# Claude Code for Finance + The Global Memory Shortage: Doug O'Laughlin, SemiAnalysis Latent Space · 2026-02-24 Doug O'Laughlin, SemiAnalysis founder, explains his Claude Code awakening in December 2024: Anthropic's Opus 4.5 one-shotted tasks that used to take 24 hours, leading him to predict Claude Code will write 25–50% of all GitHub code by year-end (already 4% of commits in two weeks). He argues this marks the death of Excel and Bloomberg for analysts—'you can just do things'—but warns of a 'hygiene' crisis as junior analysts lose meta-learning. On the semiconductor side, O'Laughlin details a severe memory squeeze: HBM demand consumes 4× the DRAM capacity, DRAM prices could double again, and CXL is reviving as a workaround. He compares AI infrastructure buildout to the railroad boom (25% of U.S. gross fixed capital investment) and says Microsoft faces an innovator's dilemma—renting GPUs to 'barbarians at the gate' while its own office franchise is disrupted by Claude Code. Google's TPU v7 enjoys a temporary TCO advantage, but Nvidia's supply chain dominance (Jensen doing shots with SK and Samsung) will reassert with Rubin. O'Laughlin also reflects on his 2,800-mile Continental Divide Trail hike as essential self-mastery. ## Questions this episode answers ### What percentage of code on GitHub does Doug O'Laughlin predict Claude Code will write by the end of the year? Doug observed Claude Code's signed commits surging from 4% to 5% of GitHub in just weeks. He predicts it will reach between 25% and 50% of all code by year-end, with 25% as a sandbagged, 95% confidence lower bound, calling it the fastest exponential trend he's ever seen. [12:46](https://addtry.com/7f836530-b209-4616-a551-a0bb519ad458?t=766000) ### Why does Doug O'Laughlin believe a global memory shortage is imminent for AI hardware? Doug explains that HBM memory, critical for AI, has a 3:1 to 4:1 trade-off ratio with conventional DRAM, consuming massive capacity. The previous memory market downturn left manufacturers without investment in new fabs for years, and now surging AI demand has created a supply squeeze that could push DRAM prices up 100% and force demand destruction. [1:41:53](https://addtry.com/7f836530-b209-4616-a551-a0bb519ad458?t=6113000) ### How does Doug O'Laughlin see tools like Claude Code replacing traditional analyst IDEs like Excel and Bloomberg? Doug argues that Excel and Bloomberg are human IDEs for information work, but Claude Code's ability to handle data analysis, charting, and research with direct coding makes them obsolete. He says he will never make a chart in Excel again, and believes Microsoft has the most to lose because its software suite is the horizontal human IDE target. [1:12:20](https://addtry.com/7f836530-b209-4616-a551-a0bb519ad458?t=4340000) ## Key moments - **[0:00] Intro** - [0:00] "This crap makes mistakes all the time. It is still just like a junior analyst" – Doug O'Laughlin on current LLMs. - [3:40] Doug O'Laughlin fell in love with ASML in 2018 and dove into semiconductors, finding it all science fiction. - [6:44] Doug O'Laughlin's thesis that Moore's Law was ending led him to predict Nvidia's dominance. - [9:03] Doug O'Laughlin's 2020 essay 'ChatGPT 3 and the writing on the wall' predicted scaling laws and Nvidia's benefit. - [9:47] Doug O'Laughlin met Dylan for SemiAnalysis because they were both early on the semiconductor thesis. - **[12:31] Claude Code** - [12:31] Doug O'Laughlin's Claude Code awakening came on December 27th when Opus 4.5 started one-shotting MVPs. - [16:38] Claude Code 4.5's ability to one-shot MVPs made Doug O'Laughlin feel AGI-pilled. - [23:35] Doug O'Laughlin scraped GitHub commits to find Claude Code grew to 4% of GitHub in weeks, an exponential trend. - [26:19] "You can just do things" – Doug O'Laughlin's catchphrase on Claude Code's capabilities. - [26:53] Doug O'Laughlin: Excel is a huge abstraction; if coding is automated, Excel will be automated too. - [28:44] Context rot in Claude Code is likened to the 'Of Mice and Men' meme, requiring careful hygiene. - **[46:53] Augmenting Experts** - [47:28] "This whole thing is a game of hygiene now" – Doug O'Laughlin on managing AI-generated slop. - [48:27] Meta-level learning is not yet present in LLMs; expert humans are needed to guide AI output. - [50:35] Entry-level data analysts will be 'murdered' by well-thought-out agentic systems, predicts Doug O'Laughlin. - [51:46] Most analysts use AI daily, but full Claude Code adoption as a base level will take 24 months. - [53:43] Doug O'Laughlin's Claude Code moment felt like GPT-3.5/4, a tool he'd 'pry from his dead cold hands'. - [54:22] With Claude Code, you get multiple turns at the wheel; the human becomes the reviewer, not the doer. - **[59:41] AI Economics** - [1:00:09] GDPVal benchmark shows models consistently outperform industry experts at white-collar tasks. - [1:01:46] Just as farming went from 90% to <1% of workforce, AI will massively reshape all work, says Doug O'Laughlin. - [1:03:25] Doug O'Laughlin's crackpot theory: AI may be massively deflationary, challenging how GDP is measured. - [1:04:58] Railroad build-out saw multiple boom-bust cycles and 45 years of CapEx; AI might follow a similar pattern. - [1:08:01] Q: Where will the money for AI CapEx come from? Doug O'Laughlin worries about funding such massive investment. - [1:09:06] Doug O'Laughlin values Claude Code at $20-30K per year, equivalent to a perfectly compliant junior analyst. - **[1:11:32] Death of IDE** - [1:11:32] All IDEs — Excel, Bloomberg, coding IDEs — are dead; agentic interfaces will replace them. - [1:13:17] Microsoft has the most to lose from AI because Excel, PowerPoint, and email are human IDEs. - [1:14:55] Doug O'Laughlin is moving from Bloomberg to FactSet API with Claude Code, saving $10-20K. - [1:15:49] Claude Code will generate 25% of GitHub code by end of year, predicts Doug O'Laughlin, likely between 25-50%. - [1:16:43] Codex 5.3 is 'coding-pilled' and struggles with general information work vs. Claude 4.6. - [1:18:52] OpenAI's next pre-train RL model could flip the coding agent race back in their favor. - **[1:36:46] Memory Crunch** - [1:37:26] InferenceX will eventually benchmark TPUs, showing their price-performance advantage. - [1:38:21] Jensen Huang personally does shots with supply chain partners to secure Nvidia's chip supply. - [1:39:13] HBM4 vs HBM3 is why TPU v8 won't be as competitive as Nvidia's Rubin architecture. - [1:40:06] Nvidia owns the entire supply chain, from HBM to connectors, ensuring first access to new technologies. - [1:41:53] Memory Mania: HBM trade ratio of 1:4 to DRAM creates a supply crisis; DRAM prices could double. - [1:43:44] DRAM prices may rise 100% again; hyperscalers will destroy demand by ordering less when power constraints hit. - **[1:57:43] Wrap-Up** - [1:57:43] Claude Code increases software creation, and RL gyms boost CPU demand, causing a possible CPU shortage. - [1:58:13] Doug O'Laughlin's writing process: read extensively, then one-shot ideas in the morning with a fresh context window. - [2:00:18] Doug O'Laughlin recommends 'On Writing Well' book; he turned it into a Claude skill for his writing. - [2:03:12] Doug O'Laughlin hiked the Continental Divide Trail for six months, a journey of self-mastery. - [2:06:25] "Self-mastery is your most important tool use of all" – Doug O'Laughlin. ## Speakers - **Swyx** (host) - **Doug O'Laughlin** (guest) ## Topics Coding Agents, Hardware ## Mentioned ASML (company), Anthropic (company), Google (company), Micron (company), Microsoft (company), NVIDIA (company), Oracle (company), SK Hynix (company), Samsung (company), SemiAnalysis (company), TSMC (company), Azure (product), Bloomberg (product), Claude Bot (product), Claude Code (product), Codex (product), Excel (product), GitHub Copilot (product), Kimi (product), TPU (product) ## Transcript ### Intro **Doug O'Laughlin** [0:00] This crap makes mistakes all the time. All the time. It is still just like a... Like, I think of it, once again, as like a junior analyst, right? The analyst goes and does all this, like, really pain in the ass information, and you bring it all together to make a good decision at the top. Historically, what happens is that junior analyst, who I once was, went and gathered all that information, and after doing this enough times, there's a meta level of thinking that's happening, where it's like, "Okay, here is what I really understand and how this type of analysis I'm an expert in, actually. I'm very good at. I consistently have a hit rate. Now I'm the expert," right? I don't think that meta level learning is there yet. Um, we'll see if LLMs do it, right? Everyone who's spending $1 quadrillion in the world thinks it will. It, it better, it better happen- ... right? If you're spending, you know, a trillion dollars, and there's not meta level learning. But for me, in our firm, that massively amplifies everyone who is an expert. 'Cause, like, you have to still do something, that you can't just, like, slop it up. It's very obvious to me when it's slop. **Swyx** [1:00] Doug O'Laughlin, welcome to LeanSpace. **Doug O'Laughlin** [1:02] Yeah, thank you for having me. **Swyx** [1:03] Yeah. **Doug O'Laughlin** [1:04] Um, I, after all this time, I just... Is it okay if I just call you Swyx? I feel- **Swyx** [1:08] Yeah, of course **Doug O'Laughlin** [1:08] ... yeah, it's the... That's, that's where my brain is, 'cause I've known you for so long. You can call me Mule if you want. I'm not care. Um, you know, yeah. Yeah, I mean, it's been, it's been a long time. **Swyx** [1:17] It's been a, it's been a long time coming. I think I first met you at, like, New Orleans or, like, one of the, one of the NeurIPSes. **Doug O'Laughlin** [1:24] Yeah, yeah. I met you at one of the NeurIPSes in per- I think it was Vancouver. Right? **Swyx** [1:28] Was that? **Doug O'Laughlin** [1:28] Yeah, no, no. I think it was, like, some after party. **Swyx** [1:30] It was, like, some after party. **Doug O'Laughlin** [1:30] Yeah, yeah, yeah. **Swyx** [1:31] And you were like, "Hey," like, "Who's this tall dude?" **Doug O'Laughlin** [1:33] Yeah. **Swyx** [1:33] I'm like, "Whoa, okay." **Doug O'Laughlin** [1:34] Yeah. Yeah, well, I mean, it's just like I, I knew about you, and we, we've, like, been internet, you know, pen pals for a long time, so it was, like, cool meeting in person. **Swyx** [1:42] Yeah. **Doug O'Laughlin** [1:43] Yeah, I think that was the first time I ever met you in person, so yeah. **Swyx** [1:45] Amazing. **Doug O'Laughlin** [1:46] I d- I didn't go to the New Orleans one. I really wish I did. I love New Orleans, honestly. **Swyx** [1:49] Yeah. **Doug O'Laughlin** [1:49] Yeah, so. **Swyx** [1:50] Go to New Orleans just in a row, and, um, yeah, honestly, we should go back there. **Doug O'Laughlin** [1:54] Yeah. **Swyx** [1:55] Um, are you guys going to Melbourne or- **Doug O'Laughlin** [1:57] I- **Swyx** [1:57] ... like, the Australia one this year? **Doug O'Laughlin** [1:59] I, um, have... I don't even think that far out. **Swyx** [2:01] Yeah. **Doug O'Laughlin** [2:01] But on a, uh... But that sounds pretty interesting to me. I think, um... I can't remember which one. There's a, there's something in, in, in Korea this year, right? **Swyx** [2:09] Yeah, I think, um, ICML. **Doug O'Laughlin** [2:10] ICML? **Swyx** [2:11] Yeah. **Doug O'Laughlin** [2:11] I think I'm gonna try to go to ICML, uh, in Korea. **Swyx** [2:14] Yeah. **Doug O'Laughlin** [2:14] And then I know iClear is... I don't know, man. There's so many conferences. I honestly, I hate to say it, I'm not much of a travel guy. **Swyx** [2:21] Well, yeah. I mean, I'm, I'm glad to catch you. Um, I mean, I, I, I am traveling to you. **Doug O'Laughlin** [2:25] Yeah. Thank you. I really appreciate it. **Swyx** [2:26] Yeah, absolutely. Yeah, that's fun. **Doug O'Laughlin** [2:27] Yeah, yeah. **Swyx** [2:28] Uh, I did not know that I would be caught in a snowstorm. **Doug O'Laughlin** [2:30] Yeah. Uh, it's, it's funny. I feel like people recently have been coming, and they keep getting stuck in these snowstorms. So yeah, uh, first blizzard in four years or something like that. Thank you for coming. **Swyx** [2:40] Yeah, yeah. It's a pleasure. Um, and so you and I go back. You, you used to be anonymous. You used to be Value Mule- **Doug O'Laughlin** [2:45] Yeah **Swyx** [2:46] ... which is how I know you. **Doug O'Laughlin** [2:46] You know what's funny is that Value Mule is, like, the very first one. **Swyx** [2:49] That's the be- **Doug O'Laughlin** [2:50] I- **Swyx** [2:50] Yeah. The... I don't know how, I don't know how I noticed you. I, I was just like, "Oh, this guy seems smart." **Doug O'Laughlin** [2:53] Yeah, I, I don't know, dude. I mean, I- **Swyx** [2:55] Yeah **Doug O'Laughlin** [2:55] ... I remember noticing you, too. So it's like, you know- **Swyx** [2:57] Yeah **Doug O'Laughlin** [2:57] ... this was in the early, like, the primordial days of Twitter. **Swyx** [3:00] Yeah. **Doug O'Laughlin** [3:00] Um, honestly, I miss those the most. **Swyx** [3:01] Yeah. **Doug O'Laughlin** [3:01] I think it was like 2017, '18, something like that. **Swyx** [3:04] Yeah. **Doug O'Laughlin** [3:04] But yeah, yeah, I remember. From Value Mule, so if you... That's, like, the deepest cut. If you are even aware of what that is, that is, like, the deepest cut that you possibly have. Um, and then yeah, I have another account, and I actually have a third account, which is my, my main account these days. **Swyx** [3:17] Yeah. **Doug O'Laughlin** [3:17] Um, so- **Swyx** [3:18] Wait. Oh, which one is it? **Doug O'Laughlin** [3:19] That, I don't want to- **Swyx** [3:19] Which one? Okay. I don't wanna dox your other account. **Doug O'Laughlin** [3:21] Oh, it's okay. It- **Swyx** [3:22] You can semi, semi-dox it. **Doug O'Laughlin** [3:23] Yeah, yeah, yeah. **Swyx** [3:24] So yeah. **Doug O'Laughlin** [3:24] So, so, so it is there. That's- It's, it's not... That's okay. That's like my oldest finance account. I think of it as my legacy account. **Swyx** [3:30] Okay. **Doug O'Laughlin** [3:30] Um, I, I, you know, I wanna have some privacy, I feel like, uh... **Swyx** [3:34] Yeah, yeah. So, so, so now you've gone all in on the brand and everything. **Doug O'Laughlin** [3:36] Yeah, yeah, yeah. **Swyx** [3:37] Okay. **Doug O'Laughlin** [3:37] Got the brand and everything, yeah. **Swyx** [3:38] Cool, cool, cool. **Doug O'Laughlin** [3:38] Um, same profile pic, you know, so. **Swyx** [3:40] Yeah. So, um, le- let's, let's do a little bit of the Doug story, 'cause a lot of people hear about Dylan, uh, and I wanted to just make this the Doug story, make, make the Fat Knowledge story. You used to be a value investor. That's kind of how you, you were Value Mule. **Doug O'Laughlin** [3:53] Mm-hmm. **Swyx** [3:53] And you had a mentor or something that nerd sniped you into semis. Is that the, the story? **Doug O'Laughlin** [3:58] Um, no. Actually, I solo nerd sniped myself. **Swyx** [4:01] Solo. Solo. **Doug O'Laughlin** [4:01] Um, so I, I wouldn't say value because, uh... Well, well, for, for everyone who's listening to this podcast, might as well be value, right? Maybe quality focus back in the day, but we had this whole thing where we wanted to buy quality compounder companies, and, um, the one I found that nerd sniped me, all the... like, single shot me, is I found ASML. **Swyx** [4:19] Yeah. **Doug O'Laughlin** [4:19] And then I, like, fell in love with it, and then I, like, after ASML, I just, like, read about all this stuff, how complicated it is to make these, who are the people who are able to make them. And then I, you know, semicapters, the whole downstream's all from there. But it started with ASML in 2018. I really fell, fell in love with it, and then I read, like, textbooks, and I just, like, kept going deeper. And my favorite part about doing that- **Swyx** [4:38] I was gonna pull out the Asianometry video. **Doug O'Laughlin** [4:39] Yeah, yeah, yeah. That's perfect. That's a perfect one. John. **Swyx** [4:42] Fucking amazing. **Doug O'Laughlin** [4:43] John's a monster, honestly. **Swyx** [4:44] This one, right? Uh- **Doug O'Laughlin** [4:45] Yeah, yeah. **Swyx** [4:46] Yeah. **Doug O'Laughlin** [4:46] Um, I mean, the thing that's crazy is he has, um... I don't know. He has a whole playlist about it, every single aspect of what goes into it. And what, what's truly great about it, it's all science fiction. Like, that's my favorite thing, is, like, science fiction exists other than, you know, the talking perfectly intelligent robot, whatever information LLM. Yeah, ASML is all science fiction. So the semicapter stuff's always been science fiction. Always loved it, always thought it was cool, thought it was the most important thing that we ever made, and, uh, yeah, kind of followed from that. Um- **Swyx** [5:13] Yeah. I don't know if you know, but obviously, you know I used to be a inve- **Doug O'Laughlin** [5:17] Yep. Yeah **Swyx** [5:17] ... uh, an analyst myself. **Doug O'Laughlin** [5:18] Yeah, I didn't know. **Swyx** [5:18] I covered TMT- **Doug O'Laughlin** [5:20] Mm-hmm **Swyx** [5:20] ... which is a freaking huge sector to cover. It is absurdly huge. **Doug O'Laughlin** [5:25] Yes, very large. **Swyx** [5:25] It, like- **Doug O'Laughlin** [5:26] I was- **Swyx** [5:26] ... I was covering Sprint- **Doug O'Laughlin** [5:27] Yeah, yeah **Swyx** [5:28] ... and got, like, uh, you know, Viacom. **Doug O'Laughlin** [5:30] Yep. **Swyx** [5:31] And then there's ASML. And- **Doug O'Laughlin** [5:32] Yeah, yeah. Yeah, the, the M I feel like, uh, the T and the M and the T are actually three completely separate industries. Um, but once upon a time, I think in 2000, they were kind of really close together. **Swyx** [5:43] Right. Yeah, yeah. **Doug O'Laughlin** [5:43] Uh, but, but ever since then, it's really split off, yeah. **Swyx** [5:46] Yeah. Well, I mean, I... It exists, my reflection is, like, I used to be in, I, I used to be, I guess, our Tech sector guy, and, like, I did the flights to Taiwan- **Doug O'Laughlin** [5:56] Mm **Swyx** [5:56] ... and I took those meetings with, like, Credit Suisse and all those, all those guys that would- **Doug O'Laughlin** [6:00] Mm-hmm **Swyx** [6:00] ... you know, tour you around and all those. I never really felt I got it, uh, because I was always being fil- filtered through, like, investor relations and all that. **Doug O'Laughlin** [6:07] Mm-hmm. **Swyx** [6:08] And, um, I think you have to do what you did, where you sort of go muck mode into, like, textbooks and stuff, and, like, actually learn about the, the tech. But then you, you hard- it's, like, really hard as an investor to, like, make the connection to, okay, well, what does that mean for this, this quarter? Like, or at least this, this year- **Doug O'Laughlin** [6:25] Yeah **Swyx** [6:25] ... right? Because, like, one, there's, like, just so much foundational knowledge. **Doug O'Laughlin** [6:29] Yes. **Swyx** [6:30] And then, and then, then you're like, "Well, okay, everything here is taken for granted. It's already priced in." So, like- **Doug O'Laughlin** [6:34] Yeah You, you, you assume that all the Taiwanese people who are buying and selling the rumors of capacity are pretty well-informed. You assume all the people who are TMT investors in the United States are pretty well-informed. I think the thing that was, like, the foundational difference for me is, like, you know, real thesis around, um, one I think being young and brash and believing in yourself to be like, "No, this is something that's really matters, and everyone else doesn't see it," really helps. But for me, the thing that was, like, I guess, radicalizing was, um, I really believed Moore's Law was dead. Um, and I was like, "Oh my God, not only is it this cool new technology that's super hard to make and very interesting and technologically very fun to understand, and, and, like, I get it intuitively, but, um, also everything... all the old playbook is about to be thrown out." Because it's been like this is super mature industry. You read these primers about it. Like, that's how you learned about things back before ChatGPT knew everything- ... or, um, you had to go and read these primers of all this information. They're like, "Oh, it's a very mature industry. It matured. It used to be really immature in the '80s and '90s and 2000s, but now you're consolidated. Growth doesn't go up a lot." And everyone kinda had this old playbook from the early 2000s. A lot of people hated hardware. Um, there was just this perception that semiconductors weren't valuable, weren't as valuable. Actually, software was the most valuable thing. Now software is getting shit on, but that's, like, outside of the scope of this. But, um, people just thought it was this old, mature business that had nothing new under the sun. Meanwhile, every single day, just making a new chip was like science fiction. Um, people took that for granted. And when the science fiction ends because you can't make the chips as small as you c- you could, all of a sudden, all those free gains you got go away, and you have to think about it. And what happened for semiconductors specifically is it created a lot of pricing power or value for everyone who knew how to make a good chip. So Nvidia's probably the best case. You could talk about parallel computing and all that stuff. But it's not just... Like, they know every aspect of it, from the chip to the networking, to the design, to the scale-up, the whole thing. It, it is, like, you know, versus it w- in the past, it was just CPU gets better, go brr, right? **Swyx** [8:35] Mm-hmm. **Doug O'Laughlin** [8:36] And so I think that I had a really deep belief that- **Swyx** [8:38] In this case **Doug O'Laughlin** [8:39] ... that, uh, M- Moore's Law was, uh- Moore's Law was ending, and everything would change. And so coming in with that, like, thesis at the top level just, like, made me wanna attack every little assumption. And something that really changed as well... I, and dude, this is honestly my, my favorite post I've ever written. It's like 20- uh, it's like a ChatGPT 3 and the writing on the wall. In, like, 2020, you know, my early- I get a early pitch to, um, for fabricated knowledge. And I'm like, "Hey, you know, I'm gonna make a really s- you know, Moore's Law's over. Scaling laws seem like a big deal." If you simplify it all the way through, it's like, okay, supply, uh, you know, supply divide, uh, you know, demand, right? **Swyx** [9:15] Yeah. **Doug O'Laughlin** [9:15] Demand is growing a lot because of scaling laws. Uh, supply is actually slowing down because Moore's Law is completely screwed. Um, that's probably really good for semiconductors. And parallel compute is gonna be a big deal, blah, blah, blah, blah, blah. Um, my conclusion then was, um, you should just, like, Nvidia's pretty much the only one who's gonna benefit. Um, and, and you know, so that's my m- my, my, my, like, good long range prediction, I feel like. I- **Swyx** [9:35] Just like- **Doug O'Laughlin** [9:36] Yeah, I just don't think- I don't think I would've expected the magnitude. I think that that's been the, kind of the craziest part about this whole story is, like, I had all these beliefs and thesis, and, like, I really, really, really believed. The reason why I met Dylan is he was the only person who was as semiconductor-pilled in the entire world as me, is how I felt. Um, so I remember, like, yelling at him, arguing about all these kinds of things in, like, our, our DMs and stuff like that. **Swyx** [9:59] Was it just online, or did you also- **Doug O'Laughlin** [10:00] It was onl- on- online. We met in person- **Swyx** [10:02] In Taiwan and- **Doug O'Laughlin** [10:02] No, no, no. And I've, I've actually only been to Taiwan with him one time, I think. **Swyx** [10:07] Yeah. **Doug O'Laughlin** [10:07] Um, so, so, I, I mean, like, look, we, we just met in person. We yapped. We went to conferences. Um, but I think that that's, like, kinda we were both really early to the thesis, kinda have a different background and perspective. Uh, Dylan is technology first, and you know, obviously technology matters. I have a little bit more of a financial background. But always around him, and it was just like, you know, he's the only wa- guy who, like, cared the, to the same level. Um, so yeah, this, the, the thing that's crazy is, like, we called it. We were right, blah, blah, blah, blah, blah. But, like, the thing that I think that still shocks me all the time is the magnitude of how right we are, you know? Like, we'd be like, "Oh, Nvidia was good," right? Nvidia's pretty good. Um, and then it's like, no, Nvidia's now the most valuable company in the world. And I think if you had me read that and, like, truly, "Hey, I wrote that. I believed it"- **Swyx** [10:50] Yeah **Doug O'Laughlin** [10:50] ... um, I still wouldn't have put that together, or, like, I wouldn't have believed it if you- **Swyx** [10:55] This was one of many thesis at the time. **Doug O'Laughlin** [10:56] Exactly. Yeah, yeah. **Swyx** [10:58] Right? **Doug O'Laughlin** [10:58] Like, there's so many things- **Swyx** [10:58] Like, what else are you writing at the time, right, that didn't