Ep. #89, Software Is the Killer App for AI with Bryan Cantrill

Ken Rimple: With everyone doing everything in the cloud these days or the perception is everyone's doing everything with scalable cloud computing, Oxide Computer Company selling rack mount systems to customers is a lot different. And it's a bold move when it was created in 2019, right?

Bryan Cantrill: Definitely bold in 2019. Yeah. And we did have one prospective investor be like, you know what I love about you guys? You're anti-cloud. And we're like, yo, yo, yo, yo, yo. Wait, wait, wait, wait, wait, hold on, hold on, hold on. We are not anti-cloud. Just like, let me try to get that one before it gets out of the park.

Charity Majors: We're the Pong of systems.

Bryan: Right. I mean elastic infrastructure is extremely important and the ability to have API driven infrastructure, to have infrastructure as code, the ability to be able to view the infrastructure as part of the larger system is extremely important. Our belief was that shouldn't be confined to a rental model, that you should be able to do more than than rent. You should be able to own it.

And this is what I mean. Charity, back in the day through Parse, when you were at Facebook, Facebook had its own elastic infrastructure that it built. And you know, for us looking at the kind of systems that Facebook, now Meta, had built for themselves, you can't buy those systems.

I mean, they're gorgeous. Circa 2019, you look at like Tioga Pass or Bryce Canyon or what have you and you're like, where do I get one of these? And it's like, no, you've got to be inside of then-Facebook in order to--

Charity: It's a Ferrari. You only get one if you work for the Ferrari manufacturers.

Bryan: It is true. You know, actually, if anyone wants to really hear very wealthy people sound very aggrieved, ask them about the way Ferraris work. Because, I mean, I'm not a car person. You know, I drive the Kia or whatever. But you can't just go buy a new Ferrari. You have to buy someone else's Ferrari, who buys someone else's Ferrari, who buys someone else's Ferrari, and that person buys a new Ferrari.

And I do kind of love the Ferrari dealerships because the, I mean, of course people are like, "no, fuck that, I'm rich. I'm going to go buy a Ferrari." It's like, you think the Ferrari dealership doesn't deal with entitled rich people all day long? All day long. Like, that's literally all they deal with. People that aren't rich are afraid to go in there. So it's like you're only dealing with rich people.

Charity: Oh, God, can we even imagine?

Bryan: It's kind of a similar thing. Because we were at Samsung, and it's kind of like the Ferrari dealership, it's like, "fuck it, I'm rich. I'm going to go buy a Tioga Pass." It's like, no, you're not. Nope, it's just not for sale. And it doesn't matter.

In fact, I was with my now co-founder, Steve Tuck at the time. We're at this OCP Summit and he's like, this is gonna be great. Because he came up on the go to market side and he's like, "I'm at Samsung, I have like unlimited checkbook. I am gonna be like the sales prospect that I always wish, I'm gonna go jump in someone's boat."

And so we're at the OCP Summit and he's trying to go around these systems. They're like, "hey, go do this unlimited budget. I wanna buy a couple of these. And they're like, nope, just, sorry. Which group on Facebook are you in? Like, I'm not familiar with you. Where do you, who do you roll up to? Like, where are you in the org chart?"

Like, okay. And the other thing it was really interesting is that we hit other folks who are at OCP summit looking to buy them. And they weren't able to buy them either, of course, because they weren't for sale.

So what you begin to realize is that the cloud model, ask infrastructure, of course is very important, but people want to have different ways of consuming that. Some of that will be a public cloud rental model, which is still great. But there's a lot of reasons why you may want to own your own infrastructure.

And so what we set out to do is really inspired by those systems, is take a real first principles from scratch approach and build the thing. We did our own boards, we did our own switch, we did our own rack design. Then we did, importantly, all of our own software.

Ken: Wow.

Charity: I don't think I've ever heard the Oxide origin story before. Democratizing data centers. That's beautiful.

Bryan: Yeah. And it's been fun. I mean, I think that felt very iconoclastic in 2019. We had one of these problems. I know Charity, you'll appreciate this where you can get, especially at the kind of the dollar figures we were looking to raise, because this is not a cheap thing to go build, you kind of have to get an entire partnership on board. And we found that we could easily get a single VC partner to fall in love with us.

