CoRecursive: Coding Stories - Notes: The Universal Paperclip Clicker

Episode Date: February 4, 2026

  Multiple VS Code windows. "Agent stopping" in a robot voice. A laptop stand on the treadmill so Claude can keep working while I run. The Big Rich sitting unread by the fireplace while I check if the migration's done. Somewhere along the way, I started reorganizing my life around keeping the machine spinning. Claude Code had become my universal paperclip clicker. This is me trying to figure out the difference between real work and just feeding it tickets.   This is some field notes, a shorter, rougher than a normal epsidoe.  Episode Page Support The Show Subscribe To The Podcast Join The Newsletter

I've been thinking of quitting. Not podcasting exactly, but sleeping. The past six months have been, I don't know, I've been doing the least amount of coding that I've ever done in my career as a developer, but also the most. The most lines of code-ridden for sure by a huge metric, but also the least. I think I've already quit recreational reading to a certain extent. And that can't be good, can I let guys.

I like to read. So today's not a normal episode. There's no interview. There's no hour long story. This is some field notes, a shorter, rougher, a little bit more personal, trying to capture a moment in time. Regular episode is on its way, but I've got something I want to share while it's fresh. And the thing I can't shake is this question. In a moment where expertise has a half-life of months and shrinking all the time, what does it mean to learn something, to invest in learning? When everything's churning, how do you know if you're building skills or just spinning? Let me show you what I mean. Right now, I'm in my office, I'm sitting at my standing desk, got my mic in front of me.

My multiple VS code windows open, and the way I tend to work is my VS code window is kind of split in two with the IDE stuff on the one side and on the other side is a terminal with Claude code. And actually, usually more than one. Right now I see two. Now I have another VS code open for Agent Corps demo. And then another one for AWS Houston meetup. And it feels incredible because, you know, I'm thinking at whiteboard speed. I'm just describing things and producing the code. Describing features, describing edge cases, describing what I want the end state to be and the system is working to make it real. But it's also chaos, right? I have mass. sessions and I'll hear them stop because I put stop hooks in. So I'll hear momentum agent stopping. And that means the momentum agent that was working on my AI running thing fixing the problem where

it's handling dates wrong, need some input for me. And then while I'm figuring that out, I might hear Azure Workshop, agent stopping. And I'll go over and see what's going on there. And it feels like I'm getting so much done, but it also feels very stressful and like I'm not keeping up like I Love Lucy with the chocolates on the conveyor belt. So in a way it feels like productivity. Like I finally got an intern, a very fast intern that runs 24-7. But then there was this moment, this tiny kind of stupid moment I had that made me suspicious. So it was the shower. I had the thought, you know, before I jump in the shower, I might as well get Claude, code working on this next ticket. And I remember thinking like, well, that's kind of a weird sentence,

isn't it? Because it wasn't like, oh, I really want to solve this specific problem or add this feature. It was like it should be running. Like I feel like it should be running. It should be doing something. It's my intern. It has a fixed cost and I just want to keep it going. I bought this stand to put next to my treadmill so that while I'm doing a run, I could have Claude Code running there. But it felt so productive that I found myself reorganizing my life around keeping the machine spinning. And to be clear, it's fun and it's exciting, and I'm having fun. But my wife thinks I'm stressed out, and I am, but I'm also exhilarated, and I'm fixated on a bunch of different things. I'm tied to my laptop. It reminds me of something that I haven't thought about in years.

Did you ever play Universal Paperclips? The title sounds like this scary AI doom scenario. AGI takes over the world and turns everything to paper clips. That's the premise. But the actual experience of the game is way simpler. It's compulsive. It's a clicker game. You have to keep clicking. You start by literally clicking a button to make paper clips, and then you buy an upgrade that makes clicking faster, and then you automate the clicking, and then you start optimizing the systems that optimize the systems. And the lesson I learned wasn't about the AI apocalypse. The lesson I learned was about addictiveness. The addictiveness of seeing the number go up of productivity happen. The thing is running, and you need to watch it, and then you need to do more stuff, and then you need to build stuff that lets you do more. more stuff and you're making more and more progress and you can't look away. And this is where I've been lately, I think.

And it's been so much fun, but sometimes I'm not clear on the value. Claude has become my universal paperclip clicker. That's the paradox I feel. Because on the one hand, in this frothy time is when you can leave your dent in the world. You know, I had that Evan You interview where he created View. You know, it was in that churning period of JavaScript frameworks. That's where there's opportunity. People build amazing careers by stepping to the front in these high turnover moments and creating something unique.