work out, you know? **Doug O'Laughlin** [11:01] Um, yeah, yeah. **Swyx** [11:01] Yeah, we can look, we can look back, but- **Doug O'Laughlin** [11:03] I, I, I, I'm pretty happy with my m- with my long-term track record. **Swyx** [11:06] Yeah. **Doug O'Laughlin** [11:06] I really am. Um, but yeah, I'm just really surprised the magnitude of how everything happened. Like, it's crazy to me that, like, CoWoS is, like, a, a, not a household term, but, like, we- relatively well-known. It was, like, an exotic technology. **Swyx** [11:18] Mm-hmm. **Doug O'Laughlin** [11:19] So all this stuff has been this, like, learning journey, really believing where technology is going, uh, why chips are so important, and then obviously understanding the big scheme of all the things putting it together. And so that's the... Yeah, that was, like, the early days, and it's, I think it's all been downstream of that, like, you know, one goated insight pretty much. **Swyx** [11:37] Yeah. I mean, and, uh, uh, probably, like, a career maker right there, you know? Like, and, and I just, like, I love those guys, like, like, sort of quarterbacking those career decisions for other people who are also weighing a bunch of things, right? Like, I have ADD, and, like, I, I just chase, like, whatever's interesting. But at some point you just have to, like- We use twos. **Doug O'Laughlin** [11:55] Yeah. I, I think one of my skills have always been, like, trend following and trend watching. I think when, you know, if we're talking, like, on my account, like ValueMi or, you know, full time, like, I was always pretty good at trends, like being relatively early. **Swyx** [12:08] Mm. **Doug O'Laughlin** [12:08] I remember loving and being obsessed with TikTok in, like, 20- uh, 2019. And everyone's like, "Why are you so obsessed with the dancing music show?" Like, stuff like that. I feel like I've always been decent with the trends. But I think the thing was, um, when you see a really big wave that you have a lot of conviction in, it's worth going all in. **Swyx** [12:23] Yeah. **Doug O'Laughlin** [12:23] And that's, that's kinda what it came down to is, like, wow, I see this really big wave. It's worth going all in. **Swyx** [12:28] Yeah. **Doug O'Laughlin** [12:28] And so, um, I reoriented my life around it. **Swyx** [12:30] Yeah. **Doug O'Laughlin** [12:30] Um, yeah. ### Claude Code **Swyx** [12:31] Cool. Um, we're gonna talk about, uh, other trends that you've spotted, primarily, like, the sort of memory cycle which, uh... But also, uh, optics, which amazing story. Uh, but we wanted to sort of focus this, uh, for the Claude Code launch, uh, Claude Code anniversary. And you've been a big Claude Code show. **Doug O'Laughlin** [12:49] Yeah. I, I am no- **Swyx** [12:51] Where's the chart with the four- uh, 4% of code? **Doug O'Laughlin** [12:54] Oh, it's, it's... Actually, go to the top left. **Swyx** [12:55] This one. **Doug O'Laughlin** [12:56] Yeah, yeah. **Swyx** [12:56] That's it right there. **Doug O'Laughlin** [12:56] Oh, you know what's really crazy is we've updated that chart. I think it's, like, five now. I mean, and, and, like, as you know, it's really easy to generate code now, so, like, that, that number will continue to climb, but it's, like, just staggering the rate at which this is happening. **Swyx** [13:09] So, so let's recap for people who... Let's say, like, I think this is one of the most important pieces I, I've read in a long time. Um, and, uh, y- you know, you, you led it, and it's... A- and it's, it's weird because I, I think of you as, like, a, an analyst, right? Like, um, you... Like, one of SemiAnalysis alphas is that you're kinda like the fun millennial semiconductor firm when everyone else is super boring and old. **Doug O'Laughlin** [13:30] Yep. **Swyx** [13:31] Um, uh, but, like, what are you doing, you know, getting so into Claude Code, right? You know, I, I... Shouldn't you be reading reports and stuff? You know, like, tell the story of your, uh, Claude Code psychosis. **Doug O'Laughlin** [13:42] So yeah, I think here's the thing is, um, if you wanna be good at any game... We're, we're tool users at the end of the day, right? **Swyx** [13:49] Mm. **Doug O'Laughlin** [13:49] If you are good... If you wanna be, like... And, and obviously this is, like, outside of my job at SemiAnalysis. Like, I have all these other things I need to do to, to grow and make SemiAnalysis the best research firm ever. But, like, let's say you're a fund manager or an analyst, right? Your job is to find information edges and, like, new ways to put information together that no one else has done. Um, and so, like, I've always thought it's really important to know the most important weapons-grade, uh, tool that you can do all the time, which is essentially ChatGPT, Anthropic, all this kinda stuff. And I've been pretty... Like, I'm a, I'm a early adopter in tools as much as I can be. And, um, like, for example, I've been running the... our, our case study that we have into Claude Code since it first came out. Like, you know, I think over a year. Like, you know, I wanna say March, April I started to do- **Swyx** [14:32] Sorry, which case study? **Doug O'Laughlin** [14:32] Uh, so the case study for people when we're, we're hiring, like, a financial analyst, like, a... our core research seat or something. **Swyx** [14:37] Okay. **Doug O'Laughlin** [14:38] "Hey, um, you know, can you, can you take this company and do some analysis, blah, blah, blah? Give us this format back." Um, and I've been running it through, like, the agentic things. And I'm like, "Hey, what, what..." When agents really come around, they should be able to one-shot multi-step hard things to do, things that would take a human 24 hours to do, right? And I always wondered, 'cause I... You know, there's some good submissions and there's some bad submissions. We pride ourselves in the case study in being good. And, and honestly I always joke, like, well, you know, they're gonna start to beat the worst submissions. And so, like, that was our... that was always my base level. I have a base level of ca- is it better than a ChatGPT agent mode, or Anthropic's Claude Code- **Swyx** [15:12] Yeah **Doug O'Laughlin** [15:12] ... or Gemini CLI, whatever. And so I started running these benchmarks a little bit, and so I was very familiar with how good it could be. But then I was like, oh, it isn't quite there. Um, I've web coded some stuff on Opus four for sure, but, like, you know, it was, like, kinda interesting projects on the side. It was really hard. It took a lot of feedback. They would, they would mess up. It just didn't... And then, um, you know, everyone was freaking out about Claude Code 4.5, and I, like, took it for a spin, especially around the holidays. I had some free time. And then I was like, okay, well, like, how good is this? And it just, like, one... It started, like, one-shotting everything, right? Like, all these MVPs that, like, you know, you have to be like, well, the UIs, whatever. It's like, no, just one-shots it. And then you ask it to do something better, and it explains what you're doing, and you're like, "That's actually really good." And so I was like, wow, generalized, e- easy one-shot MVP of these, like, projects, and able to, like, really build things on top of it 'cause you can trust what it's doing to a certain extent. **Swyx** [16:02] Mm. **Doug O'Laughlin** [16:02] And it felt like some level of capability was beaten. It was very different than what I'd done in the past. Oh, I also tried Codex 2 before this, like, um, like, when it was 5.2. Never really got it to work in the way seamlessly, agentically, like- **Swyx** [16:14] Oh, of course. But that was af- this, this was recent? **Doug O'Laughlin** [16:17] Oh, no, no, no. So, so, so, uh, this, uh... My re- most recent when I was like, oh man, the, the awakening, um, probably December 27th. **Swyx** [16:25] December 27th. **Doug O'Laughlin** [16:26] Yeah. **Swyx** [16:26] You know it to the day. **Doug O'Laughlin** [16:27] Something like that. Something like that. I'm thinking, 'cause it's between the days, and- **Swyx** [16:31] Oh **Doug O'Laughlin** [16:31] ... I, I got home from Chri- Christmas, and I was like, oh, um, my fiance wasn't feeling so well, so I had some time to mess around just by myself. **Swyx** [16:38] Yeah, yeah. **Doug O'Laughlin** [16:38] And, and then also there's 2X usage limits. Oh my God, I miss those days. But I mean, now I'm addicted to Fast. Um, but, but look, I, I was playing around with these coding agents, just like everyone else should, or d- or should in this space, and, like, Claude Code versus Codex. I was, like, doing, you know, simple testing to see if they can make a thing, and it never really, like, one-shotted, like, a total idiot's thing. And then 4.5 just started one-shotting stuff, and that to me was, like, a huge difference. And so, um, I was like, wow, it could just, like, one-shot stuff. I have all these interesting ideas I can pursue. **Swyx** [17:08] And, and can I... Uh, is it Excel, uh, sheets in- primarily? **Doug O'Laughlin** [17:11] Uh, no, no, no, not Excel shee- uh, sheets primarily. I would say it's usually a mix of, like, a dashboard, or Excel, or something like that. But a good example where I... Like, I, I think Excel, it's moderately okay at, like, let's say one-shotting a basic financial model or, like, just taking and, and putting information from one place to another. It's not at human level, but honestly, i- if you know much about investing in the, being in the business, it's like, is your model, you know, being 5% more accurate really gonna ever make a good investment decision or not? No. Never, not once. Like, no one's saying, "Oh yeah, my estimate is always one cent more tighter than everyone else- ... and that's why I'm good at stocks." No, it d- it doesn't matter. **Swyx** [17:47] It's why sell side is ridiculous 'cause, like, everyone's like, "I'm bullish 'cause, like, my EPS estimate is, is, like, 10% higher than, than the street." **Doug O'Laughlin** [17:54] Yeah. **Swyx** [17:54] And I'm like, "Oh, who cares?" **Doug O'Laughlin** [17:55] Well, I mean, as you know, sell side, if we're gonna do this as, like, shot guns a- across the bow on so- on sell side, I mean, look- One of the reasons why SemiAnalysis has such a like a successful business is because I think sell-side as a concept is very broken. If you're talking about waves and things that are changing, uh, sell-side in a lot of ways is this hereditary child of like, let's say, 30 or 40 years of banking, where you had, um, you know, a company go public, so you needed someone to talk about it, to issue securities, and sort of have- **Swyx** [18:22] You're selling the stock **Doug O'Laughlin** [18:23] ... you're, you're literally selling the stock. You have the-- But you have to be independent-ish- **Swyx** [18:26] Yeah **Doug O'Laughlin** [18:26] ... so your ratings, buy, sell, hold. Uh, one of the biggest sales you could do is like when your, when your company IPOs, um, we'll talk about you so people know who you are. **Swyx** [18:35] Yeah. **Doug O'Laughlin** [18:35] That's the, the core original part of the sell-side, right? And the problem is, like, all the research kind of has this like really kind of fallen apart. It's just not different. A lot of banking regulations has changed, and so like the primary information process, it's like a 40-year-old business model on its last legs. And so I mean, that's one of the reasons why SemiAnalysis is so good, is because we are not focused on being a 1 cent EPS thing, which I would argue isn't exactly skill. It's just mechanical maintenance. **Swyx** [19:02] Mm-hmm. **Doug O'Laughlin** [19:03] Um, we are really good at understanding when technology changes and how that impacts everything, right? Um, because it doesn't really matter if one EPS is slightly higher or lower. It does matter if like... I'm, I'm just giving an example. If AMD's Helios rack is super on time and is like out of the gate ready to make tokens on this day, because that's gonna be billions of dollars of difference in revenue for AMD, right? **Swyx** [19:24] Yeah. **Doug O'Laughlin** [19:24] Or some networking technology or something like that, some bottleneck. Being really right on the timing and the magnitude of those inflection points will make a huge difference in the stocks. And so that's our business. We're a research firm. We're independent, and, um, we've had a really good hit rate, and we, you know, we care deeply about the tech- technology. **Swyx** [19:43] Exactly, yeah. **Doug O'Laughlin** [19:43] Yeah. **Swyx** [19:44] You know, I, I didn't mean to characterize you as like- **Doug O'Laughlin** [19:45] No, no **Swyx** [19:46] ... you are young and fun, but al- also you're extremely damn good. **Doug O'Laughlin** [19:49] Yeah. **Swyx** [19:49] It's like a, it's almost like a triple threat, and I als- always wonder if it's like, okay, it's like, one, you have like deep understanding of the tech. Two, maybe you're like sort of financially sort of, uh, literate. But also t- three, there's like this like X factor that is like, well, focus on things that matters, fuck everything else. And, and I don't know what that is, but that, that obviously is the alpha. **Doug O'Laughlin** [20:09] Yeah, 100%, 100%. That's, yeah, that, that, that's always been the, the analyst PM conversation. It's like, "Hey, you know, there really is only one or like three things that actually matter," right? **Swyx** [20:17] Yeah. Find me those three things. **Doug O'Laughlin** [20:18] Find, find me those three things, right? And then there's all this information. What's actually what... You know, that's the hard part. But yeah, we-- I think the thing is like we're really focused on finding the things that actually matter, right? Like the things that like, "Hey, this SerDes is better than this SerDes. This case doesn't matter. This one actually matters because now you have a giant opportunity." Um, and so that's, that's what the game is all about, I think, in terms of the research- **Swyx** [20:39] Yeah **Doug O'Laughlin** [20:39] ... and like a, a, you know, finance perspective. But on top of that too, it's just like when you do so much research, all these different little industry parts are so hard to understand, man. **Swyx** [20:49] Mm-hmm. **Doug O'Laughlin** [20:49] Like you go to some networking conference and you're talking to a guy who works at a company with... They're talking about their new email versus whatchamacallit laser. Um, you know, I can't even remember what it, what email it's replacing, blah, blah, blah. And you're like talking about all this stuff, and they have PhDs and you don't, okay? Everyone has a PhD at the deepest level, and they're all doing... So you have to understand all these deep understandings of these parts of these, um, supply chains. But you also have to have a big understanding too, because you know this little part at the bottom of this supply chain is actually gonna impact this giant, you know, business at the top. Because it's all interconnected, but it's so complicated. Just paying the tuition to show up is very expensive. **Swyx** [21:26] So I, I think one way I'll bridge this for listeners is that, um, this is the complexity of the problem domain. That's ex- There's extreme depth, there's extreme width, and, uh, you have to kind of throw human attention at all of it- **Doug O'Laughlin** [21:39] Yeah **Swyx** [21:39] ... to find what matters. And you're, you're saying you noticed some kind of breakthrough in December where it was suddenly clicking for you. I, I just really wanted to figure out like the, the tasks, the task that was nailing and the task that is still- **Doug O'Laughlin** [21:50] Okay **Swyx** [21:50] ... is not great at. **Doug O'Laughlin** [21:51] Yeah, so let me specifically talk about my use case, 'cause hey, I am still a stock guy. I can't trade or do anything in semiconductor or, or, uh, AI world, but you know, I do still really enjoy stocks. It's one of the reasons like I'm passionate about it, and it's probably my, my defining skill, what makes me good or bad at stocks, quote, unquote. You, you know the people who are really like stocks that are like lifers, they just love this shit. Um, it's, it's like an addiction, okay? Um, so I, I'm like, "Hey, you know, um, here's like all my positions, and like here's some like thoughts on it. Can you just like kinda like start copy-pasting some notes over and putting it all together?" It's like, yeah, it does- **Swyx** [22:23] That's why you give Claude Code? **Doug O'Laughlin** [22:24] Yeah. **Swyx** [22:25] Okay. **Doug O'Laughlin** [22:25] So I started doing this, and then I'm like, "Okay, but like add it, make the portfolio, run some basic risk stuff." And it's like, "Yeah, sure, fine, whatever." And then also like everything you do is perfect. I'm like, "Okay, well like actually, can we like make an investment framework for my investment style and start to grade all this stuff, and then like attack it and do stuff like that?" You can just do like iterative work. And then I was like, "Whoa, whoa, whoa. This is like a crazy useful tool that systemized how I think really quickly." Like, okay, what else can I do with it? And the answer is like fucking anything, right? And my joke on the, the podcast is it's all a skill issue now. Um, and so I, I've been, I've been doing this systematically for every aspect that I can think of. Like, hey, now it's so much easy- easier... Like I was, um... Actually, a perfect example is this, is this chart, right? Hey, Claude Code is a really big deal. Everything's one-shotting. I'm reading everyone going into psychosis like me at the same time on the internet. How do I actually know what's real and what's- **Swyx** [23:16] Well, I wonder, right? **Doug O'Laughlin** [23:17] Yeah. **Swyx** [23:17] Yeah. **Doug O'Laughlin** [23:17] I wonder, right? So I'm like, okay, uh, I heard about the fact that the Claude Code has the commits, right, o- onto the public r- uh, onto your, your commit. It says, "Hey, signed off with, um, Mushraf Khan." I'm like, "Well, why- Claude Code, scrape me all the commits," right? And, uh, you know what? Lo and behold, it pretty much did. Like and it's like- ... okay, well, like I'm looking for this signature right here, copy-paste. It was like how would you systematically go about doing it? Did like a big query pull for all the stuff, pulls all the, uh, like every single day. The API's relatively open. And then I'm like, oh my God. Let's see how much this is growing. And it's like, okay, chart go up. And you're like how big is it as a percent- percentage of GitHub? You're like chart go up. It's a huge deal, and I'm just like watching your... Uh, you know, I have like a cron job updating it every single day, blah, blah, blah. And I'm like this is a huge deal. Like, this is the, the biggest deal. I, I love watching trends. I love watching exponential trends, and I've never seen one even remotely at this rate. You would ar- you know, 4% in, like, two weeks or so- **Swyx** [24:13] Do, do you know about PR Arena? **Doug O'Laughlin** [24:14] Uh, yes **Swyx** [24:15] It, it's the, it's the previous attempt, uh, prior to you. But somehow they didn't, they didn't, uh, they didn't talk about... They just talk about merge rates, but didn't, didn't, they just they don't, they don't plot it as nicely as you do. **Doug O'Laughlin** [24:27] Yeah. Well, the... And also you wanna... Okay- **Swyx** [24:28] 'Cause you, you asked, you had to te- well you, you had to, uh, you asked the question of what is this as a percentage of GitHub, and this, this guy didn't. **Doug O'Laughlin** [24:35] Yeah. **Swyx** [24:35] That's it. **Doug O'Laughlin** [24:35] Yeah. And, and also, um, I mean, the other thing too is, yeah, I have a lot of those as well. **Swyx** [24:39] Yeah. **Doug O'Laughlin** [24:39] Um, but, but I thought the Claude Code 'cause I'm just trying to really, really, really focus on that. So typically, yeah. Well, and also you want to ge- give an example. Um, bro, I didn't make that chart. Uh, Opus 4.5 did. **Swyx** [24:50] Yeah. **Doug O'Laughlin** [24:50] Uh, or, or I think 4.6. I'm like, "Hey, um, I want you to do it in this style. This is the SemiAnalysis color scheme." **Swyx** [24:57] Yeah. **Doug O'Laughlin** [24:57] "This, uh, re- I have, like, summarized books about visualization- **Swyx** [24:59] Yeah **Doug O'Laughlin** [25:00] ... and, like, put in here are little style tips. Yeah. Here's some style tips." **Swyx** [25:02] Still a toughie. **Doug O'Laughlin** [25:03] Yeah. Um, I don't, I don't even know, man. It has, like, it has, like... I had it go read, like, 70 books or something. I'm like, "Give me, like, you know, the, like- **Swyx** [25:10] It's probably a waste. Like, you know, you're- **Doug O'Laughlin** [25:12] I know. I, I know **Swyx** [25:12] ... defining a part. **Doug O'Laughlin** [25:13] It, it is a waste. Look, tokens are free. The, the cost of doing this is nothing. That's the part that's so amazing. **Swyx** [25:19] Yeah, yeah. **Doug O'Laughlin** [25:20] The cost of doing this is nothing. The information gathering and synthesis, like, hey, if it costs effectively the same doing 70 as 3, who cares? **Swyx** [25:26] Yeah. **Doug O'Laughlin** [25:27] Right? And so I, like, whatever. And the answer, I'm like, "Oh, this is too many tokens. You better, like, really summarize this into, like, 90 tokens or something like that." A really basic whatever, and then you have all this skill. But, like, okay, now you can put all that into a skill of how to make charts, um, in the SemiAnalysis format using any kind of data, and then you can systematically just push this out again. I'm like, "Hey, data analyst, uh, please consider all the relationships you can- **Swyx** [25:49] Yeah **Doug O'Laughlin** [25:49] ... and generate information." Like, I think it's, um... That one was not ChatGPT. That was not generated. That was not generated, which I hate, honestly. I don't like that much, that one as much as, like- **Swyx** [25:59] Yeah. It doesn't have the guidelines as the other thing. **Doug O'Laughlin** [26:00] Yeah. **Swyx** [26:01] Yeah. **Doug O'Laughlin** [26:01] And, um, and so you can just... That was, that was generated. And so you can just... What you can do is you just ask it to do is, like, "Hey, here's all the dates that we have. Can you, like, visually brainstorm with me a way to better represent this information?" It's like, "Yeah. Actually, I'm gonna generate you a timeline." **Swyx** [26:14] Okay. **Doug O'Laughlin** [26:14] Um, you can just do things. Um, and I- I, I mean, it's- **Swyx** [26:18] That, that is your catchphrase, right? **Doug O'Laughlin** [26:19] Yeah. It- that is my catchphrase right now. You can just do things. **Swyx** [26:22] Yes. **Doug O'Laughlin** [26:22] And, um, so people were looking at this from the perspective of people who are coding, and they're like, "Hey, just c- programming is a- automated," right? **Swyx** [26:29] Mm-hmm. **Doug O'Laughlin** [26:29] But, like, all information work is... You know, I would argue coding is a big subset of all information work. I think there's a, a Brian Hobart tweet or something forever ago. He's like, "You know, coding and financial, you know, finance people actually are very, like, different types of abstraction," but, you know, you are doing abstraction. Excel is a ginormous abstraction. You're building these relationships, and you're describing what you think a financial thing is worth, right? **Swyx** [26:53] Mm-hmm. **Doug O'Laughlin** [26:54] Um, I think coding's a little harder if I'm being honest with you, and you're telling me the hard one got automated. Why can't the easy one get automated? So I started to ask myself, how much can we do? And the answer is it feels like a skill issue. It makes issue-- it makes errors on the, on the margin, but you can kind of force it into, like, for me, I love using rubrics, right? Hey, um- **Swyx** [27:12] Yeah **Doug O'Laughlin** [27:12] ... I care about XYZ. Uh, out of 10, uh, score this, and then you can really do multiple things. It helps with the stochastic nature. **Swyx** [27:19] Do you, do you put it all in one prompt, like, the, the, the, the task and the rubric for the task, or do you put the rubric after all the task is done? **Doug O'Laughlin** [27:26] I, I actually have two versions of this. **Swyx** [27:28] Okay. **Doug O'Laughlin** [27:28] I'm like, "Hey, you can pull all this stuff together. Just ru-run the pr- **Swyx** [27:31] Yeah, yeah **Doug O'Laughlin** [27:31] ... uh, the rubric or whatever." **Swyx** [27:32] Yeah. **Doug O'Laughlin** [27:32] Or you can, uh, do the task and the rubric. It just depends on how you wanna do it. Um- **Swyx** [27:36] Yeah, because obviously if you put it task and the rubric, then it can iterate itself. **Doug O'Laughlin** [27:39] Mm-hmm. **Swyx** [27:39] But if you put it after, then it's probably more likely to pay attention to the, the rubric. **Doug O'Laughlin** [27:43] Yeah, exactly. **Swyx** [27:44] You know? **Doug O'Laughlin** [27:44] And well, and the other part, part of it too, um, yeah, it, it will iterate, but, like, the context rot doesn't ma- I kinda like it to be separate because the thing is it's like, okay, it needs to be this, like, fresh look at it. You have to think of it kind of like it would perceive anything anywhere, right? It just, each context window is just opening it up, and I think sometimes, um, if you have done-- if you do it together, it commingles the information to the point where it becomes biased or sus- susceptible. Opus 4.6, as you know, is, like, super sycophantic. Like, it loves to, like, say, "Yes, okay, yeah, I'll do this for you." **Swyx** [28:16] Yeah. **Doug O'Laughlin** [28:16] Um, I think having it separate keeps it, like, keeps some of that drifts kind of away, and that's, like, one of the things that I've really... Personally, I like the results better, but it's, it's just complicated. Like, part of this is really weird because I am, I'm weirdly now opinionated on taste in terms of how you should design things because you can, like, for example, the context rot thing, until someone explained it, I was like, "Oh my God." I was just, "Thank God someone said it." This is a huge deal. Um, there's this, like, meme where it's like, uh- **Swyx** [28:44] Oh, yeah. **Doug O'Laughlin** [28:44] Yeah. **Swyx** [28:44] These guys. **Doug O'Laughlin** [28:45] Um, well, do you see the meme? It's like, um, of Mice of Men and at the end of, um, you know, at the end of the book, of... I can't remember which character shoots who. **Swyx** [28:53] I've never read it. **Doug O'Laughlin** [28:54] Yeah. Like, so, so one ch- character shoots the other guy, and it's like some guy made a meme about it being like, oh, this is after your, um, after your Claude Code is garbled, you know, 5 million tokens. You're like, "Okay, it's time to put you down." Um- ... because the context rot is huge. Um, so yeah, this, yeah, this is example where- **Swyx** [29:10] So what, what are your compacts practices? Do you s- sort of aggressively compact manually or...? **Doug O'Laughlin** [29:14] Um, so I personally, with the one- new 1 mil, it, I feel like I try to do it at all in one compacts window. Um, I'm not doing ginormous projects. **Swyx** [29:23] And the, the 1 mil is very new, right? Like- **Doug O'Laughlin** [29:24] Yeah. 1 mil- **Swyx** [29:24] ... 