The Standup with ThePrimeagen - Recovering from AI Psychosis

Episode Date: May 22, 2026

The crew dives into “AI psychosis,” coding with GPT vs Claude, doom scrolling tech Twitter, meme culture, burnout, obsession, and what happens when developers start letting AI run wild. Equal parts hilarious, insightful, and slightly concerning.

Okay, this is a really important blocker. TJ, I have a blocker and I need to understand something. Yeah, yeah, yeah, yeah, yeah, yeah, yeah, yeah, yeah, yeah, yeah, yeah, yeah, yeah, yeah, yeah, yeah, yeah. Uh, sorry. I really liked this meme, kind of inverting it to where the first guy is how it always happens with chat GPT being like, dude, I got the bug. And then literally the next line's like, actually, I need to investigate some more. So that's kind of how I looked at this as I thought that was funny. and then you told me that's not how this meme works.

Right. I mean, why not? First off, why are you the gatekeeper of memes? Well, first off, this is, I,

did you see my version of this meme? No, I did not. Just go, okay, just quick, why did you go peek out over to my profile? It'll be relevant to this conversation.

Okay, I'm on your profile. Okay. Yeah. What, can you read it for the audience, for audio listeners. CEOs directing employees to their tweets.

To like their tweets. That's what you're doing. That's what you're doing literally right now. You're the guy, Brian. I'm confused. Okay, that's all right. We'll get you there. We'll get you there.

This is not like a Drake yes, no panel kind of meme. This is more of like, well, I was just saying you, that's, I'm just telling you, that's not how the meme as it stands currently works. it's fine like i mean you might be able to break ground this was what i was going to say means are fluid part of it okay part of it is i i don't have as many followers as you you can set you can't set so i'm not trying to police it i'm just trying to let you know bro to bro i'm not going to let you go out there and use a meme wrong and just let you use it wrong for like a year and then you find out like oh bad luck brian is actually about things being bad luck and not good things and i just misread it because it described like my high school experience or something

like that. Like, I wouldn't want you to have that happen. Okay. That would be very funny, though. I would, I actually, I love this idea of actually misunderstanding bad luck bride to the point of it being good luck, Brian. That actually makes me really happy. So anyways, the, the, it's supposed to be about more of a vibe of what he's giving off here, less of like a yes, no kind of panel, but I mean, you could try and switch it up to a yes, no panel. Yeah, I was just saying it also fits as like a two-part thing, someone that's like really into something. Like, because picture one is a very confident picture. Picture two is also confident,

but it's also whoopsy poopsies, right? Very classic kind of meme setup. I think Adam is probably really rethinking his decision to come on to stand up. I just miss you guys. Did you know that it was just going to be discussing the finer points of whether the meme image supported the text and the meme for like 10 minutes? I'm not very into memes. I don't have a lot of knowledge on this topic, so I'm just going to sit here and smile.

We should have done a... Oh, I can't believe what we didn't do. We should have done a meme quiz for Adam and got his take on various... We'll get that for next time, Adam. Yep. Hey, is that H-CDP? Get that out of here. That's not how we order coffee.

We order coffee via SSH terminal dot shop. Yeah, you want a real experience? You want real coffee. You want awesome subscription so you never have to remember again. Oh, you want exclusive blends with exclusive coffee and exclusive content. Then check out Cron. You don't know what has a... S-H is?

Well, maybe the coffee's not for you. Terminal coffee in hand. Live in the dream. I do love this idea of meme quiz for Adam. Yeah. Like, I don't know who the Brian one is, whatever you just said about Brian. I've never heard of it either.

Bad luck, Brian? You don't know. Oh, I tried to pick an OG meme that you guys would be like. Maybe if I saw it. Next, you're not going to know Philosopter. I don't guess I know Philosopter. Do know Philosoeraptor. Okay, yeah, that's like a pretty classic old school meme.

Are you saying it like Flossiraptor? Yeah, philosophy. Oh, I know that one. Yeah, okay. You know bad luck Brian too. Search it out. Yeah, I looked it up. I know that image.

Yeah, yeah. That's what it's called. I know that image. I'm familiar with that. It's got some all-timers. Like, has a face only a mother could love? Two dads. You know, like it's got a class.

Oh, I've never seen bad luck, Brian. Oh, really? Oh, dear. Okay. It has so many funny one-liners. It's old internet memes, and you can just keep on going down them for quite some time, and you'll laugh at each one of them. Scumbeg Steve, also a good one. There's like a lot of, like, the old memes I thought were really good because they just had so much legs comparatively. All right. All right.

Okay, so apparently we need to have an Adam versus Casey meme history. Oh, my gosh. It's like the final score is two to one out of 50s. Well, no, no, no, Casey, I'm not going to... Casey, to be clear, I'm not going to let Began be the one that chooses what memes he considers relevant. I'll try and pick one instead of the game. Right, you're trying to pick one that somebody other than Began has seen. It's like, Began is not a meme.

It's the only person who's ever posted the image. He's like, I posted this six months ago, and I laughed really hard. And I showed my girlfriend. And then she said, can we just watch the show, please? Yeah. All right. We should probably get going for the standup now. Okay, yeah, let's.