Charity: Yeah, yeah.

Bryan: And then their partnership would be like, "what is this that you brought home? Like, no, no, no, you can't keep it. Like, that's a skunk. No, it goes outside."

It's like, oh my God, did it spray on you? It's like, no, get that thing out of here. So that happened over and over again. In fact, we joked because partners would fall deeply in love with us, and then they would not be able to get the partnership over the line. And we had people that end up leaving their partnerships.

Charity: Oh, my God.

Bryan: We joke about the pelts on the wall.

Charity: The seedy underbelly of--

Bryan: We're like VC homewreckers. It was really kind of delightful, but I think it highlighted how iconoclastic it was and how tantalizing it was. And yet, if you really get into it, it's really hard. And it took us three years to bring our first product to market. Three and a half years. Which, I mean, to those people who know what we did, you're like, that's like record time.

Ken: Yeah, it is.

Charity: Yeah. In VC years, that's like three decades.

Bryan: That's like three decades. But I think the thing that we have seen, I mean, it was also funny. I then have VCs that cruise, firms passed on us, but, you know, where the partner kind of fall in love with us. They would then, like, send me things that they would find on the Internet, talking about how, you know, repatriation workloads or the necessity for on prem data centers. And I'm like, why are you sending me this? You should be sending your partnership this. Like--

Charity: I don't need your convincing, dude.

Bryan: Exactly. So I think we've seen a lot of trends like that where, I mean, I think that. And we knew this was coming in that, like the thing that we definitely knew-- I mean obviously there were a lot of trends that we didn't predict and I, I know it's true for you Charity and you all, like there are certain things that you can predict that will break your way and there are a bunch of things that won't break your way that you, that you think like this will happen and it like it does happen--

Charity: And there are things that seem obvious. And we're still using metrics, logs and traces in different storage systems. And you're just like, I thought a decade ago this Berlin Wall would come just crumbling.

Bryan: And that's a good analogy. Right? Because I think like the Berlin Wall stays up longer than you think possible, then comes down faster than is fathomable. So when these things do switch, they can switch very quickly and you've got to be in the right position. And we, you know, we got unlucky in any number of dimensions, but then got lucky and a couple of big things broke our way.

And our need for compute has skyrocketed with generative AI. With AI in general, I mean, we kind of felt like, well, we don't have a GPU based sled. So it's like that's not a market we're going to play in. Which is fine because we got plenty of other things to go do.

But what has happened, of course is with the rise of agentic AI, generative AI, it's not just GPUs. Right? And if you are sitting there at your LLM of choice and you've asked it a question and it's searching the web or it's writing a program for you, or it's doing these kind of active things, it's not a GPU that's doing that. It's a CPU that's doing that.

Ken: It's running tools. Yeah. It needs the CPU.

Bryan: Yeah, yeah. And so the CPU needs to run somewhere. And if it runs on the public cloud, it's kind of economically ruinous. I mean that was the big--

Charity: And Ken, are you trying to interrupt Bryan and be like, "this would be a good time for you to tell us who you are, Bryan?"

Bryan: Hey, yeah. I'm the loudmouth here. Or at least one of them. Charity, I think you and I can go down together on this one. Bryan Cantrill, CTO and co-founder of the Oxide Computer Company.

Charity: I first became aware of Bryan when we argued on the Internet over, I believe, DTrace and observability. And Bryan was very aggressively, like, DTrace is observability. And I was like, what are you talking about? It runs in the shell. And ultimately, I think we converged on very similar things.

Bryan: Yeah. And I actually found a deck of mine from, like, 2003. Because when observability became a more mainstream term, I'm like, didn't I? I'm like, I-- Because I don't remember the reason I started using the term-- I don't remember someone using the term observability before I did. The reason I, and Charity you'll appreciate this, I was doing DTrace no matter what. This is at Sun back in the day.

I needed to have a way to fit into a VP's deck where they, it had to be like a roadmap item, and I needed to make it a loftier idea, basically. So I'm like, I gotta come up with, like, some term. And observability is what I came up with. I'm like, this is what I'm improving. Like, this goes to the observability of the system.

Charity: Oh my God. Yeah.