1 mil- **Doug O'Laughlin** [29:24] Very new. **Swyx** [29:25] Okay. **Doug O'Laughlin** [29:25] Very, uh... And, but it's a big deal too because- **Swyx** [29:28] Yeah **Doug O'Laughlin** [29:28] ... 'cause your skills and whatever, your Claude MD is a percentage of 1 mil is so much smaller, so you just get so much more oomph, right? Because, um, the s- the 200Ks are just wiping over and over and over. Um, that's a big deal. I think it's a huge deal. And, and also with how the agents are working, the sub-agents will have their own con- context window, and then the pasting kind of, like, really saves that, that big, you know, the 1 million. You just want a really high-quality, uh, project within that. That's the best in my opinion. Um, compacts just kind of start the compression of the noise. So- **Swyx** [30:01] Oh, yeah. **Doug O'Laughlin** [30:02] Yeah. **Swyx** [30:02] Uh, mentioning sub-agents and multi-agent. Uh, so first of all, I wanted to give a shout-out to this thing from Anthopic Research where they were like, "Here's our production traffic," um, and they, uh, they did a, they did a report that was kinda like their, their equivalent of the meter chart. Um, and there's a lot of people saying that- Oh, you should ... You know, software engineering has PMF, but here's- here's the- the next list of everything else. But what if they're all also- also just software engineering, right? Like **Doug O'Laughlin** [30:26] Yeah. **Swyx** [30:26] Like, software engineering is like 50% right now, but what ... Like, there's nothing stopping it from continuing to go to 80. **Doug O'Laughlin** [30:32] I think maybe what's gonna happen, um, m- ... This is like a maybe a giant brain take- **Swyx** [30:37] It has, like, data analysis in here, which- **Doug O'Laughlin** [30:39] But- **Swyx** [30:39] ... that's what you were doing. **Doug O'Laughlin** [30:40] I, I ... Yeah, that's ... In my opinion, that is downstream of so- Like, that is d- **Swyx** [30:44] Yeah. **Doug O'Laughlin** [30:44] Um, so- so I think how we should think about it is software engineering might all be downstream of chips, which is downstream. Like, like, chips is upstream, and then it's AI, and then it's software engineering. It is all the extension of that same compute hierarchy, and I think the, like, you know, teaching where machine and, uh, code kind of inter- or, and the world intermingle right now is code. And so that's just gonna be the bleeding language that's used to- to figure out everything else. Um, that's m- that's my belief. Like- **Swyx** [31:12] Yeah **Doug O'Laughlin** [31:13] ... it, it, it doesn't make sense to build... Like for example, this is a perfect example of this, is like Excel, Claude for Excel is much worse than Claude Code using Python to use the Excel skills to then deposit into. It's all l- **Swyx** [31:26] Much worse **Doug O'Laughlin** [31:26] ... it much worse. **Swyx** [31:27] Even- even all the work they're doing on it. **Doug O'Laughlin** [31:29] Yes, 100%. **Swyx** [31:30] Okay. **Doug O'Laughlin** [31:30] Because if you think about it, it's- it's a legacy. Why make a car engine fit into a horse carriage? It should just be in a car. Like, it's like- **Swyx** [31:37] Mm-hmm **Doug O'Laughlin** [31:37] ... it's like a backwards compatibility thing where it does work because LLMs are, like, relatively generalizable like this, but why bother? Because that same abstraction of information on Excel, it's just in that because it's human formatted for us to understand, and I think that that's the important distinction. **Swyx** [31:53] Mm. **Doug O'Laughlin** [31:54] All of this information stuff, all this software stuff is just to be consumed by humans. Doesn't matter. **Swyx** [32:00] Yeah. **Doug O'Laughlin** [32:00] If they're just as good at- at putting the data together, we should be much more concerned about machine-focused of, like, software consumption. And so they can, like, you know, um, the- the LLMs and the agents can put and synthesize all the information and deposit God knows however you want it to be. I don't need to make a chart in- in PowerPoint or Excel. It will just deposit it, the Matlab pl- uh, the Matlab pl- **Swyx** [32:22] Matplotlib. **Doug O'Laughlin** [32:22] Yeah. **Swyx** [32:23] Matplotlib, yes. **Doug O'Laughlin** [32:24] Matplotlib, um, in a chart to me in a image. Fine. That- **Swyx** [32:28] Oh, oh, you're trying to use Matplotlib? **Doug O'Laughlin** [32:29] Yeah. **Swyx** [32:30] Wow. **Doug O'Laughlin** [32:31] Why? Why? You know, it's better, it's better- **Swyx** [32:33] I know **Doug O'Laughlin** [32:34] ... understanding that code. **Swyx** [32:35] Yeah, yeah, yeah. **Doug O'Laughlin** [32:35] So why ever make a chart again? **Swyx** [32:37] Yeah. That's- **Doug O'Laughlin** [32:38] If it, if it's better- **Swyx** [32:39] It's just, like, it could be inconsistent with, like, the other charts that you do- **Doug O'Laughlin** [32:42] Yeah **Swyx** [32:42] ... which I don't think you would care that much about. **Doug O'Laughlin** [32:44] I, I don't think we would care that much, but I think, one, our new charts are better than our old charts. **Swyx** [32:48] Yeah, yeah. **Doug O'Laughlin** [32:49] And number two, I think, uh, if it increases the speed of information, that matters a lot. **Swyx** [32:53] Yeah. **Doug O'Laughlin** [32:53] Um, and so I think we're much more ex- So pretty much the new charts will outweigh the old charts because they'll just grow. Um, so yeah, I think it, it, it is a little inconsistent. We have the same watermarking. Honestly, I think it's better than our old for- formatting anyways. Um- **Swyx** [33:07] Well, the first thing this looks ... reminds me of is Bloomberg. I was like, "You guys are just, like, you know, becoming Bloomberg"- **Doug O'Laughlin** [33:13] Ah **Swyx** [33:13] ... which is a nice similarity- **Doug O'Laughlin** [33:14] That's not bad **Swyx** [33:14] ... that it is the beginning of your company. **Doug O'Laughlin** [33:15] Yeah. **Swyx** [33:16] Uh, couple things I wanted to sort of double-click on because the- this is just a Claude- Claude Code, like, brain dump- **Doug O'Laughlin** [33:21] Yeah **Swyx** [33:21] ... from one of the- the biggest sort of Claude Code, uh, shows in the world, uh, which is subagents and agent swarms. Maybe if you ... I don't know if you've tried- **Doug O'Laughlin** [33:30] I have tried them **Swyx** [33:31] ... any. Pick- pick either one. Uh, whatever you want. **Doug O'Laughlin** [33:33] I have a controversial opinion that Claude does not do RL on agent swarms or agent teams. **Swyx** [33:38] Yeah, it's just an experiment. **Doug O'Laughlin** [33:39] It's- it's just an experiment. Um- **Swyx** [33:40] Yeah **Doug O'Laughlin** [33:40] ... thank you. Thank you, 'cause no one- **Swyx** [33:42] No, not he could **Doug O'Laughlin** [33:43] ... we exactly, 'cause the pro- It's just via prompt, and it's actually very bad. Um, I think subagents are okay because they usually have a Claude MD to go do whatever. Um, but the agent team is h- is actually really bad. **Swyx** [33:53] Okay. Well, well, you know- **Doug O'Laughlin** [33:54] Is- it's- **Swyx** [33:55] We can't, we can't knock it 'cause it's experimental, so- **Doug O'Laughlin** [33:57] Yeah. Yeah, no, no, it is a g- **Swyx** [33:58] What did you, what did you try it on? **Doug O'Laughlin** [33:59] Um, uh, well c- it was, like, some big data analysis of, like, many, many different companies with different KPIs into a dashboard all in one. **Swyx** [34:07] Mm-hmm. **Doug O'Laughlin** [34:07] Um, I was like, "Hey, can you just make this all whatever, split up the teams?" You know, speaking of that though, you say that, but Kimi, Kimi 2.1 agent swarm is actually good. **Swyx** [34:15] I have also tried that. **Doug O'Laughlin** [34:16] It is a- that is actually really good. So I did some, like, in- oh, example of things, like, I was never available to me, like internal benchmarking of these models and be like, "Hey, here's a set of problems I would like you to do 20 times. Um, can you do them? And then I can measure the performance between them and then, like, do qualitative like what's the difference between X and Y?" **Swyx** [34:33] Yeah. **Doug O'Laughlin** [34:34] That, that was completely out of the hands of me, a normal guy, like three months ago. Okay? Now it is completely available to me. That's awesome. Like, I am very ... I care about this stuff, and now I have the tools that's able to automate and do a lot of this stuff because, hey, all of software engineering is, like, partially automated. And so, I mean, my experience is the 2.5 swarm actually improves the model's performance meaningfully. Um, the agent team makes it meaningfully worse because there's clearly not RL done. So it, it isn't context- **Swyx** [35:03] Yeah **Doug O'Laughlin** [35:04] ... aware of what's the best- **Swyx** [35:05] Yeah **Doug O'Laughlin** [35:05] ... thing to be done. And yeah, so I think, I think it's interesting. I like subagents 'cause it's usually a little bit cleaner on a task, uh, to go do it and then come back, but the agent team is just very- **Swyx** [35:16] They had some post about how they did stuff- **Doug O'Laughlin** [35:18] Yeah **Swyx** [35:18] ... uh, where it was ... Yeah, there's, there's a bunch of RL for, for this, and I tried it myself. I thought it was, I thought it was, like, pretty ... It's cute how they do all these, like, little- **Doug O'Laughlin** [35:26] Yeah **Swyx** [35:26] ... games and stuff. **Doug O'Laughlin** [35:27] Yeah, yeah. Also, it's crazy how, like, the setup. You have to ... It's a lot of compute. **Swyx** [35:31] Yeah. **Doug O'Laughlin** [35:31] To just run the swarm, I think it's like a 16 node of H100s. Um- **Swyx** [35:34] Okay **Doug O'Laughlin** [35:34] ... and you're just like, "Dang." So you and I are not gonna be running... And this is just a run, and I'm sure there's concurrency available, but, um, yeah, I think it's really cool, and that's like ... I think that that's the sign of what's next because, you know, these agents are gonna get better to a certain extent. They're, they're, you know ... It's another benchmark and bench- like, a, another benchmark to hill climb, right? But then it's gonna be how many of these together in a bigger chain can you get to work that you could argue it's kinda like a scale-out of the reasoning problem too. Hey, how do you get these, like this one agent to essentially get a verified whatever, put it into a bigger process, and do more information work? That, that's the next thing, and it's important to have context windows that, uh, that don't, um, garble up into random stuff and is able to do just, like, good enough with token efficiency I think is a huge part of that. **Swyx** [36:20] Yeah. **Doug O'Laughlin** [36:20] Um, so yeah, that's, that's kind of what our experiments have shown, at least in terms of, like, the agent swarm versus, like, not. I think it's very clear the agent team out of Claude is an experiment. Um, but Kimi shows- **Swyx** [36:32] They'll do better. **Doug O'Laughlin** [36:32] They'll, they'll definitely do better. But the Kimi 2.5 tells you that this is already, boom, perfectly great new places to do more work on completely available to us right now. I think that's huge. **Swyx** [36:42] Yeah. **Doug O'Laughlin** [36:42] Um, because if these agents get any better, like, I don't know, I'm never gonna sleep again So. **Swyx** [36:48] Uh, honestly, like, uh, I... It's very interesting, this, uh, sort of moonshot AI. And, and this is a tangent. We're not, we're not really gonna focus on this very much, but, um, you know how, like, the, the sort of AI tigers out of China were, were DeepSeek and Qwen and- **Doug O'Laughlin** [37:01] Mm-hmm **Swyx** [37:02] ... then you were like, "Well, w- who are these, like, Kimi guys?" And, um, and these, these sort of newer names like, um, I guess, I guess Minimax as well- **Doug O'Laughlin** [37:09] Yep **Swyx** [37:09] ... would, would be in there. **Doug O'Laughlin** [37:10] Minimax, yeah. **Swyx** [37:11] And, uh, Zei has been, been around longer, but only recently much more active. **Doug O'Laughlin** [37:15] Yes. **Swyx** [37:15] So, like, I, I notice that Kimi is much more in the productization phase, like, as, as seen, as opposed to, like, the Qwens of the world, the DeepSeeks of the world who don't really care that much. **Doug O'Laughlin** [37:24] I mean, Qwen, because of how it's, uh, it's attached to Alibaba, right? Like- **Swyx** [37:28] Yeah, yeah **Doug O'Laughlin** [37:28] ... um, they, they have a way to productize it, but it's like, it's like kinda like the Gemini version. They have so much stuff to do elsewhere, right? **Swyx** [37:35] Yeah, yeah. **Doug O'Laughlin** [37:35] Um, but yeah, Kimi, Kimi's pretty interesting. **Swyx** [37:37] They're pushing so hard. **Doug O'Laughlin** [37:38] They're pushing- **Swyx** [37:38] They've got everything. **Doug O'Laughlin** [37:39] I know. **Swyx** [37:40] They got Kimi Manus, Kimi Claude. Kimi Claude. **Doug O'Laughlin** [37:43] Yeah, I know, Kimi Claude. I haven't... Yeah. Dude, I was gonna say, um, have you messed around with OpenClaude? 'Cause I did. I- **Swyx** [37:48] Yes **Doug O'Laughlin** [37:48] ... oh, gosh, I remember, um, what was it first called? Claude. **Swyx** [37:51] Claude Bot. **Doug O'Laughlin** [37:51] Claude Bot, yeah. Dude, I was gonna say, it was really, really euphoric. I was, like, having it read all my emails and my calendar and do all this stuff, and then I was like, "Wait, wait, wait. This is really, really, really prompt injectable." And I was like, "This is pretty secure and important stuff," so I, like, I was like, "You know, Claude Code psychosis is good enough for me at this point in time." **Swyx** [38:07] Yeah. I mean, so what I do is I just have multiple emails, right? **Doug O'Laughlin** [38:10] Yeah. **Swyx** [38:10] And there's, there's a safer email to give to bots- **Doug O'Laughlin** [38:12] Yeah **Swyx** [38:13] ... and I can let it use that, and if it impresses me, then I can upgrade it. Uh, but Claude Bot didn't impress me at the, at- **Doug O'Laughlin** [38:18] I, I'm gonna be honest with you, I'm gonna be honest with you, I wasn't impressed either. **Swyx** [38:21] Yeah. **Doug O'Laughlin** [38:21] That was the reason why people were freaking out about this whole book. I was like, "Bro, have you actually used this shit?" 'Cause it's not... Even right now on Claude Code in a relatively focused terminal it will be like, "Oh, blah, blah, blah." I'm like, "Dude, in the .env there is an... Like, in the .env there is an API I told you to use for this sub-case of problems, and it's in your Claude MD. Like, please focus up." Like, it still is, like, making mistakes. **Swyx** [38:43] Yeah. **Doug O'Laughlin** [38:43] This is not, like, truly AGI, and there is harness... You still have to wrangle this thing. But, um, it's not like a perfect skill follower, and the, the context in each attention window is gonna, like, change, and sometimes it'll be lazy, sometimes it won't be, but it's definitely good enough to do a lot of information with. **Swyx** [38:58] Yeah. I was gonna... Uh, so I, I use, I use our, um, I use our Discord as, uh, basically, like, a, a way to just bring information in and out. I, I, I just saw, I saw this too where, where basically, like, a lot of people are just setting up things that they could have done in Zapier- **Doug O'Laughlin** [39:12] Mm-hmm **Swyx** [39:12] ... with Claude Bot because they're like, "Well, you know, now I'm, like, AI pilled." But actually they just done it more securely with Zapier. Which I think is kind of interesting. **Doug O'Laughlin** [39:20] I guess. I, I do think it's kinda interesting, but I think there's... But the, the difference though is Zapier... I mean, I remember I've tried to use Zapier before. **Swyx** [39:29] Yeah, it's, and it's also not very good. **Doug O'Laughlin** [39:30] It's not, also not very good. The difference though is, like... And that's okay. Like, it's okay to be early to something and just wrong because you weren't the one that made it happen, right? Claude Bot, the Claude Code, Claude Bot, whatever, all this stuff, the reason why it's so powerful is it gets to completion, right? And, and, like, okay, Zapier, maybe you can get to completion all the time, but, like, man, it probably took you, like, eight hours of clicking through things and, like, copy-pasting crap to make sure it all works and it's all secure. And it's like, well, Claude Bot did it or Claude Code did it in, like, you know, four and a half minutes. **Swyx** [40:01] Mm. **Doug O'Laughlin** [40:01] And that's good enough for me. You know? That, that's a faster achievement. And so, like, it's totally okay that they were, they're right, but they were just not the right mechanism. **Swyx** [40:11] Yeah. **Doug O'Laughlin** [40:11] Right? You, you see this happen in information, like, in the history of, like, compute or- **Swyx** [40:15] I, I think there's also, like, an innovator's dilemma, I think, where Zapier, as a preexisting business, had this view of the world of automations as, like, very strict sort of on rails workflow- **Doug O'Laughlin** [40:25] Yeah **Swyx** [40:25] ... type things that their giant user base already uses. They couldn't, like, really pivot that much. **Doug O'Laughlin** [40:30] No. **Swyx** [40:31] So that's why, like, I think, like, you know, one of the co-founders left because they were like, "Well, I can't exist within this, like- **Doug O'Laughlin** [40:37] Yeah **Swyx** [40:37] ... like, tight constraints." **Doug O'Laughlin** [40:38] Like, like, yeah, you, you end up becoming w- you know, the, the, the box will control you. **Swyx** [40:42] Yeah, yeah. **Doug O'Laughlin** [40:42] You, you, you are, uh- **Swyx** [40:44] It's your, it's your golden handcuffs. **Doug O'Laughlin** [40:45] Yeah. It's just like your cage, you know? You're gonna act like how you are in the cage. And so yeah, that, that sucks for... Honestly, that, I feel like that sucks for Zapier. **Swyx** [40:50] The framing I have is, like, your priors become your prison. **Doug O'Laughlin** [40:53] Ooh. That's pretty good. That's pretty good. That's pretty good. Your priors become... Yeah. **Swyx** [40:57] I haven't vlogged that yet, but I should. **Doug O'Laughlin** [40:58] You should, you should. Your priors become your prison. I like that a lot. **Swyx** [41:01] Coming back to Claude Code, just, I, I also wanna make this, like, the sort of Claude Code- **Doug O'Laughlin** [41:04] Yeah, yeah. I'm sorry, sure. **Swyx** [41:05] Uh, no, no, no. That, like, I wanna indulge because, like, that's how natural conversation goes- **Doug O'Laughlin** [41:09] Mm-hmm **Swyx** [41:10] ... and I think people, like, enjoy that, right? **Doug O'Laughlin** [41:11] Mm-hmm. **Swyx** [41:11] And probably that's the only time we'll talk, we'll talk about Kimi. **Doug O'Laughlin** [41:13] Yeah. **Swyx** [41:14] So, like, do you use hooks? Do you... Like, give me, like, the, the Doug O'Laughlin Claude Code setup. **Doug O'Laughlin** [41:20] I have just, like, essentially a few base skills, uh, and then I have a lot of APIs. And then we've also made sure to work, and this is, like, all work in progress as well, to have APIs for some of the SemiAnalysis information out as well. **Swyx** [41:32] Yeah. **Doug O'Laughlin** [41:32] And so that way we have- **Swyx** [41:33] Like an internal server that just has- **Doug O'Laughlin** [41:34] An internal server that is, that, that is accessed by people with an API- **Swyx** [41:39] Ooh **Doug O'Laughlin** [41:39] ... so that, like, a- all the SemiAnalysis researchers are able to hit, like, some basic level of context, 'cause I think the context is really what matters. **Swyx** [41:46] Yeah. **Doug O'Laughlin** [41:46] Um, I'm too, like, too dumb to be really smart in, in order to have... Well, I guess, I guess I do have some hooks, if that makes sense, um, in terms of, like- **Swyx** [41:54] I think hooks are very under- underrated, right? **Doug O'Laughlin** [41:55] Yeah. I, I do think- **Swyx** [41:56] 'Cause you can do, like, a Ralph Loop- **Doug O'Laughlin** [41:57] Yes **Swyx** [41:57] ... with a hook. **Doug O'Laughlin** [41:58] Yeah. **Swyx** [41:58] Like- **Doug O'Laughlin** [41:58] Yeah. I feel like I underutilize hooks. I, I think- **Swyx** [42:01] Yeah **Doug O'Laughlin** [42:01] ... that is true, but I do, I do run some version of them on, like, skill calls effectively, like, "Hey, on this," then you have to start pulling all this stuff. But I think in the beginning I tried to do all this, like, hook stuff, and, like- Compounded, stuff like that. And I found that, like, you know, the Gastown, Ralph Loop era, it's like it, it is a sign of what will come, but I just don't think there's enough fidelity to, like, make crazy multi-turn something happens. So like, okay, actually less is more. Try to have, like, a strong set of smaller skills with a good amount of context, um, information to be pulled in, and then at the beginning of every session, ask and focus on what you wanna do so that, like, it prompts the ... like, not like, you know, a Claude within a Claude whatever. So here's the goal to finish within this single context window, and then get it done. And this is, like, my generalized research thing. "Hey, I want to look at the price of NAND since 1984," or something like that. This is what I wanna do. I wanna ... Uh, so, like, the ... Oh, actually no. Let me just give you the best example. That is probably not gonna work. I would like to fine-tune a time series foundation model to predict NAND and, and DRAM prices, okay? I'm gonna first start by gathering as much information as possible from all this stuff, blah, blah, blah, and then we're gonna fine-tune it, evaluate which ones we're gonna do. I chose Chronus 2 because of covariates, blah, blah, blah, blah. Try to set this all project up. We'll make it a Vercel dashboard internally for, for SemiAnalysis. Maybe we'll external if we want if it's a good enough product, okay? So then it, like, does all this stuff, and then I just, like, start planning away. Hey, can you go research Sears to search API, Serper or X or whatever you wanna use to go look for all these different, um, information sources and then bring it together, right? So this agent goes and gathers all this information. This agent goes and, like, works on, like, considering the fact that the price isn't perfect to do all this fine-tuning on, and then we, like, throw it in. I also had it of, like, oh, what do I use? It showed me which GPU, whatever we're renting on a hourly basis. And so yeah, we just pull all this stuff together, then we fine-tune it. And I'm like, "Okay, cool. Um, how did this work?" And then we just have this constant iterative loop until I try to finish something. I got to the point where I was like, okay, this, this time series, uh, LLM is