Transistor Radio: WFE and Doug's Claude Code Psychosis

Summary

ChinaTalk Podcast – "Transistor Radio: WFE and Doug's Claude Code Psychosis"

Date: January 8, 2026
Host: Jordan Schneider
Featured Guests: Doug, Dylan (and cameo commentary/rap from Sam)
Theme: Exploring the intersection of China tech, US-China policy, semiconductor industry developments, the operationalization of AI agents like Claude Code, and deep dives into the workflow and culture of industry analysts.

---

Episode Overview

Main Purpose:
This episode interweaves two primary threads:

1. Announcing and discussing ambitious new plans at SemiAnalysis to build a comprehensive China-focused research arm—including deep tracking of Chinese wafer fab equipment (WFE), chip companies, and the energy/industrial tech landscape.
2. Doug’s “psychosis”-level adoption of Anthropic’s Claude Code (Opus 4.5) AI agent for all facets of information work, from coding to workflow automation—framed as an inflection point as momentous as the advent of deep learning or the earliest “deep research” deployments.

The tone is rapid-fire, geeky, self-deprecating, and highly conversational, hopping from hard tech and hiring plans to rhapsodizing about AI habits—and always in the context of US-China tech competition.

---

Key Discussion Points & Insights

1. SemiAnalysis Launches China Research Practice

• (03:20–06:28) Dylan lays out the 2026 roadmap:
  • SemiAnalysis is building a new vertical devoted to tracking all major Chinese WFE makers to the same granularity as their American/Japanese peers.
  • Plans to cover Chinese chemical, precursor, photoresist, and etchant suppliers, as well as memory, logic, and chip fabrication players.
  • Japan is included in the “Western” research bucket (“we are annexing, so my annex is Japan” – Dylan, 03:32).
  • Early next week, they’ll post detailed breakdowns of China’s WFE share by fab type—noting the confusion around acronyms: “Can we just settle on one? […] I hate this.” (Dylan, 04:25)
• Hiring a superstar team: Dylan shares excitement over a Chinese national grad student who called him out on errors in his analysis and deeply impressed him on niche Chinese DRAM/memory players.

2. The Naming Problem (Chinese Company Names)

• (06:28–06:49) The team discusses the awkwardness of only using Westernized short forms (e.g., CXMT) versus Chinese names (“Changxin”), agreeing to lean into learning the actual Mandarin terms, both for accuracy and deeper immersion.
• “You're about to meet us at our most Chinese point in our life. That's the plan.” (Doug, 06:38)

3. Doug’s Total Claude Code Psychosis

• (07:35–14:29, 15:01–16:16, 20:01–31:25)
  • Doug details his “conversion” to coding and workflow via Claude Code (Anthropic’s AI agent, specifically Opus 4.5), calling it a “step change”:
    • “Opus 4.5 is a step change. I was messing around with it before, but now it one-shots everything. It's way better.” (Doug, 07:35)
    • “It’s all a skill issue now.” (Doug, recurring theme)
    • Doug uses Claude for charting, reporting, coding bots, automating daily and compliance tasks—even calendaring and IPO tracking—outputting everything in Obsidian for persistent, searchable storage (“I’m finally converted to Obsidian,” 10:38).
    • “Every chart in my year end outlook that was made by Claude. Every chart that was in the slack today that was made by, by a Claude. Bro. Like I'm not joking. It's a skill issue now.” (Doug, 08:48)
    • “I vibe coded five kalshi bots.” (Doug, 08:48)

Memorable Workflow Quotes:

• “I’ve completely outsourced all of that to Claude code. I don’t have a single thought. It does everything for me, bro.” (Doug, 09:39)
• “If you open up too big files, too big of files on PDF sometimes that, that creates issues. […] But honestly, man, I don’t know. At this point in time, I think everything that is a problem is a skill issue on my end.” (Doug, 19:18)
• “It’s like, it is completely on me. It’s a skill issue. I, I deeply believe that.” (Doug, 19:38)
• “Opus 4.5 is AGI.” (Doug, 19:10)
• “I’m Claude Codepilled. I feel generally euphoric about its potential.” (Doug, 20:36)

Persistent AI Research/Note-Taking

• Obsidian as a persistent Markdown “second brain”, with Claude driving input and context (“Everything I’ve ever thought of,” Doug, 39:01).
• Comparison to past “note app hype cycles”: “I had a Rome phase. I had an Obsidian phase. This is not. It’s not the same thing. Put some respect on the name. It’s totally different.” (Jordan, 39:13–39:19)

The Skill Issue Mantra

• Throughout, Doug repeats “it’s all a skill issue now”—meaning that the constraint isn’t the AI anymore, but whether you as user know how to employ it, grant it access, and articulate complex tasks.

Bottlenecks and Wishes

• What would make the biggest difference? “If Token throughput could magically 100x, I would be euphoric.” (Doug, 21:06)

4. AI in Real Life (Age of Empires, Custom RTS, AGI Anxiety)

• (12:07–14:29) Dylan describes group Age of Empires II matches with other AI researchers, leading to a roommate (Sholto) coding an entire AI-themed RTS in a week—entirely prompted via Claude.
• (14:54) “He’s just prompting the model constantly… during this game, he’s like, dude, I became a zombie. I just type my ideas and I just click yes.” (Dylan)
• The new game features factions (US vs. China), competing through the industrial to AGI future with unique modifiers and hyper-optimized agentic gameplay.

5. Generative UIs and Workflows – The Near Future

• (16:25–18:07) Dylan shares that Lambda’s founder believes “UI is going to be generative. It’s going to change according to what you prompt and want. Token streaming into your fucking eyeballs.”
• Doug predicts:

  “I am excited for the true maxi future, where you buy a computer and there’s no software at all. It just starts with a Claude Code API and you build your own software... All software, whatever OS, it doesn’t matter.” (Doug, 15:51–16:16)

6. Market Commentary, TSMC Expansion, Bubbles & Industry Gossip

• (22:38–27:27)
  • Quickfire takes on: Trump’s on-again, off-again stance with LBT; Oracle’s “dumb” debt round; Meta Super Intelligence rumors and staff drama; Nvidia’s role “saving Intel” by buying up all TSMC N3 wafers.
  • On TSMC scale limits: “If they were to double their capex, that means they would have to... employ 200% of Taiwan. You run out of construction workers.” (Doug, 26:39)
  • "If you're not a frontier node, you have child labor or Filipinos and Malaysians. I think that's the answer." (Jordan, 27:19)

7. SemiAnalysis Growth & Hiring

• (42:19–43:03, 51:27–52:00)
  • Massive ramp-up: hiring for China analysts, Singapore analysts, AI analysts, technical staff—anywhere in the world. Open calls to “slop sons” (hardcore followers).
  • China research: “I think the world needs better China research. I don't think people will pay for it…” (Dylan, 43:08)
  • Dylan jokes about whether five Singaporean juniors are even necessary: "What if we make them as good as we can, then we give them nuclear power? Then you have five roaming nuclear power bombed arm robots..." (Doug, 37:50)

8. APIs, Tech Stack & Automation Dreams

• Intense bullishness on APIs: “APIs are priceless is my vibe… the data is more valuable than any of the form factor.” (Doug, 36:44)
• Speculation that tools like Bloomberg would be better if they just provided clean APIs for Claude to query versus their current model.

9. Cultural Meme – “Meeting Someone at Their Most Chinese”

• (45:08–46:33) The “you’ve met me at a very Chinese time in my life” meme, symbolizing the rising trendiness of deep China knowledge and industrial culture.
• Doug on the Ishowspeed China tour being an inflection point for China’s cultural trendiness.

10. Personal Life, Morning Routines, and Parenting with AI

• (31:37–32:52, 46:47–47:37)
  • Morning: Doug begins days with Claude-driven workflows; Dylan blends biohacking shakes and sauna time.
  • Jordan’s daughter escapes her crib to demand attention at 6:10 AM; the team jests that Claude Code will raise the next generation.

