2026-05-15-daily-debrief

Daily debrief

Recent Conversation Debrief

• Exceptional performance integrating research with AI system (“the rig”)

  • Separate reading preparation + real-time AI access during conversation
  • Demonstrated compounding gains of systematic knowledge capture

• New DRAM report showing AI system referencing previous conversations

  • System pulling from transcripts and detailed context
  • Building cross-conversation connections (e.g., referencing Nihar’s observations)

Strategic Direction Insights

• Steve Blank feedback: “pick one” - financialization vs supply chain are different businesses

• Two viable paths identified:

  • Direct financialization approach (not building on supply chain data)
  • Data center servicing business (disposal/repair chain focus)

• Financialization considerations:

  • Compute-level vs commodity-level trading (why people focus on compute vs semiconductors)
  • Market thickness differences between compute and semiconductor markets
  • Bullwhip effect in demand chain (AWS to Nvidia delays)

Memory Market Analysis

• DRAM boom/bust cycle: 18-24 month capex lag creates price volatility

• Current oligopoly behavior: 3 players (Samsung, SK Hynix, Micron) limiting capex to maintain margins

• Memory chip margins now higher than TSMC margins

• AI demand for HBM reducing DDR supply, driving up all memory prices

• Capacity information as “most tightly guarded secret” suggests managed market

Upcoming Meeting Prep

Josh (Shift Insurance) - Tomorrow:

• Focus: Data flywheel business lessons, why not scaling faster

• Entry point: “You built compelling data flywheel business - what does it take?”

• Key question: Structural vs execution reasons for 10 years to $80M revenue

• Insurance market constraints: slow buyers, cost reduction (not growth) product, limited customer base

Max (IMEC) - Monday:

• Approach: Humble, learning-focused, relationship building

• Opening: “Our objective is to build relationship and learn from your experience”

• Focus areas: Operational pain points across semiconductor portfolio companies

• Avoid technical R&D topics, focus on logistical/operational friction

Next Steps

• Bliss: Read assigned materials on flight, start using “the rig” for research instead of Claude projects

• Dustin: Provide reading assignment for Bliss

• Both: Launch cold outreach sprint next week, block day this summer for Bliss rig onboarding

