Short answer: Jordan Metzner says he hosted OpenClaw on DigitalOcean in about five minutes for around $6 a month and expects work to move to cloud desktops, not on-prem. Dov Gertz agrees most of their computing is in the cloud.
Watch this answer from Episode 32 (starts at 22:30) · Episode 32 notes and transcript
From Episode 32: From DNA to Drugs: How AI Is Rewriting Human Biology, starting at 22:30.
Dov Gertz [22:29]: Just, I saw that, like, a lot of people are buying MacBook Pro MacBook Airs now because they don't wanna run it on their own, computer, OpenClaw. Did you do it on your own computer, Jordan? Or
Jordan Metzner [22:39]: No. I just put it, like, on a I just hosted it on, like, DigitalOcean or something. It took me, like, five Minutes install. I don't really understand why everybody thinks that they need, like, their own piece of hardware to run this thing that yeah. I just, like, ran a terminal command, and it was hosted on DigitalOcean in, like, five minutes. And I've updated it multiple times without a problem. So and I think it costs, like, 6 ball $6 a month. So I haven't found the need why you'd need a an on premise device. And I know, a few weeks ago, heard Chamath on All In Podcast say on prem is the future, then kind of it's coming back. And I actually think the opposite. I think, like, in the past few months, we've seen, at least for myself, like, my MacBook is getting burnt. Right? I'm just, like, maxing out the hard drive. I'm maxing out the RAM. I'm maxing out its energy consumption. And to me, it seems like it would be much better to have more of a cloud desktop. And the more and more I get, I actually want a Mac that has almost nothing on it, and I wanna, c SSH into my cloud desktop or multiple cloud desktops or multiple agents and try to make my machine as dumb as possible. So while he says, like or he believes maybe that, kind of removing more and more to on premise, I actually think the opposite where, running these models is gonna be almost impossible to do yourself.
Jordan Metzner [23:50]: You're going to need a large management of cluster management in order to do that. It's never going to be as efficient as sharing them with foundational labs. And I think that over time, these computers are not gonna be able to keep up with how big servers are going to get into the cloud. And so I think you're actually gonna get a machine that's very lightweight, that's very cheap to produce, that doesn't have a lot of RAM or memory locally, but is able to produce is able to communicate with s h SSH through your server. So I actually see us, like, kind of going away from on prem and moving more and more to cloud and even away from the desktop itself. But, yeah, curious how you guys see kind of even just, like, internally how your engineers are developing software.
Dov Gertz [24:26]: Yeah. I think very similarly, like, there's no question that I believe very similarly to you in terms of on prem versus cloud. Like, I think we're at cloud vector definitely. I mean, some of our computers have a 128 gigabytes RAM, like, to run, like, locally. I'm not sure I agree with the logic behind that because, like, some folks rather sometimes run things locally, but the raw the general majority is is all computers on cloud. And, yeah, there's there's but sometimes little experiments people run on their desktops. But yeah.
Jordan Metzner [25:02]: Yeah. I remember when I first joined Amazon, I think it was, like, 2016. I was in the personalization department, and they just have, like, a box in the corner of some, like, NVIDIA gaming cards. That's what they were using. You know? So I definitely see it for, like, this playtime, but, I mean, imagine, every engineer or company needing, 256 gigs of RAM on their computer in order to operate. I mean, it just seems like a non scalable model versus, kind of cloud compute. But I guess we'll we'll see what, Apple launches re new, and which way the market goes. But, again and as well, like, as the foundation models go as well. Right? So whether they even allow you those abilities to run them yourself feasibly locally as well.