---

Notable Quotes with Timestamps

On China Research, Ambition, and Hiring:

• “We are going to track everything. All the Chinese wafer fab equipment manufacturers to the detail that we track American ones… All the Chinese chemical suppliers, photoresist, etchants, precursors…”
  — Dylan, 03:32

• "You're about to meet us at our most Chinese point in our life. That's the plan."
  — Doug, 06:38

On Claude Code Workflow & AI Agency:

• “Cloud code is delicious and it's a little sycophantic and it's super addictive and I have completely decided to give my life up to it… it's all a skill issue now. Like I'm really telling you it's all a skill issue.”
  — Doug, 08:05–08:48

• "Opus 4.5 is AGI."
  — Doug, 19:10

• “I have daily calendars. Be like, hey, I needed this by this day. And whatever's like, hey, I need you to give this to me… For example, the reason why I knew what IPO today is because I put it in. Like when I did the filing, I like did. I had Claude code look for the filings for Minimax and gpu… Made it all of it, exported it all to Excel... It’s pretty good.”
  — Doug, 09:39

• "Honestly, it's all skill. It's all skill issues now. I'm very, very, very Claude code code pilled." — Doug, 11:24

On Workflow Transformation:

• "It's just a home for markdown files. And we have the AI that lives on top of it, which is what we all dreamed of."
  — Jordan, 11:38

• “It's the first time where I feel like—I just don't think I would let you rip it out of my cold dead hands… Put some respect on the name. It’s totally different.”
  — Doug, 39:13–39:19

On Generative UIs, Tech Stack, and the Future:

• “I'm excited for the true maxi future where you buy a computer and there's no software at all. It just starts with a Claude Code API and… you build your own software…”
  — Doug, 15:51
• “UI is going to change according to what you say and what you prompt and what you want. And, and, and it's just going to be token streaming into your fucking eyeballs.”
  — Dylan (on Lambda's founder), 16:25

On the “Skill Issue” Mantra:

• "At this point in time, I think everything that is a problem is a skill issue on my end… It’s like, it is completely on me. It’s a skill issue. I, I deeply believe that."
  — Doug, 19:38

• "The more important, most important part about this is like having a persistent store. So I've been using Obsidian. I'm finally converted to Obsidian." — Doug, 10:38

On Industrial China’s Cultural Moment:

• “China’s hot… I think the Ishowspeed China tour… that was one of the first things that really made it trendy and cool. And then all the influencers started to go, and then they're like, oh my God, it's so futuristic.”
  — Doug, 45:14

---

Fun & Memorable Moments

• AI breakfast shakes and sauna: Dylan’s quirky morning stack—a Brian Johnson protocol–inspired shake consumed in the sauna, accompanies Slack/text catching-up (31:46)
• Age of Empires II as AI Researcher’s Playground: Dylan hosts epic AOE2 tournaments with researchers from Anthropic, OpenAI, etc. (12:07–13:05)
• Meta Super Intelligence drama: The rumored “factions” and staff rumors in the Meta AI bubble, complete with 996 Chinese labor jokes. (24:05–24:35)
• Venezuelan turbines as a meme: Dylan’s X post joking about America invading Venezuela for “underused turbines” leads to real policymakers pinging him. (52:08)
• Sam’s satirical rap outro:

  “White on black is the truth / Claw’s my only gun / I’ll call him Maxi / Pill you every chart I ever posted / didn’t touch a single key...”
  (Sam, 54:04–54:50)

---

Suggested Listening (Timestamps)

• SemiAnalysis China Practice Launch: 03:20–06:28
• Doug’s Claude Code Manifesto: 07:35–14:29
• Workflow Meltdown and Automation: 20:01–31:25
• Market Bubble/Gossip Round: 22:38–27:27
• Hiring / SemiAnalysis Expansion: 42:19–43:03, 51:27–52:00
• "Skill Issue" Mantra and AI Psychosis: Throughout, but especially 19:10–21:25, 37:50–41:34
• China Cultural Trendiness: 45:08–46:33
• Sam’s Rap Summary: 54:04–54:50

---

Conclusion

This episode is a riotous mix of inside-baseball industry gossip, bravado about AI “changing everything,” and sincere, detail-rich discussion of the challenges (and cultural fun) of tracking China’s tech industry.
Doug’s “Claude code pilling” is not just a workflow upgrade; it’s treated as a lived, euphoric paradigm shift—one with real stakes for the future of information work, competitive analysis, and even personal agency at the computer.

Final Quote:
"I'm Claude Codepilled. I feel generally euphoric about its potential. And personally, I think everyone should try it because I think it will change your mind of what's possible." – Doug, 20:36

---

For more: Check out ChinaTalk’s newsletter. If you want to be part of this “slop son” analyst generation, open calls for China-savvy, AI-literate researchers are open!

Transcript

[Doug]

Today's. Hello, welcome to transistor Radio. Today we're going to start with a sponsored post by Dylan and Jordan talking about shaving their beard. Anyways, continue.

[Dylan]

So the beauty of this all is right before we got on, Jordan's like, did you do this professionally? I'm like, whoa, whoa, who. If you're going to riz me, let's record.

[Jordan]

Oh, my God, I can't. Oh, no. I can't compliment Dylan on air. This, like, defeats the whole purpose of this podcast.

[Dylan]

Dude, come on. You mean my ego's already too large?

[Doug]

I'm worried it'll become overinflated, bro. Come on.

[Jordan]

All right, how did we like my. How did we like my gimmick to kick us off?

[Doug]

What?

[Jordan]

I thought it was rather clever. 2026 abduct annex or pardon. But there.

[Doug]

Aren't they all. Aren't they all, like, kind of bullish, isn't it? Bullish, Bullish and bullish. I don't get it.

[Dylan]

Abduct. Who do we.

[Doug]

He's trying to do kill. What is it?

[Jordan]

Yeah, you know, 2026 style.

[Doug]

But it's, like, Trump themed. Because. But I don't which one is which. Help me with that.

[Jordan]

All right, so, yeah, no, pardon is like, it's gonna disappoint you, but, like, oh, whatever.

[Doug]

Okay, so. So it's. So we're. Okay. So is that okay?

[Jordan]

So, like, you want all of it and. And abduct is like you also want a little thing or like, we're just going to, you know, we want, like, a 30%.

[Doug]

Okay, so that's. So there's no kill is what you're telling me. And they're all. They're all. They're all bullish.

[Jordan]

You were very lyrical in your post today. Like, the drinks are still popping. It's, you know, 1926.

[Dylan]

Like, we're. We're.

[Jordan]

We're doing all right. What's there to.

[Doug]

What's there?

[Dylan]

Wait, did you post today, Doug?

[Doug]

Yeah, dude, I fucking did on fabricating.

[Dylan]

Oh, Outlook, bro. This is pretty good. I like these. I should read this.

[Doug]

Don't even worry about it, man. I've been grinding so hard. I'm so exhausted today. Actually, if you know what I do want to talk about, I want to talk about Claude code. I've had a re. Recent conversion. Okay.

[Jordan]

Whoa, wait. Can I.

[Dylan]

Can I.

[Jordan]

Can I just give you your lyrical chops for a second?

[Doug]

Sure, Please do.

[Jordan]

Can I read the final paragraph? All right. This is history in the making of the biggest capital cycle of my lifetime. Probably we're not at the end. We're drunkenly dancing somewhere in that timeless night. Maybe that itself is an omission that it ends this year, but I think it continues.

[Doug]

See you next year.

[Jordan]

Okay, clock, take us away, Doug.

[Dylan]

Okay, where is this beautiful writing and core research, bro?

[Doug]

Dude, I'm. It's. Dude, there's some. It's no value to be beautiful writing. It has to be like, stunk good, stunk bad.

[Jordan]

To be fair, that's what Dylan. That's what Doug did for like 2000 words before. He was like, oh, maybe I'll put a sentence together.

[Doug]

Yeah, right at the end. I was like, yeah, maybe I'll do something. I was like, what does the vibe. Oh, sorry.

[Dylan]

You know what I, what I do like about core research is people are always like, man, this is like the one thing the weekly. They're like, yeah, if I don't have time, this the one thing I read always. Because it's so valuable and it's dense, which is nice.

[Doug]

This is an advertisement, bro. Welcome to the others Paid.

[Dylan]

Sponsored all right, so what I wanted to talk about is our plans this year at semianalysis for 2026. Right. Jordan has refused multiple acquisition offers. And so we are kicking off a China research practice.

[Jordan]

Great.