Welcome to the stand of today we have with us, Adam.dev, if you do not know the legendary vegan from the Ozarks, he has single-handedly helped make something compete against Claude by himself. Adam is a fantastic dev, and we've made fun of him many times for being a vegan and going to Korean barbecue with us. But the more important thing about Adam is that he went from coding everything by hand, not really into the AI, going to full AI psychosis. And apparently he is back.

out of AI psychosis coming into reality again took a month off the internet and when we saw this we said okay adam come on talk to us we got to hear the story we got to understand it because i think a lot of people are probably feeling they're probably in the month long adventure or they wish they had the month long adventure or even more so they're probably in full AI psychosis right now not even realizing they need whatever you got whatever you had so we'd love to hear the story and all that tj interruption now Oh, I was just going to say I'm building my own programming language So

That's That's Hey, if Versel Labs can do it I can do it boys, okay? So I'm I got my agents working right now, right over here Separate computer If you're trying to compete with Versel

I think you're going to have to like one up them On the Loserist name Because like they came out and this call, look, it's called zero Like we're literally, we're just going to call our own things Zero. That's our level of self-confidence. You should just call it suck. Like the language is just called suck or something. Casey, Casey, mine's called one. Everyone's always talking about going from zero to one. Okay. Nice.

So that's, everyone's going to be using it immediately. Think about the SEO. Zero to one, baby. Zero to one. There's your upgrade path. Okay. By the way, TJ, you may not know this, but the reason why it was called zero is because it's the language which that introduces zero value. Oh, damn. All right, well, your Versal sponsorships are all canceled now. Adam, so can you give those of us who don't know, like, a little bit of background on, like, where this whole thing started? Because, like, I also don't know if everyone knows that you worked on open code.

I mean, that might only be passing from open code. So could we get, can we get, like, the deep, the starting point of the story? From zero to one, shall we say. Yeah. Yeah, I guess it starts like April of last year. So Claude Code came out and maybe like February, something like that. And I started working on Open Code in April. And I would say like that was around the Claude 3.7.

I'm going to reference a lot of model numbers because it's just kind of like the timeline. It helps me remember everything. That was like 3.7 where I feel like more it was kind of picking up, more people were playing with the whole agentic thing. So not just like going to the browser and asking, questions and then copy pasting code but like letting it run on your machine cloud code style so I started working on in April and I would say like until September maybe September October I wasn't really using the stuff I was building it mostly by hand and I would just like forget and I've talked about

this a lot we would just completely forget to use it for weeks or for a week and I'd be like oh man if I'm going to make this and expect people to use it I should be using it right so mostly like through the summer just kind of doing everything by hand still like playing with it but not really sold and not really impressed with what it was able to do uh i guess it was sometime in the fall when i think it was opus four it could be four dot something but i think it was four when i started trying to lean harder into it over the the fall and again it just like and i've always found this with the clod models it just like took so many shortcuts and did so many stupid things that i had a hard time like handing over anything to it without just feeling like I wasted my time uh yeah I wish

trash was here I feel like he would be able to relate more with me like I think when I'm saying this I'm just assuming Casey's thinking it's still a waste of time we all wish trash was it with us okay it's okay here I'll I'll be trash okay okay you get burritos yeah I agree also now it's time free to tell embarrassing story. If you're going to be trash, you've got to finish it up with a quick embarrassing story. No, his always end up somehow being something that I'm like, I cannot believe you say that knowing that there's a chance your wife will listen to this podcast. Right.

I'm like, I care about you trash as a person. Right. And I'm like, bro, your wife seems like a really nice lady. don't mess that up dude like there's only it's not going to get another shot trash you know so it's like I just worry for him as a friend a friend you know what I mean not even just co-host to co-host but uh to your point though Adam I 100% agree I feel like Claude it loves to be lazy

loves to be lazy yeah yeah I never really and I still don't really use the cloud models I get it was like really through sometime I think it was December that the GPT5.2 model came out somewhere around there. And that was the first time, because I had been reviewing every line that Claude would spit out. That was the first time that I was using GBT. There was a few people on Twitter very actively, like praising it and very loud about it. It was like reviewing the GPT output compared to the Claude stuff. It was just so much more clever. It felt like it was taking its time and like it took forever to run. Like it was much slower. But I felt like, hey, this model output is good enough that

I only use it more. But that was the first time I think that I really started leaning hard into it. Can I ask a question there actually about that? Because this is interesting. I feel like, so as someone on the outside of this sort of stuff, I feel like if you'd ask me, like, from what you see online, which of the two models do people prefer? I would have assumed it was anthropic, because that's what people always talk about.

Is that fake news? Like, is that not really the case, at least in your experience? I mean, I think this year, GPT has kind of taken the mantle. And I'm speaking more for like the extremely online crowd. I don't think I have any idea what the broader like 80% of developers who just clock and clock out what they're doing. But the people who talk a lot online, I think GPT models since 5.2 have kind of slowly taken it.

And now to the point, I think Codex is kind of the default. Very interesting. Our team at anomaly, I mean 20 of us, we by far use the GPT models. I would say, Casey, the thing I've seen is a lot of people who don't, want to look at the code ever really like it Claude code. Not to say that there aren't people who like looking at code who aren't serious. Like I'm there are a lot of people who are serious engineers who still like that. But in my experience, it's been like, cool. I can pay $20. Anthropa gives me $4,000 of tokens. I don't ever look at the code and I'm like, I have 35 new dashboards.

And you're like, but do the dashboards do anything? But I have 35 of them. So not to put two. point on it here, but like it sort of sounds like you guys are saying that for the discerning customer you're going to pick chat GPT. Is that sort of what you're saying at this point? I mean, I don't want to put words you mouth. So I'm not a chat GPT guy. Though I do, I do throw

it over there. I personally, every time I generate something with chat GPT, clod, or other models, I find myself always refactoring and fixing them. And I keep on trying to make this better and nicer and I keep trying to get to this golden land of you can just produce code and be happy. But I have yet to be happy.