But on the other hand, it's also a moment where it feels like there's the least to be learned. Everything you learn will quickly change. Everything that's hottest today will soon be gone. Not the AI coding agents are going away, but just that they're changing so fast that it's not possible to keep up. It's not possible to learn persistent skills. The things you learn now won't be useful in six months. You can see this with all the little techniques that appear and then expire. Like Ralph Wiggum loop.

The Ralph Wigam Loop is basically you run Claude Code in a while loop and give it as a prompt this big file that it can edit with a clear end condition and it can add in tasks or complete tasks. And so instead of that ding where I need to prompt it to the next thing, its whole memory is wiped out and it starts a new session and it starts on the next to-do item. It's super useful right now, but better orchestration will make it. it irrelevant. It doesn't make any sense that something like that should be needed, but it very much is and gives you a lot more power. But that's the vibe, right? You'll learn a trick and the existence

of the trick is proof that it's temporary and will go away. That's the paradox. You can sprint to the frontier now because there's no experts. Everything is being figured out. So you could make a difference. You could matter. But also, you could wait because everything you learn today will be churned will be gone tomorrow. You can catch up in several months just as easily as you can catch up right now. Both are true at the same time. And there's a cost to staying at the front, right? You have to keep up with everybody. And I started noticing the costs showing up in my home life. My wife and I have this pair book reading club. Well, he's got a copy of a book and we read it and it's fun. And the book we're supposed to be reading right now is the big rich, the rise and fall of the greatest

Texas oilman fortunes. And I haven't been reading it. It's cold and it's wintry here in Canada. And there are these evenings where we'll sit around by the fireplace and we'll read, except I won't be reading. Right, I'll have my laptop or we'll be watching a show and I'll feel the need to see if the migration's done or if the agent's stuck or if I can give it one more nudge because this 24-7 intern is hard to keep going. And so it seems like I'm always worried about something. Like, I'm super busy and that I'm not all the way present because I'm excited and stressed at the same time. But I think it's not a tech problem. It's an attention problem. It's about when is this real work? And when is this the paperclip clicker? Because here's the distinction

I've been noticing. There are good Claude sessions where I get a lot done. But there are bad ones, right, where I'm just chasing motion. When I have a moment before a meeting, and so I say to Claude, like, hey, why don't you start going through and fixing all the compiler warnings? Because it can chew through them, and it's better not to have the warnings. Except, you know, it can't fix all of them, and it ends up breaking something, and then, you know, it has a question, and then after the meeting, I need to poke it some more. And now I'm busy wasting my attention on some unimportant task

because I just want to make forward progress. So that's my takeaway. If you don't know what done is, you're not really delegating work. You're just feeding the clicker just because I can build something now that I couldn't before. It doesn't mean I should. So yes, things are changing. The tools are getting better. The interfaces for me are shifting from typing to talking often, you know,

from physically writing the code to discussing the outcomes and the tradeoffs and figuring out end states we can verify. and my calendar is proof of that change. I'm speaking at a bunch of things this year, and they're all AI-based. I'm going to be at scale in Pasadena giving a talk about my AI running coach. I'm going to be at MLCon talking about agents. Also, you know, teaching workshops,

making videos for work, helping with Palumi's own coding agent, Neo-infestructure coding agent, that's pretty cool. This froth of AI agented coding has consumed me. But that's not required. You can sprint to the frontier and live there, try to keep up, maybe leave your dent in the world, but probably not. Or you can just wait for things to stabilize.

Like the front end after, you know, React became dominant. And you want to miss your chance to build useful things. Everything is moving so fast that it's actually easier than ever to catch up later. Both are true at the same time. So the investment is learning how to aim, choosing that end state, deciding what matters, deciding what you're going to do with your life, with your time, and being willing to step away even when that machine could be spinning. So that's my update.

This is what I'm feeling. February 2nd, 2026. I'm working on a new episode, actually, more than one, and I just wanted to say this out loud, partially because I think a lot of people are feeling some version of this too. Take a deep breath. We live in interesting times, but despite everything that's happening with AI, despite everything that's happening on the news, life is good.

Okay, I am going to go try to read a chapter of that book, The Big Rich. And until next time, thank you so much for listening. New episode coming.