[Dylan]

Where we are going to track everything. All the Chinese wafer fab equipment manufacturers to the detail that we track American ones. All the Chinese chemical suppliers. Right. Photoresist, etchants, precursors, et cetera, to the detail that we do the western firms. When I say western, I include Japan. Because as you said, Jordan, we are annexing. Right. And so that my annex is Japan. So I'll keep on track for both. And then for the other, it's going up the stack, right. The fabs. There is so much on the side of memory and logic, both trailing edge and front end. So early next week we're going to have a post showing what we think China WFE share is across, you know, different types of fabs and where it's going by year. Because. Because recently there was this. Oh, all fabs must be 50%. But the asterisk. Chinese WFE, semiconductor manufacturing equipment. Which fab equipment. But by the way, can we just settle on one? Like, what the fuck? I hate this.

[Doug]

Why is there SME or SME?

[Dylan]

SME and wfe.

[Doug]

I just like w. Because SME has two. It's like almost like swe. Like, what is a software engineering Software, bro.

[Dylan]

That's dead. Doug, come on, tell us about cloud code.

[Doug]

Okay, sorry. Honestly, dude, I'm going to Go hard on cloud code. So let's keep going on the WFE thing while we're at it.

[Jordan]

Well, my question is, like, wait, are we going all the way up to Stack? We're doing the models in the labs.

[Doug]

Tangents.

[Dylan]

Tangent, yeah, of course. You didn't let me go up the stack. You know, you're trying to do tangents. You know, the next. Next layer up from fabs is, you know, So I interviewed a candidate this week, and I'm like, holy fuck. I want this person, you know, Chinese national doing grad school in America. She's amazing. I want to hire her because she. She said I was wrong. Which I love when my candidates tell me I'm wrong. And they're smart about it, right? They're like, oh, you're wrong about xyz. And I'm like, huh, Tell me more. And then we talked about 25 minutes about this niche ass memory player that I'd never heard of. And I'm like, holy fuck. Yes. And also cxmt. So we talked about both of those things for, like, 25 minutes. Anyway, so we're assembling a badass team.

[Doug]

Also, ironically, they're going to have the best data on public models From China. The IPO for JeepU. Is that. How do I say it? How do I say it in Chinese? Jordan, I feel like you'd know. Gerpool. It's Jirpu. Uh, Dripu is today.

[Dylan]

So I thought, I thought it already went up, like, a lot.

[Doug]

Uh, no, it went up. No, no, no, it. It, it filed. Let's see what it is.

[Jordan]

It filed like a week ago, I think. I think it's the 10th or 9th or something.

[Doug]

I think 8th is when the filing will debut on the Hong Kong exchange on January 8th and 9th. Okay. And mini matches on the 9th.

[Dylan]

Know. You know, what we. What we really need to do, Doug, is actually just know the Chinese names. Like, we got to stop saying cxmt and we got to be like Changxin or whatever the heck the name is, right? And we gotta like, dude, we're gonna create that in your heads.

[Doug]

Oh, even better. You're gonna start to. We're gonna become more Chinese. You're gonna meet us at our most Chinese point in our life. That's the plan. You're. You're about to meet.

[Jordan]

Semi analysis there and then that. That'll probably be the peak.

[Doug]

We need you at your most Chinese part of your time.

[Jordan]

Yeah, Dylan, the. Your Xiaohongshu profile is not gonna save you, but maybe.

[Dylan]

Wait, why not? Whoa.

[Doug]

What'S been working on.

[Dylan]

I think, I think my Shao poor fellow is gonna save me.

[Doug]

Oh, oh, oh, oh. I mean, I mean he. I think save means when Xi Jinping finally. Finally.

[Dylan]

Yeah. When Xi Jinping puts the gun to my head, I'm going to be like.

[Jordan]

Moving on Doug. Claude code.

[Doug]

Okay. Speaking of which, dude, I am slow. So, so, so, so, so so. Claude code pill. Like all the hype on on X is real. And I was vibe coding with Claude code before. Opus 4.5 is a step change. I was messing around with Opus 4.5 when it first came out, but I don't know what they've done since then. It's way better it one shots. Everything I know is supposed to be, you know. I know Codex 5.2 is better according to Dylan. That's what he said.

[Dylan]

No, no, no, no, no, no, no. Claude code is better at different things.

[Doug]

COD code is delicious and it. And it's a little sycophantic and it's super addictive and I have completely decided to give my life up to it. And so currently, and I'm. I'm going to do. We're doing internal webinar on like the best pract because I've been doing it every single day for the last week. I hit my usage limit every week now, which is pretty ridiculous.

[Jordan]

You're not Max.

[Doug]

I'm on Max. Yes, on Max.

[Dylan]

What? It's really cool making bro.

[Doug]

Yeah, you'll see. Did you see all those charts today? Do you think I made those? No.

[Dylan]

Wait, what? I thought you were trolling. I was like hey, change the, change the sourcing, change the graph. I'm like telling him every single chart. I think he's joking.

[Doug]

No, no, no, dude. Every chart in my year end outlook that was made by Claude. Every chart that was in the slack today that was made by, by a Claude. Bro. Like I'm not joking. It's a skill issue now. Like I'm really telling you it's all a skill issue. I. I vibe coded five kalshi bots. Like bro, I have like I just vibe coded all of my notes. I'm not gonna tell you. I don't want to get my.

[Dylan]

Here's this problem at semi analysis, right? Like for me I don't want people like trading stocks. So we obviously like semiconductor companies are off limits, right? But I keep wanting to make our like restrictions even tighter and tighter so people don't look at the stock market. There was this channel where people were sharing degen ideas. I got added and I shut it down and everyone was so sad because Doug was sharing all his like, like xyz. I don't want to say Ozempic longs, but whatever the fuck is like thing is nowadays I have some.

[Doug]

Well, I've. I've completely outsourced all of that to Claude code. I don't have a single thought. It does everything for me, bro. Everything. It. Dude, I copy, put, paste, standup notes. I like fucking everything, man. I have daily calendars. Be like, hey, I needed this by this day. And whatever's like, hey, I need you to give this to me. Like, I have ca. Like, for example, the reason why I knew what IPO today is because I put it in. Like when I did the filing, I like did. I had Claude code look for the filings for Minimax and gpu. Gpu. We're just gonna. I'm fucked it up. And then I had it like download it, found the English prospectus, made it all of it, exported it all to Excel, gave it to like ins. It was pretty good. And then like did that for the CXMT IPO was like, hey, make sure you do on these days, please, please tell me the day it actually comes out. Everything. It's all a skill issue. Like, guys, you have to think like coding is probably harder than doing, than writing notes and. Okay, but the more important, most important part about this is like having a persistent store. So I've been using Obsidian. I'm finally converted to Obsidian.

[Jordan]

Jordan.

[Dylan]

No.

[Doug]

Yes. But the thing is I haven't like, I don't write the structure. I don't do any of that. I just open a daily note. Maybe I'll like review some notes, maybe I'll add some note, like note changes. But Claude code is my primary, is my primary driver. And so like every single time I do a debrece. Also like ghetto bootstrap. My own deep research. And then like every single time it pushes out an output, it goes into a cloud code into Obsidian markdown note. Because markdown's like really go to for this. And then I can search. I can be like, hey, what was that? Deep research, blah, blah, blah, blah, blah. And importantly, it's like having context across all the things. And you'd be like, hey, I want you to specifically do a deep research from the conclusions of these five deep researches. Blah, blah, blah, blah, blah. You should reference this. And I have like all these fucking tools. Honestly, it's all skill. It's all skill issues. Now I'm very, very, very Claude code code pilled. I I. I mean, sorry.

[Jordan]

Remember all with the obscene coach.

[Doug]

I.

[Jordan]

Maybe we're missing something. We paid some college student 50 bucks to talk to us for half an hour. And then I think neither of us ever opened Obsidian again. We're just like. Because so dumb. And here we are three years later. Like, it's just. It's just a home for markdown files. And we have the AI that lives on top of it, which is what we all dreamed of.

[Dylan]

That's.

[Jordan]

That's really.

[Dylan]

You also do this, Jordan, or just Doug?

[Jordan]

Well, I think I'm going to start now. I'll report back next transistor radio.