Bryan: And I was just using that a lot to explain it more broadly. And then, it's kind of funny that we're on the o11ycast podcast here, right? When that kind of became an industry term, I'm like, is that? Is that coming from my need to justify what I was doing to Sun's like, middle management? I think it is.

Charity: It might have. So I had heard the term because Twitter, for a while, had an observability engineering team.

Bryan: Yeah. And I think Twitter got that because Twitter was one of the first places that we used DTrace in production, was at Twitter.

Charity: Wow.

Bryan: Yeah. In 2007. Well, they were a joint customer way back in the day.

Charity: Well, thanks, Bryan.

Bryan: There you go. Yeah. Well. Or thanks Sun middle management, I guess.

Charity: Man, you've been dropping a lot of phrases that I'm not sure have ever been uttered before.

Bryan: Yeah, you know, I get that a lot.

Charity: Someone sent me that deck, I guess you posted it recently. I have been trying to do social media again, not because I want to do social media, but because I realize that it is my job and I have not been doing my job. But someone sent me that deck and I still have it in a tab somewhere. I'm like, I want to post about this.

Bryan: Yeah, yeah, that was a deck that I-- And I can't remember why I was in the crypt looking for something. And that was a deck from way back in the day. That was at a Sun internal conference that Salesforce also presented at. Salesforce was very early on this. Do you remember their like "no software" thing?

Charity: Yeah.

Bryan: They really were, I mean even it's kind of weird because they're so-- Like people look like, oh my God, Salesforce is like a total pioneer. But they kind of were. They were a pioneer in-- They were SaaS long before we had a term for it.

Charity: They did SaaS before there was SaaS. Yep.

Bryan: Yeah, right. I feel like they were-- And so they had presented at that Sun internal conference on what they were doing. And I remember thinking like, wow, that's really interesting. And the thing that's amazing is like I didn't really put two and two together with respect to cloud computing.

And what I thought is like, oh, okay, so a lot of people are going to do what they're doing which is buying these big honking Sun servers and then having software available as a service. But again, the term SaaS. I know I did not hear like I don't think the term SaaS existed then. I think the term.

Charity: I don't think so either. But I have heard Marc what's his face take credit for inventing it.

Bryan: I think he's not wrong.

Ken: I think it was the first big app everyone used.

Charity: Yeah, I think I'm a little bit younger than you. I don't remember when it was happening, but I've done some research a few times in recent years just like, okay, what are these timelines? Like what is this sort of-- You know? And yeah, that's what I've read on the Internet and God knows the Internet is never wrong.

Bryan: Yeah. And it's kind of funny to me because you know, we-- And we still do this, Charity, you've been on my podcast and we do a "predictions" episode every January. Which is always like, I love predictions.

Charity: They're humbling. Predictions are always humbling.

Bryan: They really, really are.

And I think it's really helpful for people to write down their own predictions. Because you're exactly right, Charity. They're very humbling. Predictions tell you more about the present than the future.

And it is amazing to me this stuff. So we would do this every roughly January 1st, get together as we go into this restaurant in Palo Alto, the Kernel Group, and we would make these kind of predictions. We never predicted anything approximating cloud computing, even when it was getting like closer and closer and closer.

Like cloud computing, if it feels obvious, it wasn't, it really wasn't. Because I think that people-- Part of the reason we wouldn't predict that is like, who would run their infrastructure somewhere else where you would like, of course you need to run your own physical infrastructure. This idea of--

Charity: Software crashes. Somebody needs to go fix it.

Bryan: Totally. And I think also because another kind of thing I wonder about, because Sun, ultimately, once it was clear that cloud computing was developing, sun was in a position to do some really interesting things there.

I think that cloud computing needed a non-computer vendor to pioneer it.

Charity: Oh yeah.

Bryan: That even if you could rewrite history, you're never going to have a Sun or an HTTP or an IBM.

Charity: You need someone who's zoomed out, who can be like, "that is problematic. I don't like that. I want to make it not that."

Bryan: That's exactly right. So I don't know, the cloud computing, it didn't necessarily need to be Amazon, but it needed someone who could step outside of computing and yet still be a technological leader pioneer. And it kind of made sense that it was a retailer and you know--

Charity: It was bringing technology to consumers. Like we're, as software engineers, we are consumers too, you know, and it was making it consumer quality in a way.