probably not gonna work. Um, unfortunately — Um, unfortunately, the- **Swyx** [44:13] You said it was because of regimes or something else? **Doug O'Laughlin** [44:16] I think so, it's regimes. **Swyx** [44:16] Yeah, yeah. **Doug O'Laughlin** [44:16] Yeah. There's no way. **Swyx** [44:17] But it's for, for- **Doug O'Laughlin** [44:17] This regime is so messed up. Um- **Swyx** [44:19] For, for a lot of people who are, like, new to finance, this is why I have, I have an issue with all these kids doing, like, stock trading games with LLMs. **Doug O'Laughlin** [44:27] Mm-hmm. **Swyx** [44:27] They have no idea. They, they've never studied finance. **Doug O'Laughlin** [44:29] Yeah. **Swyx** [44:29] And like and like, the, you know, some- like, a past does predict the future a lot until something fundamental change, and, like, the macro shifts, and, like, risk on versus risk off. **Doug O'Laughlin** [44:39] Yeah. **Swyx** [44:39] They've never heard, they've never heard those terms. **Doug O'Laughlin** [44:40] Yeah. **Swyx** [44:40] I had to explain it to, uh, people at Cognition. And, like, yeah, like, the, the rules invert, like, completely invert. Like, w- what used to work is exactly the opposite of what you need to do in- when you have a regime change. **Doug O'Laughlin** [44:51] Exactly. And it's very, very, very hard because ... And the other thing too is, uh, you, you realize, okay, each of these, each of these are almost like a one-off onto their s- onto their own. **Swyx** [45:02] Right, which reduces your sample size. **Doug O'Laughlin** [45:03] Yeah, which reduces your sample size. And so then at the end of the ta- at the end of the day, you end up being like, well- ... it kind of just like, I guess it's, um, here's some heuristics. Good luck, have fun, right? Here's your checklist to see it might be over, but you really don't know anything until then. Um, so but, but like, okay, an example of where this project was helpful and it's like, okay, I'm not gonna have the magic LLM tell me what the price of memory is gonna be. Hey, it was a good weekend project, and I did burn quite a few tokens. But I do happen to have, after all this, like, information synthesis and analysis, all of the memory prices of everything I could possibly find, plus the things up behind API that we've paid for, plus, um, you know, en- enhanced data sources, and I have all the covariates. So like, hey, WFE, what was the consumer sentiment? Every macro thing of all time. And you know what's really interesting is I am gonna just be like, okay, well, now can you go make a summary of each and every memory regime and what it looked like and what, what, what created the beginning, middle, end, and put that in a dashboard so it's relatable and, like, e- easy, shareable, consumable within my firm and company? Yes. Um, I'll probably be done with that today. And that ... Okay, so that you're like, well, that's just gathering, doing information stuff. Like, you don't understand. No one's ever done that in the history of time. **Swyx** [46:13] Mm-hmm. **Doug O'Laughlin** [46:14] I know for a fact as the guy who, like, is, like, the cycle semiconductor guy, I've written and done more work on the cycles than I think anyone else has at this point, especially for, like, the older ones, like the '80s and '90s and 2000s and 2010s. And, like, when I did it first time, the human grokked way brain is I went and I read these old annual reports and I put it together and I try to string a narrative through it. **Swyx** [46:36] Oh, my God. **Doug O'Laughlin** [46:36] And I, and I brought through all ... I'm like, okay, what was GDP this grow- what this year? What was all this stuff? And you have to, like, make all this giant sheet to com- to whatever, and then make the narratives. No, none of that shit, dude. Like, I mean, this is, like, too much information to gather. It's like a lifetime of work. It's like a PhD project. I did it in a day, two days. ### Augmenting Experts **Swyx** [46:53] Yeah, I, I mean, I think the, the kind of pushback would be that then you don't have enough expert information to criticize the reasoning that- **Doug O'Laughlin** [47:03] Yeah **Swyx** [47:03] ... went into the report that you're s- slopping out and- **Doug O'Laughlin** [47:05] Yeah **Swyx** [47:06] ... you know, that's- **Doug O'Laughlin** [47:06] There, there is some slop. I, I- **Swyx** [47:08] But- **Doug O'Laughlin** [47:08] ... definitely agree with the slop **Swyx** [47:09] ... you know- **Doug O'Laughlin** [47:09] So, so I think of it once it ... So right now- **Swyx** [47:12] By the way, that's also existential for you guys if you get caught do- like, putting out some slop to your clients, right? **Doug O'Laughlin** [47:18] Yeah. **Swyx** [47:18] Like, you, you have to- **Doug O'Laughlin** [47:19] Yep **Swyx** [47:20] ... at, at one point be, like, extremely AI pilled and, like, you know- **Doug O'Laughlin** [47:23] Yeah **Swyx** [47:24] ... you number one in the world at applying AI to your productivity, great, but also, like, you gotta- **Doug O'Laughlin** [47:28] Yeah. You have to ... So, so I think, I think the thing that's really interesting is this whole thing is like a game of hygiene now. **Swyx** [47:34] Yeah. **Doug O'Laughlin** [47:34] Because I, I think it's like this is really hard, and I think about it all the time. I feel very comfortable with doing all this work because the thing is my- at the end of the day, since- **Swyx** [47:43] And has done the work **Doug O'Laughlin** [47:43] ... and I've done the work. I have, like, a lot of, like, embeddings in my brain- **Swyx** [47:47] Yeah **Doug O'Laughlin** [47:47] ... a lot of information. The vibes that have got me in here is actually, like, tons and tons and tons of information, set up scenarios, and, like, pattern recognition, right? But, um, yeah, you're right. This, this crap makes mistakes all the time. All the time. It is still just like a ... Like, I think of it, once again, as like a junior analyst, right? The analyst goes and does all this, like, really pain in the ass information, and you bring it all together to make a good decision at the top. Um, but the problem is, um, historically what happens is that junior analyst, who I once was, went and gathered all that information, and after doing this enough times, there's a meta level thinking that's happening where it's like, okay, here is what I really understand and how this type of analysis I'm an expert in, actually, I'm very good at, I consistently have a hit rate. Now I'm the expert, right? I don't think that meta level learning is there yet. Um, we'll see if LLMs do it, right? Everyone who's spending one quadrillion dollars in the world thinks it will. It, it better, it better happen- ... right? If you're spending, you know, a trillion dollars and there's not meta level learning. But for me, in our firm, that massively amplifies everyone who is an expert, right? And we are a firm filled with experts. And so it's this hard part where I wonder if new people, we will be less lenient in terms of, like, how much AI tools you- **Swyx** [48:58] New, like, junior- **Doug O'Laughlin** [48:59] Yeah **Swyx** [48:59] ... or new to the firm? **Doug O'Laughlin** [49:00] Junior. **Swyx** [49:00] Oh. **Doug O'Laughlin** [49:01] Junior to the firm. **Swyx** [49:02] Yeah. **Doug O'Laughlin** [49:03] Or junior, and like, like, 'cause like you have to still do something. You can't just, like, slop it up. It's very obvious to me when it's slopped. **Swyx** [49:09] Yeah. **Doug O'Laughlin** [49:09] Right? When it's slopped and there's no cognition, then it's like, like, whatever, the artisanal last 5% is, like, that really matters. **Swyx** [49:15] Yeah. **Doug O'Laughlin** [49:15] But for me, I know inherently what the 5% is. I can, like, write it away with some really easy heuristics and time, and, like, be like, "Okay, well, this is the last 5% you fixed. This is what I believe. Just fucking make up these assumptions instead. Press enter. Okay, cool. We're good to go." You know? **Swyx** [49:30] Yeah. **Doug O'Laughlin** [49:30] Um, and so that's kind of the hard part. That's a real hard part. There is still a human in the loop right now. One day, someday, it'll be superhuman, but I definitely believe the... Where we're at today, um, where we're... There, it's not there. Like, you just compound all this noise and it becomes just, like, garbled, just, like, all context, uh, rot. But in terms of, like, the capability that is over hit, like, you know, the human CPU in these, this agentic swarm is very, very powerful now. **Swyx** [49:56] Yeah. **Doug O'Laughlin** [49:57] You know, a huge, huge, huge multiplier of what you're able to do. And for me, that was enough to be like, "Th- I feel AGI pilled," honestly. **Swyx** [50:03] Yeah. **Doug O'Laughlin** [50:03] Because if, if I define AGI as many common jobs, not like... I'm not, I'm not doing ASI that's like religion. Can it automate or change or, or take or, you know, completely shift a lot of the information work? Yes, 100%. **Swyx** [50:17] Yeah, yeah. **Doug O'Laughlin** [50:17] Like, data analysis is a perfect example. "Hey, every quarter, I want you to just find me some examples of some information that might be interesting." I just can't imagine if I was an entry-level worker doing data analysis that a 22-year-old, an average 22-year-old would, would murder the hell out of a relatively- ... well-thought-out agentic system. And so you're like, "Yeah, that job actually does seem at risk." And so that- **Swyx** [50:40] Yeah **Doug O'Laughlin** [50:40] ... that, the 4.5 capability enough like that, that we hit some level agentically where it seems to work and do bigger information work, that's when I'm like, "Okay, yeah, this, this does change everything." And so yeah, there's, there's all kinds of mistakes. I... It's a new level of hygiene that we have to do. You're gonna have to understand what the absolutely of, uh, agentic work is back to you, right? I catch it making errors all the time. It doesn't always pull skills. Like, you can definitely tell... Like, context windows def- like, it gets dumber over time. It's not AGI today, but it can do these crazy long tasks, and as long as you finish it at the end and deposit it as information work, that's very valuable. **Swyx** [51:16] Yeah. Amazing. So you do a lot of, like, uh, c- client visits obviously. **Doug O'Laughlin** [51:20] Yeah. **Swyx** [51:20] Um, I... By the way, Transistor Radio, amazing for, like, understanding, like, what your world is like. **Doug O'Laughlin** [51:26] Yeah. **Swyx** [51:26] Uh, uh, are you also Claude Code pilling your analysts and your- **Doug O'Laughlin** [51:31] I- **Swyx** [51:31] ... you know, on the other side? **Doug O'Laughlin** [51:32] I've definitely Claude Code pilled the analysts. Um, everyone in the New York office, I'm like, "You must try it." I, like, really tried to, like- **Swyx** [51:39] No, not... No, I mean, not your, like, not the SemiAnalysis- **Doug O'Laughlin** [51:41] Oh **Swyx** [51:41] ... but your customers and, and all that. **Doug O'Laughlin** [51:43] I don't think- **Swyx** [51:44] You know, like, I... So my perception is they don't adopt any of this stuff. **Doug O'Laughlin** [51:46] Okay. So yes and no. Some people are interested, but you have to remember it's relatively more conservative. But I think... But if you ask any analyst if they're using AI, every single one of them will tell you, "Yes, I use it every single day." **Swyx** [51:59] Yeah. **Doug O'Laughlin** [51:59] "Of course. How could I not?" **Swyx** [52:01] Yeah. **Doug O'Laughlin** [52:01] "This is, like, an, an, a vital skill." And so the, the, the basic, the basic inference that I'm doing is I am a bleeding edge adopter. I'm a relatively smart dude who knows what he's doing and if a tool is useful or not, and I've evaluated the tool and I'm like, "Wow, this is an amazing tool that I literally, like, pry it out of my dead fucking cold hands," okay? I'm... Like, this, even if it's, like, makes mistakes, I will be using this for all kinds of work forever. Then I look around to everyone else and being like, most of these guys are enough like me that if they have an opportunity and an edge, they will obviously apply it. And they look at this tool and they start to use it. If, if they start to use it and they're thinking like me, they're gonna obviously adopt it. I'm like, "Well, I don't understand why everyone doesn't adopt it." I would argue, well, we'll see in the 24-month view, it will be a base level, I think. I think- **Swyx** [52:47] Yeah **Doug O'Laughlin** [52:47] ... Claude Code, CoWork, whatever is gonna be a base level of all information work very soon. **Swyx** [52:52] Yeah. **Doug O'Laughlin** [52:53] And you know, you see one... Um, my, my friend was telling me how his portfolio manager w- found CoWork and he's, like, getting it to read his emails, and he's like, "Oh my God, I love this," right? Everyone's moment is gonna be a little different, but I think my moment, it feels like GPT 3.5 or 4 for me, where there's that first time where you're like, "Okay, I know it made some shit up, but, like, this is better than, like, if I went for hours searching, putting information together." It can... And then also it's like the analogy power, you know, where you can say, "Hey, this is the setup. Can you com- describe it in this?" These, like, really strong pattern matching skills that are really powerful. I just think it hits some level of capability. I can't tell you what it is. It is, like, my ta- my personal taste where I'm like, "Oh wow, this is completely over the, the, the, the chasm of what needs to happen for it to be a very, very powerful tool." And so yeah, that's my Claude Code moment, I think. Um- **Swyx** [53:43] There, there's some kind of automation chart, um, that, you know, XCD has this automation chart. **Doug O'Laughlin** [53:47] Yeah, yeah, yeah. **Swyx** [53:48] And I think I, we need a version of this that is the Claude Code, like- **Doug O'Laughlin** [53:51] It needs to be much... But it- **Swyx** [53:52] Yeah **Doug O'Laughlin** [53:52] ... but what's crazy is this, the Claude Code thing, like, murders the axis. **Swyx** [53:56] Exactly. It just strips everything, like, le- right or something. **Doug O'Laughlin** [53:59] Yeah. **Swyx** [53:59] But, uh, but also, like, it, the, um... What I was trying to figure out is, well, okay, it is maybe dumber, less, less human attention, but because you can spin it up so quickly and it can sit in parallel so quickly, uh, and it, and it gets done, you get more turns at the wheel. **Doug O'Laughlin** [54:13] Yes. **Swyx** [54:13] Whereas in, as a human, you, you get one turn. **Doug O'Laughlin** [54:15] You get one turn. Yeah. **Swyx** [54:16] But it, with, with, with Claude Code maybe you get three turns, and th- the, the sort of review process is the thinking. **Doug O'Laughlin** [54:22] Yeah. **Swyx** [54:22] And you just need to get very good at review or- **Doug O'Laughlin** [54:24] Yeah **Swyx** [54:25] ... or hygiene. **Doug O'Laughlin** [54:26] Uh, yeah, I think of it as hygiene. The thing that's, like, really gonna be painful though is, like, a lot of my expert opinion has been built by, like- You know, it- it's like pre-phones and not, right? Like, your attention span- like, you know, the children are cooked, okay? Like, you know, their attention spans are really bad, all this stuff. Like, oh, I don't- I read this, like, really sad thing where they're like, "Oh, and we're getting dumber," or something, first generation. I don't know. I'm not gonna ... Maybe that's like- **Swyx** [54:48] Well, did you see the, the Coinbase, uh, earnings call? **Doug O'Laughlin** [54:50] Yeah, I saw the current pricing. So, so like you had this thing where it's like, okay, and it's cute and all, but, like, it's such an addictive technology that, like, I feel very grateful that I'm like, well, I understand what I'm doing, have this history of doing stuff, and able to apply a tool. But, like, people who are riding this curve, it's gonna be very dangerous. It's like giving everyone- **Swyx** [55:06] I know, that was true. **Doug O'Laughlin** [55:07] Yeah. That's so funny. **Swyx** [55:11] I, I think you should just do that. **Doug O'Laughlin** [55:13] Yeah, he- well, we als- we, we, we do, we do with some of the SemiAnalysis memes, you know? And, and the thing is, you say some of this brain rot is, like, so bad, which is- it is terrible, but some of it is also, like, you know, it is hitting some attention mechanism in my, in my, my deep primordial monkey brain. **Swyx** [55:28] Stimming you. **Doug O'Laughlin** [55:29] Yeah, it's stimming me, and you're like, "You know what? I can't look away- "... from the, the Subway Surfers," so- **Swyx** [55:33] Wait. Yeah, you, you couldn't look away. I was- I had to pause it. **Doug O'Laughlin** [55:36] Yeah. Yeah, I was like, literally, I- well- Well, hey, there's like ... Have you ever been at, like, a bar when they play, like, these, like, weird ... Like, there'll be like, like, TikTok videos for lack of whatever, and you just watch. And- **Swyx** [55:47] There's TikTok bars in New York? **Doug O'Laughlin** [55:48] No, not TikTok bars. Not TikTok bars. **Swyx** [55:49] Okay. **Doug O'Laughlin** [55:50] It's like there- there's, like, essentially a B-roll channel that they'll, they'll, like, sometimes play in public spaces. **Swyx** [55:54] Sure. **Doug O'Laughlin** [55:54] And you will just find yourself, like, being engaged with it. Like, there are certain things that just ... It works. So, yeah. Sorry, that's completely off. Uh, but, but I wonder ... This Claude Code pill is very powerful for me. I believe it will change- **Swyx** [56:09] Yeah **Doug O'Laughlin** [56:09] ... how it all works. It'll shift all of that over massively, the, the chart. But it's just really weird because if you didn't pay any, like, human cognition to get there, I don't think you're gonna be a great reviewer. One of the reasons why you know what, what makes that, that human feat, that loop well is because once upon a time you did that, and you could make the- to me like, yeah, yeah, idiot, you, you're not thinking about this problem in this way. **Swyx** [56:33] Yeah. **Doug O'Laughlin** [56:33] You're missing this, this ... Like, you know, whatever. You're not considering this 90%, you know, like the 10% tail, something like that. **Swyx** [56:41] Yeah, yeah. **Doug O'Laughlin** [56:41] And so it's like, yeah, I know you said this, but, like, you know, the ... I- I know g- I know that I told you the valuation is the only thing that matters, but, like, it's also fraud. You can't do both, right? Like, if you think about, like, the analysis stuff, you have to know when your own in- personal embedded model is like, yeah, actually this one overwrites this one. And that, that's through learned experience, and I wonder if we're just reviewing, we won't be building and embedding those assumptions to understand judgment. **Swyx** [57:07] Right. Right, because you're just checking for mistakes rather than trying to do original thought by just doing the work. **Doug O'Laughlin** [57:15] Yeah. **Swyx** [57:15] Yeah, I think that's, that is, that is a danger. **Doug O'Laughlin** [57:17] Yeah. **Swyx** [57:17] Uh, yeah. **Doug O'Laughlin** [57:17] That, and that's what hygiene sounds like to me. You know, like, hey, it's really addicting to be like, you know, whatever, press the button over and over and over, but sometimes you do actually have to, like, think. Um. You know? Uh, so- So I think that that's, uh ... It's gonna be really interesting- **Swyx** [57:32] I, I mean, have you tried, like, uh ... So, so, I mean, the, the s- the way to model the sort of meta learning as, as element is, like, once a night you do a batch job of, like, look over everything I've done, like, like, extract some learnings. Um, you know, and OpenClau, I- I think one of the interesting things I really liked about it was this Heartbeat.md- **Doug O'Laughlin** [57:50] The Heartbeat, yeah, Heartbeat.md **Swyx** [57:51] ... and I feel like people aren't, like, excited enough about this 'cause, like, well, this is the first instance where, like, the agents are just always on, are, like, always living, always reflecting. **Doug O'Laughlin** [57:59] Yes. **Swyx** [57:59] And- **Doug O'Laughlin** [58:00] Like, what is it? Sold.md too? Or- **Swyx** [58:02] Sold I think is much more for character and like- **Doug O'Laughlin** [58:04] Yeah **Swyx** [58:04] ... uh, whatever. But, like, yeah, Heartbeat is- **Doug O'Laughlin** [58:06] Yeah. **Swyx** [58:06] Heartbeat. **Doug O'Laughlin** [58:07] Heartbeat is the con. Yeah. I mean, I think, yeah, that's a good way to put it. Yeah, and so, like, that's the powerful thing about all this stuff, is that, like, okay, yes, we know that the conte- like, it gets garbled. We know that Open, OpenClau doesn't always do everything you ask to it, ask it to do, uh, initially. But you can see the design patterns. Like, the Heartbeat.md's a perfect example. Can see the, the design patterns where it's like, well, you know, is all of our tasks every single day actually us having this, like, genius thing, or do we, like, sit down in a single session, finish a single project, get up and get some coffee, then come back? **Swyx** [58:39] Hmm. **Doug O'Laughlin** [58:40] If it's that, and you could just fuck- you can make the Heartbeat.md consider the, like, the session to session, and like, hey, meta learnings, all this stuff, and it's only specialized and focused on one form of doing something. So it actually does have a context of all the ... Like, let me ... I'm thinking, like, a customer service agent or something like that. It does have the context. In fact, it can look at every single time it's ever happened. That's actually information and context no human could ever hold. You're like, wait, that, that feels like AG- like, like, that's, uh, effectively good enough to do a huge information task- **Swyx** [59:11] Yeah, yeah **Doug O'Laughlin** [59:12] ... and have enough context, and be able to fetch it, and maybe, like, there would be some verification to make sure it doesn't just totally mess it up. But that, to me, feels like a design pattern that you can build something on. And so that's the, that's the vibe, is that we've hit some capability that you can, you can do ... You can build these much bigger blocks now, and those bigger blocks are not just, like, this single line of code. It might actually be a business. It's kinda crazy. Like, I, I wouldn't have put myself as AGI pilled. I think 4.5 is, like, actually- **Swyx** [59:40] I think my own timelines have moved up a lot. ### AI Economics **Doug O'Laughlin** [59:41] Yeah. **Swyx** [59:42] Um, are you guys watching GDPVal? **Doug O'Laughlin** [59:44] I ... To the best that I can, but I'm feel like I'm mostly just trying to- **Swyx** [59:48] No, no, no. So d-d- to me, when GDPVal came out ... So, uh, for, for, I, I, I mean, I, I, I'll just- **Doug O'Laughlin** [59:54] Yeah. **Swyx** [59:55] GDPVal is, like, an, basically a, like, a, a broader suite bench, let's call it, where it's, like, applied on every disci- every profession that is white collar that you can model, and it is above, like, something like 2% to 5% of GDP, something like that. That's why it's called GDPVal. And they, they had human experts do the tasks and, a- as well as GPTs, and here's the results, right? Like, um, and where 50% is parity with industry expert. **Doug O'Laughlin** [1:00:19] Yeah. **Swyx** [1:00:20] Um- **Doug O'Laughlin** [1:00:21] Coin. Yeah. **Swyx** [1:00:21] C- coin flip, exactly. Where ... So, like, you can see the, the nice, uh, increase from 4o to Opus 4.1, and since then, obviously 5.2 and, uh, Opus 4.5 have already exceeded. We're at 70-something now. **Doug O'Laughlin** [1:00:33] Yeah. **Swyx** [1:00:33] Which means models are consistently better than industry experts- **Doug O'Laughlin** [1:00:36] Yes **Swyx** [1:00:36] ... at c- at these things. **Doug O'Laughlin** [1:00:37] Yeah. **Swyx** [1:00:37] So to me, like, this is the AGI. **Doug O'Laughlin** [1:00:40] Definition, isn't it? **Swyx** [1:00:41] Yeah, yeah. This is, this is- And so, like, I think, I think the problem though, yeah, I would say that that is the definition, sure. Um, so the thing that's crazy is 'cause there's, like, this ASI element that people are, like, really, really focused on. **Doug O'Laughlin** [1:00:51] Wait, we're moving to goalposts. **Swyx** [1:00:52] Yeah, we're moving to goalposts, but I'm like, bro, I ... The, the goalposts, like, I, I, I mean, we'll see if this is actually the machine god, and Shogoth will come and talk to us and vibrate on our same- **Doug O'Laughlin** [1:01:01] But what if, if I do think so? **Swyx** [1:01:02] I wasn't ... Okay. Uh, I, I, I don't ... I, I'ma be honest with you, I'm very open. I will change my mind often. I'm not ... This is not something I feel intuitive in my gut today. Maybe it's the next, next X thing, but when it comes to, like, the, the, the GDPval version of this, yes. **Doug O'Laughlin** [1:01:16] Yeah. **Swyx** [1:01:16] This is, this is- **Doug