[Dylan]

I'll tell you guys about my cloud code pilling. So over the holidays we got absurdly into age of Empires 2. For reference, I've been falling asleep to age of Empires two videos from T90 for a while. But specifically we started playing. It's a bunch of AI researchers. So it's like anthropic people, OpenAI people, GDM people, et cetera. Um, we're just playing Age of Empires. We've done like a handful like four or five of these sessions where everyone comes over to my house. We do 5v5s. It's a ton of fun. And a couple people there have just gotten so, so into it. Specifically one of them being Sholto, my roommate. He. He's. He. He created an entire RTS from scratch in one week. About like AI. Like you start off in the industrial.

[Doug]

Age and then you're not shipping.

[Dylan]

Well, he had the week.

[Doug]

She'll get back to work. Sure. Every week off to do more work. Bro, I had a week off to make a super semi analysis giga brain and he's making an rts bro.

[Jordan]

You guys, I did for the holidays, I did Buddhism. It was great. No AI. I was just reading my.

[Dylan]

So mindful.

[Jordan]

I know.

[Dylan]

Wait.

[Jordan]

So good. Next time just you wait.

[Dylan]

So, so, so this RTS is pretty sick because it goes all the way through like industrial age, all the way through like the AI timelines. Like like you know, using AI to accelerate your research. Using AI. You have the. You have your. You have an entire like tech tree of like researching models and like implementing them in society. You have robotics, you have like, you replace all the people with robots eventually you have Internet, you have hacking, you have like all this shit. Um, it's. It's pretty fucking sick. And you have two factions, US and China. And they're fight, they're racing against each other. They have different modifiers, right? Obviously Chinese are good at production and you know, these things in power. And so there's like power grids. It's, it's. It's an incredible, incredible unit.

[Jordan]

Do you move around on the board? Is it like text space?

[Dylan]

And it's more performant? It's more performant than any RTS ever because it, it can handle battles of up to 50,000 units versus 50,000 units.

[Jordan]

Oh my God. I gotta, I gotta.

[Doug]

And so was it Claude coded, I presume?

[Dylan]

Cloud coded.

[Doug]

Dude, I'm telling you, it's all a fucking skill issue. I'm not joking. I've come to that conclusion too. Like, dude, everything is a skill issue. It is how much agency and how.

[Dylan]

Much time you want to get into. Two months. He hasn't written a single line of code. He's just prompting the model constantly. And, and, and at least at work, it's like for him, he's like, oh, you know, the quality of the code I must submit is like really good, right? Because it's like he's, he's leading RL infrastructure, right? You know, basically everyone's work is built on top of his shit. But during this game, he's like, dude, I became a zombie. I just type my ideas and I just click yes.

[Jordan]

Actually, like some Enders game shit is sholto. Like, just like. These are all just people around the world.

[Doug]

Okay. Okay, So I do during work. So. So, okay, this is also a little bit of confession. 1. I only really started this hardcore stuff maybe on the 29th, and then like obviously took whatever. But I've been. I've been obviously having psychosis since then. Um, but pretty much what I do every night, every day, I go to work and I like git. Cause it's synced via git. And I'm like, hey, update the thing and whatever. Then I work, work, work, work, work. Last thing I do, I say git, git sync. And then I come home and I open up my. My fucking Mac. Because Mac is way better on. On COD code. Cause like fucking WSL is total trash. And. And then I just do git sync and I just keep. I just. It's like an addiction, bro. I have like six tabs open and I just keep going to each tab in each thread and I'm like, oh, wait for it to be done. Okay, done. And then I'm like, oh, wait for it to be done, then be done. It. It feels. It feels so ridiculous.

[Jordan]

Terminal. You're just on like. Claude.

[Doug]

No, it's all. No, it's all on Terminal. Terminal's better, dude. I Tried to do the Claude web thing. It's. It's trash, bro. Terminal and also like CLI is just like aesthetically pleasing. I'm excited for the true maxi future where you buy a computer and there's no, there's no software at all. It just starts with a Claude Code API and all, and then you build your own software based on your own computer. Like, I earnestly believe that I would take.

[Dylan]

When is Doug going to go down the Apple OS route?

[Doug]

Dude, I will take. I will take a bet. I will take a bet that's like 20 years out. Like for real. Like all software, whatever os your bullshit. It's like, no, it doesn't matter. You whatever.

[Dylan]

Lambda the Neo Cloud. So Lambda the Neo Cloud, their founder is like super pilled on this generative ui. He. He's like, obviously they're running a Neo cloud. But he's like, yeah, like what? Basically every, every UI is going to be generative. We're. We're gonna. It's gonna. UI is going to change according to what you say and what you prompt and what you want and, and, and it's just going to be token streaming into your fucking eyeballs.

[Doug]

And it'll be like, it'll literally map to your little. It'll be like a sycophantic little mapping where it like maps to how you consume information. Like, Dylan, Blind Dylan's a colorblind little fuck. So it'll be only just the little colors that. Max, wait, no, I'm not color.

[Dylan]

I'm not colorblind. I'm partially. I'm bad at colors. I'm not blind.

[Doug]

Okay, but. Okay, but it's going to make it.

[Dylan]

I am ready.

[Doug]

His will be like, his will. His will be like insane. Like, the colors will be like really wonky, but it'll be like, ooh, this is what gives me the dopamine. Like, and then it'll just be. You'll be addicted.

[Dylan]

You'll be like, little Doug's will be grayscale because he's Monk mode.

[Doug]

No, dude, it's fucking.

[Dylan]

It.

[Doug]

CLI is kind of grayscale, bro. I'm telling you, I'm so. Claude Code pill.

[Dylan]

Did you know the. Did you know the main guy behind Claude Code uses white terminal, not black?

[Doug]

Yeah, bro.

[Dylan]

Wow.

[Doug]

You know, do you know that there's. There is.

[Dylan]

He's Russian.

[Doug]

There is this OG advertiser guy. I think it's Oligiv's advertising method. And pretty much, no, it's Oligivi. Then there's tested advertising Methods. Two different books. But tldr, there's been a lot of research done on this and I'm sorry to break this to you, Dylan, but black on white is the most readable, most highest throughput, most effective. And it's always been that way. It's always been that way. But here's something.

[Dylan]

I'm with you. I like black. I mean, like, look at my website. Like, come on.

[Doug]

Like, it's black on white on white, bro.

[Dylan]

Oh, sorry. Black on white. No, no, no, sorry.

[Doug]

Yes, white with the white background. Anyways, but that's a tldr. Honestly, maybe, maybe I should try. Maybe I should try white. White. Cli. Dude, I'll try that tomorrow. I. I name all my tags. Also. I named. I named. I named the system too, because I want it to be interpersonal. I also have been thinking about on like. Because every morning I say, hey, like, like, you know, make my daily note effectively. I was thinking like, maybe can generate a little ace, like a little art. I'm like, I'm having psychosis, guys. Like, I feel I. Okay, I'm not having psychosis, but I also understand that. Or like, rather I feel like for the first time in my life I can understand what LLM psychosis looks like. Because I'm like, oh, it's. If it just kept sycophantically agreeing with me and just kept like hitting my dopamine receptors, I could go completely batshit crazy. Um, anyways, that's all to say. Opus 4.5 is AGI.

[Jordan]

What's a. What's the limit you've ran into? How has it disappointed you? It's gotta. You've gotta run into some walls with 10 days of non stop usage.

[Doug]

If you open up too big files, too big of files on PDF sometimes that, that creates issues. Um, I reported that bug. I don't know, it's just. And. But then it gets like. It gets like stuck in these random loops. Um, but honestly, man, I don't know. At this point in time, I think everything that is a problem is a skill issue on my end. I, I believe that.

[Jordan]

I.

[Doug]

It's like. It's like driving in a car. And for the first time in my life I'm like, there is nothing like. It is, it is completely on me. It's a skill issue. I, I deeply believe that. And I think, I think people who haven't played around with it heavily is like, it's like making a mistake. You're. You're like watching the future right in front of you. It has Already happened. It is cloud code and, like, the derivative of whatever. And it's like, you shouldn't even call it cloud code because this isn't code.

[Dylan]

It's.