Bryan: Totally. And now I think there are other companies that could have done it that you could kind of in the abstract, but yeah, no, it's kind of wild. And anyway, so we, it's, it's just amazing to me. And I kept, you know, going back in time talking about observability, about like the things that we did see and didn't see.

And there were a lot of things that we just didn't see or would mispredict. Right? And I think that the world has a way of unfolding that kind of, that defies predictions. I mean, as Yogi Berra says, "predictions are hard, especially about the future."

Charity: Especially about the future. Which right now. What I think is so interesting about this moment is that we are in a period of compressed change. I described it to someone the other day I was like, it feels like over the past three years we've gone from machine languages to Ruby on Rails and at the same time we've gone from FreeBSD racked servers to the cloud, both at the same time, all in three years. But we're not to the other side of it yet.

Like all we know is that we don't know anything. Like, oh shit. All this stuff is really exhausting to humans. It's like we just found that out a month or two ago. Oh shit. Like we know that what we were doing doesn't work anymore, but we don't know what does work. And so we're all just kind of like--

All right, let's do a prediction. Bryan, how long do you think it's going to be until we look back and go, "oh, oh well that was obvious. This is of course the future of development with AI."

Bryan: Yeah, that's a really good question. I think that there are a bunch of things that we will look back on and it will be obvious now, but are you know, just like I've got at this point my kids, I've got a 21 year old, an 18 year old and a 13 year old. And you can always, as the kids get old there are always properties they had as an infant that like, you know, my now 18 year old kicked the shit out of my wife in the womb.

I mean that kid was just like, had, didn't need a vital organ that he didn't just want to like, just-- And you know what, that still is kind of his personality, still kind of an organ kicker. Love that kid, you know. So I think it's kind of things like that where it's like there are going to be things that when we look backwards like, oh, absolutely. Like we can clearly see that now. But I agree with you, it's really hard to see how a bunch of this unfolds. I think it is also like there's a bunch of reductive thinking that's not terribly helpful.

Charity: Yeah.

Bryan: You know, where people are extrapolating in ways that like. Well, we actually, I'm not sure that that's a helpful way of thinking because that's not very actionable like you're talking about, you know, whether it's job loss and so on.

Charity: Maybe so, but is it going to help us to worry about that right now?

Bryan: Yeah, I really don't think that that's constructive. And I also get, especially for young people who are earlier in their career. I think that like they, and I would just say like, people have always been very concerned that like the young people are never going to find-- I mean, you know, we know that Charity, from, from back in our day. It's like there's always a concern and you kind of need to just like ignore the adults and go like you just got to go find something to do.

And also, by the way, it may mean the other thing that probably going to see a return of, it's like, yeah, you're going to take some like, some shitty jobs and like, that's okay.

Charity: We're all fretting about how are we going to train the next generation of engineers. And I think we should fret about it. But honestly, I think it's much more likely that they will figure it out.

Bryan: They will figure it out, a hundred percent.

Charity: If anyone is willing to hire them.

Bryan: Yeah, well. And if they're not, I do think it's like. Well, I actually do think we're kind of at a time of like it's really cheap to start a company. One thing that has happened is that the-- And I, because I've been trying to think about like, what are the historical analogs for this. I do agree with you. It's a compressed time frame and like there's no one analog that fits because it's a lot of stuff changing.

But you know, I gotta say it does remind me of open source. Open source was a really deep revolution and it's kind of easy to sleep on it because it feels like, oh, of course that's the way it happened. I'm like. But it really. There was a time we went from an all proprietary world in the mid-90s to a world in which you could not build--

Amazon could not have built the public cloud without open source. Google could not have built Google without open source. Facebook could not have built Facebook without open source. If Facebook was buying a Windows license for every one of those Tioga paths, like you couldn't do it. Right. It's like it literally is. It wouldn't economically stand up.

And open source, because you kind of go back to that revolution happening in the 90s, there was some kind of similar rhetoric, you know, about like the. Oh well, like open source means that like, like the world, the communities just kind of develop it. I just get to kind of use it. Or does open source mean, like, am I going to download something off the Internet and run it on my car?

And the answer is yes, that's exactly what you do. Like, if you want to entertain yourself, go to like the licensing screen in your car or where it's forced to give you all the licensing disclosures and it's like, yes, your car manufacturer downloaded something off the Internet and ran it on your vehicle. I mean, that's reductive. But if you want. That's also true.

