, you're, you're extending sort of like the model philosophy to the, the ch- trading cluster, like, yeah. Yeah, and we're, we're not going for, obviously, like, large scale pre-training runs.

33:04 Uh, the thing that we've built multi-node training for is, uh, we see a lot of, um, smaller scale post-training. Uh, like, uh, people are post-training like medium-sized fund models, uh, so they can, uh, get higher quality on inference. Uh, this is a perfect fit, uh, for something like that. Yeah. That is my impression of how a lot of these labs explore branches in post-training- Mm... and then eventually merge whatever they find in. Yeah.

33:31 The, the other use case we've seen for multi-node training is even if you have a big cluster, your researchers are still doing small runs, and- Yes... having elasticity there- Right, right, right, sure... matters a lot more. Yeah. Uh, the, like, this is actually like the current limiting factor for auto research, which is like you, you basically need to give your model some GPUs in order for it to completely run.

33:51 We have a blog post on auto resource, and model is, uh- Yeah... yeah, like, uh, turns out to be pretty good substrate for that. So my impression is au- auto resource means many things, like- Yeah... with anything that Andrej coins. Right now, still science fair, right? Like, not, not, not actually like- Men... I don't know how many people are actually doing this. We're having a golf. [laughs] Yeah.

34:09 I thought the same thing. Yeah, you would know. We, like our internal both training and inference teams actually use this sort of the general shape of this quite a bit. Uh, like we have this one internal repo called Auto Inference, which essentially we've automated our own FDE efforts using, uh, this harness, which is, um, the agent will just spin up a suite of different things.

34:34 It'll even run like, um, NVIDIA Nsight Profiler, and it'll, like, tweak configs, and it'll arrive the right thing. Uh, it'll change your GPUs from H200 to B200. It actually works really well. Nice. So yeah. By the way, I enjoy that your FDE is so technical that you have to do these things. [laughs] It's very different from FDE from other people. [laughs] Yeah, yeah.

34:54 For our FDE team is, uh, essentially they're like applied inference researchers or applied training researchers. Someone told me, like they have to be able to build, but they also have to be able to sell. Uh, do they have to sell, or are they like they're good, they're just like post-sale type of thing? It does, uh, being able to talk to a customer and engage effectively with them- Yeah... matters a lot.

35:14 Um- They want the same thing. Yeah. You know? But it's, it's not really a sort of sales, uh, thing. We, we pair them with... We have solution architects as well that are more on the pre-sales side. Okay. Let, let's spend a bit more time on auto research. This is a big focus for me- Mm-hmm... for, for, for this year. Where does this go? You know, like, um, have, have people explored enough?

35:33 Like, you know, there's all these beautiful charts of like improve, improve, and then they sort of level off a bit, and then you find the next thing. Is this basically sort of one abstraction up from normal training? Is that how we think about it, or do you think about it differently? Like model level training versus basically high, like AI-driven hyperparameter search.

35:50 Yeah, like, uh- Uh, someone, some people call it like neural architecture search or whatever, right? Like... Yeah, I mean, I... So the, the stuff I've seen people do with it is nowhere on the architecture level. It's pretty much tweaking parameters, but it's, it's basically a hyperparameter sweep that's guided by some sort of model intuition, so it's like much more efficient than, um, what our other, uh, sweep would have.

36:12 Uh, yeah, I mean, you know, it's just, it's just a question of where you want to spend your compute, you know? Right. 'Cause yeah, you can just throw infinite amounts of money on this and somehow you'll bang out Shakespeare, you know. [laughs] Yeah, infinite monkey. Uh, yeah, I mean, so like the very good for model.

36:26 Uh, and I, I think it's also very important that agents can spin up other agents, can spin up their own- Mm-hmm... infrastructure. Like very, like very good for you. Uh, how good are LLMs at generating model code? Like, you know, like the benefit of existing pre-LLMs is that you are in the data. Yeah. They're, they're actually surprisingly good.