[Doug]

It's cloud computer. Someone told that to me. And it makes so much more sense because it's like, hey, it's an LLM that has full access to your computer and it can do whatever it takes, like in whatever, like, file markdown, whatever the fuck you need it to do. It is. Yeah. And. And if you think about it, dude, what are you doing all day? I'm on my fucking computer all day. All we do information work wise, is being on our computer all day. They are perfect information agents, and they're an opus in its sycophantic little manipulative way, is ready for it to be shaped into your perfect little, you know, agent thing. So, anyways, that is all to say. I'm Claude Codepilled. I feel generally euphoric about its potential. And personally, I think everyone should try it because I think it will change your mind of what's possible.

[Jordan]

Well, so how much further does it need to get? Or, like, what is the form factor that needs to change for it not just to be the Doug o' Laughlins of the world, but, like, I don't know, everyone working everywhere and every small business to have this, like, change how they.

[Doug]

I think. I think we're. I think people. I. I think it will this year. People will figure it out this year. But, like, it's still. We're still early on that S curve of whatever. Honestly, what would be my biggest ask is, like, if Token throughput could magically 100x, I would be euphoric.

[Dylan]

I think that's why you acquired Grok, I guess.

[Jordan]

Dude, I. Dumbest, dumbest story of 2025 had to be that, right? I mean, dear Lord, come on.

[Doug]

All of 2025.

[Dylan]

Okay.

[Doug]

Yeah, all of 2025. It's a big year, bro.

[Dylan]

What was the dumb story?

[Doug]

I thought tariffs were really dumb, but that also did own me with the whiplash. So.

[Dylan]

I don't know, I feel like.

[Doug]

Let's think about, like, semiconductor AI stuff.

[Jordan]

Oh, the. Well, I think LBT going from Trump saying he needs to be fired to being best buds the day later, that was pretty hysterical. Buying 10% of the company. I like that.

[Doug]

I. I think. I think Trump was just, you know, playing hot and cold with a girlfriend.

[Jordan]

You know, just trying to drive a bargain. See where. I see where he's coming from.

[Doug]

I think maybe not the dumbest thing, but I think the Oracle stuff, the last round of debt has been really stupid. I think they shouldn't have done it. The last round of debt, because like all the cds, that's, that's literally. Because why is it dumb?

[Dylan]

I mean, don't they need the debt?

[Doug]

They could have waited like literally three months and I think it would have been okay. Yeah.

[Dylan]

And then they would have had the revenue from the beginning.

[Doug]

Yeah, that's what I'm saying. Yeah. They literally, they, they're like, oh, big, big number. And everyone's like, wow, that's a big number, but probably next year you're okay. And they're like, bigger number. And everyone's like, I don't really believe that. And then CDS kicked off all this stuff. Everyone's like, is it a bubble? Like all the really, like, fear of the of. And also, if I'm being honest with you, it's completely like, it's like an own goal. Oracle, Oracle could have completely prevented this. Everyone else is totally fine and everyone's freaking out about just Oracle. And they could have just waited six months.

[Dylan]

Yeah, but it's a bunch of finance bros who don't understand that OpenAI is going to be able to pay for it and so they need to stop coping.

[Doug]

You say that, but look at the fucking stats, dude. They're losing share. Code red. Code red.

[Jordan]

Well, I mean, dumbest in progress. It's a little too soon to tell, but Meta super intelligence, I mean.

[Dylan]

Whoa, whoa, whoa, whoa, whoa, whoa, whoa. Meta super intelligence is a fantastic firm. I love all my top five China team.

[Jordan]

It's got revenues, got a couple.

[Dylan]

I love all my top five customers.

[Doug]

I. Oh, actually have you heard this like random tea, the, the Chinese, like the Chinese X tweet or whatever being like there's factions within Meta SSI and one of them was like, yeah, this Alex Wang stuff. And then they like scrounge enough money together to buy Manus and it's like a Chinese only speaking team. They're like fighting for their lives. I.

[Dylan]

They're not Chinese only speaking, bro.

[Doug]

I, I believe it's fake, to be clear. It's fake. I'm going to say fugazi, but I like, I like the drama in my brain that There are some 996ers fighting for the lives.

[Dylan]

My friend told me yesterday that she actually was the, was the one doing the DD on them.

[Jordan]

Acquisition more than I do Grok, because like, they just need some people to make products. And that's not going to be your fancy AI researchers who would Just quit if you made them do actual work.

[Doug]

Let's actually do some rephrasing of Grok because I think and Grok with a Q. Okay, so what is in a bigger shortage than even DRAM right now or optics? People who can make accelerators and gen.

[Dylan]

What's in an even bigger shortage? N3 wafers.

[Doug]

Ooh, I don't think so. There's a simple little problem to solve for that, bro. You buy asm. You buy. You buy a semicap. We're going to see one.

[Dylan]

No, for. Not for Nvidia. Nvidia can't just buy semicap.

[Doug]

No, no, yeah, you're right. You're right. TSMC can do that. But TSMC is being. I mean dude, intel is going to work because Gigabrain Nvidia is actually saving intel because they bought all the entry.

[Dylan]

Is a thousand twelve hundred bucks.

[Doug]

I told you bro, we've been bullish man. But like.

[Dylan]

Oh, I know we've been bullish but.

[Doug]

Like, but still like. Yeah, no, because dude, because I'm. Because the orders have fucking moon anyways. No, but if you think about Giga brain. Gigabrain gigabrain WFE 140B.

[Dylan]

WFE 140B.

[Doug]

I. We'll see. I'm not going to do. Sure. Well, but, but think about this way. Jensen bought all the N3 wafers, right? And because of that, Apple with their scrappy little seconds and no volume growth smartphone has to go to Intel. So Jensen saved Intel by stealing all of Apple's wafers. Gigabrate. He's very benevolent.

[Jordan]

And, and tsm, they're just not going to expand because they're like Taiwanese and want to run a country for company for 100 years. Like what, what's the deal here?

[Dylan]

They're. They're expanding. They're expanding faster than they ever have. They're doing record capex. But you know, what's the phrase people say on the playground when someone doesn't want to do something? No balls. No, I'm just kidding. I'm just kidding.

[Doug]

I think, I think it's just hard to deploy that much. Like you know, they already employ everyone in Taiwan. If they were to double their capex, that means they would have to hypothetically employ two people. 2. 200% of Taiwan. You know, we're gonna run out of construction workers. We'll have to. We'll have to couldn't. You know we'll have to sees the, the bike manufacturing giant giant ever. All the employees have to go work at tsmc. They, they graduate high school and they directly work at the fab. They, you know, they literally get the diploma.

[Jordan]

Yeah, we're lowering each requirement.

[Doug]

Yeah, exactly. Exactly. When you. They're actually lowering to 15, their child labor will be.

[Jordan]

If you're not a frontier. If you're not a. A frontier node, you basically, you have child labor or Filipinos and Malaysians. I think that's the answer, right?

[Doug]

Yeah, that's the. That's. That's what Taiwan needs to do. I mean, that isn't all of it, but I think really, it comes down to, like, there is a friction of deploying that much capital. And just in terms of shells also.

[Dylan]

They just don't believe.

[Doug]

Yeah, they don't. They're not maxis. But this is where they should be listening to this beautiful transistor radio, and they should try cloud code.

[Dylan]

I hope they're not listening because I just said they don't have balls.

[Doug]

Well, Dylan, that's your problem, not mine, bro.

[Jordan]

Maybe they just need more people to insult their masculinity. Like, they're.

[Doug]

If you feel. If you feel really upset about Dylan saying you have no balls. TSMC Executive, please tweet at Dylan for feedback.

[Dylan]

Please buy more data.

[Doug]

Well, I think. Yeah, I think I'm being serious. Quadcode, I think, can do it. I really feel like it's good enough to make a bubble. Like, we could have a whole bubble on. In my opinion, it's like, that powerful. I'm. I'm. I'm. I'm never. I'm such a believer. Like, this is a big. It's a bigger moment than 03 for me. Than O3. Deep research. It's bigger moment than all of that for me.

[Jordan]

Bigger than, oh, one. Yeah, bigger than 01. All right.

[Dylan]

I mean. I mean, Deep Research was the biggest moment, but now this. That's crazy.

[Doug]

No, I'm getting serious.

[Dylan]

Do I need to get on this, Doug, or do I need to just, like, read all my Slack messages and reply to everyone?

[Doug]

Dude, you can actually have it read to read your Slack messages. This is all skill.

[Dylan]

I don't. I don't think it can, because I have to, like, conduit the information.

[Doug]

Yeah, that's true. I mean, you could just control A, bro.