O'Laughlin** [1:01:17] Do white collar work **Swyx** [1:01:18] ... literally the white collar work, which is- **Doug O'Laughlin** [1:01:19] And most of the, most of the **Swyx** [1:01:21] ... very boring **Doug O'Laughlin** [1:01:22] ... world knowledge more ... Like, like, actually it's almost all, not almost all, but it's a huge portion of all of work in the world. **Swyx** [1:01:28] Yeah. **Doug O'Laughlin** [1:01:29] It's like now we just made ... Like, my favorite stat is, like, once upon a time 90% of people were farming, right? Now today less than 1% of people are farmers. It's kinda like this crazy shift where technology's gonna massively change the relationship with all of that, and it's gonna be, like, this 99.1 thing. I don't know if it'll be quite that drastic or whatever. Maybe, you know, everyone's just doing leisure. So far my experience is everyone just works harder. That's been my experience. But it just, it just feels like a massive moment's happened, like, the, the steam engine's invented, and, you know, the, the trains are here. And, and- **Swyx** [1:02:02] Yeah **Doug O'Laughlin** [1:02:02] ... everything's gonna change in knowledge work, and it's kinda crazy. Yeah. **Swyx** [1:02:07] The ... This, this is an economic cycle, uh, from my macro days that I'm ... I c- I can't remember the name. I can look it up. But it's basically, like, there's the stages of economic development where, like, your, your economy starts out majority agriculture. Then it discovers, like, manufacturing. **Doug O'Laughlin** [1:02:22] Mm-hmm. **Swyx** [1:02:23] Then it discovers white collar work. Then it discovers ... They, they build, like, a very mature financial sector. **Doug O'Laughlin** [1:02:28] Mm-hmm. **Swyx** [1:02:28] And, like, though these are, like, like, a layer cake that all declining over time- **Doug O'Laughlin** [1:02:31] Yeah, yeah **Swyx** [1:02:32] ... with, and then the new thing's increasing. So at ... My, my theory is, like, there's this, like, fifth layer that's, like, has to open up that starts to happen because I do fundamentally believe we'll just invent new work. **Doug O'Laughlin** [1:02:41] I do believe that. **Swyx** [1:02:42] Yeah. **Doug O'Laughlin** [1:02:42] 100%. Like, um, humans are very adaptable. That's, like, my favorite thing that I've learned. Um, you're able to adapt to gone, like, coldest, coldest place in the entire world, the warmest place. Humans are in every latitude. That's in a physical sense, but I think we're gonna find a way to make utilization go up. Um, but we'll d- we'll, we'll invent more work for sure. But, um, I think the thing that's crazy is just, like, things change so quickly, and that five to 10-year period, like, 10-year gap can be drastic and crazy, and that's just societally wild. **Swyx** [1:03:15] But yeah, it's, it's happening in our lifetimes. **Doug O'Laughlin** [1:03:16] It's happening in our lifetimes. Like, it's, like, happening, like, right now. Like, it's, it's just, it's really crazy. It's, like, very ... And, like, h- No, so this is, like, a complete side task on- **Swyx** [1:03:25] Yeah, yeah **Doug O'Laughlin** [1:03:26] ... I'm, like, really curious of when we start to see it in a much bigger way in the real economy. That's, like, my, my, my pet- **Swyx** [1:03:31] Yeah. Where ... Why is it not showing up in GDP yet, right? **Doug O'Laughlin** [1:03:33] So there's gonna be ... You know, some people are gonna be like, "Oh, you know, the facts. The facts, the internet, same thing, information transfer," whatever. I think I'm actually scared for a third worst thing, which is, like, now, now this is a complete crackpot theory. Please don't hold me to this, um, internet. Um, but what if AI is mashl- massively deflationary? **Swyx** [1:03:51] Yeah. **Doug O'Laughlin** [1:03:52] And, and, and also I think, uh, one of the more interesting conversations I've had in a bit is, like, what w- GDP was invented once upon a time as a way to figure out how much we could divert, you know, normal economy away just to war during World War, like, one or two or something like that. **Swyx** [1:04:08] Okay. **Doug O'Laughlin** [1:04:08] My spiciest take is I feel like GDP itself is gonna be very, very challenged by AI because information work ... Yeah, so how we, how we capture it effectively is all of an economic good, and then the service hours divided by hours, okay? So there isn't, like, a widget to widget difference, but in theory if we could break all of information work down into units, we're gonna have a lot more information work for sure, like, more work will be done. I don't know what the value of that's gonna be. Is it gonna be so much increase in supply it's deflationary? That seems to be, like, a real concern. **Swyx** [1:04:45] It, it's possible. **Doug O'Laughlin** [1:04:46] Yeah. **Swyx** [1:04:46] And then we'll figure out how to use it. But, like, there may be a Great Depression of AI- **Doug O'Laughlin** [1:04:52] Yeah **Swyx** [1:04:52] ... where, like, we figure it out. **Doug O'Laughlin** [1:04:53] Yeah. Well, I, I wrote this whole thing about railroad stuff 'cause it's my favorite- **Swyx** [1:04:57] Okay **Doug O'Laughlin** [1:04:57] ... my, my favorite capital cycle. **Swyx** [1:04:58] Is this on Fab or? **Doug O'Laughlin** [1:04:59] It's on Fab, yeah. **Swyx** [1:05:00] Okay. **Doug O'Laughlin** [1:05:00] Um, I can't rem- Uh, it's, like, railroad Fab. Uh, it's about all the railroad stuff over time, okay? Pretty much because, um, we're, like, everyone was first looking for the internet. We've well massively passed the internet in terms of the absolute size of the build-out. **Swyx** [1:05:16] Mm-hmm. **Doug O'Laughlin** [1:05:16] It's not even close. Like, we- **Swyx** [1:05:17] What, what numbers are, are you thinking of? **Doug O'Laughlin** [1:05:18] Um- **Swyx** [1:05:19] Like, what, what could- **Doug O'Laughlin** [1:05:19] I think a, I think a trillion was, uh, a trillion all in was essentially the real dollars for- **Swyx** [1:05:24] Okay **Doug O'Laughlin** [1:05:24] ... version, and I think we are well past. Like, we, like, whatever, this year, and it's cumulative, right? We were well past that. I think railroad, the reason why it's so interesting is 'cause, um, honestly it's way crazier. But, but prob- part of the problem and craziness of it too is, like, railroad was literally, like, one of the first added layers of the layer cake if you think about it. Before it was agriculture, and railroad was like, "Okay, well, how do we move this agriculture around faster?" **Swyx** [1:05:47] Yeah. **Doug O'Laughlin** [1:05:47] Um, and then banking got ... I, I, I kid you not, like, one of my big takeaways is banking effectively got invented by railroads. **Swyx** [1:05:54] Oh. **Doug O'Laughlin** [1:05:54] Because there's no need to- **Swyx** [1:05:55] You have to finance it **Doug O'Laughlin** [1:05:55] ... finance it. Yeah. So much money was needed that, like, effectively 85% of all paper or whatever was essentially just railroad debt. Um- **Swyx** [1:06:02] Yeah. Um, one of my favorite anecdotes was before there was a federal bank, uh, a federal reserve, Andrew Carnegie was the federal reserve. **Doug O'Laughlin** [1:06:09] Yes. Yeah, there were individuals. Yes. Um, yeah, and so all this stuff, uh, so it's, like, the whole thing, I kind of did some work on the Gilded Age, all this stuff, but, like, my takeaway is, like, that was a really interesting cycle because it was so big and took so long to deploy. It actually was 45 years of, like ... There's three cycles actually. There's three boom-busts. **Swyx** [1:06:29] Damn. **Doug O'Laughlin** [1:06:29] Um, I don't know if it'll be quite that long. All the cycles kind of collapse. **Swyx** [1:06:33] Yeah. **Doug O'Laughlin** [1:06:33] Um, are ... that- **Swyx** [1:06:34] Because, you know, information- **Doug O'Laughlin** [1:06:36] Information **Swyx** [1:06:36] ... moves around faster anyway. **Doug O'Laughlin** [1:06:37] Exactly, yeah. And so you, you have all this stuff where I think it's gonna happen faster, but, like, I would be really shocked if it was all in one go. That's my vibe. **Swyx** [1:06:44] Yeah. **Doug O'Laughlin** [1:06:44] Um, where it's like it's all in one instantaneous up down. I think it's gonna look like mult- some multiple cycles. So yeah, kinda just wrote about railroads. The... There was like a baby railroad cycle, then there was a huge railroad cycle. The modern railroad was invented out of it. That's like my favorite analogy for this because, like, I think it was like GDP percentage of CapEx each year were like high single digits for... So sustained for like 10 years. **Swyx** [1:07:08] Yeah. **Doug O'Laughlin** [1:07:08] But what's crazy is, like, that amount of spend is like we're, we're, like, well on track for that. It- **Swyx** [1:07:13] Did you do percent of GDP? 'Cause I think that's- **Doug O'Laughlin** [1:07:15] I think so **Swyx** [1:07:16] ... the, the way you make it convertible. Uh, Stargate itself, 2% of US GDP. Um, and, uh, I mean, it's gonna go up like **Doug O'Laughlin** [1:07:27] Yeah. Yeah, that's a... Yeah. And it's not all gonna be in one year, right? But, um, it's... Okay, so yeah. So, so total CapEx four- it was 4.8% of GNP and 25% of total gross fixed capital investment. **Swyx** [1:07:43] Okay. **Doug O'Laughlin** [1:07:43] So 25% of investment every year and four or 5% of GNP- **Swyx** [1:07:48] Yeah, I think we're there. **Doug O'Laughlin** [1:07:49] Yeah. **Swyx** [1:07:49] You know, Star- **Doug O'Laughlin** [1:07:49] Yeah, we're- **Swyx** [1:07:49] Stargate plus Entropic plus whatever **Doug O'Laughlin** [1:07:51] ... yeah, yeah. We're, we're right there. **Swyx** [1:07:52] XAI. **Doug O'Laughlin** [1:07:52] Yeah. So we're at- **Swyx** [1:07:53] Meta **Doug O'Laughlin** [1:07:53] ... the railroad buildup- **Swyx** [1:07:54] Yeah, yeah **Doug O'Laughlin** [1:07:54] ... which is like at one point in like... But the thing that's crazy- **Swyx** [1:07:56] And we, we, we should exceed it like- **Doug O'Laughlin** [1:07:58] Probably, yeah **Swyx** [1:07:58] ... yeah. No, not, not probably, like we should. **Doug O'Laughlin** [1:08:00] Yeah, okay. **Swyx** [1:08:01] Like this is bigger. **Doug O'Laughlin** [1:08:01] Yeah. **Swyx** [1:08:03] Okay. **Doug O'Laughlin** [1:08:03] I, I'm, I'm like I, I would like to say yeah, sure. I, yeah, we, we will do it. I, I'm worrying. I'm like, "Dude, where are we gonna get all the money?" **Swyx** [1:08:10] That's like such a, like, the pedestrian concern. **Doug O'Laughlin** [1:08:13] Yeah. **Swyx** [1:08:13] It, it, it's not a pedestrian... I mean, that's what happens every capital cycle. **Doug O'Laughlin** [1:08:16] I worry like we must- **Swyx** [1:08:17] Going hands in the Middle East- **Doug O'Laughlin** [1:08:18] We- **Swyx** [1:08:18] ... they'll flip the thing **Doug O'Laughlin** [1:08:19] ... we must, we must... The w- wh- this happens every single time. This the reason why th- like bubbles happen, right? Is like we essentially get so big, where it's like this must be built. It doesn't matter the price. **Swyx** [1:08:27] Yeah. **Doug O'Laughlin** [1:08:28] And then all of a sudden we look at it, where it's like, "Ooh, that was a steep-ass price." Um, but I think, I mean, the thing I think about this is like how I think about the big picture is there is a demand curve and a supply curve, and we have no idea when they cross. They will cross one day. And every single year the demand... The, we're finding that demand curve, and then the supply curve we're just like we're doing our best to deploy it. And I think for me, like I don't know when that number is. I'm not... I don't wanna say a number go up forever 'cause I feel like that's like g- intellectually dishonest. **Swyx** [1:08:56] Yeah. **Doug O'Laughlin** [1:08:56] But Claude Code, for me, is the first time where I'm like... And we're bringing it all back together, where you're like, "Demand go up so much." I am now guzzling as an individual. Like for example, I'm, we're, we're off... I'm off max. It's not enough. It's not even anywhere near enough. Like- **Swyx** [1:09:10] I mean, some people buy like five maxes and then- **Doug O'Laughlin** [1:09:12] Yeah **Swyx** [1:09:12] ... they rotate. **Doug O'Laughlin** [1:09:12] Well, yeah. So, so I, I... So I'm on fast- I'm on fast with one million on API, which is that is like an addiction level, if any sense. Um, but yeah, I, I, I really think, um, it's the first time where like, okay, well, actually, how much is this worth to me on a yearly basis? I think it's like 20 to $30,000 easily. **Swyx** [1:09:31] Yeah. **Doug O'Laughlin** [1:09:31] Like if not more. Like I don't understand, like what's the c- like I can't price it. I have no idea the elasticity. **Swyx** [1:09:37] Yeah. You pay for a perfectly compliant junior analyst. **Doug O'Laughlin** [1:09:39] Yeah. **Swyx** [1:09:40] Right? And so- **Doug O'Laughlin** [1:09:41] What's able to- **Swyx** [1:09:41] What's that cost? Like- **Doug O'Laughlin** [1:09:42] Yeah **Swyx** [1:09:42] ... 90K **Doug O'Laughlin** [1:09:43] ... that's able to work in parallel. **Swyx** [1:09:45] Yeah. **Doug O'Laughlin** [1:09:45] Like you can have 100 of them. It's kinda crazy. **Swyx** [1:09:47] Yeah. **Doug O'Laughlin** [1:09:47] Yeah. **Swyx** [1:09:47] Oh, yeah. **Doug O'Laughlin** [1:09:47] So it's- **Swyx** [1:09:49] It's a skill issue if, if you cannot manage a junior analyst- **Doug O'Laughlin** [1:09:52] 100% **Swyx** [1:09:52] ... that is 20K a year. **Doug O'Laughlin** [1:09:53] Yeah, 100%. **Swyx** [1:09:54] Which like, I mean, okay, like, you know, skill issue is like it's your fault. But no, like we have to learn how to do this. It's, it's- **Doug O'Laughlin** [1:09:59] Yeah, it's like- **Swyx** [1:09:59] It's at, it's two, it's three months old. **Doug O'Laughlin** [1:10:01] Exactly. It is two month- That, that's the correct way of putting it. It's like- ... it's a... it was definitely a skill issue that you didn't know how to get, like, your settings on your iPhone to work. We know at one po- one point in time. But like in the very first month of us having it, no one's gonna be like, "Yeah, you idiot, you rube. You don't know how to use your n- completely new technology that gave birth last month." **Swyx** [1:10:21] Yeah. **Doug O'Laughlin** [1:10:21] Um, I think it's just about a ti- it's a bit of time, and it's like kind of interesting 'cause you're, like, watching. I mean, what's cool is that if you're like on this absolute bleeding edge, you get to see the design patterns, like blossom in real time. And like, um, we have this like really old, uh, older guy who's like been through the history of technology since like forever back then. He's like one of the most interesting, uh, intelligent people at SemiAnalysis. And he talks about how- **Swyx** [1:10:44] Wait, who is it? **Doug O'Laughlin** [1:10:45] Uh, Tanch. Okay, so like he said this like... We, we, we had the conversation one time. He was talking about, like, early internet, how like it wasn't actually sure if the browser was gonna win. It was like a remote web file service. Some people thought like, "Well, it just... I'm just gonna reach in and play with someone else's web files remotely," right? Who knows, right? And that kind of, you know, it kind of is a remote web file. Who, who the hell knows? They were design pattern searching back then, and I think we're at that again, where all the design patterns are open and it's like really interesting 'cause there's many different ways this could go. **Swyx** [1:11:16] Yeah. **Doug O'Laughlin** [1:11:16] And we're gonna have to kinda collectively agree what's the best set of hygiene, set of design patterns, what's the level of abstraction, and then like all the rest of how much SaaS it will disrupt. All, everything else, who the hell knows? But you get to watch it, like front row seat right now. ### Death of IDE **Swyx** [1:11:32] Yeah, yeah. I, I mean, m- my biggest one, uh, and I, I'll, I, I do wanna bring it to Semis in, in a little bit, but, uh, is the IDE. Two months ago, they, uh... we had Stevie Agee f- from Gastown talk about how 2026 will be the year IDE died. And I al- like two weeks ago, three weeks ago, I recently was like, "Shit, he's absolutely fucking right." **Doug O'Laughlin** [1:11:52] It's over. **Swyx** [1:11:52] Yeah. **Doug O'Laughlin** [1:11:54] I, I'm really wondering too because like IDE... So, so I think that same... My, like my... The reason why I'm so excited about this is I get to like... Look, I never... My daily driver was never an IDE, right? My daily driver was like Bloomberg or Excel or something like that. But I have a personal belief. It's not happening yet because we're not quite there in the maturity curve, like software's just gonna be first. But the year like of, you know, Excel is dead for finance- **Swyx** [1:12:20] Yeah **Doug O'Laughlin** [1:12:20] ... and like it's, it's, it's- **Swyx** [1:12:21] Excel is the IDE for analysts **Doug O'Laughlin** [1:12:23] ... it, it, Excel is the IDE for analysts. Bloomberg is the IDE for analysts. Like, I believe every one of these IDEs are done. It's dead, over, gone- **Swyx** [1:12:30] Yeah **Doug O'Laughlin** [1:12:30] ... and dead. I just think it's... Why? Why? It doesn't like... Just imagine the concept of you... Like, I remember when I learned Bloomberg, I had to like watch videos to learn about all the random- **Swyx** [1:12:40] The series **Doug O'Laughlin** [1:12:41] ... subfolders- **Swyx** [1:12:41] Yeah **Doug O'Laughlin** [1:12:41] ... keys, how to use this, how to use this one, you know, the tactic knowledge of using this function versus that function. That's like- ... crazy to think about. That is like, that is like horse and buggy. Mm-hmm. Okay? The agent with the information that can perfectly retrieve and analyze stuff is gonna have the ability to, to s- to pull that all together in a better UI than it was with no legacy whatever. I think all of that is dead. And like, um, this is why I'm like my, my spiciest take of all is like M- Microsoft has a lot to lose. Mm. I think they have the most to lose of everyone. Yeah. Because Excel is a human IDE for information work that's generalizable. So is PowerPoint, so is email. Those are the base core level of abstraction that decided to be broadly generalable. But I, I just don't think that matters anymore. I think Claude Code or Cowork or whatever is gonna be the year that, like it will destroy all of that. All that information work that, that where you sat every single year, it's, it's over. Um, I think that that's the, the one that's like more shocking and scary that like people don't believe. Like, I believe in my stomach with conviction because I have already had that moment for me. Yeah. I will never make a chart in Excel again. I actually will not. Wow. Yeah. Um. It's hard to let go because I, I have so much like ingrained knowledge of, of like manipulating things directly in, in Excel. Uh, Bloomberg, I have ... So, um, there's no way that you would know this, but like my very first startup was a, an attempted Bloomberg killer. Mm-hmm. Uh- Oh, no ... Sentieo DK Office. I remember. I remember. Are you Remember Office? Yeah. Yeah. No, no, I- Oh, yeah, yeah. You are one of the few. You are OD- I was a Sentieo, I was a Sentieo customer. You were a Sentieo customer . I was a Sentieo customer, like I rolled in, dude. I remember- Uh- ... like how dare they acquire Sentieo. I had a, I had a patent. I, we filed- Uh-huh ... for a patent for similar tables. Anyway, one of my conclusions was like Bloomberg is just like three things. It's, it's, it's Slack, and it's the journalism, which is w- amazing, and then it's the, the data feeds. It's actually not really the UI. Yes. Yeah. It's not really the UI. Yeah. But I, I think for the first time in my life where I just think that like I just wonder if that like ... Okay, if you can get ... Oh, obviously, it's ... So you're telling me that the future, the undisputable future is just like it's IB and nothing else, and then like a terminal that like- ... types in some stuff. I think that if you are, uh, marginal and on the curious and not hyper-interconnected, which I would argue that I am at SemiAnalysis. Uh, like for example, I'm trying my absolute best to just rip Bloomberg out. We're going to FactSet API. Like all- Yeah ... all-in API with a Claude Code is my belief of the future. Hey, verifiable data source that you trust. Yeah, yeah. Hey, scale- But you guys can do it. You- Yeah. For traders, we're so sort of- For, for tr- no way. Yeah. I understand. Like there's an information network- Yeah ... that's like outside of this, it's- Then you do deals in IB. Yeah. Right? Which are tracked by the regulators. Yep, 100%, 100%. Yeah. It's totally ... I completely- But as an analyst, yes. For as a analyst, yeah. And so like, but I just think that like, okay, that doesn't really ... So you're right, the core cashflow cal thing will continue onward. But like each iteration of this AI thing, I was like, "Yeah, I'm still gonna be using Bloomberg," right? This is the first time I was like, "Actually, no, I don't care anymore." Yeah . Um, the IB's, my util- utils of like marginal value from, from IB is like now outweighed by how like clunky this is, and I wanna just make some charts. Mm. Right? And so that's- Like, like i- immediately you save 10 to 20K- Yeah ... for switching down. Yeah. There you go. It's amazing. Yeah. Uh, by, by the way, what was your Claude Code end of year prediction? 