36:45 Uh, I think like- Pre-cloud four they were not, and then now they're able to one-shot, uh, stuff out of the box. And we're playing around with releasing like a model bench for like the harder- Yeah... things, uh, that the LMS cannot do yet and, and maybe- What, what's an example of that? I think the, the things that sometimes agents struggle with, um, without right guidance and a skill is, um, how to, uh, use the rest of our observability. Like how to...

37:14 Something is failing, like d- how do you look at the, the logs and then update the right thing? It's sort of reasoning about that. But they're able- Okay... to one-shot like- Yeah. You c- That's cool... you can just add a skill to it? Yeah. So, so we have a model skill now that- Which is kind of actually why we built this model bench.

37:31 It's to find things like that so we can, we, we can address them in our- Yeah, that's true... in our training skill. Yeah. Yeah. No, no, I mean, it's, it's good. Uh, are you facing any shortages? Uh, you know, like we talk a lot about GPU shortages, but also CPU, also memory. Mm-hmm. Yeah. Uh, we have had a lot of growth, which means that, uh, there, there's a-- we've had to be much better about- Planning...

37:54 proactive capacity planning. Yeah. Uh, so we, we have, uh- Which by the way, like it's like a MBA's like dream job- Yes... is like just planning this stuff. I think last time you and I talked about something maybe about this. Yeah. I mean, we have a, a really competent team of people that we call, um, the role is called compute strategy.

38:10 Uh, so yeah, if anyone [chuckles] listening here, uh, wants to work on that. Compute strategy? Yeah, yeah. And like, uh- I think, I think the norm, the normies call it FP&A or something. [laughs] Well, it's more-- It's, it's not FP&A. It's, it's, um... There's a lot of interesting financial questions of like, uh, like what is the blend between one year and three-year reservations?

38:30 W- how do we forecast our own capacity? Uh, how do we... Basically, especially since our capacity is very fungible across different GPU types and different regions, uh, like you basically have to model a lot of it. Uh, and you also have to have an opinion on how the supply chain is gonna evolve, and then you have to like take bets, uh- Yeah... based on that. Tokenomics. Yeah.

38:51 This is like probably a not a real point, but, uh, you know, I was trying to think about like what other industries... I was trying to think about like, you know, we cannot be first to like these kinds of problems. Yeah. And what other industries have had this? And I was like, airlines with, with fuel, and like they have to hedge their fuel.

39:05 And like I think for a long time, Southwest, because they made like a hero fuel bet, they like were like super low cost because- Oh... compared to everyone else. [laughs] Yeah. I, I hadn't thought about that. Um, but- We're at a fun time too, you know? Yeah. It's... A lot of the, uh, the compute business in general, uh, for us is also about being very good about capacity management.

39:27 That is how you have great unit, uh, economics, uh, but also over time is how you can unlock more value for customers. Like, um, one of the things we're building now is like a way for customers to get, um, if they don't care about latency, uh, like get much cheaper pricing, and they'll get results back in like next 24 hours or something, uh, like a batch tier essentially.

39:47 Yeah. Uh, and those are levers we have because we control the whole stack and scheduling and whatnot to give people- Yeah... a sufficient- Yeah, I feel like- Yeah... they are not as popular. Like those, like the, the Frontier Labs have all those APIs. They're not as popular as they should be. The demand that we see for something like that is actually not for LMS.

40:05 Uh, although sometimes people wanna run evals and do- Okay... synthetic data prep and, and there it makes sense. Okay. Um, but it's, uh, from a lot of non-LM companies, uh, like people who are doing computational bio, like they, they have to run really big batch jobs, and they don't care about when they get it back. Yeah.

40:22 And like they have a reasonable... It's, it's also like a cousin to the stopping problem of like, will this finish in time? [laughs] Yeah. Well, you, you can, you can bound it. Yeah. Like you can give people- Yeah, yeah, yeah... SLAs on it. Yeah, yeah. I think what's, what's interesting is like the, the next phase of model. Mm-hmm.

40:38 Like what, you know, do people expect from you, uh, now that you're sort of established and you're like a well-known compute player among all these leading companies? You had an inference launch week, and we talked a little bit about the launches. Uh, like what else? Like what, what else should people know? We are building primitives that make our, our users' lives much easier.