• Follow-up: Re-engage Joe with more concrete approach after outreach sprint

Transcript

Them: Conversations, you know, and like this conversation we had with him today, we would not have been able to have that conversation. No. And I want to give you a lot of credit. I think you did an exceptional job of. I mean, like, I think today was like prime form bliss. And what I mean by that is, like, you did separate reading. You brought those readings into the discussions. You built an AI system to allow us to ingest the material. You then were able to on the spot, access that information, bring that information forth. Like, I thought that, like, that’s, that’s excellent. That to your point about how we could have had this conversation a little bit ago, we couldn’t have had this conversation a little bit ago for so many different reasons. And in that moment where you’re in the conversation synthesizing something that you read the night before and incorporated into AI and then were able to reference accordingly, like, that is really powerful. Yeah. And that’s like the compounding gains of the rig. Like, that’s why shit needs to be in the rig, because it just builds on itself. Exactly. I’ve noticed like this new report I did on dram, it’s like start. It’s starting to like, reference shit from like conversations we’ve been having. Yeah. Like, I’ll find this one. Like dram. Oh, the interface is sick. Like, what was the guy saying again? Right. No, no, Jason something. There was something about Jason in here. Like, it, like, you can see how it keeps saying, like. Yeah, yeah, well, like, like it’ll reference. Like Nihar’s observation is relevant. Like it’s getting. Is that the snipping process? Like, what’s allowing it to do that? Is it the snips? It. I think it’s a combination of the snips plus the fact that it saves like the detail. Deep roost for everything, which, which directly is. Pulls from the transcripts. Yeah. So no, that’s super cool. And I think at some point this summer we’re going to have to be like, we are blocking off a day. Yeah. For like. Like I need to start doing my research in the rig. Like, I, I’ve been using my, my Claude project a lot and that’s just not smart. Like, why. Like, we’re not getting the benefit here. No. And it’s not saving. Totally. Yeah. And I do a lot of one off querying, but I build off my queries. So, like, ours are our, our uses are gonna look very different. Right. For me, it’s like much more kind of like give and take. Give or take. Like, don’t. I Kind of like when it doesn’t give me too much information. Oh, I do, I do that sometimes as well. Like all aspects. Like here earlier today I was like, I was like, how do people perceive the dram exchange? And then I was like, you know, just doing like little things. Like this, Right? Yeah. No, that’s awesome, man. That’s super cool. Anyway, we don’t need to talk too much about knowledge, but like, I think that conversation was like. That conversation is exactly the way that I think these sorts of like knowledge sharing conversations should go where the point isn’t like trying to get a pilot or like trying to like do something. Like that conversation was like an archetype of like, this is someone who just knows a lot about the industry. Let’s just like pick their brain. And I thought that was a great example of that. Yeah, totally. And I really like it. Yeah. Like there’s some of those conversations you have and you’re like. And this guy’s like, I’m like, why do you say stealth? He’s like, makes people more interested in me. I liked. He just like, he had a lot of substance, but he also wasn’t like posturing himself. He kind of was just. He was so confident that he’s like, I have nothing to prove. I’m just gonna like sit there and be like, just like, ah, yeah, this is how it works. He wasn’t like trying to be like, oh, I’m a big. Whenever someone says I don’t know, it’s a sign of such confidence. And there were a few times like, yeah, I don’t really know. This is just congestion, conjecture. Yeah. But anyway, I think thematically coming out of that conversation, two things emerged to me. Yeah. As I do my own little DJing and connect the dots. When we had a conversation with Steve Blank a while ago and we were like, this is our roadmap. He’s like, pick one. Right. These are all different businesses. And what became clear to me is like, we could actually pursue the financialization or the supply chain. Yeah. Like they don’t actually necessarily build on the same foundation. I mean like, yeah, like, you’re right. Building the financialization on top of a data asset generated by the supply chain. It’s kind of a Rube Goldberg machine. It’s like, you’re doing this so you can get there. It’s like, why not just. Yeah. Go straight to the ball. Yes, yes. And the way we think about financialization would fundamentally change if we weren’t building an on top of society supply chain data. Yeah. Because like, are you familiar with energy markets at all? Not really. So I’m not much. But one of the things I was tasked with at Heinz when I was working for Doug, it was during the ESG craze and like Heinz came out and said we’d be net zero by whatever. Doug’s like, help me figure out that actually means. And he’s like, and I don’t want to buy carbon offsets because like that’s all. And I was like, well Doug, if we’re gonna be net zero, we have to become an energy person producer. Because