[Dylan]

Control A. Control C. Control Shift tab.

[Jordan]

If we still need to control A, like, we're not getting to the macro.

[Dylan]

Control A, Control C. Alt tab. Control V. Enter Alt tab. Yeah, there you go.

[Doug]

So sometimes when I'm really in a deep COD Code Research, I feel like I'm the robot because it's Doing all the work and I'm just chauffeuring information back and forth. Should think about that. That's big brain stuff, bro.

[Jordan]

Anyway, sorry you needed to. Not like you. We're in order for it to matter to every business in America, right? Like. Like, I understand. Like. Like it needs to just be able to take over your computer, right, and do everything and you would be able to trust it.

[Doug]

But I'm being serious. It's all a skill issue. I saw someone said that they got COD code to return a shoe and then reorder a larger one on Amazon with just. With just the ask. You. You re. I'm not joking. It's literally connected to Amazon. You could do. You could just open up your fucking website. Like, if you given. I've given login. Like, I'm like, whatever, bro. Here's the fucking logins. Like, here's the API key. I'm like, it's all on one credit card, so I'm like, if some. If something just blows up, I'm like, delete. But I'm just like, hey, have fun, bro. It's all a skill issue in my opinion. Like, give it more access. Say, like, say whatever. You could do whatever the fuck you want with it. I. Whatever. The MCP server gods are gonna like smite me for saying this, but like, it can just like. Okay. An example, Dylan, is I'm trying to automate the pro, and this one's pretty hard. Automate the trade request form for our compliance process. I'm pretty close, actually.

[Dylan]

I don't know if I want to automate our compliance, bro. That's not something we should.

[Doug]

No, no, no, no, no, no, no, no. Not. Not our compliance. The trade request form. It's just a form. And no, we're not automating compliance for my. No, no, no. It's just like, hey. But it also keeps me honest too. It's like, hey, what's a restricted list? Like, all this stuff, dude, it's all a skill issue. What I'm saying is everything is a skill issue. You need to reframe. Everything is a cloud code skill issue. That is. And like, I mean, someone. I forgot. Which I forgot who said this. It was on one of the frickin door cash podcasts being like, hey, all of this stuff is like a skill issue of prompting or whatever. Whatever tools you're using. And I really. I truly believe that it's. It's mostly a skill issue. It's on you to figure it out. But yeah, the great.

[Jordan]

The great Andre flip is you know, November 2025, he pops the bubble and then December, he self reinflates it saying cloud code is AGI.

[Dylan]

You guys want to know how I, how I start my mornings now? It's pretty interesting.

[Jordan]

Sure.

[Dylan]

I was like, okay, I guess not.

[Doug]

Yeah, yeah, I want to hear so.

[Dylan]

So, so wake up. You know, use slack for a little bit, check text and stuff. But then I, then I go and grab my, my big protein shake cup. I put a scoop of the Brian Johnson protocol stack, AKA bj. I put two scoops of protein whey isolate and then I put a scoop of collagen and then I use a coffee stirrer and I mix it all up with water and I take this water bottle and then I go to the sauna and I sit there for 60 minutes while I drink my shake and my water.

[Doug]

While I actually usually what I do. Okay, you want to know my morning thing? Try not to wake up Hannah because she always fucking sleeps in deeper than me.

[Dylan]

Turn on, talk to.

[Doug]

I'm not joking. That's exactly what I do. I actually open up, say hey, make the daily note. And then I'm like okay, well what's the outstanding, what's the most important thing to work on today? What are the outstanding tasks I told you to do yesterday? And it pulls it all up, I'm like okay, one. And then, so then I'll go like I'll go run a task, I'll brush my teeth, I'll run a task, I'll take a shower. And then what? And then I'm out the door.

[Dylan]

I'm a bit open eyed. Deep research pilled still because you've seen semianalysis functions, right? We have this bot internally where you can at semiannysis functions and it engages OpenAI's deep research API and it deep researches. Anytime there's a support ticket it goes into this slack that's called support vip. It drops it in and then it automatically queries semianalysis function, semi analysis functions, reads the support question and then it goes through our slack, right? Has you know, all of our notes, all of our, you know, every note that we put out externally as well as all the internal communications and it creates this long ass response that's like three messages long and has links to all these threads. And then a lot of times the analysts can just like look at that and be like, yeah, I mean and then, and then they like condense it down and say the right thing. You know, there's always a couple of things that they need to fix and then they send it and it's like, what the fuck? Like, I don't know. I think. Can cloud code do this?

[Doug]

Yeah. Why? Dude? Is a skill issue, Mr. Claude. It's a skill issue, bro. Yeah, I'm sure it can. I'm sure it can. It can read every. It could read like I could be like, hey, summarize 60 of the notes that I've read in the last day, put them all into context and then do a deep research and whatever. Like, and also like you can like. Okay, so I think you could probably program whatever the integration into Slack. I think, I think SL ChatGpts is better right now.

[Dylan]

Well, we're not using it because both of them suck and we made our own.

[Doug]

Yeah, yeah, well that's. Oh, that's true. Okay, so that's all on Kyle. Yeah, I mean I think Claud co could do that, but I think it would probably be better. Just hitting the Deep Research API is probably the best in terms of doing deep research queries. But like, I don't know man, there's so much like Dylan, I actually learned that you can fine tune 4.0 models and apparently the fine tuning is much more valuable than like for like really dumb things being like, hey, this is good or bad content. And I was thinking about trying to do that and then like using the thing that's exciting about Claude code is you can do any hook you want into any API. So for example, what I could do is in theory I could be like, hey, here are 300 articles and I'll be like, here's the most valuable part for each one of these articles according to my taste. And then I could fine tune it and then I could serve these API whatever. But the important part is Claude code could literally be the one who's ripping the API agentically and be like, oh well, I did a search and then I applied. I had a, you know, I had my fine tuned Doug model to be like, this is actually the valuable things in the search and it runs all of that for you. And then part of this too that's like so crazy is like there's so many, like you could just. I mean also just for context of how. How max e'd out. I just bought a Mac mini just to run headless in my apartment so that I can just run run jobs all day for like bots. Because these, like that's actually becoming an issue is like these cron jobs. Kyle is going to kill me because barely any of this is on fucking GitHub. Your My entire life's work Is just locally on my computer, you know, whatever. But I could just make another one.

[Jordan]

Another one.

[Dylan]

Doug was like, dude, I made this awesome thing and then he links like his own IP address and we're like, Bro, it's like local host dot slash.

[Doug]

No, no, I was fucking joking. I know. Localhost8000 is. This is what it is, bro. That. That hurts, man. That was a. You should come see this really sick thing. No, I've been doing ironically, a lot of digital, so I'm very bullish APIs for this reason because APIs are really, really like. Essentially you just like guzzle, dude. You just give it the API and it does whatever the hell you want. Very bullish API. Very bullish consumption software. Very bullish. Like, I mean, like, bro, I'm paying search API. Oh, actually, Dylan, what do you think about the. So, so is it. Which, which API got. Is. Is got the Google. The Google. The Google lockdown. Is it SERP API?

[Dylan]

I have no clue what you're talking about.

[Doug]

Yep, it is. It is. Sorry.

[Dylan]

So Google did the search people that they blocked. Yeah, yeah, yeah. Well, did you see the core research note about, you know, before that, you know, we wrote a couple about.

[Doug]

Yeah, no, we didn't.

[Dylan]

Metasearch, you know, you know, we got, we got, we got.

[Doug]

Well, dude, there's a lot of. I mean, you know, like, ironically, like Brave is really good. I've been, I've been researching a lot on the search to API stuff, but yeah, it was pretty good. Dude, the search, the search APIs are really valuable. All this, like APIs are priceless is my vibe. Like, I don't, I don't even. Like, for example, if Bloomberg could just API all of its information away, I would rather pay that. I think just the data is more valuable than any of the form factor. Bloomberg's total. Like Bloomberg would be better off giving everyone, you know, a, A Bloomberg color. Like Claude code with all the data in the back end of an API and that's it. And then, and then ib. That's it. Like instant.

[Jordan]

Remember they were trying to train their own models. There's like, still are.

[Doug]

They gave up. Actually they use Claude. They have. Ask Bloomberg. It's pretty shitty. Claude code. You're not even talking to Doug anymore. You're just talking to Claude code.