41:02 So, um, I think for example, with LM Inference, thousands more companies are gonna post-train their own models and, uh, deploy open source models for inference. Uh, so we're thinking a lot about what is the best product shape for that. And, uh, that involves everything from our training gym to, uh, then, uh, endpoints that get frontier-level performance, um, you know, again, without having to talk to anyone.

41:25 It looks somewhat different on other verticals. Like, uh, we're also seeing a lot of real-time, uh, audio-video stuff in there, um, uh, which is why like, uh, we're working on things like regional routing, uh, with, with fallbacks. So you can, you can get sort of GPUs that are as close to your users as possible. Uh, so you, you get like low latency for video streaming and, and whatnot. And then on the agent side, it's, um...

41:52 we're still sort of working very closely with our customers because stuff is changing so fast in terms of what they need. And, uh, I think beyond sandboxes and persistent file systems, uh, there's a lot of other things people need from this agent stack as they build production agents. So yeah, we're thinking about those other things that fit in there. I want to ask what, what the other things are. [laughs] Yeah. I probably should share right now. I think, I think...

42:17 Okay, so, uh, I do think a lot about the principal components of cloud, and you do talk about compute storage networking. Mm-hmm. Yeah. 'Cause so far for me, it's fine. Uh, so far for... I mean, the first couple generations of cloud, it's fine. What's different, qualitatively different about agents that you, you need some new permission level?

42:38 Like a lot of people, you know, obviously, okay, and I'll just kinda spew tokens at you until, until it like hopefully sparks something. Yeah. Like the, the new level now is whatever Cloud Code does, which is dangerously scope permissions or like allow this by command or like whatever, right? And, and sometimes they're like, "Well, okay, we, we have like this adaptive thinking mode where like... just, just trust me, bro. I, I will make the [chuckles] calls for you." Is that it? You know?

43:01 Like l- basically like sort of LLM-mediated permissions. And now you're looping it with a goal and letting it roll. Yeah. I mean, I'm, I'm skeptical of LLM media permissions for- [chuckles] stuff that is at the sandbox level because you do want hard boundaries. Yeah. Otherwise, obviously someone can exfiltrate stuff. But like, like maybe, maybe that's old school thinking. Maybe we're the dinosaurs.

43:23 [chuckles] Maybe the AI OS or the LLM OS is really the kernel is a goddamn LLM. [laughs] Um- Like it makes you feel uncomfortable. Yeah. I'm, I'm cultural- But that's what trusting the LLM is. Like imagine a spherical cow, perfect LLM. Right. Let it. [laughs] Maybe. Maybe. I wanna test the boundaries, right? Yeah.

43:42 Like, and obviously I, I don't believe that, but I wanna see where I'm wrong 'cause that's, that's the non-consensus. Yeah. I mean, I, I think you always need hard guardrails when you want, um... And you compare those to softer guardrails, right? And that's gonna be a lot of mediated. The-- I'll get you-- end with a couple of your commentary on like the ecosystem outside of Model.

44:05 Managed agents, everyone has one. Gemini, OpenAI, Claude, uh, very useful for you, but also like it is their way of starting to edge into your space. Yeah. Uh, what's going on? Yeah, I mean, we're, um, very excited to partner with Anthropic and some of the other foundation labs, will not name who we're also working with.

44:27 Um, the way we see it is the managed agent thing is a great place to start if you're starting out building an agent. And, um... But, but then when you get to, um, building something more production grade, like you're a company that's like Ramp that's building their own, uh... Ramp also runs their accounting agent on us, so their external facing agent.

44:50 You need a lot more control over, uh, your compute primitive on things like, um, what sort of-- how do you persist different files that the agent has access to, and how do you snapshot and restore? How do you control the networking? Um, maybe you want GPUs. When you get to that point, you kinda want, um, a specialized sandbox provider, uh, that gives you those things, and that's the role that we are trying to play.

45:15 Um- Yeah. We don't really have an opinion on the harness, whether it runs in cloud managed agent and you hook it up to Model Sandbox, or you run the harness in Model Sandbox. We'll see where people converge with that. Yeah. Do you any opinions on like the meta harnesses? Um, is it just another layer on top of these things? You mean like the OpenPi and- OpenPi is one.