And so I find them all to be near similar and that how you prompt it probably makes a bigger impact. Okay. Except for Claude loves like adding, like GPT loves adding tests. Like they all kind of have their all like little flavors to things. Okay. I would say like this generation with the latest models, they're probably all pretty similar. I think it's more probably like the, the relations to developers from Anthropic versus OpenAI has shifted a lot. And I think there's been a lot of backlash with just the way Anthropics handled certain things.

So it's probably more of that at this point. Okay. All right. I just wanted to hear that because, you know, I don't have experience with these things. So I was just kind of, I was curious on that because it's like it seems like the public, the public perception sometimes seems to be that Anthropic is ahead in some way. But that sounds like that's not really true from the experience of people here. So that that was good Frasian.

Thank you. I do want to throw one more thing out from a, I think it was Lex Friedman who said this, which is that people develop like a personal love for the model because they solved a problem at a time. And then at some point in the future, fails on solving a problem, so they switch models. And then that new model successfully does it. So they're like, that's my new favorite one. So it's like more of like an irrational love as opposed to like a highly scientific.

This is the good one. They're not actually counting like how many times did this one do a good job. How many times did that one do a good job? They're just kind of like remembering the last big success or something. Yes. I think that that's typically work. Yeah. I also think probably like people are discounting that like the models probably do some stuff

differently. Like they're pretty big. They spend a lot of money. They've seen a lot of different stuff. They have different things. And like probably different people's brains, my guess would on average work better for different ones where like, oh, I'm an overtalker. Maybe it, you know, like there's probably some stuff too where it's like like that. You'll definitely see some and like Adam, you probably feel this too. Like if I ask GPT to make, if I make like whatever, you know, five, five, if I ask it to make something and I tell it to make some UI, it will put in the why all over the place and it will say without fail which i don't understand how they haven't gotten rid of this hello this is an app built with effect and it's running on bun also click down

here for more info and you're like what website ever yeah yeah when you get to the front page they say here's our text stack as the first big header every time you do this i don't understand why it's doing this, but it literally just says, hello, here's our tech stack, and here's some quote from you describing what you wanted, how to build this. And you're like, okay, guys, that, that's crazy. So for, anyways, I don't know, Adam, you can get to me, but I was just saying, they do some things clearly different and you feel it, like, when you're using them. Yeah. Yeah, that's definitely part of it. I think for a long time, people have complained that the Codex models, the GPT models are pretty bad at UI, and Sonnet seems to be, or Opus seem to be better at that stuff. So it is weird. And, like,

everybody has different workflows. Some people like to work with it where they tell exactly what to do and they just want to do it very fast. So people like DHHH are very vocal about models like Kimi K2.6 or whatever, the open source models. Okay. Because they're very fast and they can be very good at those kind of tasks. But if you want to like have it build broader things, scope out, you know, boilerplate stuff for like a whole project, then maybe other models are better. There's all these kind of nuances. But I think what Prime said going back to the like last success, That was really kind of what I think spun me off into what I would call the psychosis. Okay, cool.

Once you've been into that AI psychosis, you see it everywhere. Like you see so many people's posts and you just like, you know exactly where they are on this journey. Oh, wow. Like if they start talking about multiple accounts, like they have multiple codex accounts or multiple clod subscriptions, or they're talking about leaving them running overnight, like that kind of stuff, you can just go back and remember exactly the frame of mind you were in when you kind of like went down that path. This is fascinating. Yeah, say more about those things.

things Adam. I don't do any of those things because they seem very distracting to me with my family. So I've actually avoided this version. Well, I'm a very intense, like, obsessive person. So I think I ran the whole thing kind of on a speed run. Like, it maybe normally isn't so intense and it takes longer. And maybe it's not so disruptive to your life. But, I mean, it led for for me to like a month long, complete burnout. I mean, it was like, I was very intensely down this rabbit hole. And I think it started with like it's like so i played a lot of golf in my past and the thing about golf it's such an expensive sport it's such a pain in the ass in so many ways like it's just extremely difficult like you're just constantly chasing balls over the the woods and everything but you'll hit like one good

shot every round and that'll get you coming back like you'll have that moment where you're like oh i'm coming back tomorrow because golf is amazing and that i think that's what it was like 5.2 era one of those models just put out enough really like this is a cool output it It was something that kind of got me hooked, that dopamine of seeing it do something that you're like, I might have done it that way. Like that's, that was awesome. And that kind of started the, well, then I'm going to review it less. Like I'm not looking at the code as hard. I'm like doing, because those models were so slow, you start to do all these weird like parallelization optimizations.

Like you're running work trees. You're doing all this stuff because you're waiting 10 minutes for it to finish. So then you start looking at all that. Work trees for those that don't know. It's just like cloning out your workflow. into another one that's just like a different branch of Git. So you can actually be editing, you know, two different folders, two different get branches at the same time for one project.

I thought we were back on the golf analogy. So that's, thank you, Brian. That's awesome. Yeah, the ball is off. Adam, can I say for a really good strategy to enjoy golf more, this is what my mom does when she golfs is sometimes she doesn't even write down the stroke count for the, the like that hole, but she will just put a little. if it made a nice sound and got in the air.

And she says, that was a good hole. So, like, and she has fun when she goes golfing. That's what I'm saying. You're not breaking any course records, guys. Okay? Yeah, that's breaking any course records. You put a little star. Have you met T.J's mom, happy Gilmore?