25? Yeah. I want you to know, I sandbagged the everliving shit out of that. Oh, okay. I, I just believe 25 is very ... Like I, like because like I ... The rate it's on is like, whatever, 50 or something like that. But I think, I feel, I wanted to give a 95 confidence interval. Mm-hmm. I think 25 is within the 95 confidence in- interval. Sure. So it's, so it's between 25 and 50. Something like that, yeah. Yeah. It's, it's just absurd. But, um, you know- And it, it could also be Codex. Code- Are you also watching Codex? Yeah, yeah, yeah. So, so to be clear, uh, I, I'm actually even willing to comment on that 'cause like I know we've gotten a lot of shit of being Codex haters. Yeah, I, I think, I, I ... And by the way, when I put Claude Code, Codex, Agent, whatever all in percentage that we can publicly see, I would argue the fr- the ratio outside of that's probably gonna be higher too. But whatever. Yeah, I think together. Um, yeah, we're watching Codex. Um, I actually think Codex is ... Codex is pretty good. 5.3 I think. So we had the whole thing, I, 'cause like I wrote most of the articles. Like, oh, token efficiency, the context rot, all this stuff. Is, is it the same one or ...? Um, yeah, yeah. It's in the bottom. It's in the, it's in the, the paid section. Okay. Like I said, but like TLDR, I was like, "Well, you know, the reason why Claude Code is so good, Anthropic is so good is 'cause all of this token efficiency, the token efficiency is better than ChatGPT," all this stuff, blah, blah, blah, blah, blah. And then like 5.3 Codex came out, and it's like, yeah, that, that completely doesn't matter anymore. They're like, they're, they're so back. Um, I really think 5.3 Codex is awesome, um, in, in coding though. But you can watch it. Like the reason why I like Opus 4.6 so much is because when I'm using it, I'm using it for like coding in is the way I interact with it, but I'm using it for broad generalized information work, right? But I think the difference is Codex wants to code 'cause it's RL'ed to be so good at coding to win on SWE-bench, that like you're trying to use it for general information. Be like, "Hey, I'm trying to ... Can you go research and search all these websites?" And like I don't even think they have web search in it or whatever. Maybe, yeah, maybe you can give it to API or whatever. But it's like great, I'm ma- I'm scrap- I'm creating a piece of scraping software to go look at these websites. I was like, "No, no, no, no. Just like, just ingest tokens of what's on the websites." Like, okay, great, I'm still ... Like it's so coding pilled in the RL that I think, um, it isn't generalizable- Mm ... in the way that- Mm ... that 4.6 is, where it's like, oh, I could have it, I could have it make some rubric or do some research or do something like that versus Codex. It's very, it's very co- Yeah. It, coding Codex is coding pilled. And so that's what ... Uh, but I, I am very optimistic actually on Codex, and we do track them, um, quite a bit. I, I ... You can ... They have a meaningful amount of thing. Share, you could see the Bloomberry. They have the, the sh- the, the chart bloomberry.com. So the Claude Code definitely is in the lead. Um, but I think the part of it too is like the, the like to like comparison. There's a ratio of Codex that's not available 'cause it doesn't sign off every commit. It does sign off on pull requests. That, that ratio is much closer, so all the OpenAI people like Rune will tell like, "Blah, blah, blah, we're not accounting for it." Yes, we didn't account for it, but like- Okay ... I, I, I think Codex is better. I think there is some real problems and issues, but I bet you the second that they have a new pre-train with the RL, 'cause the RL stack on Codex 5.3 is amazing, like it's very coding code. Um, that's when, that's when the, it, it flips over. And yeah, look at the other players here. It's just like- I mean, my favorite thing is that how GitHub Copilot is, like, number one and, like, I've never heard of, like... Do you know anyone who uses GitHub Co- Copilot? **Swyx** [1:19:05] Yeah, t- look, look. Okay, that- that's, that's a bubble talking, right? **Doug O'Laughlin** [1:19:08] Okay, that's a bubble? **Swyx** [1:19:08] Yeah. **Doug O'Laughlin** [1:19:09] Yeah, that's, that's the, that's the SF bubble, like the- **Swyx** [1:19:11] Yeah, yeah **Doug O'Laughlin** [1:19:11] ... the- **Swyx** [1:19:11] Like, you know, there's, there's, like, all these Windows users, and you don't talk to them, right? Like, uh, you do, but we don't in, in San Francisco. And, like, that's just, that's fine. That, that's definitely bubble talking. But yes, Copilot has a billion in ARR, I think, at least. **Doug O'Laughlin** [1:19:24] Yeah. **Swyx** [1:19:24] Uh- **Doug O'Laughlin** [1:19:24] But what's crazy is Claude Code is, has a ratio... They, their attribution of Claude Code, uh, in ARR is 2.5. **Swyx** [1:19:31] Yes. **Doug O'Laughlin** [1:19:31] So that, on this, the daily install counts, right- **Swyx** [1:19:35] Mm-hmm **Doug O'Laughlin** [1:19:35] ... is an order- **Swyx** [1:19:36] Which, which is, by the way, just the VS Code extension, right? **Doug O'Laughlin** [1:19:38] It's just the... Yeah, I know, I know. **Swyx** [1:19:39] That's not even the default way to use- **Doug O'Laughlin** [1:19:41] Yeah, yeah **Swyx** [1:19:41] ... Claude Code. **Doug O'Laughlin** [1:19:41] Yeah, you're right, you're right, you're right. Well, CLI, NPM, DOM is another way to track it. **Swyx** [1:19:46] Mm-hmm. **Doug O'Laughlin** [1:19:46] But I think they have, like, their own cust- their own installer now. Anyways, all in all, definitely heard, understand it's very hard for us to, like, actually track it. But, like- **Swyx** [1:19:56] I, I'm not, I'm not gonna say it. I'm just, like... I, I think Codex, uh, a big thing I'm watching as well is Codex back, uh, because they- **Doug O'Laughlin** [1:20:03] Cod- **Swyx** [1:20:03] ... they reported, like, Jan to Feb they doubled users. **Doug O'Laughlin** [1:20:06] Yeah. **Swyx** [1:20:06] Okay. **Doug O'Laughlin** [1:20:07] So, so I have some s- not skepticism, just 'cause they have such a big ChatGPT portal that could be like tri-atlas. Like, the modal that pops up- **Swyx** [1:20:16] Mm **Doug O'Laughlin** [1:20:16] ... can really move big users. Like, they're not quite a google.com in terms of having so much ability to, like, siphon off users off, but I, I wonder, like, the like to like. But, but that's, like, my skepticism. But, like, great- **Swyx** [1:20:29] Oh, uh, I, I have an answer for that. Uh- **Doug O'Laughlin** [1:20:30] Okay **Swyx** [1:20:30] ... Alexander Ambracos was just on the Lenny Pod saying that they actually haven't invested enough in the web experience. So, like, I, I, I think, I think the attribution for that is zero. **Doug O'Laughlin** [1:20:39] Okay. **Swyx** [1:20:40] Yeah. **Doug O'Laughlin** [1:20:40] I guess I just saw a modal, be like, "Oh, try Co- try Co-" I mean, but the, but a modal isn't- **Swyx** [1:20:45] Yeah. **Doug O'Laughlin** [1:20:45] And then to be clear, Codex in the, in Mac is great. I'm actually... I mean, like- **Swyx** [1:20:50] Yeah, yeah, it's... They, they actually with the app. **Doug O'Laughlin** [1:20:52] The, the- **Swyx** [1:20:52] The, the app launch **Doug O'Laughlin** [1:20:53] ... the, the app launch is actually pretty good. So, so yeah. And, and I think I'm pretty bullish to that, honestly, especially for coding, 'cause it's, like, very coding code. I just can't get it to work as well for non-coding stuff. **Swyx** [1:21:04] Then my, you know... Uh, you use Conductor? **Doug O'Laughlin** [1:21:06] Uh, no, I've not used Conductor. **Swyx** [1:21:07] Oh, okay. Uh, I, I thought I heard you say on a podcast that you use- **Doug O'Laughlin** [1:21:09] No, I've not used Conductor. **Swyx** [1:21:10] But so, so basically, like, the, the, their argument for any la- any first party app is that they're only gonna prefer their own first party piece. **Doug O'Laughlin** [1:21:18] Yes, 100%. **Swyx** [1:21:19] Which, like- **Doug O'Laughlin** [1:21:20] They're already- **Swyx** [1:21:20] ... no play **Doug O'Laughlin** [1:21:20] ... they're already doing it. Like, like, I feel like this is how they're gonna differentiate, right? Like, they're going to- **Swyx** [1:21:25] Well, then, then you have a Conductor where you can use Codex and Claude Code for different tasks a- as you see fit, and so this is the, the clean superset. No? **Doug O'Laughlin** [1:21:34] In theory, yeah. But, but, but, I mean, this is, like... Okay, so, so, so then you can argue this is the clean superset. It feels kind of like, um... I guess my design pattern on that is really skeptical- **Swyx** [1:21:45] Mm **Doug O'Laughlin** [1:21:45] ... of building on top of something that is growing very quickly and has- **Swyx** [1:21:48] Oh, yeah **Doug O'Laughlin** [1:21:48] ... all the money and whatever. **Swyx** [1:21:49] Yeah. **Doug O'Laughlin** [1:21:49] Like, I just think my favorite one is, like, platform as a service, if you remember that one. It was, like, infrastructure as a service, platform as a service, SaaS, software as a service. **Swyx** [1:21:57] Mm-hmm. **Doug O'Laughlin** [1:21:57] And then like, "Oh, this platform as a service." And it's like, it always just ends up being in the middle, so it just gets eaten by one or the other. **Swyx** [1:22:03] Mm. **Doug O'Laughlin** [1:22:03] I, I think of that, like, middleware layer, unless if it's a really, really, really compelling case, often dies. But that being said, in this moment, I agree. I actually u- I like to have them, like, review each other, like, having them yell at each other is really great. I might actually try this soon. I haven't used, I haven't used Conductor personally. **Swyx** [1:22:21] No? **Doug O'Laughlin** [1:22:21] Uh, I've mostly just been, you know, going deeper into the psychosis. **Swyx** [1:22:24] Yeah. And this is, as a c- former cloud analyst, very typical of, like, do you want a multi-cloud or do you want to go all into one cloud? And the, the, the classic argument for multi-cloud is, well, then you can use the best of everything. **Doug O'Laughlin** [1:22:35] Exactly. **Swyx** [1:22:36] But if you go all into one cloud, you can exploit- **Doug O'Laughlin** [1:22:38] Yeah **Swyx** [1:22:38] ... uh, the, the sort of minor features of everything. And, um, uh, you know, the, it makes a market, and you, there's no right answer for everyone. **Doug O'Laughlin** [1:22:46] Exactly. Yeah. Yeah, I mean- **Swyx** [1:22:48] It's- **Doug O'Laughlin** [1:22:49] Yeah **Swyx** [1:22:49] ... tha- that's one, that's one of those things where, like, even the really small percentages in AI still really matter because they're, they're huge, and, like- **Doug O'Laughlin** [1:22:55] Yes **Swyx** [1:22:56] ... people are very happy, very productive- **Doug O'Laughlin** [1:22:57] 100 **Swyx** [1:22:57] ... make money. **Doug O'Laughlin** [1:22:58] Okay, it's good to be an analyst in the space because, um, it's fun to keep up with it, right? Like, I agree. Like, I, I think everything, everything- **Swyx** [1:23:05] Like, we like the horse race, right? **Doug O'Laughlin** [1:23:06] Yeah, I like the horse race. **Swyx** [1:23:07] Number one, number two. Ooh. **Doug O'Laughlin** [1:23:08] Yeah, yeah. But- **Swyx** [1:23:09] You know? **Doug O'Laughlin** [1:23:09] Yeah. No, no, I know. But then you have to... Your brain also have to be, like, number two is really big, too. Um, and then I, I just think, like, for me, someone who likes the history of all this, like, like, likes history of innovation and competition and disruption and stuff, likes new technology, it's like a very fun time to be following this stuff altogether. **Swyx** [1:23:27] Tech during the, like, 2017 and 2020 years, so boring. **Doug O'Laughlin** [1:23:32] Yeah. **Swyx** [1:23:33] Yeah, at least for me anyway. **Doug O'Laughlin** [1:23:34] I thought it was pretty boring too. **Swyx** [1:23:35] Yeah. **Doug O'Laughlin** [1:23:35] Sorry. Uh, I was, I, I interrupted you in mid- **Swyx** [1:23:37] No, no, no, I, I, I remember what I was talking about, bro. **Doug O'Laughlin** [1:23:40] Okay. It's just fun time. It's a fun time to be- **Swyx** [1:23:42] It's, it's a good- **Doug O'Laughlin** [1:23:42] Yeah **Swyx** [1:23:42] ... thing. **Doug O'Laughlin** [1:23:42] Things are happening. **Swyx** [1:23:43] Okay, I wanted to transition to a little bit of a spicy thing, where you were on TBPN, and their title that they chose for you was, uh, "Doug O'Laughlin thinks Microsoft is out of AI." And woo, did you, did you not see this? **Doug O'Laughlin** [1:23:57] Okay, so, so, so I wouldn't say out of AI. No, I did s- Okay, so I didn't watch it. I never re-watch these things. **Swyx** [1:24:02] It's- **Doug O'Laughlin** [1:24:02] Okay, so how I think about it is- **Swyx** [1:24:05] But, like, it, you, you said things like- **Doug O'Laughlin** [1:24:07] I- **Swyx** [1:24:07] ... Microsoft is scaling back investment and- **Doug O'Laughlin** [1:24:09] Yep **Swyx** [1:24:09] ... yeah. **Doug O'Laughlin** [1:24:09] So, so, so it was the previous conversation I was talking about- **Swyx** [1:24:11] Yes **Doug O'Laughlin** [1:24:12] ... how Microsoft has the most to lose. **Swyx** [1:24:14] Mm. **Doug O'Laughlin** [1:24:14] They have the most to lose of everyone in the entire world, if you think- **Swyx** [1:24:17] 'Cause they're the, they're, they are- **Doug O'Laughlin** [1:24:18] They're, they're the software company **Swyx** [1:24:19] ... they're the horizontal software company. **Doug O'Laughlin** [1:24:21] If ChatGPT is dead, they're fucked. **Swyx** [1:24:22] Yeah, exactly. They're the horizontal software company that humans use their software to do information work. Okay? No... Like, I cannot paint a bigger target, okay? I cannot paint a bigger target. Um, and my, my- **Doug O'Laughlin** [1:24:33] So- so is Source. **Swyx** [1:24:34] Yeah. Uh, well, so, okay, that's another two... No. **Doug O'Laughlin** [1:24:36] Well, Microsoft is automatically too big for Source. **Swyx** [1:24:38] Yeah. **Doug O'Laughlin** [1:24:38] Yeah. But the other thing too is they have this Azure business. **Swyx** [1:24:41] Mm. **Doug O'Laughlin** [1:24:41] Um, I don't think they're completely out of the race. I'm, like, you know, it's a really great clickbait title, but the, the problem is the Azure business with OpenAI, right? You're essentially renting barbarians at the gate. You're, you're like, you know, this is r- ancient Rome, and you're like, "Hey, we need some extra guys, so we're gonna, we're gonna pay money for these barbarians to protect-" **Swyx** [1:24:56] The Golden Army. **Doug O'Laughlin** [1:24:57] Exactly. **Swyx** [1:24:58] From Game of Thrones. **Doug O'Laughlin** [1:24:58] Yeah, yeah, the Golden Army. And the problem is, like, each year they become more powerful, and then, and then at some point they're just like, "You know, we could just, like, you know-" ... scale these, these shitty walls. So like the, the... So that's the problem is the moat, the, the wall and the moats every year are getting more dilapidated as they continue to rent GPUs to, to the barbarians. **Swyx** [1:25:17] So it's just like Google, Yahoo again? **Doug O'Laughlin** [1:25:19] Yeah, yeah. **Swyx** [1:25:20] Like **Doug O'Laughlin** [1:25:20] It, it is exactly like that. And so it's just like this weird process where... That's a terrible setup too because what happens in the history of that is you have to choose one or another, okay? If you do either poorly, you're, you're like, you're like somehow in a third worst place. You either all-in become Azure, like maybe in the telecom era, right? 'Cause you're team T guy. You become dumb pipes, okay? That is the, the Azure becomes Ca- uh, what is it? Charter, right? **Swyx** [1:25:44] Ooh. **Doug O'Laughlin** [1:25:45] Yeah. Uh, but then, or the other version of this is you say, "No, no, screw these guys. I have to like reinvest back in and like essentially steal, copy their, their features and build up my moat." That means I need to stop investing in Azure for the stock. That really sucks because the stock is very much weighed on out-year Azure revenue. And meanwhile, if we actually had to value Microsoft ex Azure, the multiple would be really low right now. I think about that all the time. What would this trade ex, ex Azure? So this is just Microsoft. **Swyx** [1:26:15] Yeah. **Doug O'Laughlin** [1:26:15] Eight times earnings, 10 times earnings. Like, it was trading like that before actually. **Swyx** [1:26:19] Oh, geez. **Doug O'Laughlin** [1:26:19] And like, like remember the 2010s era when it went all the way b- down to like 10 times earnings in the Steve Ballmer era, and then it inflected outward as it did O- Oak- **Swyx** [1:26:28] With Azure **Doug O'Laughlin** [1:26:28] ... yeah, yeah, Azure and O365. Yeah, there we go. That's right. **Swyx** [1:26:32] Yeah. **Doug O'Laughlin** [1:26:32] Yeah. **Swyx** [1:26:33] Okay, I don't think you have the answer, but I just like... This is probably the most bizarre... I wanna call it , but I don't know if it's a or not even- **Doug O'Laughlin** [1:26:40] Yeah **Swyx** [1:26:40] ... because it's a clear decision where they were the lead investors in OpenAI. They had the deal, and they consciously obviously stepped back. They're still good partners, but like what happened? Like **Doug O'Laughlin** [1:26:52] So I, I think the biggest blunder of all time that... The part that like is kinda crazy to me about that one is like, yeah, I, I, I definitely think there was a financial decision because when you look at it, it looks like a conversation of shareholders, ROIC, and how much are you willing to burn cash? Because like, you know, effectively you look at all the other peers, and Google I would argue is going to free cash flow zero. I think Meta will go to free cash flow zero. Microsoft is still like, you know... Satya did not make the company. He is a professional manager, and there is a board, and there's a comm- **Swyx** [1:27:25] He's being responsible. **Doug O'Laughlin** [1:27:26] Yeah. He- yeah, he's being responsible, right? But the problem is that responsible, this is like an innovator's dilemma, right? Like do I maintain maximized shareholder value and cash flow today, or do I have a, a deep belief that AI will kill the hell out of my core business and I need to all-in invest in... You know, am, am I ready to bet the entire company on, on a trend? And it seems like Satya is not a believer. You know, we've been talking about AGI. He is not ASI pilled, okay? He doesn't have any fear of the Shogoth. **Swyx** [1:27:53] Mm. **Doug O'Laughlin** [1:27:53] He thinks it's just like a new- **Swyx** [1:27:55] It's a tool **Doug O'Laughlin** [1:27:55] ... it's a new Lotus- **Swyx** [1:27:56] Yeah **Doug O'Laughlin** [1:27:56] ... or it's a, you know... Lotus and Excel came around, right? Like it's just a new tool. But I think at the same time this, this, um, conflict between renting GPUs to the barbarians who will disrupt your business or, you know, your actual core business, it's clear how they're feeling. In the call of earnings, they talked about they could grow a lot faster if they wanted to, but they're trying to reinvest back into the internal capabilities. That to me sounds like we are not gonna hire as many barbarians. We're gonna pull... You know, we're reinvesting these walls, pull in together, and try to defend the core moat, right? Because the, the dream of this, and in theory you're like, oh, remember in '23 when they did the first big deal? You're like, "Wow, Microsoft's gonna win it all"- **Swyx** [1:28:33] Yeah **Doug O'Laughlin** [1:28:33] ... 'cause they already have all the distribution, and they're gonna have the perfect product, and boom, they're gonna have this giant business that makes them, you know, whatever, 100 billion, $100 trillion, okay? Whatever number you wanna say. But reality is Claude for Excel, Claude for PowerPoint is literally exactly what it's supposed to be. **Swyx** [1:28:50] Microsoft should've built it. **Doug O'Laughlin** [1:28:51] Mi- Microsoft should've built it. **Swyx** [1:28:52] Yeah. **Doug O'Laughlin** [1:28:53] And so now you see the barbarians, and this isn't even your primary barbarian issue. The guy who... You know, this is like, uh, these- **Swyx** [1:28:58] This is the, the- **Doug O'Laughlin** [1:28:58] This is like the tribe over the hill **Swyx** [1:29:00] ... clone- **Doug O'Laughlin** [1:29:00] Yeah, yeah **Swyx** [1:29:00] ... barbarian. **Doug O'Laughlin** [1:29:01] Yeah, this is the tribe over the hill, you know? And tho- the tribe over the hill is like, like, you know, on a, on a nightly raid, easily sack the hell out of your castle. And you're like, "Dang, this is an issue." So, so Microsoft now is super stuck in the middle, and so how they're gonna have to do this is totally different. I think they're gonna keep, I think they're gonna keep pulling back in. Um, we're starting to see that. Like, they're gonna do internal training. They're gonna try to do more foundational models. They're gonna try to use the weights that they have access to. **Swyx** [1:29:24] Is MEI? **Doug O'Laughlin** [1:29:24] Yeah. **Swyx** [1:29:25] Okay. **Doug O'Laughlin** [1:29:25] Yeah. But I'm very skeptical because their execution has been kinda dismal. **Swyx** [1:29:30] Well, you know, it remains to be seen. They, they, uh, they're, they do have inf- uh, you know, they are one of the big, uh- **Doug O'Laughlin** [1:29:35] Biggest companies **Swyx** [1:29:35] ... biggest companies in the world with all these resources. **Doug O'Laughlin** [1:29:37] Yeah. **Swyx** [1:29:38] Uh, I, I, I just wanna push back on the, the sort of ir- responsibility part. Like, you know, so Oracle picked up the slack. **Doug O'Laughlin** [1:29:44] Yeah. **Swyx** [1:29:44] Is Oracle being irresponsible? You know? **Doug O'Laughlin** [1:29:46] So, so I, I'm actually... If we're gonna talk about Oracle, I think so. Let's talk specifically about Oracle- ... 'cause this is where we're gonna go. I think Oracle was irresponsible because the magnitude- **Swyx** [1:29:54] Oh **Doug O'Laughlin** [1:29:55] ... of what they did. **Swyx** [1:29:55] Okay. **Doug O'Laughlin** [1:29:56] It, it like... The thing is like I think the slack, they should've done it, but like the whole setup in my opinion on Oracle is o- own goal. They messed up the messaging. They messed up the fundraising. And in my opinion, if they were not... Like, like one of the things that happened is they went so aggressive out the gate, did the quarter where they said like $400 billion. Right? They, they said- **Swyx** [1:30:14] Promised the world **Doug O'Laughlin** [1:30:15] ... they promised the world, then they pr- proceeded to raise as much money as possible. And like this is the first time they've ever done these giant build-outs, and so now there's delays. Uh, everyone's like, "Whoa, whoa. You did this much?" Right? Capitalism is kinda like, "Hey, hey. Pump the brakes." **Swyx** [1:30:30] Hm. **Doug O'Laughlin** [1:30:30] And, and seriously, I think that if they just tiered it out better, meaning that they didn't do it all in one periods, played a little bit of expectations management, this year's revenue from the deployed GPUs should partially help start to keep self-funding, and that's how you make this work in a g- glide path without going up, down- **Swyx** [1:30:49] Right **Doug O'Laughlin** [1:30:49] ... up, down, and big bang. And so they... I think what really happened is the big bang that really screwed them up was the debt side. They, they just offered so much debt. It's kinda funny because in high yield TMT, it's such a big part of the entire index. Like, the issuance is so big, it's like- **Swyx** [1:31:07] Uh, debt index. **Doug O'Laughlin** [1:31:07] Yes, of the- **Swyx** [1:31:08] I, I have zero familiarity of this stuff **Doug O'Laughlin** [1:31:10] ... of the... Hey, I'm, I'm pulling some numbers up. I did the numbers- ... forever ago, I'm like, I, I hallucinate and whatever, forget all the precision. Let's just say all of investment grade, uh, TMT is, like, 500 billion. Okay? Um, I think Oracle is, like, 135 of it. So that's like, that's so big. And so each time you have to- you put up a huge new issuance, you have to give someone an incentive to go buy your debt instead of someone else's. And so you just kind of like, they're screwing up the liquidity because these issuance are so big, diluting the whole pie. It, it makes all the terms a little better or more favorable for investors. So it lit- the entire index is selling off because it's like- **Swyx** [1:31:44] Yeah, it's supply **Doug O'Laughlin** [1:31:45] ... yeah, it's a supply thing, right? And that's the thing that's, like, crazy to me is, like, so they, they massively overshot. And I think that we're, like, a weird bottleneck I never ever, ever, ever, ever thought of, uh, thought we would ever, ever hit. And I think you could appreciate this uniquely as, like, one of the bottlenecks is, like, supply of debt into the market. **Swyx** [1:32:04] Mm-hmm. **Doug O'Laughlin** [1:32:05] Like, like f- capitalism cannot, like, absorb that much capital demand. **Swyx** [1:32:09] Mm. **Doug O'Laughlin** [1:32:10] Um, because the order of magnitude, it's totally different. These hyperscaler businesses have been completely self-funded since the history of time, had never gone out and issued anything. First time they wanted, they turn around and they're like, "Hey," instead of like, "Can you give me a, a $10 billion loan limit?" Like, we've never done that before, right? So the absolute size is kind of screwing it up, and I think that Oracle specifically was way, way, way too aggressive into a relatively illiquid market. And so, like, you have to do this, like, you have to kind of leg yourself into it if it's gonna be like that. But they, they like super jolty did these big, huge incremental ads and kind of flipped the whole thing. Oracle CDS, people all freaking out. I think a lot of it's mechanical, specifically on how badly it was done from a supply-demand perspective. And I think they can pay for it. Hey, you wanna, wanna hear you do all right one? **Swyx** [1:32:56] Microsoft could have just internally funded this and, like- **Doug O'Laughlin** [1:32:58] Yeah. Microsoft could have internally funded this. It would've been totally fine. 100% agree. And, like, this example where it's like, yeah, I think that that's a blunder. It's a perfect example of a blunder 'cause Microsoft's cost of debt is the same as the United States government. It's, like, the cheapest you'll get anywhere else. **Swyx** [1:33:11] Jesus Christ. **Doug O'Laughlin** [1:33:12] And, like, just from a, like a, like a P time, like the math perspective, no one else ... They, they're better than Oracle. They, they just, by their credit rating, they have a 2% more profitability at a, at a capital basis. That's, that's, like, you can't beat that. **Swyx** [1:33:26] Mm-hmm. **Doug O'Laughlin** [1:33:26] Um, I don't know why they decided not to, but now they're in this weird thing where they're like, "Uh..." They're, they're, they're kind of wavering. Like, like, to win, you have to be, like, really bold, right? And they're kind of, like, doing this one thing over here, being really defensive with Copilot. Satya is now, you know, the, the product manager of Copilot, and then they're also pulling back from Azure. Meanwhile, the competitors are pul- are pushing in for the supply. It's a really weird game. I think Microsoft has to choose, choose a direction. **Swyx** [1:33:54] We'll see. We'll see. **Doug O'Laughlin** [1:33:55] We'll see. **Swyx** [1:33:55] Um, and it's okay. **Doug O'Laughlin** [1:33:56] That's what's gonna make it fun. **Swyx** [1:33:57] Yeah. **Doug O'Laughlin** [1:33:57] I mean, I'm, I'm more than happy to change all of my opinions when new information comes around. **Swyx** [1:34:01] Yeah. And I'm sure it will, it will- **Doug O'Laughlin** [1:34:03] It will **Swyx** [1:34:03] ... we'll have more information as it emerges. I wanted to touch on TPUs and then go into memory. **Doug O'Laughlin** [1:34:07] Mm-hmm. **Swyx** [1:34:08] TPUs will hopefully... I, I don't know, maybe, maybe a short one, but like, you know, uh, for a long time you could not buy TPUs, uh, at least like current gen TPUs externally, and now you can. And Google's open as a, as a, as a supplier, I guess. **Doug O'Laughlin** [1:34:22] I think Sergey doesn't wanna lose. **Swyx** [1:34:25] Yeah. **Doug O'Laughlin** [1:34:25] And I think the thing that happened was, um, up until... Like, you know, he wasn't... No one was there. **Swyx** [1:34:29] A part of the whole DeepMind story was we will hoard all the TPUs- **Doug O'Laughlin** [1:34:33] Yes **Swyx** [1:34:33] ... because we were first. Uh, we'll, you know. **Doug O'Laughlin** [1:34:34] Yeah. **Swyx** [1:34:35] And, and