like buildings consume energy. And we were like, well, on the one hand that’s ridiculous. On the other hand, like energy is not that different from real estate, right? It’s like very capital intensive. You have to find an off taker for energy, which is someone to buy your energy. You find a leaser for a lessor for real estate that someone occupy your space. Like there’s some parallels. So I did some diving then. And what’s very interesting to me is his whole thing was there’s a few different themes that emerged. One, financialization at the compute level instead of the actual commodity level. So it’s like, is the commodity, is the thing that you financialize? It’s like trading oil versus like trading energy. And like people trade oil in general. But it seems like there’s more conversation. Like Alex Lopoli is connected me to like a bunch of people who are financializing computer. And I think that’s very interesting. Like why, I think that’s a real question to understand. Like why are people focused on financializing compute? We don’t really trade. We don’t have a futures market for like gigawatts. We have a futures market for oil. Is it, is it because for compute, like the consumer, like it’s not like an oligopsony. It’s like this broad range of customers buying compute. Whereas like GPUs and like hardware, like it’s like a much more concentrated pool of buyers that are turning that into compute. And therefore you don’t need this like super transparent market. Like I, I think more. I think maybe. I think it’s more just like you’re not in Mohammed. Are you in any marketplace classes besides pimpikis? No. So like, which is a complete joke, but like the idea of like market thickness, like how thick is. Yeah, yeah, yeah. Like it’s very. It’s much more thick for compute than it is for semiconductors. So maybe. But either way, one thematic thing is like we could go straight for the financialization. Another thematic thing is like where do you financialize? At the compute level or at the inputs to the compute. And then the third thing is there was a lot of intrigue to me about the bullwhip effect in the demand chain. Yeah. Right. So it’s like how much delay is there when demand for compute kicks up? How do you move from Amazon from AWS to Nvidia? Yeah. And then you could also map that to like. And what was he saying? The time required to take a data center online. Right. Like yes. Which I don’t think is quite right. Right. Like there’s. Because even in a functioning data center, like there’s. I think that that’s like a whole statistical problem. I mean I was reading a lot about like the demand or the boom and bust cycle for memory, which is really interesting how the capex works. What will happen is there’ll be a run on memory prices. Firms will pour a lot of money into capex. 18 to 24 months later when these new firms like fabs come online, all of a sudden the supply goes up like this. Prices go down. Companies start to scale back their production and reduce the capex in new fabs. Scarcity happens, prices go up. But what’s happening now is like the firms are learning from this and there’s a price boom right now. And they are purposely like the fact that there are only three players in dram. Really it’s a oligopoly. They’re kind of like colluding not to increase their capex too much. And that’s why they think that there’s going to be like a raise in like the margins of memory chips. So I wonder. I’ve been learning all this tonight. It’s great, it’s super interesting. But it is similar to something I was thinking about when Minsuk said. When Minsuk said like the most tightly guarded secret is the capacity. Yeah. Like it does feel like this, this market is being managed totally. Which has kind of an oily feel like, like there is like cartel like behavior going on in the memory chips. Right now there’s three players and they’re all basically kind of colluding not to increase their capex so much. Like pump out more of these chips so they can keep this wave going. Do you know that memory chips right now have higher margin than tsmc? Is that true? Yeah, as of like the, like as of Q4 25, not AG$, right? Yeah, yeah, yeah. Percentage gross margin is now higher on I think Hynix than it is on tsmc. Yes, that is interesting. This is just my finance brain that like, knows how numbers can be manipulated. I am curious what the aggregate dollar margin is like, if it costs a penny and they make five. If it costs $0.01 and it makes $0.05. Yeah. Whereas TSMC, it costs 10 bucks and they make 20 bucks. Like that’s a 500% margin as opposed to a 200% margin. But like it’s. Yeah, it’s still astonishing for sure. But you know, there’s, there’s a missing. Yeah, I got some, I got some stats on the. For later. These type of stats on unit economics and margin data. Wait, wait, let’s. Let’s not dive yet because I want to get there. Yeah, we can talk about this all night, but. Yeah, but that’s also where like, you’re doing so much reading. I want you to assign me something. Yeah. I will print it out before my flight and like read it on the flight back from New York probably. Totally. But yeah, so I think that that was a big takeaway. That’s kind of like financialization side. And that’s very interesting what you’re learning about the memory market. Like, and I do think financialization could be a fascinating avenue for us to pursue. Yeah. And play into our strangest. Keep