Uh, really, really, really fun to golf with. Yeah. Anyway, sorry. Continue, Adam. Uh, yeah, where was I? I guess like, uh, that's, that's when it started so it was like had a few outputs that really compared to what I was experiencing with the anthropic models just felt like this thing so much smarter and then you just start to like

assign so much intelligence to this thing that maybe it just got lucky maybe it was just the exact right thing uh that it needed to solve and it solved it really well anyway it yeah it led to like i mean i would be running you know four or five of these things at time and i'm sure there's people that have gone deeper than this and they're like rookie numbers or whatever uh you'd have like all these work trees and you're just doing like endless refactors that are completely pointless like or like doing audits like for performance stuff I'm sure people are going to resonate with this like you are in AI psychosis if you're hearing this and you're like wait I do that every day it's called AI psychosis okay you just like it's like the same props recording an ad and it's like a

Sarah McLaughlin style like this if this is you you're in AI psychosis and I remember seeing other people in the in the height of it for me like through the winter December January I remember I remember seeing other people say things online and you'd be like, yeah, he gets it. And like what he gets is he also has like, I suppose. But they'd be saying things like, they'd say things like, I can't sleep. I just can't like, and that was me. I couldn't hardly go through a night. I have my laptop in the bedroom and I'd wake up because I want to start another prompt.

Like I literally couldn't think of it like going the next 30 minutes without doing more work. And it's totally nonsensical and like just insane to look back on and think that I was in my right mind. I wasn't. Can I jump in? By the way, if you don't know, Casey, I just watched the Mark Andreessen clip, and he's talking about AI vampires. Like this apparently is super popular in the valley

where you are effectively ruining your entire sleep because all night you're reprompting your agents and going back to sleep. And then waking up in 20 minutes to make sure they're all still running and then going back, just totally destroying your brain in your life over being productive.

It's like people debating whether they should get into the polyphasic sleep and stuff. Like, I'm going to sleep five times a day for 20 minutes each. Like, it was the exact kind of stuff. If you do that, it's kind of like you have six days. And if you think about it, you stack that up over one month. I'm going to

kick your butt. Yeah. Okay, sorry. Put the clip in. Josh, put the clip in. So I measure time. I've compressed and condensed time. I bent it. My day is 6 a.m. to noon. And I'm not crazy. You're crazy for thinking it takes 24 hours, just like some dude in a cave

did 300 years ago. My second day starts at noon and goes till 6 p.m. That's day two. And then the next day is 6 p.m. to midnight. What I've done now is I have changed a manipulated time. I now get 21 days a week. Stack that up over a month. I'm going to kick your butt. That sounds excellent. I have no idea. He wasn't. I don't think he was joking. I don't think he was joking. So I do want to say on the like the AI psychosis part of like waking up and all this stuff. This is like not that different from a bunch of other things where people like people do insane stuff to like play slot machines or do a bunch of, you know

what I'm saying? Like, it sometimes feels like, back when I was trading it in like 2012, 2013, I was losing sleep over the line markets. Yeah, right. Sounds a little bit like World of Warcraft too. Originally, a member of people had problems with that

where they would just kind of, they would stop leaving the house and they would just, okay, yeah, like a, yeah, yeah, yeah, it was a pattern. Yeah. So, so Adam, like, how does this start exactly, though, like, because is it, were you seeing particularly good like improvements in productivity or was it really just it was just the fact that you liked seeing that the AI could do something and so you felt like there was something to like to go after there and you just couldn't help yourself but do that even though it wasn't necessarily like oh my god I'm getting all this much more work done I think it's the latter I think you would you would convince yourself of the former to justify it like this is so much more productive I'm getting so much done I see but I think it really is just like that

prompt and dopamine cycle. You would, you'd submit a big prompt, you'd come back and see the results, and it's just the slot machine thing. It's like, did it do a good job. And then you get into, and I'm sure there's people to get into these with LLMs, even outside of coding. Like, I'm sure, I mean, you hit the people falling in love with models and stuff. I'm sure it's a similar phenomenon or just a sick of, what is it, sick of pants, whatever.

Yeah. Yeah. Sick of pants. Yeah. Sick of pants. I don't know. I love shorts, though. Yeah.

Yeah. they're clearly like they're clearly engineered I'm from the Ozarks okay we don't use big words they're clearly engineered to like give us that

reward cycle like it's probably no different than any other thing you can get addicted to and I get addicted to things very easily and then very hard when I do so that's really interesting so what was like the worst

point of this like like when so at some point you decide you're like you self decide you have a problem i guess like were you the first person to go i have a problem uh or did somebody tell you you know i mean my wife probably would have first noticed uh what was a sign that you were having a problem like what did she i mean the sleep stuff yeah what she said the yeah the sleep stuff is the big thing for me because i i've struggled with staying asleep for work just before all this you know for my entire career i would it felt like christmas morning every day like i'd sleep four or five hours

night and want to get up to work on the thing because like I get a big something out of of programming and building things. So I've always struggled to sleep and we had to in the last couple years really like build some boundaries around I'm going to sleep eight hours a night or I'm going to stay in bed at least. So when it started interrupting my sleep, I think that's probably the first sign that drew attention from my wife. Nobody really like nobody I work with. I don't remember Dax or anybody that works close with me with me saying like, hey, I think you're you're losing your mind. Uh, I think it was probably just the burnout. Like I, the first real like something has to change is I didn't want to wake up in the morning. And that's like very rare for me. I'm just, I'm a morning person. When I start sleeping until like 7.7.30, I know something's wrong. And yeah, it was it had just driven me into this kind of like very hard burnout. But nothing, I've never experienced anything like at my career. Not to this extent. I've had burnout where for a couple days I'm like, I want to work on something else. But not where I don't even want to go. it up or I don't want to work like it's just a it's always been like an obsession of mine so