So I think that open source really, really changed things. And one of the things it did is it lowered the cost of developing software in a way that benefited us all societally.

Charity: Yep.

Bryan: And on the one hand, like, people, you can complain about the amount of software in your car, and I get that. But go drive a car from 2000. If you want to get something to really complain about, go drive a car from 2000 and be like where do I plug in my Spotify here? It's like, oh, no, no, no, no, no. Yeah, sorry, do you have a-- Oh. If you've got a fancy car from 2000, it has a CD player in it. If you've got a super fancy car, it's got a CD changer in the back.

Charity: Yeah, five platters.

Bryan: Five platters. Otherwise you'll be introducing yourself to a tape deck or FM radio, my friend. And it's like, oh, look, who wants their software back in their car. Okay, okay .

The reality is that we have increased our quality of life by lowering the cost of developing software. And I think that open source did that. I think that now we're doing it again, and we're doing it again in some really interesting ways. We are lowering the cost of doing all sorts of things.

Charity: Software is not going to be a priesthood anymore, which is a good thing.

Bryan: I think you're exactly right. And that's another way in which open source is an analog. Open source did break the priesthood.

Charity: I agree. It is not a priesthood, but there's still, I think a lot of software engineers hear that, and quite reasonably they worry about. I don't think so. Like, the more software we have, the more broken it is. Like, come on, there's always going to be a job for people like us. As long as there is software, it will break. It's the only thing we know to be true.

Bryan: It is true that if your job is to put out kitchen fires, and let's face it, that's our job, the idea that, like, there's going to be a thousand kitchens in every home. That should be good news for the fire department. Don't worry, like, there are going to be plenty of kitchen fires. You do not have to worry about it.

Charity: I think that, yes, the pace of change, but it doesn't mean you don't have to learn new tricks. You know, Adam Jacob was, you know, he was working on System Initiative, another one of the very cool, you know, startups that everyone was really excited about, and recently spun that down. And I just think he has been doing the Lord's work by talking on LinkedIn and just being like, "all right, infrastructure folks, it is time. This is changing. The time is now."

And I wrote a blog post a couple weeks ago and I was just like, you know, I-- I was giving a closing keynote at SREcon less than a year ago. Fred and I were standing up in front and we were just like. The reason we were telling SRE that they should learn AI was because this was a big pitch. Your complaints are going to be heard unless you're inside the tent. Like, that was it. And I'm just like, that was less than a year ago. I can't connect to that now.

Bryan: Yeah, I think you're right. And I think, you know, these are fewer and harder to find, but I think it's-- I've told every, I mean, for folks at Oxide, like, everybody should be using this stuff to at least know where it can be used and where it's not gonna be used, because it is remarkable. And there are some domains in which it is extremely powerful. I think it's also, it is so powerful and it is a very-- It's a power tool and it also allows you to do some things that are not a good idea clearly.

Charity: People need to run towards the waves, though. Don't wait for the wave to smack you, because that is not going to be a good time.

Bryan: I think so too, because I also just think, and maybe this is because of my fundamentally optimistic disposition, because I think that, like, we want to use technology to improve our lives writ large. And I think that people are going to do that. I think that where these things don't work, people won't tend to use them. I mean, there's going to be some natural settling.

I do think that we will look back. I mean, I think some of the things we will look back on we'll be like, oh my God. I mean, I absolutely think, and I think that, you know, I had predicted this a couple years ago, this idea of AI doomerism and P(Doom) was going to drop out of the lexicon. I think that's kind of happened. I think that like, you know, I think that people are talking about the much more mundane effects of, of like, I'm worried that junior software engineers are going to be able to find a job much more so than I think the killer bots are going to destroy humanity.

Charity: Because it's just religion.

Bryan: It is religion.

Charity: It's just flimsy ridiculousness. I recently read this book called More Everything Forever by Adam Becker. He is a science journalist with a PhD in astrophysics and he just relentlessly demolishes the Silicon Valley, like sort of transhumanism and all altruism, effective accelerationism. And it is so satisfying. He's like, look. And he doesn't say this, but I think software is going to turn out to be the killer app for AI because it is something that we made to be reasonable about.