talking. The other thing that became very interesting to me is the two themes that are emerging is one, financialization. And two, it’s less like supply chain. Supply chain and more like servicing of data centers. Right. So it’s like, I thought a lot about like the disposal chain that you were talking about. Right. So it’s like, how does a chip go from. From like piece of silicon manufactured by TSMC goes to Foxconn goes to. I need to understand this better. Nvidia to Dell to Meta, it breaks, Nvidia gets it back, they refurbish it, it goes back in Meta, retires the chip, and then somehow like that old chip is ending up in this street in Shenzhen. Yeah. And it’s like, what does that whole thing look like? And then what does the actual servicing of that look like in terms of like, who is on the hook for what Warranty, how? Again, my financial brain is like a warranty is a financial product. So like financialization, how are they then picking up the chip? How are they getting it fixed? I gotta get your imessage so I can start hitting on this. But like, that’s the other big thing that I started picking up on. These things are breaking a bunch. Right. The anecdotal stuff he was sharing of like they break A lot. I’m like. And that’s a great thing to talk to Jonathan Heidegger about because he ran data center. What actually breaks? So that’s the other. Yeah, that’s the other big theme that. Yeah, I found a lot of the, like, operational details really interesting there. I just, like, I’m curious how you can build a business here that isn’t just like a software platform. It just like automates service because it just doesn’t really seem like there’s much of a moat there anymore. It’s so funny because my brain went the opposite direction. Really. My brain is like, I don’t see how you build a business that’s not a literal services business. Like, we are either driving trucks to pick things up or. That’s exactly what I mean. This is exactly what I meant. Like, my brain originally went in like, okay, how do you build software to integrate all this? Which is basically what Lonnie asked for the other day. And I kind of agree with Michelle and like, there’s no moat for that. Like, Nvidia could vibe code that immediately. When they realize it’s easy, they could hire Palantir to do it within days. Like, you know, whereas, like, if you. Palantir is going to charge you a lot of money. That’s the thing. If there’s any sort of growth for us, like, us being like, we’ll build it for free is very different than like, what is Palantir going to charge for that ridiculously high $15 million. Yeah, they’re not giving Nvidia free pilot. Like, so I do think. I don’t know. But anyway, keep going. Yeah, I just, I like, I agree with you. Like a services business. Like, yeah, like, like a 3 PL that specializes in just chips. Like, where I get scared is like, I, I’m like, if Amazon does this right, they are your business, you know, like, they’re like, dave, this is a solved problem for Amazon. Yeah, Reverse supply chain. But I do think there’s a lot of, you know, like, you hire a bunch of plumbers and like your plumbers come and fix the water coolant. Like, it’s a classic. Where my brain went when he started talking about that model was more like Peai roll up style play. Right? So it’s like you buy some services business, you implement AI, you start building some data models that help you just do better in that business and then you have a good margin. Yeah, exactly. But you’re not just providing an AI tool that you’re selling to the businesses. You’re using AI to enable existing hard businesses, which is a fundamentally different business. Right. That’s a private equity business. That’s like a. Dustin and Bliss are making money immediately. Yeah. And not, like, taking a bunch of venture money. And I don’t know, part of me is okay with that, but I’m, like, not opposed to that either. I just want to work on interesting problems and create impact. I’m not trying to have to bind myself to some Silicon Valley orthodoxy if I need to raise from this firm and do Y Combinator. Exactly. Exactly. So, yeah, that’s so. Yeah. I don’t know. I. Like, I. I just. I think how can we have more conversations like that, that just challenge us to keep, like, learning and pushing? Because, like. Oh, see, I think those are the easy ones. Exactly. That’s what I mean. But, like, I think they’re. I think they’re energizing and they get us exciting, which is, like, an underappreciated functionality again, an interesting way that our brain moves in different directions. I’m like, how do we have more Lonnies? Because that whole conversation was built on the back of Lonnie. Like, we brought a unique insight to the table that made, like, we. Like, we can have a way better version of that conversation with Jonathan. Yeah. But the more unique insights we can bring to the table for him to react to, the more interesting it’ll really be. That’s what I mean. And conversations like the one we had today are, like, useful mostly in their ability just to kind of keep leveling, mess up. Yeah. And then make the Lonnies more useful. Make the, like, real conversations better. Oh, I forgot to tell you, Avery got back to me super funny. She’s all excited to help out, but also just, like, doesn’t answer texts, which is a good thing to know. No one answers texts. Like, people are horrible at answering texts. She said. My old