quick question was just just so I really understand it for you it was the it was the pursuit of making the AI have the whole in one and it just didn't happen so you just kept pursuing this like feeling of like the perfect commit the perfect prompt the perfect no or was I think I just I don't think it ever stopped doing what I was looking for I think I just physically mentally wore myself out to the point that I had to stop. Like, I don't think anything clicked for me. When we were talking about it, Adam, I thought, like, one of the things that I remember you saying was just like, the company you were at before had like less of, like, direct

competitor. So it'd been like a while since, like you're working on open code. There's like, you know, a trillion dollars floating around of people trying to take every user, of course, and things like that. So some of it, if I recall, you were just like, oh, there's like a trillion things for us to build. And I want to build this and I want to build that. And we need to build all the things at once. And oh, my goodness, I can build all the things at once because I have AI.

That was the impression that I got when we were talking about it a while. Yeah, it's a very competitive field right now. I mean, a lot of people trying to figure out how they can be the ones that shape programming henceforth. So, yeah, I'm not used to that. I'm not used to having any competitors, let alone 100 of them I can look at on Twitter every day. So yeah, Twitter was, or X was a big part of also mental health drain. All of it kind of came together with the AI psychosis to kind of drive me into a hole. And it forced me to stop.

I mean, I didn't work for a month. Like I've never done that in my career, five weeks even, I think. And that was enough time to get some perspective to not interact with LLMs for a while. It's kind of like you start looking at it all very differently once you get kind of snapped out of that. Can you tell us what you did during those five weeks? So just no Twitter, no LLMs or no programming? Well, he was, he was, he was tired of, he, he was affecting him negative psychologically to go on X and see all these competitors. So he moved to Blue Sky where he wouldn't be scared of competitors anymore.

No, I actually, yeah, I was pretty, I was pretty off. Sorry. Pretty off of using that. No, it's good. I do have a Blue Sky account. Okay, there you go. Partly true. Partly true.

Get off the podcast. I don't use I don't use it I don't know I have one I try to cross posts on there sometimes I'm just kind of lazy about it

I don't I want people to be able to get stuff wherever they want it right so yeah anyway yeah I was for that I mean for that five weeks the first couple weeks

I was really sick and my whole family we just get sick because my kids are in school and that just happens and the veganism true in the veganism

yeah so I was a couple weeks just like hold up in bed watched Netflix for the first time in years that was interesting Breaking Bad is pretty good I'd never watch Breaking Bad

and like everybody talks about Breaking Bad I was like gonna watch Breaking Bad anyway I watched some Netflix and then I would say the last few weeks a lot of time golfing ironically got back into golf

just hanging out with my wife we played a lot of tennis while the boys were still in school just a lot of time offline it was it was the most offline I think I've been yeah in my career so in 15 years how are you feeling now

I feel amazing. Yeah, so I came back. I'm just now getting started working on some stuff again with the OpenCode team. And it was like, I feel like I came back right as the team. The team won just grew a lot. So we just hired a bunch of people. And I've kind of seen how a lot of the team members have found a healthy way to work with AI. And like just seeing their perspectives on how they use it to kind of like,

to like generate better code and focus on like spending more time. trying to drive toward a given outcome as opposed to go really fast, cover a lot of ground, like all the kind of stuff that I got into that was so unhealthy. So kind of like going more into the craft and trying to go back to we're just going to get to really good outcomes and use these tools to get us to better outcomes and we would have taken the time to get to before. So yeah, it feels like most of the team has kind of converged on this very healthy way to approach AI coding.

And that's been helpful. Like I can kind of reframe how I view this stuff. It's still a little scary. Like I don't know like am I just one prompt away from just like crazy obsession again. I hope not. But yeah, it's felt good. It's been like a week and a half now. I've been back.

Is there anything that starts to kind of trigger that feeling? Like is there a specific type of response or something that you can feel yourself drifting towards? Or is it just all general usage? I don't know. I don't really feel like I'm going to go back at this. point, I feel like it's some kind of hype cycle thing where once you're on the other side of that peak, you're probably like, yeah, the perspective, you're not going to fall back into it. But yeah, I don't, I can't say for sure, I guess.

I don't know really what triggered it the first time. I just think it was kind of like feeling that this thing is more intelligent than the previous models. And so maybe that's it. Maybe some new model comes out and all the hype that happens and all the excitement and it does something, some unique thing that the other models always failed at. And you're like, oh, my God, this one, this one's it. It's here to save us all.

Walk us through. You get a new ticket, Adam. How are you going to fix? How does that go now? I'm interested to hear on the... What's a ticket? You know, an issue on GitHub or something that you want to go fit. Yeah, not like for speeding.

Adam, it's like... For developers are like Teage where when they ship code, they get a lot of these issue reports on it. You're probably not as familiar with that. You guys do get a lot of issues. Big fan. Big fan. I get a lot of tickets. my boss keeps sending them to me.

He says we have AI. He's making 10X the tickets. He's coding now too, so that's really fun. But people seem to think it's like traffic tickets. But no, it's like raffle tickets. The more you have, the more you win. That's how it works. It's a 50-50 raffle.

Exactly. Yeah. Well, I think for one thing, like, the models have gotten so much faster. So the GPT models, which I only really use those. I don't really use the anthropopic models have always been fast. But now they're fast enough that you don't really have to do the whole paralyzed thing. Like you don't, you don't feel like you're wasting 10 minutes if you prompt and then just wait for it. Like used to it took so long you kind of needed something else going on.