[Dylan]

Doug, are we making a mistake by hiring five juniors in Singapore?

[Doug]

No.

[Dylan]

Okay, well, you're saying Claude code can do everything.

[Doug]

Okay, so, so, so here's my take. Okay, but Dylan, Kate, you got to think even more maxi. What if we take These five juniors in Singapore, we make them as good as, as good as we can, then we give them nuclear power. Then you have five row roaming nuclear power bombed arm robots from like the, I'm thinking like Civ 6 or whatever, like the, you know, the nuclear age robot giant robots. And they're just roaming and doing awesome work, I think. Okay, so I think, and it's kind of back to where I was saying, it's like a skill issue. It's like the people, like having a really high agency person being able to like still understand the scope of problems, break it apart and then like have the resources to put it all together. Like that's what matters. Because I'm serious, it's just a skill issue now. So.

[Dylan]

Sorry.

[Doug]

God, I'm, I'm, I'm. I, I want you to know I've been holding all this in for our, our internal cloud code webinar. And then also I, I, I have.

[Dylan]

To, you gotta organize your thoughts a little bit more instead of ramble mode though, you know?

[Doug]

I know, I know. I'm, I'm, I'm going to do one actual project that like, I'm going to do a sample project and kind of show my daily workflow, but I'll have to sanitize my daily workflow.

[Dylan]

Makes sense. Makes sense. Well, you don't have to sanitize it.

[Doug]

Well, no, because I have like personal, like my personal notes are on there. It's all one big Doug brain. Everything I've ever thought of.

[Jordan]

Do you really think you're gonna, you're gonna keep doing this a month from now?

[Doug]

I mean.

[Jordan]

Cause I've, I've, I've felt like I had a Rome phase. I had an obsidian phase.

[Doug]

Like, okay, so, so I had a Rome face too, for sure. I think we both had a Rome face together. This is not, it's not the same thing. It's like, it's like get, get, get Rome's name. Like, get Claude Code's name out of it. Like, put some respect on the name. It's totally different. Like, I think it's, I don't know. I, it's the first time where I feel like. I just don't think I would let you rip it out of my cold dead hands. I think my cold dead hands would still, would actually be the last thing it would do. It would press 2 to make sure all, all edits allowed so that you can't rip cloud code out of my dead hands. I'm just, I don't know. And like, okay, I did Try codecs during the codex hype phase. But Codex just never like it never finished. It never. It never like it just never got it done. And maybe I'm a shitty programmer because I am. I'm not doing shit. I'm just like, I'm just telling my little slop child to do whatever the hell I want to. But yeah, I'm forcing everyone in the New York office to try Claude code. Oh wow. And being like, hey, you have to do a project.

[Dylan]

Wait, you know Kimbo is like super against. He's not super against it, but he just straight up codes in vim and doesn't even use like coding assistants that much.

[Doug]

Yeah, he's a first.

[Dylan]

Which is funny because the other person, our technical team is like we're not hiring anyone until we get a thousand dollars a month of coding spend, dude.

[Doug]

I, I guess, I mean the history of compute tells you that the one that overwhelms you with slop with more transistors and more packages and more support wins in the end. I believe.

[Dylan]

All out frontal barrage of slop, dude.

[Doug]

I literally believe like like software. How it's just like this compounding thing where it's just like transistors allowed for like more and more and more slop software to these ginormous packages that obviously still got the job done but were completely bl. I think that that same pattern is what AI is going to allow into like the physical world. Like we're just going to just, we're just going to consume, consume, consume information. I mean I'm. Yeah, I don't know, maybe I'm. Or, or it could be real. Opus 4.5 has one shot at me and I'm having full psychosis right now. You know, there's only two options. Two ways.

[Jordan]

My western, next month's transistor radio. We'll see, we'll see how it does.

[Doug]

Maybe next month's transistor radio will be.

[Jordan]

How's house?

[Dylan]

Wait, we should do a live one. Jordan, we should do one live when I'm in New York.

[Jordan]

Okay. Oh, you're here. When are you coming?

[Dylan]

Last week.

[Doug]

Last week of January.

[Jordan]

Yeah, sure.

[Doug]

Great.

[Jordan]

I'll buy another mic. I'll buy the like little four person thing. Is there anything else you want to talk about besides cloud code?

[Dylan]

I mean, I guess, I guess we didn't finish the semi analysis plans, you know, so Keep going, keep going. Right.

[Doug]

Sorry for those, sorry for the long segue.

[Dylan]

We've got the chip companies, right? So we're going to do all the Chinese chip Companies we're going to analog chip companies, real chip companies. Right. You know, because analog isn't real. You know, keep going down the stack or up the stack. We're going to do all the server companies, we're going to do all the data center companies, we're going to do the entire grid, we're going to model their grid, we're going to do all of the AI companies that we're going do AI adoption there. And as such we're hiring analysts for China. We're hiring multiple analysts in Singapore as well. And then we're also hiring for AI analysts. Right. So. So any of those things that you guys are interested in to check it, check out Dover. Dover for semi analysis, our job port. Or just reach out to one of us if you're a true slop son. You know a better way to reach out to us than that?

[Doug]

Yeah.

[Jordan]

What's the market for the China analysis?

[Dylan]

Altruism. Like I mean like I think, I think the world needs better China research. Right. I don't think people will pay for it but I think, I think it's, it's. You build things before people want to pay for it. Right. And then, and it turns out people want to pay for them. We'll figure it out. Right. I'd rather just invest and, and have it, you know, and have the best China information in the world. The most objectively factual, best research. Not any of this think tank policy of people who don't know the first thing about any and not any of this like coper, you know, bullshit industry association that's funded by China or by companies that hate China but like just objective real shit.

[Jordan]

Well, I am really excited for this. I think it needs to happen. Yeah. Selling, selling the thing where people. Well, I mean it's like if they're, if it's like competitive analysis then people start caring. But if it's just like. And I guess there's like, like kind of you know, SME purchase projections and then there's like competitive. Then there's like impact on western firms from a competitive perspective. But yeah, selling China data and China analysis is hard when you, when it's not like a market entry thing. But I wish you guys the best and I believe in you Dylan. If anyone can make it happen was.

[Doug]

Was the rts, was the RTS game. Was that what pilled you? You're like holy crap, we don't know anything about them.

[Dylan]

No, no, it's just, it's just like I feel like every team in the company needs someone who's focused on China also.

[Doug]

Is it because Dylan. You've met Dylan at a very Chinese time in his life. Sorry, I don't know if you guys are familiar with this entire meme.

[Jordan]

Oh, it's a great meme. We're doing a whole podcast about it.

[Doug]

Stay tuned.

[Dylan]

Really?

[Doug]

I love. I love that there's China you want to join, actually, in China. This depends on the time. I love that there's people in China being like, oh, you've met me. Like, they're like, howdy, I'm trying to become more American. And it's like, you're watching the east meets west on TikTok and Xiaolongshu. Yeah, it's pretty cool, actually.

[Dylan]

I don't know the meme. Can you explain it.

[Jordan]

Right now?

[Doug]

China's hot.

[Jordan]

Yeah, China's trendy. China's in. If I had to, like, Japan. The thing is, like, Japan is played out. Everyone's been there. Everyone gets it. It's old. It's. It's like there's no youth anymore. Kind of same thing with Korea. Like, they got a little more juice, but still, like, what are you gonna do?

[Doug]

Do you know what? The moment, I think, actually broke it. I think the Ishowspeed China tour, I think that that's the. That was one of the first things that really, like, made it trendy and cool. And then all the influencers started to go, and then they're like, oh, my God, it's so futuristic. And then a lot of just like, you know, the traditional, like, industrial propaganda about how much production they have, which is so much. They have so much more production than we do. And then being like, wow, it's really cool. And then it just. And then the memes. But the meme is just like. It's like, I. I'm claiming that I'm 100 Chinese, man. Next time you talk to me, I'm gonna say, nihau, woman, whatever. Like, I don't know what. Whatever. I'm just start talking to you. And then so there's this whole meme feed about being like, hey, man, you've just met me in a really Chinese time in my life.

[Jordan]

It's like my favorite bellies smoking cigs, sitting on plastic, eating hot pot. I mean, I think it's good. Well, China is really freaking cool. I think it is cooler in Japan. Japan, like, wow, they don't even have kids anymore. What are you gonna do with that?