One of the things he says is all of the compute power in the world can't predict the weather two months from now. It can't predict the outcome of a social interaction with three or four people. Like, that's just complexity. Biology is not bits, right? So like this whole AGI, you know, here's my prediction. In 2027 we're going to start to learn that we had AGI all along. That sometime in 2020-- Because it's so poorly defined.

Bryan: Yeah, no, I do think you're definitely right in that there is going to be-- And I actually wonder if this was going to happen even last year because I think OpenAI has certain financial triggers with Microsoft that kick in if they have achieved AGI. So they've got like even an incentive to be like, you know, mission accomplished and be like, okay, we've got the AGI banner up on the aircraft carrier, so we're going to call it AGI Accomplished.

No, I absolutely agree with you and I think that we will-- Our disposition is, and I think we will look back on it. Oh no, actually it would be interesting to know if we do because I think sometimes we just forget about the fear. One thing if you really, I'd be, and Charity, you remember Y2K, there was a lot of fear around Y2K.

And it's kind of like hard to find because I do think people just forget. They want to forget about the Hysteria because it's embarrassing. In hindsight it was kind of stupid. But these hysterias are really interesting because to me, the kind of the omnipresence of the hysteria and the fear tells us something about our own humanity, when we have technology that we are afraid of how is it going to force me to change.

Charity: All the way back to Prometheus in the fire, right?

Bryan: 100%. The Greeks told us everything. Right? And I think you're exactly right.

Charity: There is a rate of change that people can tolerate that I think is being very stressed right now.

Bryan: Yeah, I think you're right. I also think that that rate of change--

Charity: It's not sustainable.

Bryan: So is this a bigger change than going from an agrarian society to an industrialized one in the span of like two and a half decades, three decades, where you know, you look at the adoption of like radio or the adoption of, I mean, or even faster television, where television goes from basically something that barely worked on a benchtop to being ubiquitous in like a decade. And like that's a pretty big change.

Charity: And for that to happen, electrification had to happen.

Bryan: Electrification. And I think that you know, for every one of those changes you also can find those folks that are like, I don't like this change, like I know I don't want this change. And I think you can also find people that really tacking into it. I do think that like historically young people have been excited for-- The change feels exciting and not terrifying.

I do think one thing that's a little bit different is I think that we made a bit of a mistake. I mean "mistake" is too strong. But I think a lot of people have studied computer science because mom and dad are paying a lot for college and you know, the college is way too expensive and so like I need to know I can get a job at the end of it. So I'm going to be very, very pragmatic, focused and I, I'm going to get the job at Google or Meta or what have you.

And it's like, it's probably the wrong reason to go into an academic discipline. And I do think like that is a shift where people are like, no, no, I, I, I did this because I'm going to get the job at Google. It's like, well that's going to be, that's going to be shifting a bit. And that I think is going to be-- So I think that is a difference.

Charity: That said, I will say there is something to be said for being taught to work out a problem systematically. Anytime I see a marketing leader, I'm just like you can tell the ones who have learned to debug, you know, in a systematic way.

Bryan: Totally, totally, totally, totally, totally.

Charity: Just like you can always tell the ex-philosophy majors who learned logic. So it's really the same thing.

Bryan: There is a lot to be said to kind of like learning how to think and you know, I do think that like a little bit of desperation is good for everybody and the--

Charity: I think the difference right now though is that-- So like I'm not worried that AI can replace-- I mean AI is a great, like it brings you to the midpoint, right? And if you get a shitty mediocre-- But I am worried that we do not have especially wise governance around and that that might. It doesn't matter if AI can actually replace a human, if a CEO-- The rash of layoffs that are happening. Nobody's like oh well we overhired, right?

Bryan: "We overhired,"right? No one wants to say that. No, no, no, can't say that.

Charity: So darn productive that we just had to lay off half of our people. Like you can see the tidal wave coming.

Bryan: Yeah, totally. I mean I think it's ridiculous and obviously with that the big Block kind of blaming it on AI. It's not AI. You over hired. You overhired. And I do think that like there's going to be some settling in where people realize okay, wait a minute, these things are actually extraordinarily powerful and we will kind of begin to understand how they can be powerful and how they can be useful.