So I've stopped the parallel stuff. I'm not trying to do anything like that. And just kind of focusing on a single task. Getting on Discord, like we've just got so much going on in our team channels now. We've grown that we're 20 people or whatever. So in between prompts, I can kind of catch up with the team or do stuff that I need to do on the open source side. That's been the healthy approach, I think, is just not trying to, like just having a perspective that like more code being slung into the code base is not more productive.

That's not the goal because you start to kind of feel like how much can I output or how much can I churn? How big of a diff can I create every day? That kind of, you know it like in your right mind, you know that that's not it. But like in the middle of it for some reason, that starts to become this like game. You feel like there's a super pet. It's Gary Tan on Twitter. Like you just start to you can relate with, or I can relate with all the things he says where you can just see the AI psychosis so clearly, like, start to feel so powerful. Like, I'm amazing. I'm going to need a new USB C cord. That's how, I'm going on this thing. A lot of tokens. That was going to be my follow-up question, actually, Casey, was how many USB C cords did you go through during that?

I don't even know if Adam gets the joke. I don't actually. Yeah, I feel really embarrassed. I don't think it was while you were offline, but there's a great. which we still are trying to determine. We cannot figure out. Is it satire or not? We don't know. But Gary Tan made a post about how like guys, burning through so many tokens with Claude Code over here

and my USBC cable burned up. Does anyone know what the rating is for tokens on these bad boys? I have, I luckily have it actually on quick pull-up on the app that I, of course, vibe-coded. I have my own Gary section, and this is it right here. Is it possible I use quad code so much? Somehow my USBC connectors burned out on my MacBook Pro. Two of mine are dead and won't charge, and now the third max is out of 15 watts.

Now I'm afraid my code Tomogachi is about to die. So, you know, Adam, maybe you didn't go far enough into... Maybe not. Maybe you only saw the first circle of the psychosis. Gary Chan is down the center. Yeah. The only problem with... my current perception of Gary is,

I don't think anybody is smart enough to write satire that good. So I don't understand. I don't understand. Anyways, whatever. So, there you go. There's only few world famous comedians, and I just don't know if he is,

he's one of them. That's what I'm saying. Like, I just, yeah. He showed up in the lobster suit, dude. That's true. I mean, I don't know. Maybe he's, like, way further ahead than we think in terms of the comedy cycle.

Yeah. Anyway, so here's a stupid question for you, Adam. Like, are you guys going to add something to open code that's like, if you would like to turn this on, this is an AI psychosis detector? Like, it's like, we will put up a thing that's like, warning, you have freaking five instances of this thing running. And it's been going for over 24 hours. And we are very worried that you need to like see. a psychologist right now just, you know, no judgment, just

trying to tell you. Yeah, it's like the old we pop up. Like, maybe you should go outside now. Yeah. Take a break. I'm not really joking. I was just, like, in all seriousness. It's actually brilliant. Yeah. I don't think we've talked about it, but given my history with it and the damage it did

to my mental health, it's not a terrible idea. I kind of like it. It seems like, you know, as long as it's opt in, I can't imagine people would complain, right? Like, it's like, hey, you turn it on. there needs to be a way to vote for your friend that it gets turned on so like if you know like nine

of your friends like suggest that it gets turned on for your account then it turns it on for you yeah okay so open you have to put in like your your ex handle and Twitter it'll go like crawl and to see who your friends are and it'll look to see if they start saying things

and it'll go like okay we're turning on the AI psychosis detector for you yeah okay have we thought about just asking the AIs if they have? Yeah. So is it hard working on a

AI product when you have to detox from AI? Because like it's kind of like it's like, it's kind of like, oh, I was a cocaine addict, but I have to work at a cartel or the DBA or someone who's like, cocaine's around all the time, right? And you're just like,

I wanted to try and get off this stuff, but people are asking like, is this pure? You know, like so. Yeah, you know, that's part that I kind of I feel like I've forgotten that that was definitely part of the psychology of that initial getting so hard into it because I would just tell myself like if I'm not using this why would I expect anyone else to use it so I kind of like convinced me yeah even if I wanted to write something by hand I would be like no I should try and do it with the eyes so I think I did kind of get forced into the reliance and the daily usage of it maybe that led to the obsession yeah

I don't know I'm a week and a half in like it's a little scary uh that I'm going to continue to use these things and I don't know exactly what it was, you know, that, that triggered it in the past. Are you, are you now doing things more by hand? Is it, or is it still, like, do you still effectively operate the same or has any sort of change happened? I mean, the change is, is mostly just slowing down and realizing, like, the goal and like, what, you know, what am I measuring week to week? What's important? It's not, I don't know, how many big PRs I can merge. It's, it's more, like, outcome-based. So I've definitely slowed down, but I am, I'm in, I'm, mostly just generating stuff.

I'm not writing a lot by hand. I'm definitely reviewing like I was in the beginning. I kind of got out of that in the middle of the psychosis. I wasn't even looking at the code anymore. You're just like, there's no time to read code. We must keep going. Okay. So, yeah.

Oh, go ahead. You can go if you're still going along life because that's going to slightly shift. I was going to. I was just going to say like in, in my experience is probably the person on this call with the second closest thing to AI psychosis at any point. or at least like I use the models a lot more than Casey for sure and more than Prime. Like if you stop reviewing for like a week, sometimes it feels like okay for something. But then you look back and you're like, why is there 500 instances of is record in this code base?