45:34 Um, I think Vercel had one, which I can't remember the name of right now. Fredshot had one. Uh, and then, uh, to me, most recently was Data j- Databricks that had Omnigen. All these are sort of meta harness. Like it's kind of pseudo agent cloud type things. I personally have not played around with them. Yeah. Built agents with them. I mean, everything is bullish model as long as it consumes more infra.

45:57 That's why we're focusing on the infra layer. It's, um, somewhere where our, um, relative competence is and, um, also it's a hard problem to solve. Yeah. I mean, I, I will say like just generally reflecting on, I don't know if you-- if there, if there's other topics on Model, but like just generally reflecting as an infra person, not as intense as you, but in that field, this has like been the most exciting time in infra. Like it was boring actually for, for, for a while, and you couldn't really get people excited about data infrastructure.

46:29 Like Eric would get on Data Console, everyone just watched the video and like say, "Look at how many sandboxes I can spin up," and no one gave a crap. [laughs] Yeah. And like now everyone gives a crap. [laughs] That's true. It is a very exciting time, and I think a lot of that's driven by just the, the amount of scale all of this stuff needs.

46:50 I think the, like a lot of your initiatives or a lot of your like product directions make sense in retrospect, which is like the best kind, but I wouldn't necessarily have thought about it myself, which, which... [laughs] We need, we need the predictions, you know. [laughs] I mean, I think there's a lot that you just don't even see, right?

47:05 Like you have the batch, you have the voice, you have the multimodal, um, but what, what else, you know? What else is coming up for us or- Yeah. Where, where do you see things going? Yeah. I, I mean, in general it's, it's clear that there's a-- obviously there's a huge shift happening. I, I think one thing that's not as obvious to people because LLM inference gets talked about so much and is also we work a lot of companies that are, uh, doing things like drug, drug discovery and computational bio, like the chart discoveries world. Big things are probably gonna happen there.

47:37 Uh, we work a lot of robotics companies that are actually putting robots in, in like active deployments and getting good results out of them. Is there air gap model? Is there, is there a version that is like on-prem air-gapped whatever? No. We, we- You're usually cloud only. Yeah. Yeah. Okay. But yeah, I mean, uh, so what you're saying is like because you're focused on primitives and they're good primitives, you find use cases in all these kinds of things. Yeah.

48:01 Actually probably diversifies you a little bit away from LMS all the time. Yeah, absolutely. We, we're-- our, our goal isn't to only serve the LM inference market. Uh- There are a lot just on the website, the audio, the- Yeah. We've had, we've had both on the, on the pod. Computational bio images. Yeah. I mean, there's a lot here. There's QTI TTS, customizing. Oh, Chatterbox.

48:21 Um, you know, there was a customizing Whisper. Okay. Yeah, yeah. This screen reminds me of a fallen competitor, uh, which Replicate. Mm-hmm. Um, what's your postmortem on what happened? This is one thing we've kind of stayed away from is providing an API for models, because I think providing model APIs is some of it ends up serving like a really hobbyist market, which is much less sticky. Yeah.

48:50 And we've always wanted to build for companies that are building sort of products and need sort of more flexibility that's not just an API. And- Which you, you can build an API for a model and, and this is clearly what it is. But you can-- but what you're saying, you can wrap it into a more fully functioning back end that you, that you run. Yeah.

49:07 So actually all, all of our examples, it's not that spin up this model, here's an API token, use it. They're actually all code. Okay. And so the point is that this is just an example. Starter code. Yeah. But you can, you can tweak it however you want. Yeah. And if you're like a company building a product, like a computational bio, whatnot, yeah. I guess I'm trying to tease out for listeners- Yeah...

49:28 when does it stop becoming, "Oh, you're just an API call and you're just a wrapper on API," to becoming what you call a product, right? Mm-hmm. Like, what is that layer? Like what-- Like, [chuckles] you know, obviously more lines of code, but like beyond that, what, what is the substance that people add that qualifies it to be something more?

