Short answer: Jordan Metzner says it is shocking Apple and Facebook have not launched clouds, buying everyone a $10,000 computer makes less sense than cloud spend, and developers must be willing to change how they work almost weekly.
Watch this answer from Episode 32 (starts at 25:42) · Episode 32 notes and transcript
From Episode 32: From DNA to Drugs: How AI Is Rewriting Human Biology, starting at 25:42.
Dov Gertz [25:41]: So do you think, Jordan, the hardware providers like Mac will be inclined to opening their own clouds?
Jordan Metzner [25:50]: The fact that Apple hasn't launched its own cloud has, like, still shocking to me. The fact that Facebook hasn't launched its own cloud is, like, still shocking to me. And it just seems like all of them have some sort of compute and opportunity to sell that downstream. I mean, Amazon's been monetizing Mac computers and AWS for years doing, device farming and things like that. Right now, obviously, we're in a shortage of RAM and, hardware prices are going through the roof. But, right now, if you want a Apple computer with more than one terabyte, an Apple laptop with more than one terabyte, it requires a custom order. They're not available in any store. And a terabyte I mean, my computer hard drive has a terabyte, and I'm full. So, I think, like, definitely, the hardware manufacturers are far away from being able to keep up with, like, the amount of compute power we see on, either, GPU based or even CPU based servers. I've even seen some people say that they're they're buying, like, old HP servers, and, like, sticking them in their office because they have a 120 gigs of RAM on them already, and they're pretty much, like, free.
Jordan Metzner [26:52]: So but, yeah, I think, like, cloud in general to me has seemed like it was a good solution back then, and it kinda solved a bunch of problems. And it doesn't seem like it's a logical thing to go buy everyone a $10,000 computer. It seems like it's much more likely to buy a thousand dollar computers and put that $100,000 into the cloud. But, again, I don't know. I the only thing I can say for sure is that there's going to be a lot of change. And, I tell all my developers, like, if you're not willing to change the way you work almost on a weekly basis, like, then you're not ready for this world. Because, like, if a new model comes out next week that we see a step function improvement, then, like, we immediately move, and that actually is a bigger dictator for us in how we work. Right? So, if OpenAI launches, Codex 5.4 or 5.5 and it's only available via APIs in the cloud, then, like, we're never gonna go on prem because they're not gonna allow us to. Know?
Jordan Metzner [27:42]: You If Anthropic does the same thing, we're never gonna go on prem. Now if we start to see some of these open source models actually catch up or even, surpass the closed source models, that might change the behavior. But I think then we'll just see the cloud providers providing these at the lowest cost possible. So I think every turn, it just goes back to, like, it's better to have a professional run a cloud, and you worry about your computation locally, but, it's it's ever changing. And, honestly, like, who knows what's these agents are going to invent and, maybe we can become a lot more efficient, and maybe, a 100 gigs of RAM isn't necessary because we can run things with, 36 gigs or something, even going back to 16 gigs or something that. So but it's super exciting and a lot of fun. I think, like I mean, I know myself and, like, most of my developers haven't been sleeping the past few weeks just because, like, as soon as the job ends, you wanna, like, tell it to do one more thing. You know? So it's probably like a weird vicious cycle, but it's been like, it's crazy how, the virality of building has been able to expand beyond developers to kind of other people inside of the R enterprise, and then their excitement about building stuff when previously they were never allowed to. And, Sam mentioned something to me the other day, kind of you know, engineers were always kind of siloed in the sense that, like, their brains were focused exclusively on writing syntax based code and kind of, don't disrupt the engineers that are writing this code, and, you kind of do everything you can to prepare a PRD and a document and all these things so that when their time comes, they'll be able to pick it up.
Jordan Metzner [29:16]: And now just the democratization of everybody we'd be able to build kinda just opens that up to the point that, there is, like, not that type of silo anymore, and it's become kind of
Sam Nadler [29:26]: Great episode, Dov. Thank you so much for joining. Anything to wrap up, or otherwise, we'll we'll call it the end of the episode.
Jordan Metzner [29:34]: Yeah. Dov, tell us where
Dov Gertz [29:35]: Sorry, Jordan.
Jordan Metzner [29:35]: Okay. Tell us where people can find out about you.
Dov Gertz [29:37]: Oh, so our website is converge-bio.com. Please do that or follow our page on LinkedIn. We share a lot of interesting scientific content. Yeah. And this has been really good. I need the game. Please send me the cell game.
Sam Nadler [29:51]: I will.
Dov Gertz [29:52]: In our Next conference. Great.
Jordan Metzner [29:54]: Okay. Awesome, Dov. Well, thank you so much. Thanks for joining us. Thanks, Sam. This was a great episode. Thanks, everyone, and hope everyone will see you soon.