What? And it's just literally a function. Function is record type unknown returns is record type. And like, and you're like, okay, if it did that, what else did I not notice in the last week? Like if it couldn't even figure out that we have a function called is record, which already we banned from the code base. And I wrote like some of these I literally have like ESLint style rules that automatically check. The model does it so often that I like made something like, this function name is banned. We don't do that function here.

There will be no, even if there's an actual reason to check for is record, we don't do that here. Well, if I'm wondering The goal here, I think, originally, was they're trying to provide a tool that does what programmers generally do. And so they looked across the code bases and were like, you know what? Most production code bases have about 15 functions called Is Record and they all do the same thing. Only some of them may have slight differences so that you don't get exactly the same depending on which one you accidentally called. So here you go. guys, we've delivered to you exactly what you wanted, something that produces code just

like you run. Yeah, it does, in a way, it kind of does feel like the AI is the most human when it's retarded. Yeah, it's like, that probably happened in the code basis of this thing red, right? Yeah. But so, but the, so like from my, you know, from my side, like Adam, I'm wondering if a lot of it now is more like, it's just about, and like as a human, I can't focus on 10 things. deeply at the same time. Like, right?

So like when I thought I could do, like, if you thought you could run 10 agents simultaneously focus all them deeply, like that doesn't really, that doesn't make sense with my experience or like the people that I know. So then it's like, is it just that now you feel like your brain, it stays turned on the whole time. You're focusing on one thing. Like even though maybe the LLM still writing a lot of the code for a certain task, like, that's the main difference.

Like that's, I'm trying to sort of piece out where you think like that. that's the distinction. Man, I don't know. The tell for me is like, I sleep fine. So I don't know exactly like when that would go, if, you know, what would be the first signs I would notice that I'm slipping back into it? You guys have me a little worried, to be honest now that I've talked to the whole thing. Like I'm a little on edge that I can slip back into this.

We'll be on the lookout for you. Don't worry. Yeah, please do. That's the other thing. I don't know. Like, nobody really said anything to me. I don't remember people being like, hey, I think you have a problem with this AI coding thing. Well, to be fair, we couldn't see what your sleep,

schedule was. If you share me your if you share me your Google Fit Analytics or something, I can be on the lookout for you. That's right. We'll monitor your sleep schedule and email you as soon as we to check the nominees. I'm just going to do a quick I'm just going to take a quick 15 hours to build a quick agent to be able to monitor you

if that's that video for you. So actually I do have some real questions which is you know. What was? Wait, what were our questions? Our questions. I was asking real questions. Okay. I have a realer Questions, buddy. All right, these are some real hard-hitting questions.

Okay, real question here coming in hot. With your life, your life is a bit unusual. And I'm not just talking about being a vegan. Like, if you look over a lot of the people between the ages of 22 to 30 right now, they don't have, they're not married, they don't have kids, they don't have kind of, this kind of well-established kind of road. They don't have muscles. They eat red meat regularly.

They're probably on Adderall. they don't have like a lot of um who is this person this is like 85% of the valley right I'm like I'm like spot on calling all these people so far this is a real question this is a real question my questions were just these fake

loser questions and the crime comes in with this description of the average person I don't even know where I don't know where is that is Nebraska is it Silicon Valley where is that person eating red meat no has no muscles, they're just a fat blob that oozes along the ground.

You have, like, what is this person? Anyways, can somebody please take what he just said, feed it into like mid-jurney and see what comes out as the drawing, and so we could post it here right to see what, what is Prime talking about? Okay. I want it, I want Prime's question, though, just if someone can in the comments, leave the LinkedIn version of this question. Okay.

I want the LinkedIn. Okay. We're all saying they want versions. I want like a cartoon version where it's like the drawing in the guitar music, like drawing the person as I speak. Okay. So like the average developer right now does not look like your life. Like I'm just going to throw that out there.

I don't think the average young person typically has this type of life, a house stability, family, kids, all that kind of stuff, right? Just all, all by statistical purposes, everyone is getting married significantly later, having kids way later. how do they, and also friendships are completely like cratering right now. So how do they have the same people that can speak into their life or how do they recognize that they're in psychosis? Like what is your advice for somebody who right now is jacked out of their mind,

Silicon Valley vampiring right now, but they have nobody to speak into the life. Like what would you say or how would you say for them to recognize that they're in a bad place? Man. I mean, to recognize it, I would say, yeah, it's looking at your health. Are you abandoning habits you had before? This is basic, like, addiction stuff.

If you're, like, choosing to do this to send prompts instead of, you know, working out or whatever you were doing before that was healthy, that's a problem. In terms of, like, how it plays, I mean, I didn't have somebody intervene. I just kind of, like, it ran its course. But you said you had your wife be like, bro, you're not sleeping anymore and you're bringing your, you're bringing your laptop to bed. I don't really like this. Yeah.

And once we stopped that, once that became like, no, we're not going to have, we're not going to have that going on. I think it did start to get a little better. But it didn't really prevent the burnout. It didn't prevent me crashing pretty hard. Because, I mean, I stay here for eight hours day. It's the, it's the cocaine thing. It's like, I am going to go to work today. And, like, that's going to be an exposure that is going to keep the thing going.

So I don't know if, like, if intervention is a thing that's all that. Because, like, for every one friend that, if you did have one that told you, hey, I think you're going a little hard on this, you can find, like, 10 people online that are like, no, man, you're not, you're not going hard enough. Like, you got to go harder. Because it's very much a culture right now. And I think like that typical, especially if you're talking about a developer from Silicon Valley, like I think they're immersed in a group that thinks this is, this is how work looks now. Yep.