49:46 I think there's a little bit of like a selection effect of like a lot of companies who do wanna get deeper into that level are probably building something that's more differentiated. And, um, I think, um, an example is like we-- with LM and friends, originally, we, uh, worked with companies that were building their own post-training frameworks or they were, uh...

50:09 Ramp actually early in the day was training their own tokenizer and like swapping out the tokenizer in Llama and whatnot. I'm not saying that's- Mm-hmm. [chuckles]... that was successful, um, in, in that case. But a better example is like, um, like, uh, let's say Suno. Um, because Suno, um, does not use Modal for training. Mikey on the pod, yeah.

50:27 But they, they use Modal for all their inference, and that's because they have like a custom-- They have completely custom model architecture, and that means that they have to be at the code level and tweak things that are not, um, you know, just an API. It's interesting as well, like we had, uh, Ethan, uh, most recently on the xAI Grok team- Mm-hmm...

50:46 make a prediction that actually like the next tier in video gen is not a better video model, it's a bet- better model or agent that orchestrates video models. Oh, interesting. Language model backbone that can use tools- Right... and write code. Like, yes, I can make my six-second video or my ten-second video from Grok, but actually I want my six-minute video. Mm-hmm.

51:06 And I'm not going there through normal video gen. Yeah, that's interesting. I actually ha-- So we, we have GPU sandboxes, and recently have seen a few companies doing s-sort of agents that do video manipulation or, um- Yeah, give it FFmpeg and just do whatever is- Run, run FFmpeg. [laughs] But like that, that's not enough. Yeah. You need to give it Adobe. Yeah, I hadn't put it together with like it would actually be a video production thing.

51:32 Uh, in, in my mind, these things were going more towards editing and- Yeah. You're learning a lot about models. Well, I, I think about this a lot. No, obviously. [laughs] Yeah. Sorry. Um, Luma. Luma Agent is a version of this for video production, but you know, it's a one-off. I was gonna get your quick takes, uh, on, uh, on some other stuff that happens- Mm-hmm...

51:50 in recent news and just, just see if you have anything interesting. Gitpod, very like somewhat like, you know, different market. They, they're in like sort of like the CI/CD market, but actually technically very impressive. I don't know if you've like taken a real look at them. Yeah. Uh, we've, um...

52:05 People on our team have talked to the Gitpod team, and they've-- they're technically very strong. Yeah. I, I actually am-- We're, we're very bullish at Modal on the CI market as well because- Okay... there's, there's more agents, uh, coding agents. Yeah. Uh, they're gonna run a lot more CI, and the primitives there can be much better. I think there's a lot of wasted CI. Yeah. So is it just like let's filter?

52:28 Like what, what is the highest order bid here in improving CI for agents? Well, there, there's a lot of wasted time in CI on like- To batch it... preparing your artifacts and like, you know, getting you to the basically, um, uh, preparing your dependencies and whatnot. Oh. And um, obviously, like build systems help with that.

52:48 But like if you have primitives that are like memory snapshot and restore, can you just run CI more efficiently? Oh, oh, okay. Okay, okay. Interesting. Yeah. I mean, another form of like, you know, on-demand compute. Yeah, exactly. Yeah, yeah. Uh, it, it needs the same, again, platform. Yeah. So, so, uh, for those who don't know, Gitpod rebranded to Ona. Mm-hmm. It was like there was this whole thing.

53:11 I, I got-- Actually, I, I, you know, like sort of semi-sounded the alarm at Cognition. I was like, "You should take these guys seriously because their infra is very good." Yeah. Uh, and but, you know, then, then they join OpenAI, and, uh, presumably, we'll, we'll see Codex Cloud from the Ona team. Mm-hmm. Like which, which I think would be very, very strong. The-- To me, like teams like that, that can set up the networking and like the, the secure boundaries for like, and your like agents to have their own cloud each, uh, effectively is what you're doing kind of.

