The Exchanges

Q folks working together, so it sounds like it's great when it works well, but you also have moments that are challenging, right? Like, you had an issue with a reactor vessel at one point during a particular test. Can you talk about how you overcome some of the things that are maybe not going the way you had hoped or, you know, posing some challenges that you didn't quite expect?

A Yeah, so, um, there, There, there are always challenges, and, and, you know, they range from minor setbacks to really good head scratchers. For the engineering test unit, one, we sourced the vessel from a shop, um, in Southern California. Very reputable shop. The costing was very reasonable. Our team was integrated in the design. Like, we had a, we had a very good relationship with them. It's a, it's an ASME certified pressure vessel, so it needs to go through a series of testing, hydrostatic pressure testing, to make sure that It can actually remain intact, uh, within the design basis. Um, so they did this testing, um, they happened to do it horizontally. The vessel was laid down on its side, which was not a great thing, so talking about, like, the learnings, we have a database that gets filled with all of these lessons, so don't do pressure testing on, you know, with the vessel on its side.

AI assessment note: “we have a database that gets filled with all of these lessons”

Q I love this. Okay, you mentioned that at first it was the engineering test unit, and then it became ETU one of three. So can you tell me when and what happened with the first engineering test unit? Like, when did you know that this was actually instead not going to be your only engineering test unit?

A Yeah, so I think this is like, 2019, 20 20. Um, there's always been kind of this question, well, the, the iterative process is like, start small and then go big. What would be the merits of doing a smaller reactor first? And so now Kairos is building the Hermes reactor in Oak Ridge, Tennessee. The Hermes scale is actually derived off of the scale that we chose for the engineering test unit. And part of that was, hey, if we want to move fast on a small reactor, we should leverage the engineering at the scale that we're already working with. So that's going to make it easier, going to make it faster. The other thing, which was I say a lot of things about what Kairos has done with, with great pride and what we're going to do with a lot of confidence, but, you know, doing nuclear power is, is challenging. And the, the requirements that go into a nuclear vessel, I don't think there's any other system quite like it in terms of the chemical environment, the temperature, the radiation, and all the hardware that just needs to be packed into a really tight space. And we basically said, we're not sure that we can go straight from this smaller scale Non-nuclear system to the full-scale commercial nuclear reactor design. So, um, we're going to do a smaller reactor, and the goal of that one is going to be, let's be as prototypic as we possibly can for the reactor system. A lot of other thing…

AI assessment note: “Yeah, so I think this is like, 2019, 20 20.”

Q any, are there any moments, though, where you find yourself asking the question, you know, for this particular component, or for this particular system, Should we outsource it, or should we build it ourselves? Is that a constant tension that you're facing, or do you almost always in those cases just default to building things yourselves, or how do you think about, think about those questions as they come up?

A We have more confidence now, certainly, than we did years ago. The team understands the benefits of taking ownership over not just the design, but the build. Uh, there's always a, there's always a decision, and, um, as the needs evolve, Like, I think the same principles still apply. And one of the key, key things is try not to stretch too far. Uh, and so if there's something that's new or dramatically different, it's usually beneficial to start off at least with a partnership, um, with, with someone who has, you know, some part of the key knowledge. And so on the pump, we, we started off, you know, we have a partnership with Barbara Nichols. Kairos has absorbed more capability ourselves, but it was a really productive partnership. We have found partners along the way that are, have been great, great to work with, and if that works out, that's great. Like, we want to, we, we don't want to do everything ourselves because it adds to the list of things that need to be done. Right now, we're, we're pushing the boundaries on the civil construction site, which is, um, this is much bigger scale than, than all the other things that we've been setting up to build, and so right now, one of the exciting things that is emerging is, um, Um, the opportunity around, uh, precast concrete, which is really just, like, potentially completely transformational in terms of how we build these building…

AI assessment note: “if there's something that's new or dramatically different, it's usually beneficial to start off”

Q iteration, and then you're coming to the point where you're ready to build a more complex sort of proof point version of your system. Was it always obvious that you were going to do something that was a non-nuclear demonstration, and was everyone immediately bought in, or did you have moments where you thought, ah, maybe we should move immediately to something that actually has some nuclear fuel in it?