manager, Ryan Chen, who used to work at Global Foundry, said he’d be happy to talk. He was also at McKinsey doing similar stuff to me for longer. Let me know how it goes. Send his number. Amazing. So we got McKinsey and industry expertise. That’s good. Maybe he can get us to that guy McKinsey, like, Nick Santanam or whatever. The guy Joe told us he’d introduce us to and then never. We gotta attack Joe’s. Yeah, Joe’s been quiet. He hasn’t responded, like, our last few emails. But I think we gotta come back to Joe something, like, more concrete, you know? Like, I also think we just, like, Joe will get back to Us, we just would need to tighten up the iteration cycle or the follow up cycle. Yeah. Which I don’t know if we need to do right now. No, I don’t think he’s the most important one. No. He’s useful though. He’s smart, you know, and he knows a lot of these other people as well. I think he’s gonna be a really good guy. After our cold outreach. Sprint. Yeah, we just got it. We gotta launch those messages next week and they just have to get launched. Yeah, I think that’s the thing. Yeah. Those are the types of conversations that are going to be less like, we’re going to go in there treating them as connectors and more just going in there being like, what can we do for you? Like, what’s your problem? Exactly. Like, it’s more transactional, which I actually think is better. Because a lot of these conversations we have now, they’re kind of like relationships we want to like, kind of build as like sort of connectors and like long term, like the Max Augusts of the world and like the Michelle’s of the world and these sorts of people who like, yeah, we like in 30 minutes we want to get some insight from you, but it’s almost more useful. We’re just trying to touch base. So, you know, like, we’re on your radar. Right. Like Lonnie. Right. It’s like that call could have gone the way that like Vinay’s sister went yesterday where we talked and we’re like, ah, there’s nothing really for us here. But like there was something for us there and now we’re like, okay, we’re gonna go with you again and we’re gonna like attack that problem more. Like it’s like, it’s a little more transactional and it’s like a good thing. Totally. I totally agree with that. Even Jason, like GSB student. We’re both kind of working on things. None of us really know what we’re doing, you know, like. Yeah, as opposed to we’re gonna have these calls and it’s like, hi, you work at Nvidia in this role. I want to talk to you about what you do. Yeah. And they’re gonna get on the phone and we’re gonna ask them questions about what they do. Like there’s not, there’s not going to be like 10 minutes of small talk at the beginning. It’s going to be like, hi guys, how can I help you? Like, you know, totally. And those are the mode. That’s where I thrive. Like, that’s where I’m going to be like Prosecutor Bliss. Yeah. Cool. Okay, so we got tomorrow. Yeah, tomorrow I can start. I can tell you about the video with that guy. So he built a data flywheel business. They’re very AI pilled right now. I got cut off before I had a chance to read that last Slack thing, but I don’t think they’re doing that great. Like, I think it’s worth more than a billion dollars, but like it took them 12 years. They’re not growing that fast. They haven’t raised in a while. Like, I don’t think they’re ripping. It’s just kind of cruise control. Like, I think. I don’t really know, but like it doesn’t seem like they’re ripping. And so what I want to ask him is like, how much of the business that you’re in now did you set your sights on from the beginning versus how much did it evolve over time and what in terms of building a date? Like, to that end you built a data flywheel business. Was that your plan initially? Or did you just realize that this data was working and then as you were doing that, were you like, what are the things that got in the way of that being successful? What are the ways in which that was more or less valuable than you expected? I don’t know how to asking about being a dick, but like, basically, like, why are you not absolutely ripping? Like, you figured out a data flywheel business and insurance, it’s a massive market. Like, why are you not just that dude, right? Like, I’ll read you some snippets from this cloud output and there’s a lot, but like the very last thing that it says, which I’m, I probably would not do, is it worse worth asking Joish directly, if you had structured shift to be a rocket ship instead of a steady compounder, what would you have done differently? And he’s like, I bet Claude says, I bet the answer involves either government contracts or usage based revenue model. And his answer to that question will be worth the entire meeting. Like probably won’t ask that question. But yeah, you know, like it took them 10, 10 years to get to 80 million of revenue selling to insurers. Like, that’s fine. But yeah, it’s interesting. Like I’m actually gonna read this one. So what’s your like entry point to the conversation? Because I think this is kind of where we like. So let me, let me actually, let me actually read this to you because I think it’s interesting. I said, why are they struggling? Why are they not taking off like a rocket ship. Why SHIFT isn’t a rocket ship. The short answer Their customers big insurance companies are some of the slowest and most conservative buyers on earth. And the product SHIFT sells is a save us money tool, not a help us grow tool. Both of these things put a hard ceiling on how fast the company in the space can grow. This isn’t a SHIFT failure to structural feature of selling AI to insurance. The six reasons 1 customers move at glacial speed 2 they sell cost reduction product, not a growth product 3 there are only 50 customers that really matter and SHIFT already has most of them. 4 insurers have armies of their own data scientists. 