I think we're just going to end up, I mean and maybe again this is my fundamental optimism-- I think we're just going to end up in a much more moderate position where that seems extremely, extremely sensible. I think it'll be like how could--

Charity: That does seem extremely sensible.

Bryan: How could anyone have thought that this thing would the-- I mean it is kind of funny to talk to like my own kids about, especially a couple years ago when I feel like that this, this P(Doom), Doomerism was much higher. And my daughter who is now 13 was then like 11, 10 or 11. She just thought it was hilarious that there were any adults that thought this thing.

She's like first of all, it has no arms or legs. And I'm like, that's a very, like, that's a deeper insight than you might realize. Like that is like that's an extremely important thing. Like you can actually. And she would love to like she was, you know, when she was first playing with ChatGPT. I mean she's just like loves being emotionally manipulative. Like for her it's like this is not a person.

So like the things that I would be, would, would be taboo with my friends, I can just manipulate this thing and like, I'm just delighting myself. So she was telling ChatGPT. Like no ChatGPT, you're my only friend. So I thank you so much. My only friend. And of course ChatGPT is like I am not your friend. I am software like-- You're immediately hitting like the safety things inside of ChatGPT.

And she is like, I'm sorry, is my only friend telling me that you are not my friend? Like, and you could see like ChatGPT is beginning to like legitimately freak out. It's like a very responsible babysitter and things are getting like out of control and it's like--

Charity: Your kid's going to be president someday. I look forward to this.

Bryan: You need to like, if you are, if you need counseling, you should seek mental health counseling. It's doing all these things. And she of course is like, this is exactly the wrong thing to say to like a preteen. She is just like uproariously laughing and I'm like, "ChatGPT she's laughing her ass off when she types this in." I want to like give it this additional. There's some very important context you're missing. There's a cackling 11 year old who is having the time of her life screwing with you.

And I'm just like, the kids are fine. The kids are going to be fine. They, I'm just like not worried about them. I think that there, there is much more to be-- I just think that this is true for all this stuff. There's more to be gained, to be lost and, and if there weren't, we wouldn't do it.

And it's not like, you know, we've other like technological change. I mean you know, I came up during the Cold War, right? And where there's a real fear of nuclear annihilation and you got a lot of these things that are like this is basically like a weapon that has like limited civilian use. And that's not this, like, this is something that is so general purpose that it yes, can be weaponized and there's a lot of like safety and regulatory concerns we have about that and we should have about that.

But this is a very general purpose thing that can be used for lots and lots and lots of things and lots of those things are productive and helpful and a good use. So it's funny again, I feel like this is a very moderate middle position that I think is very common but it's not being extremely well articulated by a lot of folks unfortunately.

Charity: I feel like there's this natural tendency that humans have to extrapolate from the micro to the macro.

Bryan: Yes.

Charity: I was reading a really interesting article the other day about how GPT usage is spreading like wildfire and people report being more productive. But there is no sign of that in macro level productivity stuff which I mean jobs are not actually getting replaced, individual humans are doing more things or whatever. But I think, oh, it was in The Economist. The Economist is like maybe this is a big world changer but it has yet to show up in the data. I'm just like that's fascinating.

Bryan: Yeah. And I think there are going to be dimensions in which it will be and this is where I'm thinking maybe it will look more like open source where it's like open source is a very big deal but to see the productivity gains you got to like look a little bit. I mean it's not. I mean you have to know that like yes, a Google couldn't be built without open source.

And I think we will have things like this where we will have companies that will be built that like no, you actually needed this technology to exist or you would not have been able to build this company. But the thing that it offers improves our lives incrementally.

Charity: And creates new problems for us to solve.

Bryan: And creates new problems for us to solve. Absolutely.

Ken: So one of the things I was thinking around all this is like how are you most productive with AI agentic systems? It's by being able to explain yourself and explain explicitly, "this is what I want. This is what looks good, this is what looks bad." It makes you better at externalizing what you're thinking and externalizing what you need, like you are working with software engineers. Right. That's the best way to get things out of it, I think.

Bryan: Yeah, I think yeah.

Ken: So there's a benefit to using it in that you sharpen your skills in those areas. If you were an individual contributor a long time and maybe you know, working off of other people's requirements, now you actually have to act like you need to feed somebody else your requirements to get something done.