[Doug]

I don't know. Well, you're just gonna. You're gonna. It's like Disneyland. All the the old people will be the attendants. It'll just be a, it'll just be a creepy tour park. Kind of like Europe, you know.

[Jordan]

All right, I got a thing to close on. So my mornings now, our daughter jumped out of the crib and Kay is tall enough to open the door. So open her door and then open our door. And then at 6, 10 in the morning walks in and says baba. And I thought it was a dream and I freaked out. And then our daughter was standing there wanting to play in the morning. So quad code, do that one.

[Doug]

You, you know what you do is you buy her Mac mini, you give her, her a little, you give her a little screen, you delete everything except for one thing that opens on, on startup. Have Claude do it all for you. Fuck it. He'll make a custom ui. I don't know, I'm sure it could do it. And then it'll just be like, what do you want to do today? That's it. I'm just saying. I'm just saying start em young.

[Jordan]

We're pretty close, so. Oh, so there's only one. So none of the western firms shipped AI toys that had like something you could talk to because they were all too scared of like people jailbreaking it. A Chinese firm did and then there was this whole big scandal because it was like started telling you like CCP propaganda because it was like some Quinn model they put in it saying like.

[Doug]

And you'd be like China and you'd be, and you'd be like what happened in Tiananmen Square? And then it like, like blows up or something.

[Jordan]

Teleports back to Shenzhen and.

[Doug]

You know, rats on you to the police.

[Jordan]

Yeah. What are you going to do? Innovation's got to come from somewhere, man. Dylan, I hope you, I hope we are. Part of your application coverage is toys. It's going to be the, the beach.

[Dylan]

Toy toys we have not expanded to, but industrials, energy, AI adoption are the main areas we're pushing in China. Those are the areas that we're going do better than anyone this year.

[Doug]

That's it.

[Dylan]

I mean there's going to be a lot more, bro. There's going to be a lot more.

[Jordan]

Yeah. What non China stuff. Give us more. That's all. Yeah, come on.

[Dylan]

I just said industrials and energy.

[Doug]

Global industrials.

[Jordan]

Okay, great.

[Doug]

Yeah, great. I'm just. Do you want to talk? Oh, whatever. You'll see when it comes out more. No, no, what else?

[Dylan]

What else? Well, we launched the memory model, so that's great. If anyone wants to Buy a memory model. Actually, jk, it's only available to full service. Full service customers. What else. What else is there?

[Doug]

We'll say. We'll call that beta. I mean, I don't know. I'm excited for something hopefully that will happen the first half of the year also. Hopefully a conference.

[Dylan]

What's in the first half of. What's happening in the first half of the year, bro?

[Doug]

I'm not gonna, like.

[Dylan]

Can you. Can you DM it?

[Doug]

Yeah, but it's gonna be so awkward while we wait for me to DM you what I think.

[Dylan]

Well, you know, we can always pivot to some other topic. We can always go psychosis mode. We can always go ADHD mode. You're about my love life. You can talk about many things, you know. Just kidding. There is no love life, you know. Just kidding, bro.

[Doug]

Claudo can fix that for you, man.

[Dylan]

Did you slack?

[Doug]

Me? Slack? Yeah.

[Jordan]

Okay.

[Doug]

Okay.

[Dylan]

Oh, yeah, yeah. No, I think. Doug, I think. I think we can just literally talk about it because.

[Doug]

Cool.

[Dylan]

We cut. They cut off access to us. No, they cut off access to us. That was the only reason to be secretive.

[Doug]

Okay, cool. Yeah, yeah. Then we're good. I mean, do you want to do it? Dude, it's.

[Dylan]

We'll have our. We'll like, have a splash.

[Doug]

We'll do a splash where.

[Dylan]

Where high yield will come and we'll record a video and then we'll, like, do a post. It'll be sick.

[Doug]

Okay. I'm excited. Tbd. Big things coming soon.

[Dylan]

Bigger physical. Physical operations. Something else has Capex. We have Capex.

[Doug]

Yeah.

[Dylan]

Huge.

[Doug]

We're buying three Mac Minis. No, no, dude, honestly, if this works.

[Dylan]

Apparently I called Boomer. The Dells that we has a standard issue Boomer and old. And then Paul took that as, oh, Dells are banned. And I'm like, oh, fuck.

[Jordan]

Well, is an Apple kind of Boomer, though. I mean, I could really use a new PC format.

[Doug]

You should be you. Okay.

[Dylan]

So our newest hire wants a framework laptop. And we're like, yeah, I think he's like. He's like, I've never used Back or OS X or Windows. And I'm like, holy fuck. The level of autism we have at the same day. Wait, who straight up said, I've never used Windows?

[Doug]

Who's that? Who's that? Who's that?

[Dylan]

Harrison.

[Doug]

Okay, that's. It's pretty impressive, dude.

[Dylan]

Dude, the. The technical staff is absurdly cracked.

[Doug]

It's pretty cool. It's pretty crying. But anyways, you should be buying Max right now.

[Dylan]

We're also hiring for technical staff. We're hiring for cluster max. We're hiring for AI analysts. AI tokenomics. We're also hiring for China analysts. We're hiring for Singapore analysts and technical people. Anywhere in the world, if you're passionate, not in any of those places. Or any of those places. Still reach out because we're hiring for sales still.

[Doug]

I'm trying to.

[Dylan]

We're hiring for everything. If you're good.

[Doug]

Yeah, just go to the.

[Dylan]

And if you're a slop son, join.

[Jordan]

I want you guys to have an Africa analyst. When. Where's like the Africa story.

[Doug]

We'll get a lot of. We'll get the Venezuelan. We'll probably do the Venezuelan GPU first, bro. There's an. There's an actual Brazilian memory company, so. Oh, my God.

[Dylan]

Can I tell you guys a story? Hold on. So this Venezuela thing, right? I'm freaking out. Like, everyone's freaking out. Like we're monitoring the situation because that's what we do. And then, and then in the slack, right, you know, a couple of the guys are figuring out, they're like, holy fuck, there's all these underutilized turbines in Venezuela. And then like, that's it. That's it. That's all they say. And then I make a post on X about how and it's a, it's a fake news post. It was a shit post. I said that at the end of the like post. But most people don't get that far. So a lot of people thought it was serious, but I was like, oh, we're not invading them for oil. We're invading them for underused turbines. We're actually going to take these turbines and ship them back. Elon wants to put them in Colossus 3. OpenAI wants to do Stargate Venezuela. We'll see what happens, right? And it was like just a complete shitpost. Multiple policymakers DM'd me people in like. And they're like, oh, good idea. And it's like, what the.

[Jordan]

Oh, my God. Oh, my God. Well, there's nothing else you want from this country. It's like you're invading the wrong country for the wrong.

[Dylan]

Less influence.

[Jordan]

Who cares? It's Venezuela. God bless you. You want to, you want to influence like a, like a impoverished country that.

[Dylan]

Is with a lot of oil and minerals.

[Jordan]

Okay.

[Dylan]

And underutilized turbines.

[Jordan]

All right, I think we're gonna call it there. Welcome to 2026, everyone.

[Doug]

Happy new Year.

[Dylan]

It's good old fashioned imperialism. I'm not coping in any other way. It's just imperialism.

[Sam]

White on black is the truth Claw's my only gun I'll call him Maxi Pill you every chart I ever posted didn't touch a single key just tight make it beautiful and let the Ellen breathe My ego's getting massive but I'm saying it's all right Everything's a skill.

[Dylan]

Issue and the model got the mind.

[Sam]

I'm Claud code pill Then I'm losing my mind body, body coding everything leaving thoughts behind magn in my bedroom running b all night hope it's 45 is AGI the future's looking bright it's all a sk it's all. I don't write no functions I just whisper to the screen hey bro, fix my trading form make sure it's clean 6 tags open cont for the green My girlfriend thinks I'm crazy But she doesn't see the dream eyes overturns one prompt it's done my whole life. But I just help my singing and it handles all the death they said I need to structures Said I need a plan but I'm in psychosis Move my digital from Age of Empires to Venezuela Building my reality one token at a time. It's not a phase, mom this is the future. Now if you'll excuse me, I need to talk to my computer. Sam.