Jake Cooper: If you are writing code by hand, you are doing this wrong, right? Like your code, the tools are good enough at this point like that you can move extremely extremely quickly and like yes, there are issues and pain points and all these other things. Um, but you should be reviewing the code that you are writing instead of trying to go and write it by hand.

Jake Cooper: Like all of those architectural patterns, all of those other things like you're not just don't like you're not going to throw them in the garbage or whatever. Actually, they matter more now than than than any other time. But you just shouldn't spend your time generating code that you would write. Like if you know how to go in and write it, just like ask the agent to go in and write it and then reconcile it until it looks like you would have written it.

Allesio: Hey everyone, welcome to the latest space podcast. This is Allesio, founder of Colonel Labs, and I'm joined by Swixs, editor of Len Space. And today we're in the studio with Jay Cooper of Railway,

Swix: Hey. Hey.

Jake Cooper: conductor of railway.

Allesio: Um okay so well for those who don't know what is railway let's give people a crisp definition up front.

Jake Cooper: Yeah, railway is the easiest way to ship anything. You just go to the canvas or you talk with claude um and you say deploy Postgress instance, deploy my GitHub repository, run this code, etc., right? Um and you'll just be up and away to the races, right?

Jake Cooper: But yeah, we want to make it really easy for not just to like deploy things, but for you to almost like evolve applications over time. Like we believe that most of the tooling right now is kind of like stacked up like you're stacking entropy on top of entropy on top of entropy, right? So you have like Docker and Cube and then like Ansible scripts and all of these other things, right? Um, and if we can kind of like version all of your software for you and keep track of all of the changes, then we can make it actually trivial for you to clone environments, you know, fork into a parallel universe, get copies of like production data, get copies of like any of your services, make those changes, validate those changes, collapse it in, uh, without kind of having to just like reproduce everything across a, you know, a staging environment or all of those other things, right?

Allesio: Amazing. Uh, one thing I I was looking at your background, right? like um Bloomberg, Uber, there's nothing immediately that stands out to me as like okay this guy's going to found like the next great platform as a service. Uh what prepared you for real?

Jake Cooper: It's almost like a curiosity just like ever go deeper, right? Um and so like you know started out on like front end stuff you know like working on the like wolf from like web mathematic and like porting it over there and then you know briefly moving to Bloomberg um and then moving towards Uber and like distributed systems and kind of like taking all the jump bikes uh kind of systems and and moving them over to a distributed uh system built on top of uh Cadence like uh the yeah the pre-temporal temporal. And so like it's just been a a a continual step of like I I want this experience whether it is like walking up to like a bike and just unlocking it and like having it be like frictionless to like work or whatever and then like necessitating the like depth required to go in and make that happen, right?

Jake Cooper: Like it it a lot of the work that that I do and a lot of the team does is like it's all in service of that experience, right? and like we fundamentally don't care like how deep we have to go whatever like we will swim to the bottom of the swimming pool to go and get the experience right um and I think that's what a lot of of you know kind of the trajectory was right and so it's not like I have a physics PhD or whatever I did like an EECS degree you know it it's just it's always been about just trying to figure out that next step of like how do we get there right um and that's like what's led to you know starting railway for that experience and then like moving all the way to bare metal data centers right like you know I was adding patches to the kernel this week Right. Just to like get the experience there cuz I'm like I see it and like how much better it can be. Right.

Allesio: Yeah. You have a data center in Singapore. Singapore, we're adding a second one in Q3. So, Yep. So, so like what's it like? I mean I I' I've never built a data center.

Jake Cooper: You basically just go and you say, "Hey, listen. I want uh power and I want a cage." Uh and they're like, "Great. Here this is what it's going to be." Um and then you rent the cage for a period of time and then you have to fill the cage uh with racks servers and then hook up internet to it, right? Um that's realistically.

Allesio: and like what's the math versus obviously the clouds?

Jake Cooper: Yeah. Uh it our payback period when we go to to metal um if we rent it in the cloud, our payback period is about 3 months. It goes crazy. It's nuts. Yeah. and and that's like four years worth of like depreciated hardware, right? Um and so I think it's like you're going to see a lot of this almost like compute crunch so to speak because a lot of the hyperscalers are buying up a lot of stuff.

Jake Cooper: if you go back to every person is going to run you know dozens hundreds whatever of agents in parallel like you should spend more than you have you have like you have no conceptual idea of like how much compute is required to go in and make that experience happen. Even if you're deeply efficient, even if you're sharing resources, even if you're doing all of these other things correctly, and that doesn't even count inference.

Allesio: Is there any point at which you started prioritizing um AI developments or agent development?

Jake Cooper: Um and probably over the last like six months we've probably deep deeply prioritized like Agentic as a as a mechanism to go and build and and deploy things. uh just because we we believe fundamentally like the the curve is so sheer and like that is the way that people are going to go and build and deploy software. we've moved from assembly to C to C++ to JavaScript to now like words, right? And you're going to need to be able to close that loop, right?

Allesio: Yeah, the SDLC is changing and it's something that both of us are like super interested in exploring as well. Um, my one of my thesis or it's not my thesis, it's the pull request is dying, right? It's it's going to be the prompt request.

Jake Cooper: and then beyond that code review is also kind of dying because do you really need to uh if you have all the other systems in place? But that like pushpull rebuild thing, right, is a point of friction that we are like removing entirely from from our loops.

Allesio: cattle not pets. Right. Because like your prod like it has a name like a baby, you know, I have to keep it alive. But when it's cattle, you can just mass farm and you can like roll out and you can like, you know, uh portion out parts of them and kill them or whatever.

Jake Cooper: I actually think that maybe that's the the hot take, but I think that that's actually going to change. And I think you can move towards having pets so long as uh you have a and this is going to be a jump. So long as you have a cloning machine for your pets. If you can snapshot every single thing at every frame, then like it actually doesn't matter if you know that obliterated because you have some sort of like snapshot of it, right? What if you just had the whole file system? what if you just snapshot it and what if you lazily load the entirety of the file system? right then you can get around this problem entirely.

Allesio: Sounds fun. Amazing. All right, that's the future of cloud. Thank you. Thank you for having me. It's been wonderful.