5. Insurance is regulated low margin industry 6. The data network play is genuinely hard. What this means in plain English. SHIFT built a great business in a structurally slow market. It could be the best company in the world at what they do and still only grow 30% a year. The companies who become rocket ships sell to customers who can buy fast developers, tech companies, small SM. What’s SMBs? Small Medium Business. Small medium businesses who spend more as they grow. SHIFT sells to massive bureaucratic enterprises on multi year contracts for cost reduction product. So like and this is what it says, the brutal version of for you again, whatever. But like here’s what you need to internalize. The customers you’re targeting for Project TBD are even slower and more conservative than SHIFT company. SHIFT customers, defense primes, tier 1 semiconductor manufacturer, government agencies. These are some of the slowest enterprises on earth. If if shift took 10 years to get to 80 million in revenue the realistic comparable for project DBD is probably 12 to 15 years to get to a similar scale selling to Defense and semis two strategic levers government contracts that bypass normal enterprise sales Strategic genuinely usage based product. But like that is an interesting thing to note right? Like if you’re selling to these big slow industries, they’re going to be big and slow. So I think you were saying entry point. Was that your question? I think I do want to the conversation because I like a meta note on all of our interviews. I feel like we’re at our strongest once we kind of gain some momentum and once we’re in the line of questioning I feel like where we kind of usually are the weakest in meetings is getting into just leading off and moving entering into the substance right? Like in terms of the conversation we already we always have around like oh like we’re gonna ask you some questions and then like what are you working on? It’s always this kind of like weird dance like you know, maybe, like, it’s not actually that weird and we’re not actually that bad at it. I just. I just. I just find it’s always like this, like, funny dance. Yeah, well, I’ll think about that a little bit. My first instinct says it’s just harder. Like, people just warm up to each other. Some of that’s just structural. Yeah. Even if you ask your best question up front, like, you’re not as comfortable. Yeah. I think the reason why we’re not as great is, like, we don’t know what we’re asking for. Yeah. And the more targeted you can be, the more compelling you are and the more excited they tend to get because they don’t have to think as much. Like, they’re like, okay, targeted question, targeted answer, you know. Yeah. Whereas if you ask them to boil the ocean in a question, they’re going to sit there and be like, well, what do you want? Like, exactly. Exactly. Yeah. I just think the entry point is like, look, you built a data flywheel business that’s really compelling. Yeah. Would love to know what it takes to build that business, what it looks like to do that. Like, that’s. That’s a lot of inquiry that I would. Yeah. But it’s noteworthy that, like, that’s not a great business, you know? Yeah, totally. I wonder whether if you ask, like, Claude to do a similar analysis of coalition, what. What it would say and, like, how they’ve grown. Yeah, I’ll do that. Like, because that’s kind of the other. The other. The other comp. We have in terms of, like, the data exhaust insurance play, you know? Yep. Okay, cool. Either way, I feel like we’re, like, good enough there. Yeah, totally. What about our friend Max? Max. I think what we really want to zoom in on is the fact that, like, Max oversees partnerships for imac. He probably. He is involved probably in expand, which is their VC fund, as well as I start, which is their incubator. I would assume, based on his title, that he has some operational involvement or strategic oversight of those programs. And I think, like, we need to just really get at, like, you know, we’ve been trying to talk across different companies and identify pain points and operational workflows. One particular one we’ve had is the Lonnie thing, which, you know, we’ve seen a lot of manual work and lack of integration across different systems and the return and repairs lines. What are. What are some of the sort of stickiest operational pain points that you’re seeing across your, like, portfolio of partner companies and, like, startups that you’re working with. Let’s, let’s pick on. Yes, totally. Because I think, I think like the more we can steer it away from like R D, like technical things which were completely unequipped. So what I was gonna say, what I was gonna say is let’s, yes, totally. Let’s pick on the phrase operational pain points. It might be the right phrase but let’s talk about that for. Yeah. Is it the right phrase or is it like non technical pain points? Friction. Yeah, it’s like sort of