so, like, why so? You know? Why, why, why give anything to Anthropic? **Doug O'Laughlin** [1:34:38] I think it's because at least last year it beca- um, pre-Gemini 3, it was like, "Dude, we have all these TPUs. We're gonna hoard them all," but, like, people aren't using our products anyways. And, and like, like, hey, what's all, what, what good is all these TPUs if, um, we're getting our asses kicked in consumer? I think it's an interesting thing too, 'cause the other thing I think about is there's a lot of different ways to, to, like, break this down. One, we wrote about it in TPV8, like whatever, we think Rubin will be much more competitive. I think Ironwood, uh, V7 is the peak gap between on, between TCO, between Nvidia and, um, TPU, right? So if you are at your absolute strongest point, what do you do? There's two ways you could do it. You could try to maximize and, like, squeeze the juice and, like, make margins, or you can, you can gain market share. I think the perspective of doing this externally with, with Anthropic is to gain market share because the biggest gap you have, um, and, and one of the reasons why there hasn't been a second merchant chip, and also you can argue Nvidia's the most valuable company in the world, what's the value of TPU in Google? It's, it's huge. **Swyx** [1:35:43] You must have done the math. **Doug O'Laughlin** [1:35:44] I've done the math. It could be like... It's like- **Swyx** [1:35:46] A trillion, you said? **Doug O'Laughlin** [1:35:46] It's like a trillion. **Swyx** [1:35:47] Yeah. **Doug O'Laughlin** [1:35:47] It's like a trillion or something like that. Assuming it gets, like, 30% market share or something like that. **Swyx** [1:35:51] Yeah. **Doug O'Laughlin** [1:35:51] Everyone has been trying to crack the merchant silicon mode, right? And now they have the biggest absolute outperformance. All- a lot of the people who did the original TPU program are, like, now at OpenAI. **Swyx** [1:36:02] Some of them are at MetaX. **Doug O'Laughlin** [1:36:03] Some, some of them are... Yeah, you're exact. Some of them are Meta- They're all over, right? **Swyx** [1:36:06] Yeah. **Doug O'Laughlin** [1:36:06] Like, the, the core team that did most of the engineering have, like, since really dispersed. And so I think the gap might, might close over time. And so at this absolute period of time, they're gonna, they're gonna win market share. And then, then what happens is if you have an install base, you have an incentive to upgrade your install base. That's, like, the hugest problem with AMD, for example. No one wants to buy new AMD chips 'cause it's not like they have old AMD chips. **Swyx** [1:36:28] Mm. **Doug O'Laughlin** [1:36:28] No, they're not upgrading from anything. **Swyx** [1:36:30] Mm. **Doug O'Laughlin** [1:36:30] And so when you have that number two place, you have to, like, you have to win definitively and then also you have an opportunity to win, win again next year. I think the install base issue's been a kinda huge one. And so TPU is at the point where the software ecosystem is mature enough. The hardware is definitely mature. The networking's really mature. You have a really good external customer who actually knows how to use your product. If you want market share, now's the time. ### Memory Crunch **Swyx** [1:36:53] Yeah. That would be insane if they, if they actually sort of pump the, pump the gas on that stuff. Um, are you also hearing w- I, I don't know if this is something that affects your analysis at all, because I don't have any appreciation for the, or the, the sizes that we're talking about here, that JAX is helping TPUs win, or, or JAX is winning relatively to PyTorch, at least in, like, the academic arena, which is a leading indicator of what it's gonna be used in, in- **Doug O'Laughlin** [1:37:16] I do not have as... I, I don't have a special purview on that. The thing I'm most excited about, and, like, very much TBD we'll see, is, uh, InferenceX will have TPUs eventually. **Swyx** [1:37:26] Mm-hmm. **Doug O'Laughlin** [1:37:26] That's something we wanna do longer term. I think that that will really- Show in numbers what's- **Swyx** [1:37:30] As a benchmark? **Doug O'Laughlin** [1:37:31] Yeah, as a benchmark. **Swyx** [1:37:32] Well, how do you expect them to come in? **Doug O'Laughlin** [1:37:34] Pretty good on a price basis. I mean, our expectation is, like, they're- they're the best TCO by a meaningful amount right now. Anthropic's very clear how they feel. Like- like, everyone is very clear. I think even OpenAI would take- I think everyone would eat as much TPU v7 as possible if you had it in a perfectly unconstrained world. It would probably be, at this exact moment, like, you know, the- the hottest kid on the block until Rubin comes out. But the reality is supply chain really matters, and they're-- that's just-- that's not available. And so that TCO, um, that TCO advantage is at this absolute biggest aperture. Then, like, Jensen essentially gets his-- uh, he gets their stuff together, it's competitive, and boom, it closes. So this door only open right now. Probably TSMC's the biggest blocker, so there's no... **Swyx** [1:38:17] Yeah. **Doug O'Laughlin** [1:38:18] Yeah, what- what can you do? **Swyx** [1:38:18] It- it's this cascade, right? They-- which I think you've talked about. **Doug O'Laughlin** [1:38:21] Yeah. **Swyx** [1:38:21] Like, all the way-- it goes all the way back to the fabs. **Doug O'Laughlin** [1:38:23] Yeah. Yeah, well, it's interesting 'cause it's like even more than the fabs. Like- like, on the optical net- **Swyx** [1:38:28] I read their link I should be pulling up. **Doug O'Laughlin** [1:38:29] Um, yeah. Yeah, that- that- that's it. That's it. **Swyx** [1:38:31] Yeah. **Doug O'Laughlin** [1:38:31] TPU v7 and Google **Swyx** [1:38:32] Yeah. **Doug O'Laughlin** [1:38:32] That sort of thing. **Swyx** [1:38:33] Yeah. **Doug O'Laughlin** [1:38:33] Yeah, so- so yeah, it all goes back to the fabs. It all goes to who's making the chips. And, like, I think one of the big differences too is just, like, the per-- It's just, like, a really good cleverly designed system architecture, and it's relatively stable, and it's clear that they-- you can pre-train big models on it, which is, like, a huge, huge swipe at OpenAI right now. That being said, like, I think OpenAI will get their- their act together very quickly. And so yeah, that's kinda, like, the narrative. I think it's gonna be a good story for probably, like, a year or two, but then the real question is the, um, V8 we just don't think will be as competitive to Rubin, and that's when your special window starts to close. **Swyx** [1:39:11] What's the technical reason why? **Doug O'Laughlin** [1:39:13] HBM. **Swyx** [1:39:13] Okay. **Doug O'Laughlin** [1:39:13] HBM4 versus 3. And then- **Swyx** [1:39:15] And that's a secure-- is the strategic decision by Nvidia? **Doug O'Laughlin** [1:39:18] Uh, yeah. I think one-- Uh, so Nvidia is always-- If you think about Nvidia, they're always trying to gas it as hard as they can. Like, they- they-- Like, it is a high-performance ship. It is a f- it is an F1. Like, it is as maxed out as possible. Um, TPU is kinda like this, like, replicatable pod in a very large-- with, like, very high stability, right? **Swyx** [1:39:37] Which if you know the history of Google, that's what, that's what they do- **Doug O'Laughlin** [1:39:39] That- that's what they do **Swyx** [1:39:40] ... with infra. **Doug O'Laughlin** [1:39:40] Yeah, that's what they do with infra. Yeah. But, um, I think GB200 would have completely mogged, you know, V7 if it came out on time and stable. It came out a little delayed, and it wasn't stable. And so I think, um, there's a lot of different ways to kind of course correct that. And the one thing that's important is, like, I think on the infrastructure side-- or sorry, on the, um, supply chain side, bar none, Nvidia's the best. They own the entire supply chain. **Swyx** [1:40:06] Mm-hmm. **Doug O'Laughlin** [1:40:06] They really do. Like, um, you think all those HBM price increases, they're gonna come for TPU just like, uh, Nvidia, but Nvidia was literally in Asia. You saw him drinking with everyone, with the SK, with the-- everyone, with all the Korean guys, all the TSMC. He's doing the shots with everyone. Why do you think he's doing love shots with everyone, okay? It's 'cause he, he needs to get the chips, okay? **Swyx** [1:40:27] Uh, so yeah, this is Samsung's chairman? **Doug O'Laughlin** [1:40:29] Yeah, this is Samsung's chairman. **Swyx** [1:40:31] Yeah. Uh, and who's the, who's the other guy up here? **Doug O'Laughlin** [1:40:33] But let's put it this way, um, that's- that's a huge deal. That's a huge- ... huge, huge deal. Do you think, do you think, uh, Google was out- **Swyx** [1:40:39] Hyundai. Hyundai. **Doug O'Laughlin** [1:40:40] Yeah, Hin- Yeah. Do you think Google was, uh, what, what, you know-- Do you think Sergey was out in Taiwan drinking to, to get supply? No, 100%. There's an opportunity here, but there's only so many TPUs that can be made because the-- because of all the bottlenecks, right? And so Nvidia has all the supply chain locked up, and so they're gonna have, like, so much of that, um, kinda constraint there. And so it's, it's gonna be really interesting. They're gonna, they're gonna get the best, most performing HBM. They're gonna be first on the roadmaps for even more rack density. They're gonna have, like, the best connectors, the best... You know, the whole system will be, once again, turbo jammed again for as hard as it can be. **Swyx** [1:41:17] Mm-hmm. **Doug O'Laughlin** [1:41:17] And the people who made V7, like, they made-- The, the chip was done, like, three or four years ago. Like, the, the talent, um, dispersion aspect where people who worked really hard on this team to make this great chip has really kind of gone all over, that starts to get worse. And so if that gets better, which takes some time, I- I think our current read is that, like, the HBM specifically and the memory scale-up is gonna really go in Rubin's favor, and so that's the big difference. Um, and I think, as you know, that's what makes the context windows. That's when we're able to do bigger, bigger- Everything. Everything, yeah. And so they're gonna really jam it, and that's, that's gonna be a huge advantage in performance. **Swyx** [1:41:53] One thing I love about your analysis is it's not actually just the context windows. It's not just the KV cache. We also have to offload it to non-HBM. **Doug O'Laughlin** [1:42:00] Yeah. **Swyx** [1:42:01] Every other part of the memory, it, um, the, the-- it's, it's, like, su- such an interesting cascade, uh, waterfall of, like, just, like, a dim-- a short squeeze and, and everything. But it's not a short squeeze. It's like a supply squeeze. **Doug O'Laughlin** [1:42:11] Yeah, it's a sup- I mean, there's like, uh, I just wanna, like- **Swyx** [1:42:14] I mean, if I do one ratio of, like- **Doug O'Laughlin** [1:42:16] Yeah. Yeah. Okay, so it's a- **Swyx** [1:42:17] To quantify to- **Doug O'Laughlin** [1:42:18] So three-to-one to four-to-one ratio. I think, I think make sure-- So, so it's in the Memory Mania, um, post that we just put out of, like, the four-to-one or the trade-off ratio. **Swyx** [1:42:26] Yeah. **Doug O'Laughlin** [1:42:26] Um, scroll down somewhere and you'll see. **Swyx** [1:42:29] Yeah, so, so basically for, for listeners, it's the idea that, like, when you convert to HBM because there's a huge demand for HBM, it takes three times, uh, one HBM sort of units is, like, three times, uh, of the other sort of DDR or whatever, right? **Doug O'Laughlin** [1:42:42] Yeah. So some amount will always be lost in production because yield isn't perfect. And so effectively, you're trading some-- Like, you're trading, uh... I, I ma- I actually wrote a really funny piece, like, I, I called it, like, super oil, but let's-- this is a better one. Um, pretty much, like, you-- In order for this higher grade of jet fuel has been invented, and the only way to make it is to, like, actually, uh, get rid of all, all your other fuel, and you have to, like, massively condense it and refine it, okay? So, um, now what happens is if there's any demand here, it's an instant shortage. And so we, we hilariously enough came out of, like, the biggest shortage ever in NAND and DRAM. Like, terrible. Like- ... catastrophic, the worst one ever. Like, the, the, the last analysis I could put to it is, like, '96 or something like that. Seriously, it's like a, an, a historied one. And then meanwhile, we have all this new demand, HBM specifically. The highest end, you need the most memory. The, the trade ratio is crazy, so each, you know, each bit of, uh, of HBM is essentially a 4X multiplier onto DRAM. And then now, so we- we've completely constrained, took all the DRAM capacity. We just came out of this shortage, so no one invested in any clean rooms or capital equipment or anything like that. People got, like, massively free cash flow negative. No one's spending a cent, okay? People could go bankrupt, you know? **Swyx** [1:43:54] Yeah. **Doug O'Laughlin** [1:43:54] So you- they haven't invested in these three-year-long lead time items, and then now there's the, like, more demand than God, and it also evaporates the middle layer because of the KB cache offload. And then boom, you're just looking at the supply and demand, and you're like, "Yeah, this is not gonna catch up for, like, two years." I think the thing that's, like, so interesting is the supply chain squeeze, because these clean rooms take two years to make, man. And effectively, everyone paused, and how bad the last cycle was really forced everyone to completely pause altogether in terms of adding any new capacity. **Swyx** [1:44:25] Yeah. **Doug O'Laughlin** [1:44:25] And so now we're a few years later, and all the supply is gone. So I mean, people are... I mean, it's just, it's crazy. We... I, uh, our post, our conclusion is, like, we, we could see DRAM prices, like, go up 100% again. Like, it's, it's gonna be the point where... And this is, like, also example, like, really interesting in the whole thing. Another 100% I think is demand destruction. I think you will start to have demand destruction from- **Swyx** [1:44:46] What does that look like? **Doug O'Laughlin** [1:44:47] Um, where hyperscalers maybe purchase less or something like that on the margin. On the margin. **Swyx** [1:44:53] Yeah. **Doug O'Laughlin** [1:44:53] Right? Because they're like, "Okay, well, what if I just really focus on this energy aspect instead?" And also ironically, all the energy, um... So, like, every- not every data center in America, but, like, many, many, many, most of the data centers in America are delayed. So you had this thing that's supposed to come on in 12 li- months. It's coming on in 18. Maybe what you can do is you can play chicken with memory prices, and you can kind of push out... Of course, everything you have in the pipeline, you, you pull forward as hard as you can, okay? You pull forward, you double, triple order, and then the DRAM and the HBM guys are like, "Oh my God, how? Look at all this demand." And then at some point in time, what happens is you say, "Well, we pulled this all forward. You know, we have... The power's gonna constrain us anyways. We're gonna, like, kinda chill out the orders." And historically, that's when the memory market... That- that's what causes the crisis, uh, the- the prices to drop. Realistically, just looking at h- at the aggregate demand of how much we've purchased in term- terms of power, it just seems like the gap is just huge. It's completely off, to the point where the most obvious logical leg of the AI, the AI trade is effectively investing in memory- **Swyx** [1:45:55] Capacity **Doug O'Laughlin** [1:45:56] ... capacity, yeah. Well, not, not 100%. **Swyx** [1:45:57] Oh, yeah. **Doug O'Laughlin** [1:45:57] I mean, it's, uh... You could say SK Hynix and, and Samsung. **Swyx** [1:46:00] Uh, Micron. **Doug O'Laughlin** [1:46:01] Micron, and all the, all the semi... Like, Semicap has been ripping. **Swyx** [1:46:04] Which, like, uh, by the way, when I was in, uh, Balyasny, we, a majority of, a lot of money we made was just being on Micron- **Doug O'Laughlin** [1:46:10] Yeah **Swyx** [1:46:10] ... in, in the last, like- **Doug O'Laughlin** [1:46:11] Yeah, it's a good example. Yeah. So, like, you have all, uh, the Semicap stuff, right? Like, all the, everything that is even remotely related to investing in capacity for memory, that is, like, the ultimate bottleneck right now. **Swyx** [1:46:21] And also for listeners, it's gonna affect, like, your phones. **Doug O'Laughlin** [1:46:24] Yeah. **Swyx** [1:46:25] Like **Doug O'Laughlin** [1:46:26] Yeah. I, Apple, I think Apple's moving the goal- **Swyx** [1:46:28] I had to buy a, I had to buy an SD card for this thing. **Doug O'Laughlin** [1:46:30] Yeah. **Swyx** [1:46:30] And it was like 1,000 bucks. **Doug O'Laughlin** [1:46:31] Yeah. That's, that's nothing too. That's- And that, and that... 'Cause, 'cause, like, that, that's just the NAND side. Dude, have you looked up, like, like, I wanna say, like, 64 gigabytes of, of DRAM? **Swyx** [1:46:44] Like, like, I'm moving up... Like, I, I need to refresh my iPhone. **Doug O'Laughlin** [1:46:46] Mm-hmm. **Swyx** [1:46:46] I'm moving it up because I'm doing this research for... **Doug O'Laughlin** [1:46:49] Big. Oh, yeah, you need to do it as soon as- **Swyx** [1:46:51] Buy your iPhone now. **Doug O'Laughlin** [1:46:51] Yeah, yeah. You buy your iPhone now because what's gonna happen- ... is when, um, iPhones go into the spot market- **Swyx** [1:46:57] Mm-hmm **Doug O'Laughlin** [1:46:57] ... it, prices are gonna go up 100% on them. **Swyx** [1:46:59] That's insane. **Doug O'Laughlin** [1:46:59] And so they have to pass the- **Swyx** [1:47:00] We're gonna be buying, like, old iPhones and then taking them out for the memory. **Doug O'Laughlin** [1:47:04] There's... No, no, that's actually ha- There's a, there's a whole... Super-duper deep in the weeds, there's this whole, like, technology that was very focused on cloud era, um, called CXL, which is memory expanders for CPUs in order to have, like, whatever, just, like, elastic pools of compute of CPU and DRAM and whatever, memory attach and whatever. It never really took off because essentially HBM was, like, the way. That really crushed it all. High performance, best of breed wins. But this CXL technology that kinda never really took off is gonna take off, just because what they're gonna do is they're gonna take DDR4, they're gonna take the oldest, every bit of spare memory they can find, and they're gonna put them into racks, and then they're gonna attach them via CXL. So, like, this- **Swyx** [1:47:42] Oh, okay **Doug O'Laughlin** [1:47:42] ... yeah, yeah. **Swyx** [1:47:43] So exactly that. Yeah. **Doug O'Laughlin** [1:47:44] Yeah, it's exactly that. But the thing that's so crazy is, like, this dead technology is, like, having a shot on goal because of how bad the storage cont- or how bad the memory constraint is. **Swyx** [1:47:53] Yeah. **Doug O'Laughlin** [1:47:53] Like, yeah, that, that, um, you know, I- **Swyx** [1:47:55] Cool **Doug O'Laughlin** [1:47:56] ... I was, like, a CXL bull for once upon a time, and then it became very clear it was gonna die, and now it's like it's back, but only because the entire express intent is to have these D... Like, old chips, pool the old chips, attach it to something new. That's what it's gonna be like. The, the memory shortage is just, like, it's crazy. So- **Swyx** [1:48:14] Yeah, it's incredible. **Doug O'Laughlin** [1:48:14] Yeah. **Swyx** [1:48:15] So obviously this is lower level than I usually go to, which is, which is why I'm having so much fun. One thing I, I do tell people about is, like, well, you know, everyone i- including Sam, by the way, is, like, predicting longer context windows. We've been kind of effectively stuck at a million for two years now. **Doug O'Laughlin** [1:48:31] I've actually been thinking about that a lot. **Swyx** [1:48:33] And, like, this is not gonna go to 100 million context windows. It's not gonna go to trillion. Like, we're f- this is it. **Doug O'Laughlin** [1:48:39] Yeah. **Swyx** [1:48:40] This is it for, like, five years, 10 years. **Doug O'Laughlin** [1:48:42] Pretty much. Okay, so the question is, will c- yeah, I mean, yeah, probably, actually. Will capitalism work? Will we, will there be a way for supply to show up? Probably, but on top of that, I wonder if there's gonna be, like... Like, I, hey, his- his- history of compute, what happens is you have to, like, you have to, like, make a curve of the sup- of the context windows. Like, does free context windows go to, like, 1,000? Hey, you can use ChatGPT free now, but you c- your context window is, like, 1,000 tokens or something like that. And then you can just like somehow do a tiny, like a tiny parcel for that just so that you can then charge like, you know, 100X more for 1 million. The 1 million context window is like a mansion, you know? That's the real- **Swyx** [1:49:20] You live in a mansion right now. **Doug O'Laughlin** [1:49:21] I live in a mansion right now, yeah. **Swyx** [1:49:22] Uh, oh my God. Uh, the, the word just context rationing just came to me. I'm like, "Fuck." Like, we're gonna have like vouchers for like, "Okay, you can have this amount of context today." Like **Doug O'Laughlin** [1:49:31] That's like... Yeah, you're gonna have to, you're gonna have to learn how to use it well- ... because of the DRAM, yeah. Okay, so I actually have a question. I know 'cause like, okay, long context to me makes a lot of sense, right? Hey, that's like, like that's like the memory scale-up version, like if you're thinking about chips but in the AI world. I just am like always been curious 'cause it does feel like at least in my stated experience, really long context, like you see in the papers they kinda like drop off. They like actually don't use all the context. So that's like kind of the thing I've been most interested in is like does the 100 million context actually matter if, if, if it's not possible to use it all? **Swyx** [1:50:05] They, you know, versions of the 100 million do exist today. They're- they just suck in various ways. They're- they're not actually applying full attention, right? **Doug O'Laughlin** [1:50:11] Yeah. **Swyx** [1:50:11] You can, you can use a state space models or even like a LSTM to like, uh, you know, uh, to, to process 100 million tokens, but you're, you're not paying full attention to all those 100 million tokens. **Doug O'Laughlin** [1:50:21] Yes. **Swyx** [1:50:21] And so I think like the way that we have context models today and those curves, they will improve over time- **Doug O'Laughlin** [1:50:25] Yeah **Swyx** [1:50:25] ... and they have, they have been- **Doug O'Laughlin** [1:50:26] Yeah **Swyx** [1:50:26] ... improving a lot, but we're, we're just not, we're never gonna use all of them, but we'll, we'll improve like on the algorithm side. **Doug O'Laughlin** [1:50:32] Yeah. **Swyx** [1:50:32] I think for me what matters is you, you represent the physical constraints that us, the software side, can never surmount- **Doug O'Laughlin** [1:50:39] Yeah **Swyx** [1:50:39] ... because it's a physical constraint. **Doug O'Laughlin** [1:50:40] Yeah, they- **Swyx** [1:50:41] And well, I mean, it just, it just like physically we cannot double... We can't even- **Doug O'Laughlin** [1:50:45] Yeah **Swyx** [1:50:45] ... we can't even double the number unless it's 10X. **Doug O'Laughlin** [1:50:47] Yeah, yeah. **Swyx** [1:50:48] Like what, what, what's the point of talking about any of this? **Doug O'Laughlin** [1:50:50] Yeah, what's, what's, what's the point? Yeah. I was gonna say, uh, uh, like we could, we could in- we can invent a lot of things. Context rationing's pretty good. I really like that one. Context frugality or like budget or something, I feel like everyone's gonna be like, "Whoa, whoa, whoa. You're running out of context wi- uh, window today, you know?" Like maybe that's what happens next year, we're, we're, we're- **Swyx** [1:51:06] Jesus Christ **Doug O'Laughlin** [1:51:07] ... we're charged on context window. **Swyx** [1:51:08] And then the, the one of the more recent obsessions is recursive language models, which again is just reusing the same context window on- **Doug O'Laughlin** [1:51:14] Yeah **Swyx** [1:51:14] ... over and over and over again. **Doug O'Laughlin** [1:51:15] Yeah. I've been pretty interested in that, but like to be clear, I'm a total idiot. I have no idea. Claude tells me what's- **Swyx** [1:51:20] You're the, you're the semis guy, man. **Doug O'Laughlin** [1:51:21] Yeah. **Swyx** [1:51:21] Like, you're, you're really good on your stuff. One thing I wanted to, to spot check on was Talos, uh, which came out recently. **Doug O'Laughlin** [1:51:27] I have not, I have not messed around with it. **Swyx** [1:51:28] Okay. And you don't have to mess around with it. It- it's just the, this general theory of custom ASICs burning the weights into the chip- **Doug O'Laughlin** [1:51:34] Yep **Swyx** [1:51:35] ... so you