53:41 And I'm just trying to draw the analogy or, or the differences. If you have studied them, like what is the philosophical difference, you know? My sense is maybe they didn't go after the right market at the right time because we-- I guess also got lucky with like agent use cases really taking off and, um, needing, um, like more of like a sandbox shaped thing than like, um, uh, my understanding is, yeah, Gitpod probably- I mean, like sand-sandboxes work- Dev environment... like CI/CD- Yeah... is sandboxes. [chuckles] Yeah, yeah.

54:10 It's just like build time sandboxes versus runtime sandboxes, and actually it turned out runtime was better. [laughs] Right. And, and the difference there is runtime sandboxes have a different configuration surface of like how you configure images, how you like attach like pers storage and... Yeah. It's, it's fascinating. Other people, Astral.

54:29 Mm. Also OpenAI. [chuckles] Also like Python tooling ecosystem people. Are you still sort of bullish build- building on top of Python? Also recently Modular also- Mm... uh, got bought by, by Qualcomm. Just any, any, any of your takes there. Yeah. I mean, we, we had Python as our first SDK language because that was the language that people did data and ML in. I actually now have Go and TypeScript SDKs as well.

54:53 Uh, and our runtime is completely language- It is, it is written in Rust, but it's, it's not tied to Python by any means. We haven't seen... A-a-I think with like inference and training stuff, people are still very Python. And the interesting thing with like the agent stuff is people use our TypeScript SDK a lot more because they're not actually doing anything that needs ML.

55:13 I don't think we'll have to go beyond that super soon- Yeah... 'cause Python and TypeScript for us is still Dominant. The last two languages in the world. Yeah. [chuckles] That's it. Well, English and prompting [chuckles] thus far. English and prompting. I occasionally talk to people who try to build new, new languages. They're like, uh-- Even, uh, what's his face?

55:31 Brett Taylor, uh, who's chairman of OpenAI- Mm-hmm... was like, "We, uh, we need, we need a new language for, for LLM," so no one has come across one, and I keep looking. You know, Python and TypeScript are-- You have a lot of data, plus- Mm... but then also they are very imperfect as-- just as languages themselves.

55:48 Then my close is, I think Modal used to be a big bet on developer experience. Mm-hmm. And you've pivoted the team to agent experience. Is it, like, the way now, like, do you-- do, do-- can entire companies and unicorns, multi-unicorns, be built on just having better agent experience? Do you need something else? It's a big part of our identity.

56:07 Uh, it's not just, uh, you know, like the, uh, the very tactical, how does an agent use the CLI, but it's also how easy is it to spin something up. Like, what is your iteration time when you wanna spin up a new service and, uh, you wanna get something going in prod? Uh, in practice, that matters a lot, uh, to people, and, uh, I think it will continue to matter.

56:29 Like, uh, people are building stuff even faster, and if you give them ways to do it quickly and- [chuckles]... uh, not have overhead, then... The-- I think the debate for me has been, do you do anything differently that is, like, very fundamentally different for developer experience versus agent experience? Mm-hmm.

56:44 You seem to be on the side of they're-- they, they're like this. They're like cosine- Yeah. We actually also have a blog post on that. [chuckles] Cosine, cosine similarity on, like, zero point nine or whatever. Yeah. I mean, pretty much it's the, um, the, the main shift for us has been, as I said, like, we built this, uh, benchmark, Modal Bench, uh, to see where agents are lacking and- Yeah...

57:02 uh, actually literally imp- add surface areas to a product if, if they're reaching for something, like maybe this should just be a CLI. They halluc- Uh- Uh, yeah. They hallucinate their own features. Yeah. And sometimes it makes sense. Like- [chuckles]... if they're reaching for this thing, it's product feedback, like give it to them.

57:17 And then, yeah, actually moving-- w-we used to only have like logs and metrics in our UI, just moving all those things to the CLI as well, so they're accessible in, in, in that form. As simple as that. [chuckles] Cool. Thank you so much. Yeah. Yeah. Thank you. This is great. This was fun. Yeah, it was a great update and, uh, I can see why you guys have succeeded so much. Uh, it, it is really, uh, focus, but also really good execution.

57:39 Thanks. I mean, we have a long way to go. All right. Thank you. Cool. [outro music]