logistical frictions. Operational frictions. This had, AI had a good one here. I think operational pain points is the right thing. So like the question they drafted is like, like you oversee partnerships with a thousand plus different semiconductor firm. What’s the operational problems? You see companies struggling with the most, blah, blah, blah. We keep hearing about operational workflows that are surprisingly manual. Things like reverse logistics and repair. Where are these, you know, where is like the manual friction in these operational workflows? Yeah, yeah, yeah. Right. I think like, but it said too narrow for some like him. Probably not. That’s still like a massive question. I mean I think we start with like we’d love to learn more about like what you spent your time on. Yeah, exactly. Right, totally. I think that sounds great. We want to get into like I also think we’d be super deferential. Yeah. Like a little verbose. We’re grateful for you taking. So grateful for your time. Like it’s amazing what imac is doing. We just want to look, we really just want to learn from you. Like I think that’s the app approach here. Like lean humble and try to make him feel like he’s teaching us and he’s excited to be a teacher and yeah. Mentor. I also think it’s really, I noticed you don’t do this but like I think you should share your Fulbright thing like you introduce. So be like I worked at Palantir. Like, yeah, be like. And then I studied in Greece on a Fulbright. Right. Like we’re talking to a European. Yeah, yeah, yeah. I, I, oh, what’d you do a Fulbright on EU regulation and things. We can, that can be a bit of a digression with some of the Europeans. But you know, I can mention that, I can mention that with him. The euros, like these sorts of. The euros. The euros like this stuff a little bit more. Yeah, I mean I think we just lean on like very grateful, very much trying to learn. Yeah. Here’s what we’ve been thinking about. But we really want to hear from you. Can you do me a favor? Can you pull up, can you pull up the email I sent to Max? Like what are we telling me what to talk to him about? I mean he sent us the whole thing about like a semiconductor, like a Ferrari is not a Ford or like some. That was an electric response. Okay, here we go. Our preliminary hypothesis. End users of semiconductors lack visibility into the risks and regulated workflows. We sent them the idea as of a month ago. Basically. What did we say we wanted to talk to you about? It was like to discuss the semiconductor supply chain. We’ve been working to study this space and identify opportunities for improvement. Considering your leadership at imec, it’d be a privilege to connect. Wasn’t there something else? And then you sent the whole thing with the digital twin. Go up for a second. He said, Consider your expertise. Value chains. Cool. Okay. And then he said, yeah, he, he writes sentences that are the length of paragraphs. It’s crazy. He’s like Tolstoy. He’s like, semiconductor is just like saying a car. There is the F1, that is a car. The Bugatti that is a car. There is a Mercedes that is a car. There is a Toyota that is a car. And I think what we’re interested in is things related to AI, which is dram. No, no. I mean DRAM is more ubiquitous than just AI. But yeah, it’s actually probably not directly related. I mean the interesting thing. No, I think it is though, because the interesting thing about memory is that there are limited number of wafers that are able to go into the memory market. And there’s two types of memory, there’s two types of dram. There’s like DDR, which is like commoditized regular dram. And then there’s hbm, which is high memory, high balance memory. And that’s what goes into basically like the Nvidia H200, all that stuff because they’re integrated with the GPUs as part of these like Nvidia GPUs. And what ends up happening basically is because of the fact that the demand for HBMs is going up and up and up. And like these, these memory companies are producing more and more HBMs. The supply of DDRs, commoditized memory is going down and prices that are going up there too. So there’s actually this effect where like the prices are going up across both segments because there’s like trade offs in terms of like where they can allocate the wafers. So this is a really interesting Dynamic how the AI demand is pushing on HBM and then it’s also dragging up the like commoditized map leading to a reduction in supply. Yeah. Have you ever thought this is just a realization? I’m saying have you ever thought about like economic stuff? Like, like is that line of thinking something, something that you’ve done before or is that like kind of a new thing? I’ve, I’ve always like dabbled in economics. I never like studied it formally but like I only ask that because like it’s like so like that’s what the way you at Palantir are like data. Yeah. Like every investment job is like demand here is going to lead to. And I just, I’m just realizing totally I got an H in my exact accelerated micro class. You understand it. I’m good at math. Like you know like a real math for it for sure. But anyway so. But yeah, like it’s a very interesting dynamic. Yeah but, but that’s what I’m saying. The AI thing is pushing on non AI related inputs as well because of this dynamic. Totally. So anyway, like I think, I think I agree with your thing. Humble go in asking questions. How do we handle