A So there was not one mind when we're going through this decision process. So We had this, this timeline, which was like, 2018 to 2030 to get a commercial demonstration reactor up and running. And that timeline is just, it's, it's like forever in terms of, of engineering. And we wanted to create big things that could be done in more like two to four year cycles. And if we do non-nuclear systems, we can, we can go faster. That's a huge advantage. The problem is like, At that point, we had essentially a commercial design that we were working on. We didn't want to build something that big for the non-nuclear industry. So, how do you scale down, and, and what do you define as what you need to actually do in that test? And so, we basically said, all right, well, we want to, this, the scale has to be relevant. So, it's basically about half, the, you know, the length scale is about half as much as the commercial reactor design at the time was. So, um, the first point was, You know, it's an opportunity for us to go through iterations of all the major, uh, systems in the reactor. We basically trained a whole, like, generation of Kairos engineers around the whole life cycle of a major system, and in the nuclear space in particular, because there's so few build opportunities, there are a lot of people who just have experience in a very narrow band of that development cycle for the technolo…

AI assessment note: “there was not one mind when we're going through this decision process.”

Q us through kind of how this worked for you. Did you know from the very beginning that you wanted to follow this iterative process? Did you know from the beginning that you were going to do a whole lot of non-nuclear work first? How did you really get through your paces as a company in those early stages leading up to your decision to build a non-nuclear engineering test unit?

A It took us a while to kind of figure out what this was going to look like in practice. Your story about, like, those original tests, it was, there was kind of, like, this existential moment where, like, we have the infrastructure, we have this vision, but, like, we have so much to do, right? Like, to, well, you're starting off, like, we got to design a new nuclear reactor from scratch. Like, where do we start? And we've got, I don't know, we probably had 80 people at the time when we moved in, in 2018. It's like, What do we focus on? I was like, all right, well, what can we do? Like, we're going to focus on this one particular area, basically how we get pebbles out of the reactor. I mean, we're going to put two teams on two parallel tracks, and we're just going to have them work through cycles really quickly, and we're going to see how that, that process works. You must have seen one of those teams working through when you, when you came through. I think what was really crucial about that was, we're trying to, like, interpret what was happening from SpaceX with their iterations on, and through the rocket development Trying to incorporate learnings from agile development, which really has a lot of basis in software. When you're working with hardware systems, there's just going to be things that are different in terms of trying to do the iterative process. And so, like, those ear…

AI assessment note: “It took us a while to kind of figure out what this was going to look like”

Q So tell me a bit more about that. Is that challenging? Is that creating any, you know, challenges within the company, like competition for resources? Do you ever get pressure from investors who are kind of like, hey, what are you doing building all these things all at once? Let's do this more serially so you can incorporate the learnings. How do, how, have some of those tensions come up?

A Well, zoom out even further. We're, cause, and, and we're, we're building a whole lot more than that, too. So we're building a salt production facility in Albuquerque, and we're also building fuel production infrastructure. So vertical integration, in-house manufacturing is central to Kairos' strategy. It's really important both to the development strategy, but also to the commercial strategy. So from a business perspective, when we look at, like, bringing something in-house and doing it ourselves, It's not just that we're controlling costs and schedules when we do that. We're also capturing value. So our business model is the more we do, the more Of the revenue that Kairos captures from the project. But for a company that has manufacturing capabilities and production capabilities, this is actually a lesson in the iterative process that's different from, I think, maybe conventional best practice. In a perfect world, you can go through that life cycle of iteration, get all your experience, and have it cycle back through, and then build the next one, right? And maybe even in the extreme case of rockets, you know, Your entire, like, product life cycle is, like, 30 minutes, and so you can take that experience, and you can go and incorporate that back into, to the next cycle. The, the lifetime for reactors is decades, right? So you can't do that. Um, the other side of it is, when yo…

AI assessment note: “this is actually a lesson in the iterative process that's different from, I think, maybe conventional best practice.”

Q unit in Albuquerque, New Mexico. You're now building Hermes and Oak Ridge, along with your ETU-III. So what led you to decide to have these different locations, and how do you manage people across all of this and make that cross-pollination happen? Like, in my mind, I'm envisioning something like 1800 Slack channels for your company. Or some ungodly way of keeping in touch. So what does that look like?