don't need memory. **Doug O'Laughlin** [1:51:35] Yep. That's pretty good actually. I think- **Swyx** [1:51:37] Right? **Doug O'Laughlin** [1:51:37] I think, I think that that, that makes sense to me. Um- **Swyx** [1:51:40] This came just comes at the perfect time. **Doug O'Laughlin** [1:51:43] It does, but I guess, okay, so historically the question is how big does it scale, right? But like, I mean, you know a lot of the models are kind of actually smaller than you think, right? **Swyx** [1:51:52] So like that's like... Sorry, what do you mean? I don't know. **Doug O'Laughlin** [1:51:53] Well, a lot of the produc- the, the production- **Swyx** [1:51:55] Yeah, they, they get distilled to shit. **Doug O'Laughlin** [1:51:57] Yeah. **Swyx** [1:51:57] Yeah. **Doug O'Laughlin** [1:51:57] They get distilled to shit. So it's like the push and pull there is gonna be like, okay, can you just burn in a, a s- like enough efficient Pareto frontier in terms of performance to be burned in straight onto the silicon that doesn't need memory, and then boom, you can scale this forever? Uh, versus like, you know, the performance edge of the long thing. It's pretty clear to me that like Talos has a place because you're kinda seeing this market bifurcate a little bit. **Swyx** [1:52:21] Mm-hmm. **Doug O'Laughlin** [1:52:21] You could argue the prefill, uh, decode, disaggregation, stuff like that is like the focus on performance inference serving is gonna be a subset of the market and then the training and the whatever and the big production, like the... You're- we need to kind of break it in- into smaller parts in order- **Swyx** [1:52:36] Yeah, at least we're using the same terms. **Doug O'Laughlin** [1:52:37] Yeah. Yeah. **Swyx** [1:52:37] And it makes no sense. **Doug O'Laughlin** [1:52:38] Just in order for the compute to be even remotely okay. **Swyx** [1:52:41] It makes sense to you and, uh, I mean, TBD on the sort of practical implementation, but otherwise burning the weights into the chip. Why didn't Etched or some of the other guys get there first? **Doug O'Laughlin** [1:52:51] Um, Etched is pretty interesting. I don't know. **Swyx** [1:52:55] Okay, I'm asking you to speculate, yeah. **Doug O'Laughlin** [1:52:56] I mean, I mean, I'm not gonna like super speculate. I mean, the th- the thing is like their thing is like, okay, how do we have a, a big systolic array? Um, right? But like they didn't burn weights into the chip. That's a little different, right? **Swyx** [1:53:06] I mean, like look, like I mean- **Doug O'Laughlin** [1:53:08] I, I just think- **Swyx** [1:53:09] The way to speed things up is to never transfer anything. **Doug O'Laughlin** [1:53:11] Yeah. Yeah, that's, that's the fastest way possible. And so, so but the thing is the bet on, on, on this really large systolic array is effectively everything is compute-bound, right? But like I don't think that that's really the case in, in terms of like where we're actually seeing issues in production markets today. It's like you're actually seeing all the issues in the memory, uh, right? And so like I d- I just don't know if that's like gonna be the perfect solution. There is definitely a world and space and like a design space where they're gonna be very valuable and cool, but like also the reason why my hit rate for every AI accelerator chip is like very... Like I just don't believe in them is 'cause like where are they? Until Cerebras and Groq honestly they were all considered failures. And even then we're like, what are they gonna do with Groq? What are they gonna do with Cerebras? So- **Swyx** [1:53:54] Is, is Aminova out? **Doug O'Laughlin** [1:53:56] Uh, no, I think Aminova was like a much more- **Swyx** [1:53:58] Sure **Doug O'Laughlin** [1:53:58] ... interesting one, but I think- **Swyx** [1:53:59] Yeah **Doug O'Laughlin** [1:53:59] ... there's like, there's like all kinds of deal issues with that. I haven't been keeping up with that one as much. **Swyx** [1:54:02] Yeah. I, I always, I always try to mention them as, as part of that cohort. **Doug O'Laughlin** [1:54:05] Yeah. Yeah, 'cause I kinda forget about them too, but yeah, honestly I was gonna say they were- **Swyx** [1:54:09] You know, once, once a year they show up. **Doug O'Laughlin** [1:54:10] Yeah, they, they do and they're not, they're not so bad. Yeah. **Swyx** [1:54:12] Yeah, yeah. **Doug O'Laughlin** [1:54:13] Yeah. **Swyx** [1:54:13] You mentioned actually some CPU shortage stuff or CPU, uh- **Doug O'Laughlin** [1:54:16] Yeah **Swyx** [1:54:16] ... sort of opinion. What's, what's going on there? I'm, I'm less- **Doug O'Laughlin** [1:54:18] I think it's- **Swyx** [1:54:19] ... less informed. **Doug O'Laughlin** [1:54:19] I think, okay, so okay, I have, uh, one. We'll start with a conspiracy theory that I think is really funny. **Swyx** [1:54:23] Love this. **Doug O'Laughlin** [1:54:23] Have you been noticing just like I feel like web services have become really unstable. **Swyx** [1:54:28] It has been down a lot. **Doug O'Laughlin** [1:54:29] I, like and like me- okay, this is pure like, you know, schizophrenic hat brain 'cause I have a schizophrenic trend hat brain. I'm wondering if it's, um, two things: shipping vibe code slop to prod, that's number one. That's definitely something if that's possible, but it's happening to all the clouds at once I feel like. It's not just a AWS thing. It's not just a G- like a GitHub Azure thing. We are kind of right at the exact five to six-year period of the refresh cycle of COVID. So COVID, we had this big 2020, 2021, you bought, like, $100 billion of CPUs and stuff like that. **Swyx** [1:55:02] Mm. **Doug O'Laughlin** [1:55:02] And so we're right at the natural end of life for these chips. And so usually what you do is you have this big refresh of all these chips. But what- what's been happening instead is everyone has essentially scrounged all of their budget as hard as they can. But then, like, I feel like I've seen it in, like, Azure, like, hey, last night my Amazon Prime thing doesn't work. And I was like, "It'll probably work in the, in the morning, babe. Don't worry about it. I think it... I think Azure's just... Or, like, AWS is pissed tonight" You know? **Swyx** [1:55:25] Mm. **Doug O'Laughlin** [1:55:25] Something like that. Um, but I think, uh, so we have this five-year thing. Everyone scrounged every single dollar they could to essentially invest in as much as AI as possible and just do maintenance CapEx on CPU. Ironically, at the same time for all this Claude Code stuff, is actually if you have this coding agent just generate you God knows how much compu- uh, how much, like, software, where is the software gonna run? On CPUs. So I think we're gonna see some increasing utilization as well as the fact that RL is, like, actually heavily used for, like, RL gyms. You have to s- you have to simulate software, and it uses a lot of, uh, CPUs. So the or- not quite, like, the orders of magnitude of the GPU stuff, but it's just such a big trend even when it steps slightly in a, in a place, mass amounts demand. **Swyx** [1:56:07] Yeah. **Doug O'Laughlin** [1:56:07] I feel like we might actually be seeing a CPU shortage, partially 'cause of this refresh cycle- **Swyx** [1:56:12] Yeah **Doug O'Laughlin** [1:56:12] ... and partially also because, like, I legitimately believe that Claude Code... Claude Code is increasing c- software creation. Um, and then on top of that there is real- **Swyx** [1:56:20] Yeah **Doug O'Laughlin** [1:56:20] ... demand from RL. **Swyx** [1:56:21] Yeah, yeah. And just general d- uh, production agents as well. Uh, you know, we just, um... Yeah, e- every, like, RLMs take compute, and, um, you know, C- Open Cloud takes more compute. **Doug O'Laughlin** [1:56:31] Yeah. **Swyx** [1:56:31] And it- it, you know, it's just, it's just, um, different slope but at the same sort of direction. **Doug O'Laughlin** [1:56:35] But still an upslope- **Swyx** [1:56:36] Yeah **Doug O'Laughlin** [1:56:37] ... and in a slope that, to be clear, has had massive underinvestment for the last two years 'cause everyone- **Swyx** [1:56:41] Yeah **Doug O'Laughlin** [1:56:41] ... like, how do... It- it- the same problem that happened, massive underinvestment, 'cause they're like, "Screw it. We're doing maintenance only. We're j- all we're gonna do is maintan- maintain the past. We're not gonna add anything else." And then all of a sudden just a little tiny slope on top of it, you're like, boom, shortage. **Swyx** [1:56:54] Yeah, yeah. Amazing. **Doug O'Laughlin** [1:56:56] Yeah. **Swyx** [1:56:56] So semis guys say semis numbers go up. **Doug O'Laughlin** [1:56:59] Yeah. **Swyx** [1:56:59] Is- is the- **Doug O'Laughlin** [1:57:01] That's- that's one way to put it, yeah. I- I- the thing that's crazy is we talk about the demand just- **Swyx** [1:57:05] But it's like- **Doug O'Laughlin** [1:57:05] Yeah **Swyx** [1:57:05] ... it's like you're right. **Doug O'Laughlin** [1:57:06] W- like, I mean, it's- it's- **Swyx** [1:57:07] For sure **Doug O'Laughlin** [1:57:07] ... like, you're like, "Show me where I'm wrong." **Swyx** [1:57:08] Yeah, like- **Doug O'Laughlin** [1:57:09] "Show me where I'm..." I- I mean, I'm definitely not. But the thing that's crazy is, like, memory prices are gonna go up so much that we're gonna have to choose which m- which- **Swyx** [1:57:16] Exactly **Doug O'Laughlin** [1:57:16] ... to go on. **Swyx** [1:57:16] The, the, yeah. **Doug O'Laughlin** [1:57:17] That's the crazy part to me. **Swyx** [1:57:18] Yeah. **Doug O'Laughlin** [1:57:18] Historically, memory has never been a constraint like this where I said, "Actually, you're not gonna get your low-end, uh... You're not gonna get your low-end phone. You're not gonna get a GPU this year for gaming." None of that stuff. You're n- you can't do these things for you 'cause your price is half the market. That's what's crazy. **Swyx** [1:57:33] I hate to say it. **Doug O'Laughlin** [1:57:33] That is the first thing that's happened in a long time. That's gonna be really interesting to see where that shortage and how it, how it's, like, digested and felt. So- **Swyx** [1:57:40] No, that is amazing. That- thank you for that breakdown. I feel like I- I really understood it, like, talking to you. ### Wrap-Up **Doug O'Laughlin** [1:57:43] Yeah. **Swyx** [1:57:44] Uh, let's really transition to a couple personal things and- **Doug O'Laughlin** [1:57:46] Yeah, sure **Swyx** [1:57:46] ... as- as the end. How do you write? 'Cause you- you write a fuck ton. **Doug O'Laughlin** [1:57:50] Yeah, I do. I have been writing a little bit less these days now that I'm, like, in the SemiAnalysis, uh, mega mind. I definitely write a lot. Um- **Swyx** [1:57:59] And, like, you kept going with Fab. 'Cause I- **Doug O'Laughlin** [1:58:00] Yeah, for a while. Okay, to- to be clear, that was really... So- so- so look, I'm still trying to do Fab 'cause I- I- I do feel deeply connected to writing. Um- Let's just specifically talk on this a little. **Swyx** [1:58:11] Yeah, just- just- just, like, explain yourself, you know? **Doug O'Laughlin** [1:58:13] Okay. So, um, the thing, uh, before L- LMs came around, the thing I felt the strongest about, my- my number one information skill, is I was able to read and synthesize and process at, like, really high speed, really high throughput, decently high comprehension, the adjustment is speed in terms of comprehension, almost anything. Like, when my, like, when my friend gets a PhD, I go read their paper. I was like, "Oh, I have a pretty good i- idea of what you're doing." I was like- like, hey, when I was interested in semiconductor book, I literally raw-dogged some textbooks. Whatever, the comprehension was not very high, but, like, hey, whose comprehension is, you know? Um, but I was able just to, like, push through these books and learn. So I've always loved reading. That's, like, my- my number one original competitive skill set differentiator, and also something I, like, loved as a kid. Crazy reader when I was a kid, always have been. And then, um, starting the Substack, which has been really fun actually, 'cause I just really wanted to get my story out, like, the things I cared about, closed the loop for writing for me, 'cause I love, I love reading so much. It makes a lot of sense that I love writing. I think what really helped is I wrote every single week for, like, since October '21, like, consecutive streak for a long time. The streak has been a little broken as of late. SemiAnalysis plus Fabricated Knowledge is pretty hard to do. **Swyx** [1:59:24] Yeah. **Doug O'Laughlin** [1:59:24] But, like, all of '24, I think, like, we're just talking just, like, every single da- every single week I will put something out, right? **Swyx** [1:59:31] Is it, like, a hard rule, like one a week? **Doug O'Laughlin** [1:59:33] It was a hard rule- **Swyx** [1:59:34] Okay **Doug O'Laughlin** [1:59:34] ... one a week. **Swyx** [1:59:34] Yeah. **Doug O'Laughlin** [1:59:35] Uh, at least an attempt to two. **Swyx** [1:59:36] Yeah. **Doug O'Laughlin** [1:59:36] And so, uh, I think one of the best ways, all the people who write, uh, who write about writing all say the same thing. You need to just be writing. Um- **Swyx** [1:59:45] Yeah **Doug O'Laughlin** [1:59:45] ... and so that's how I, that's how I- **Swyx** [1:59:46] So writing every week, then, huh? **Doug O'Laughlin** [1:59:47] Yeah, it really helps. Um- Well, I was gonna say, what's crazy is, like, it's- it's kind of hard these days, and I... And LMs kind of have really... I don't know. I don't like LM writing. **Swyx** [1:59:56] Yeah. **Doug O'Laughlin** [1:59:57] I do like it for ideation, like, making outlines. **Swyx** [1:59:59] Yeah, yeah. Research. **Doug O'Laughlin** [2:00:00] Here- here's my un- un- organized thoughts, make it into an outline, and then, like, you know, I'll even be, like, put bullet points in the outline, and I'll literally read the outline, and then, like, ideate and write in parallel. But yeah, that's- that's how I feel about writing, I guess. Write more. I have a strong... Uh, for nonfiction writing, I really like this book called On Writing Well. **Swyx** [2:00:18] Mm. **Doug O'Laughlin** [2:00:18] Um, that's just a really good classic book. It's actually, um, summarized and syn- synthesized into a wr- into a skill for me. **Swyx** [2:00:27] Oh, yeah? **Doug O'Laughlin** [2:00:27] Yeah, yeah, yeah. So, hey, uh, please edit this. Use these... Uh, use this style guide. Use the, like, learnings from this book. So yeah. **Swyx** [2:00:35] That's awesome. **Doug O'Laughlin** [2:00:36] Do stuff like that. Yeah. **Swyx** [2:00:36] Okay. And then, uh, do you, like, have a- a topic idea list that you groom? Like, my- I've put mine in Apple Notes now, but it's- **Doug O'Laughlin** [2:00:42] Bro **Swyx** [2:00:43] ... it's- **Doug O'Laughlin** [2:00:43] No, never. Never. I'm one- **Swyx** [2:00:45] And I'm just- **Doug O'Laughlin** [2:00:45] I'm just a one-shot-ter. **Swyx** [2:00:46] Whatever's on your head. **Doug O'Laughlin** [2:00:47] Yeah, usually I one-shot the- the idea all the way. **Swyx** [2:00:49] Yeah, yeah. **Doug O'Laughlin** [2:00:50] Usually I think about it for quite a bit, so it's been bouncing around in my brain. Um, and then at some point in time, I've, like- ... condense enough information to make a really crappy outline, and that's usually when I just one-shot go. **Swyx** [2:01:02] For me, like, it's hard to one-shot and bounce because you will forget, right? And sometimes you, you have, like, really good stuff that you forget, and sometimes it's actually... Uh, so I call this mise en place writing, where you basically just have a store where you're just kind of writing or working your ideas in parallel, and then every now and then you cook. **Doug O'Laughlin** [2:01:19] Yeah. **Swyx** [2:01:19] Um, and so this is async and this is sync, right? This is, like, passive. Like, "Oh, here's a data point." **Doug O'Laughlin** [2:01:24] Yep. **Swyx** [2:01:25] "Here's, here's a quote, here's a thing. I'll just go slot it in the right thing." And then, and then I bake it. **Doug O'Laughlin** [2:01:29] Historically... Okay, so how that pre-writing actually works today is probably in the SemiAnalysis Slack. Uh- **Swyx** [2:01:35] Right **Doug O'Laughlin** [2:01:35] ... just, like, all the little things. **Swyx** [2:01:35] Then you just search it up when you need it. **Doug O'Laughlin** [2:01:37] Yeah, I search it up when I need it or something like that. But, like, I do most of the pre-writing, I think, in my brain, and I have places that I put it out- **Swyx** [2:01:43] Yeah, yeah **Doug O'Laughlin** [2:01:43] ... that I reference it later. But, um, my favorite thing too is, like, when it comes to the... Because, like, okay. Uh, well once upon a time, much more on the beat, "Oh, hey, here's earnings," read every single one and put it all together. But, like, um, my favorite skill or tip or whatever is like, hey, do the pre-writing, think about it, all that stuff, and go to sleep and wake up. The next... The fresh context window in the morning is my number one advice on writing. **Swyx** [2:02:06] Helps you decode- **Doug O'Laughlin** [2:02:06] Yeah **Swyx** [2:02:06] ... better? **Doug O'Laughlin** [2:02:07] It helps so much better. Like, literally if I'm like, "Hey, I need to write something right now," I will do... I'll write it all down, I'll make outlines, I'll do all kinds of crap except for writing it, and then I'll be like... And I'll go to sleep, and then wake up, and the first thing I do, I'll open up a new tab and I will w- write it. **Swyx** [2:02:21] Got it. **Doug O'Laughlin** [2:02:21] Um, and then so usually I... That will get me to 60, 75% of something, even if it's, like, an outline where I, like, have gotten all the ideas enough to know how to fill it out the rest of the way, and then that's, that's how I take it from there. **Swyx** [2:02:32] Cool. Amazing. Uh, last thing, hike. **Doug O'Laughlin** [2:02:35] Yeah. **Swyx** [2:02:35] Uh, so, uh, one bit of context for me is, uh, I ha- I just, I just, I've never taken a break. Never. Um, and I feel like, you know, if you take a break in this time, you're, like, just gonna be so behind. **Doug O'Laughlin** [2:02:48] Yeah. **Swyx** [2:02:48] You're just gonna so miss out. Uh, I just found out my friend from OpenAI took a break a year off to bike through Japan. **Doug O'Laughlin** [2:02:56] Recent. **Swyx** [2:02:56] And like, how, how could you? Like, you're gonna miss it. You're gonna miss everything. But he's like, "I'm good," you know? Like, like, "I'm, I'm, you know, having kids," or whatever. You did a sabbatical as well, and, like, it was pre-AI. But, uh, it was interesting. I, I, I, you did, you did the Appalachian Trail? Which one was it? **Doug O'Laughlin** [2:03:12] Uh, so there's three big ones in the United States. **Swyx** [2:03:13] Yeah. **Doug O'Laughlin** [2:03:14] There's the Appalachian, the Pacific Crest Trail, and then there's the Continental Divide Trail. So I did the Continental Divide Trail, which is the longest and most remote of the three. **Swyx** [2:03:23] Okay. **Doug O'Laughlin** [2:03:23] Sometimes considered, like, the, the, the older, ba- whatever. But, like, honestly the PC... They're all different trails. Like, I'm, I'm pretty steeped in hiking culture. Um, I think mile for mile, AT is actually the hardest. But I did the CDT as my first trail, as my first through hike. Um, you know, you learn a little bit about the three when I was choosing which one I wanted to do, and the CDT was the one that scared me the most. Uh, it was like, hey, this would be the hardest, biggest accomplishment I could possibly imagine. And I thought, if I never have an opportunity to ever do this ever again, which so far seems to be pretty correct- ... um, which one am I gonna do to feel the most like, hey, I did the thing that I really wanted to do? 'Cause I've always wanted to do a long-distance hike. And so I, I chose the Continental Divide Trail. I did that in 2021, pre-AI, and- **Swyx** [2:04:05] But after the GPT-3 essay. **Doug O'Laughlin** [2:04:07] A- after the GPT-3 essay, yeah. I felt like I was missing out a lot, and there was, like, a huge... It was a huge year for Substack. I feel like I missed out, like, a very big year of, like, the big growth. **Swyx** [2:04:16] You're doing okay. **Doug O'Laughlin** [2:04:17] Um, I'm, I'm doing, I'm doing fine. Um, but I, I just think that for me it's something that I always deeply wanted to do from an intrinsic perspective. I think something is, like- **Swyx** [2:04:25] Like life fulfillment. **Doug O'Laughlin** [2:04:26] Yeah, yeah, life fulfillment. And, and I would definitely do it again, but I probably, um- **Swyx** [2:04:30] And to be clear for people, it's, like, four months, five months. **Doug O'Laughlin** [2:04:32] Uh, six months. **Swyx** [2:04:33] Six months. **Doug O'Laughlin** [2:04:34] Six months. Uh, six months, 2,800 miles we'll, we'll call it on the route. 2850 or whatever the miles I went. **Swyx** [2:04:39] And, like, you meet people on the way, but, like, you're mostly alone. **Doug O'Laughlin** [2:04:41] Mostly alone. Did it alone. You get the trail name. **Swyx** [2:04:43] Dude, it's a whole... Audiobooks? **Doug O'Laughlin** [2:04:45] Oh, I listened to audiobooks until I hated them, listened to music till I hated it. Got bored as hell. Like, you just, you just, you go, you f- you go through all of it actually. **Swyx** [2:04:52] Yeah. **Doug O'Laughlin** [2:04:53] Um, yeah, it was awesome. Six months. I think the thing I think about is so far in most... In, in, in my life up until that point you get, kinda get kicked from situation to situation, right? You create a f- a view, a form of yourself. You think you know yourself. You have ideas of what motivates you, how do you react in situations, blah, blah, blah, blah. I think the one, the CDT amount, like, it's just like I like h- I like the outdoors, I like hiking, I'm, like, good at it, whatever. It's just something I really appealed to me from an adventure perspective. Like, when in modern life do you get to say, "Hey, I'm going on an adventure"? Never. Like, and that's what it was. It was a, it was an adventure for me. Um, and one that I got to, like, really... You, you, you know, it's like, oh, the journey is the destination or whatever. You learn a lot about yourself, in fact. I learned it didn't grow me up per se, but I feel like I am more well-defined of my view of myself. I understand how I react. I actually know where my exact line... Or it's like, you know, you're like, "Oh, I'll go do this." It's like, actually, no, I know my exact line where I'm like, "I would not do that." I know exactly where I'm not... That's too scary, too hard, too whatever. **Swyx** [2:05:54] Yeah. **Doug O'Laughlin** [2:05:55] I know my limits a little better. I feel like I know just more about myself. It is a very condensed version of a very intense life. And yeah, I wouldn't give up that experience for anything in the entire world. It was extremely personally meaningful to me. I think it's very fun to go back to the lower part of the Maslow's hierarchy of needs. Like, all this crap what we're talking about today is so abstract. It's, like, totally fake, and we were not born and built for it. We were born to, like, you know, our human evolution got us to, like, scrape a living in the mud, okay? **Swyx** [2:06:24] Hunt and gather. **Doug O'Laughlin** [2:06:25] Yeah, hunt and gather and just not die. It's kind of interesting to go backwards and to see what feels... Like, dude, I was so hungry, so scared, so alone, so, like... But also, like, super low. The, the, the phrase is, like, lowest lows and highest highs. These crazy lows where you're like, "What am I doing? What does it all mean?" Highest highs of me like, "Holy crap, it's so good just to be alive." All these things where it's like, it's just so, like... The raw experience of life is so meaningful, and you don't get to experience it without doing it that way. And so yeah, I wouldn't... I highly recommend it. It's very... I would do it when you're younger. I wish I did it after college. **Swyx** [2:06:59] Yeah. **Doug O'Laughlin** [2:06:59] Um, like, right after college and said, "Hey, like, whatever, kick us out a year." I think it's good to learn about yourself. **Swyx** [2:07:05] Yeah. **Doug O'Laughlin** [2:07:05] It's really important. Your, your self-mastery is your most important tool use of all, so yeah. **Swyx** [2:07:09] Love that. **Doug O'Laughlin** [2:07:10] Yeah. **Swyx** [2:07:10] Self-mastery is the most important tool to use. **Doug O'Laughlin** [2:07:12] Yeah. **Swyx** [2:07:12] Amazing. Well, thank you for, uh, jumping on and, like, covering everything. **Doug O'Laughlin** [2:07:16] Yeah. **Swyx** [2:07:17] I feel like I got, like, got to go through the sort of Claude Code psychosis all the way to the SemiAnalysis, all, all the way to the hiking. **Doug O'Laughlin** [2:07:23] Yeah. Thank you. **Swyx** [2:07:24] Yeah. **Doug O'Laughlin** [2:07:24] Thank you for having me on. **Swyx** [2:07:25] Yeah. So this, yeah, great to catch up. --- This library is powered by PodHood (https://podhood.com), the podcast website platform.