the like what are you guys working on thing. I think you’re very good at this. I’m like less good at that. Quite, quite honestly in these meetings. Like I think you’re much more graceful around like being like we’re not trying to hold everything secret, we’re just really early. Like you just do a much better job at like that whole spiel and I do, yeah. I’m happy to jump in with that. So like if you want to jump in with, with, with on that particular wine, I’d appreciate it just because this is a higher stakes. Sure. Conversation. What I would say is the tactic I’m going to use for this. Try to encourage you to do the same. I think what we do is like we get on the call and we be explicit. Like our ask is. I wouldn’t say quite in these words but you could literally say in these words is like our, our objective for this call is for you to be someone that we can continue to learn from. Yeah, exactly. Like that’s our goal. Just to be very clear and like yes. We’re trying to understand these things. We’re studying this space. We think it’s really interesting space. We want to figure out a way we contribute. We want to figure out a way that we can contribute that creates both practical value to the ecosystem but also economic value to us. But our objective for this call with you is to build a relationship and continue to learn from you because you have such an incredible depth of experience and perspective in this space. Yeah, right. Yeah. Like, if he could be on, like, our board. That’s what. That’s what Jillian is saying. Yeah, exactly. Like, we were really looking to learn from you. We’re looking to. For. We don’t come from this space. We recognize we’re outsiders. Yeah. We do have a lot of friends that come from the space. But, like, we really want to start learning from people that we trust and building relationships. Like, that’s another thing that’s really compelling to me. Yeah. If you get someone who has the right vibe, like, we really care about building relationships with people that we can learn from. Yeah. Like, if you just say exactly what you actually want, it’s really powerful. Awesome. Yeah. If you. If you want to open up by saying that in those words, that would be awesome. Yeah, I think you open it. Cool. I’ll time those at the right time. Yeah. But I also lend you that. I’ll be there for sure. But I lend you that also as a tool that you can totally. Yeah, I’ll. Yeah. I mean, I’m sure there’s going to be a little bit of the usual small talk at the beginning, and then he’ll be like, okay, what can I do for you guys? I’m like, I think that’s a great way to open it up. And I can just. First question is we can just throw. Say something like, we’d love to hear about what you’re working on these days and what you spend most of your time on, and then just take it from there. And then we can weave in more of this stuff around, like, oh, okay. Yeah. And other. By the way, if we want to ask that in a little more spicy way, I mean, all of this also is just, like, tools that you can take. Like. No, I think. What’s the most exciting thing about it? Yeah. What’s the most concerning thing about, like, those kind of superlatives? Superlatives are a very good way to get to engage in a way that’s like. Like, a little sharper than, like, what are you working on? Totally. Yeah. Like, you obviously have a massive portfolio, but, like, what is the thing that keeps you up at night? That’s, like, a much sharper way to ask the same question, which is like, what do you do? Yeah, totally. Great. I like that a lot. I like the approach you spilled out for this meeting. And then I’ve done some more, like, again, specific Prep on like imac and what they do as well. But what do they do? Just research. Research. They’re kind of seen as like a neutral ground between the major semiconductor firms for them to like basically do research that’s like pre competitive stuff such as foundational stuff in the semis industry and kind of serve as like a commons between like the different firms. How valuable is iMac2 potentially? What’s the best case scenario here? They’re valuable mostly in the sense of they have relationships with all the firms. Could be like a big. Some of the best case scenario versus like signal. Signal, signal. As well as like entry points and introductions to people in the. In. In the major firms. Okay. Okay, got it. But they’re not a customer. No, probably not. I mean, I don’t know. They probably have some logistics around like how do they get all the equipment into the labs and what whatnot. But like I consider that probably rather. Rather bespoke rather than, you know, they’re going to be most akin to like a Stanford type customer. Yeah. Or like, like a Los Alamos research laboratory, you know. Yeah. Which is an interesting book. Yeah. Interesting market, right? Yeah. We went after a bunch of like. Yeah, definitely white space. Totally. So I’ll open up with. With that and kind of some of the tactics you mentioned. My ask is if he does the like, what are you working on? If you can just kind of disarm that in the way that you’ve gotten like really good at for sure. And be like. Dustin, you want to talk about that? Yeah, totally. Yeah. Cool. Just like I. My goal is to like learn and like just get better at that question. I just think, think you’re. I totally get it. Thank you for being clear on that. Cool. Ready? Break. Let’s do it. Excited? I got a.