A Um, yeah, there are, there are a lot of Slack channels. Um, I mean, I think that there's, in retrospect, there's a, there's a rationale about why everything is where it is, and it was deliberate. The sequencing of sites is, is really important, and each site serves different functions. Um, in the early days, there, there wasn't an obvious place where we could find the people and, and put Kairos and just do everything. There are capabilities that we can find in people in the Bay Area. There's great software development. We have, like, um, a range of different schools to pull from, but we're, we're fundamentally limited in the size of, like, how big we could do things here, um, and so we needed to find another site to do that. Albuquerque was the place to land. Uh, we were already familiar with it. There's an ecosystem with the labs there. Um, there is kind of, like, a nuclear development world, uh, in New Mexico, uh, but then it was kind of, like, Well, we want to build and expand and do more there, and the logical place to put the next thing is either next to the last thing you built or next to the thing that you plan to build in the future, and both of the Albuquerque sites and the Tennessee sites fit that bill perfectly, so Albuquerque has really Alameda has become the hub for kind of, kind of the rapid early, early development testing for new ideas, but also it's really the …

AI assessment note: “there's a rationale about why everything is where it is, and it was deliberate.”

Q what your engagements are like on the customer side? Are you planning to, to actually produce power with this first Hermes unit? Are you planning to sell that power? How are you bringing utilities into the conversation? Because I think this is one of the Other big hangups for the clean deck sector is to figure out, like, how do you actually bring in customers, particularly on these first-of-a-kind endeavors?

A Yeah, so Hermes, we worked through the plan to do Hermes in twenty-twenty. It was in conjunction with a DOE proposal, uh, that was kind of the seeds of the advanced reactor demonstration program. Our approach was, this is going to be our plan, and whether we get, we get the DOE word or not, like, this is our plan. We were selected for a DOE award for, for the Hermes reactor. The idea was, let's do like the minimum viable reactor. And so, take out the things that don't really need to be there, and just get to the point where we can definitely, hey, we can design, license, build, and operate And, and honest to God, nuclear reactors. And we've gotten huge amounts of value out of that. Even in the, like, the construction phases now, we give us, give us real data, real cost numbers, um, that are going to project forward how much our foundation is going to cost for future reactors. We have really good, good knowledge about what those are actually going to be based on the experience we have, uh, from, from even, even to this point in time. So fast forward, we got a construction permit from the Aggregatory Commission. First advanced reactor, you know, to submit and, and get, we did it in just over two years. Um, that was a huge accomplishment and, you know, again, something that people thought was going to be really, really hard and challenging, and we were able to, to navigate that pr…

AI assessment note: “The idea was, let's do like the minimum viable reactor.”

Q to build Hermes I at the same scale as your ETU I. It's not going to produce power, but Hermes II is. So talk a little bit about, well, I guess first of all, let's talk about Hermes II and the power production opportunity there. Is this power that you're going to be selling into the grid? How are you thinking about that in the context of your customer base?

A Yeah, so with, with Hermes I, we really wanted to, to be true to the idea of a minimum viable product, and product without electricity, maybe that doesn't quite match up perfectly, um, but we really needed to just prove that we could deliver a nuclear system. Uh, for us, really the demonstration was around the ability to deliver the nuclear heat, and, um, really just focusing on what are the essentials. The opportunity for Hermes II was Um, was, if we were going to build this reactor, well, why don't we go quickly for permits to build more of them? And because our architecture means that the things that are important for safety in the regulatory space is pretty narrow, we can actually change the design in a very basic way for the power generation side of the system, and the feedback loop into licensing is almost nothing. So, the technical review on that reactor is basically, for the things that you're adding to the system, does it Does it have a feedback loop that's going to impact the safety? And, and the NRC basically concluded, no, there's no feedback loop. So the safety case looks exactly the same. Um, the original decision to pursue Hermes II was, hey, we're going to be building one of these reactors. Like, let's seize on this moment and keep building. Again, build, build, build, and do more of them. Um, turns out we've, we've been learning a lot and, you know, gaining con…

AI assessment note: “it wasn't really quite viable as a full commercial product offering.”