00:04:41 · Isaiah Taylor
[00:04:41] S1: Yeah. You know, I think that in nuclear, it's particularly tempting because nuclear operations are really, really hard. Like it's just really hard to split an atom. Not from a technical standpoint, it's actually pretty easy to split an atom from a technical standpoint, but from the perspective of operationally being a company that can build a core, can fuel it, and then take it critical, that is actually a really, really hard thing to do in terms of making sure that your analysis systems are actually good. Right? Does your analysis lead to something that becomes real in the physical world? Do you have security? Do you have MCNA? Do you have a really good operations team that understands what they're doing? Does the regulator agree with you on all of those things? Right? So because iteration is so difficult in nuclear, people have just sort of defaulted to design. Right? Because you can do design on paper. Right? And so that's really what we've been doing. The other thing is I think that people look back to an era of nuclear where it was more possible to do some things in design land. What I mean is it... There's a certain period in in nuclear history where you had sort of like the EPC driven style of building nuclear plants, and it it kinda worked. And the reason it kinda worked was because we had already done a lot of fundamental r and d in real world tests. Right? So the Atomic Energy Commission turned on 70 different nuclear reactor prototypes, the Navy ran all of these different nuclear reactor prototypes, and then we also had a very healthy EPC industry in terms of combustion plants. Right? So gas and coal plants. And so you kind of took those two things together and you could sort of throw that into an engineering context and do some plant engineering and throw it to an EPC and you get a plant out of it. None of those things are true today. Like we haven't been turning on 70 test reactors, we don't have that huge core of people who have just turned on 70 test reactors, and we're also not very good at even building normal plants anymore. Like we don't build plants at the speed that we should, at the scale that we should, We don't build bridges as fast as we used to. We don't build dams and highways and all these other works of civil infrastructure.
00:06:53 · Isaiah Taylor, Ty Morse
And so assuming that we could sort of go back to building nuclear the way that we were in the sixties and seventies just is not true. And so really how we have to build nuclear now is the way that you'd build any other deep technology category, which is it has to be iterative. You have to get a group of smart people together in a room. You have to start in hardware. You have to build as quickly and safely as you can and get real world data. And it really is the only way and you will quickly learn from that. And you'll learn lessons that no matter how smart the other guys are, they will just never learn those lessons. Right? Like without the hardware, they just will not discover, you know, the the painful lessons that we've learned in this plant. [00:07:37] S0: What are a few of the things that you had on paper, you thought things were gonna work, and then you, like, have contact with reality and everything gets fucked?
00:07:44 · Isaiah Taylor
[00:07:44] S1: Yeah. I mean, you know, there's like so many of these examples, and you know, probably some of the things I'm about to say will go try the other way, and then we'll discover that it's even more painful to do the other way, so I'm even hesitant to even say some of these things, but like, a good example here is this, even the story behind this citadel here, we had originally thought, okay, we're gonna have to go to modular construction on our nuclear shielding eventually because you really don't wanna be casting on sight. Like when we get to scaled production of nuclear reactors and we want to turn on a reactor every day, we can't be pouring concrete and doing rebar work and form work before we go and turn a reactor on. So we have to get to modular shielding. And, you know, three months into the process of figuring out how do we actually make modular shielding, we're like banging our heads against the wall, and we're like, this is impossible, like this is, we've given ourselves an impossible challenge, and so I directed the team, like, screw this, we're not doing modular shielding this time, like, let's figure out... Just pour it in place, right, like, you know, we'll tackle that later. And we spent a couple weeks, like, figuring out how to pour in place, and we realized, oh, that will add two months to schedule. Like, we'll actually just add two months where we can't do anything because we're doing rebar and we're doing formwork and then we're pouring and the reactor's supposed to be inside and we're supposed to be doing wiring. And so then we actually came back to modular modular shielding again and ran into the 16,000 extremely hard things that it took to get this right where we don't have to grout this shielding. Like, that's an extraordinary thing. The fact that we can actually stack this shielding up with the crane, we've got an overhead crane here and we just place those blocks, they don't connect together, we don't have, you know, like bolts that that bolt into each other, we don't have grout, it's just stacked. And that was extremely difficult to get right. And, you know, our engineers have, like, 15 different things, you know, we all have 15 different things in our heads that we had to run into into a brick wall and then work our way around it. And, you know, I I just think about if you were to try to design this on paper, and we did.
00:09:49 · Isaiah Taylor, Ty Morse
Right? We did design this on paper a year ago, and it looks very different from how we originally designed it. And, like, could you, in theory, have gone through, like, 10 more design cycles to try to get here? Maybe. But it would have taken like years to go through those design cycles. [00:10:04] S0: A little bit more efficient to just like shoot the rocket up and see what happens.
00:10:04 · Ty Morse, Isaiah Taylor
[00:10:04] S0: A little bit more efficient to just like shoot the rocket up and see what happens. [00:10:08] S1: Yeah. Exactly. And so in at Valar, we we have a phrase which is that steel is cheaper than software engineers. So like software engineers are actually very expensive. You think, okay, we wanna understand the nature of this problem, we wanna do a really good design, and we're gonna do that through analysis, right? If you're doing that through analysis, you're paying software engineers, and the software engineers are gonna go and, you know, come up with this perfect imperfect world model, and then they're gonna test against this world model, and you're gonna be a couple months later with an answer, and if you add up the, you know, salary cost to the people on that team over that amount of time, and then you go and say, how much steel could I have bought for that amount of money? It turns out you could buy a lot of steel for the same amount of money that you got a software answer, and for that amount, you probably could have welded it together and gotten an answer in the real world. And we find this to be true over and over again. You know, when we first started saying this to ourselves, it was sort of a joke, and then we realized, like, oh, this is really true, like, to an order of, like, five or 10 x in many cases. [00:11:09] S0: When we were just touring this facility, there was this amazing moment where you pointed to, like, a ConX, and it's got effectively the brain of the nuclear reactor in [00:11:16] S1: it. Yeah.
00:11:16 · Isaiah Taylor, Ty Morse
[00:11:16] S1: it. Yeah. [00:11:17] S0: And you said if this was bought from traditional suppliers, it would cost something like $17,000,000. Yep. And I think the idiot index for nuclear is just insanely high. How much does this cost, you know, roughly? And then what have you kind of... Like efficiencies you've been able to wring out through just doing it yourself and Yeah. Integrating?