Bryan: I think you're exactly right. And I do think that these things allow us to get better as software engineers, as thinkers, as SREs. I think that, like, these things allow us to up our game. I mean, I think, you know, Charity, you said earlier about eliminating the priesthood. This allows you-- I mean, even with open source, things like kernel development have had a mystique about them.

Well, sit down with Claude and not to have it do it for you, but have it walk you through, have it explain like, hey, I just saw this PR go up for like the operating system that I'm using, and I want to understand it better. Like, can you help explain this to me? And it's amazing at that. It's like all of a sudden we-- And I do think. I mean, one thing is definitely going to be interesting is that higher education is going to--

It's going to be really interesting to watch its transformation because we have now made this amazing tutor available to anybody, all the time. And I think the ability to improve our own understanding of these systems, I think is-- So, to me, like, that is just being able to explore topics. I think that, you know, Charity, we've all seen where people are because of rampant imposter syndrome, which everybody feels, by the way, like, to a certain degree, like, if you don't feel imposter syndrome, you're kind of on the wrong team.

Charity: What's wrong with you?

Bryan: Especially now. Yeah, yeah, you kind of need to be on a team that is solving a harder problem if you don't have imposter syndrome because you want to be in that space where you're like, wow, I'm on, on a team that--

Charity: I have to fight to keep up.

Bryan: I have to fight to, like, that's good. Like, that's good. And these, these systems can help you on that and they can help you, like, ask a question that you like-- I feel like I'm kind of embarrassed to ask this question because I-- And it's always been my experience that the best technologists are unafraid of asking basic questions because they're like, they've got the confidence to be like, I don't fucking care what you think about me. I need. I, like, I want to know the answer to this question.

So they'll ask a basic question. It's always funny to watch them. Like, everyone else is like, oh, I got that same question. But I didn't want to ask it. With these things, you can ask basic questions. So I think that there's a lot of really interesting positive developments to be had and then there's some danger too.

Charity: Okay, Bryan. Last thing.

Bryan: Yeah.

Charity: What is it that two years from now, when you're back on o11ycast, what do you think you're likeliest to look back and go, oh, I should have seen that coming.

Bryan: Yeah. Like, what are the things that really seem obvious now? Yeah, that is a very good question. I think that we are going to see much more software creation and I think that-- You know, I, I don't know that, that I quite. I mean, I clearly like investors believe that-- I did see one kind of famous angel investor VC say that SaaS is uninvestable.

Charity: Oh, yeah.

Bryan: Which, I mean. Okay, that warmed the cockles of my heart.

Charity: Everybody wants to build all their own tools. That's what, that's just what people want.

Bryan: Yeah, I think that that's right. So I think that there's going to be a-- I don't think it's quite that extreme, but I do think that there's going to be a lot of people building their own tools. I think you'll see a lot more custom software.

Charity: There are a lot of tools that are deeply hated, that are deeply loathed and that are charged way too much. Like they're treating the software like it's an asset and it's not.

Bryan: I think that that's right and I do think, and I think that that is going to be a real change. That is going to-- And that's a change that's probably a pretty positive change. I think that we are going to, on the one hand, have to get more creative.

On the other hand, these things are going to allow us to get more creative. So I think that we're going to see a more settling-- I do think that charity said earlier that software is the killer app and I do kind of agree with that.

Charity: Give us some closing thoughts. What's next for Oxide?

Bryan: Well, Oxide, you know, I've been saying for years that Oxide might make it. I think we are going to make it, actually. Yeah. So, yeah, which is great. I mean, I don't mean to sound surprised, but--

Charity: It's always a surprise. Most companies fail. That's amazing.

Bryan: Right. And I think that, you know, when you know you're going to live, it changes your disposition on things. Of course, just because I'm uttering this arrogance all out, the gods are right now plotting new ways to punish me for my hubris.

But yeah, I think, you know, so I think we're excited. There's a lot we want to go do. We're going to go. We believe that there's a bunch of need for on prem infrastructure that has modernity, that you shouldn't have to choose between modernity and being on prem. You should be able to be API driven. You should be able to, to have all that elasticity. You should be able to have it in your own data center. So we think we're going to do it. We're excited.

Charity: Yay. Well, thanks for coming, Bryan.

Bryan: Thanks for having me. This was awesome.