So, yeah, I don't know. I think it just kind of has to play its course. I don't really know that you're going to get talked out of it. Do a post tag Ryan Fleury and see if he says delete your account in the reply. That's a pretty good metric. I would be honored. You know, and like that would be my, as like, if he says that's chill, you're good. You don't have it.

Yeah, that's a good point. Yeah. That's a good point. Just put C.C. Ryan Fleury at the end of the retreat. And if he replies, delete your account, then you know you got psychosis. Done. The official diagnosis. I would say, I would say, you go to the doctor and you're like, sorry, I just, I got to delete your account from Ryan blurry today. I'm worried that I'm really in trouble. Like that my. I'll, I'll give you, I'll give you another test of maybe more serious one that I think you can do. If you find yourself agreeing with everything, someone who has a large monetary incentive reason to sell you on, right? Like,

oh, Anthropic says all of software dev is cooked in six months from now. There will be no reason to have skill. And then you look at their website and they're spending a bunch of money hiring people. And you say, but I agree with what they're saying in the pitch. That's like a good, that should be a little at least tickle in the back of your mind that like, hmm, maybe the billions of dollars that they have on the line are an incentive for them to mislead me and make me be in a bad place. You know, I think that that's like, that's an easy one to see online. If you're agreeing with everything Dario says, that's bad, guys.

That's bad. Yeah, there's definitely a flavor of it. Like, there's a flavor of the psychosis, these types that I don't understand really, like, what would motivate someone to be so, like, aligned with technology in general. It's like, if you try to say anything bad about AI or about, you know, what's going on with the big model labs, like, they're there to defend. and they're just like this random account on X. Like they clearly have something going on in their brain.

Just like technology enthusiasts or something. Maybe this predates the AI psychosis. Maybe this was going on. Oh yeah. It happened with like VIM and programming languages. Defending technology has happened forever. Like this is definitely nothing new. But I will say that the,

the ferrequency of it is unique in the sense that with like zero. Zero is a good example of this. There was somebody in the comments was like, man zero's awesome i can't wait to try it it's just like okay how can it be awesome and you've never tried it like there is like something so broken to this entire idea like this doesn't make any sense like you're so glued to whatever they send you that you're like this is the best thing ever i will now try it and it will be the best thing ever right like that is definitely some form of reality misstepping there uh i i have a question is it also is it a

real one, Casey, or is it just kind of like a throw away? No fake question. It's only a half and half question. Because I don't really know, like, this is not my, this is not my monkeys, not my circus, right? Like, this is a whole, I don't know anything about any of this, right? But I wanted

to ask, like, would it also be a good signs, as we're talking about signs if you start really getting excited about the G-stack? So would that be a good Would that be a good sign that maybe you've gone to if you're like, oh my God, the G stack? Is anyone really excited about that?

I don't know, but I feel like if you were. I mean, I've seen people. I mean, I've received not just one, but we're talking about dozens of comments being like, this bro is just angry because his skills have been reduced down to an MD file that Gary made. You know, like that kind of stuff. Yeah. All right. Well, okay, but to be fair, like, for you, pro.

No. I'm just saying for me it's okay, but like Casey, he can't reduce him down to an MD. I think, why is there no P-Stack Prime? Why haven't you made a competitor on stream? You're going about this all wrong. I'm not even joking. Why have you not made a competitor to the G-Sat called P-Sack on stream? I would watch the heck out of that.

Okay, to be fair, I am in the process, and I've been experimenting for the last couple months, making Prime Agent. Okay, there we go. That's all I wanted to hear. That's all I needed to hear. Yeah, it's a different. difference, you know, take on the whole thing. Okay. I would also say it's, uh, finding like the P stack is not as sort of like mythical as

finding the G stuff. Okay. So it's less to brag about it. It's not that big a deal. I said, I couldn't imagine people. I think I'm just not imagining programmers who are like taking Gary Tan seriously. But there probably is a group of like business pros, like right, non-technical folks that are like gary's on to something and they probably are impressed. So yeah, that would be a,

that would be a tell. So, uh, I guess like we should probably wrap up here, right? I mean, Adam, what's, uh, what's the closing? Like, what's, I guess your, your takeaway kind of sounds like you're not sure yet. Like, it's like, it's like, we're kind of just in the early stages of this and like you're going to come on in six months and tell us what actually happened, I guess, sort of. Yeah. I mean, I think, yeah, I think my particular story is still being written. It's possible. It's possible. I fell back into it.

I just imagine it, though, like the hype cycle. Like the, I just imagine that's kind of what everybody goes through. You go through this crazy peak where you're like, this is God technology. And then eventually you fall off. And now I'm on that slow, steady climb. I'm trying to figure out how to make this stuff actually help me day to day. So I think that's the long kind of healthy journey I'm on right now. That's my hope.

I think you're going to do great, Adam. Yeah. Thanks, Teage. Appreciate that. So I do, okay, can I have one more question for you? Okay. Real one, of course. Adam, you, you've always been a very, very good programmer.

Like, I've worked with lots and lots of people. There's few people who I would love to work with, you being one of them. Do you actually feel like you're getting this massive multiplier? Or are you still on this thing where you're, you're convincing yourself that you're getting this multiplier? Because when you're in like a code base, you're super familiar with and you're making changes. A lot of times it can be quite quick for you to just be like,

I know this so well. I know exactly what I want. But due to habit or like, no, I have to do it this way. You're doing these like different loops. What is like, are you actually getting that? So I would say I'm obviously biased. The cocaine factory thing. I can't escape the fact that I'm pretty invested in these types of tools for programmers.

So that that's pretty well kept me using the tools and not doing things by And I've been pretty vocal that I miss a lot of that. Like I miss the just what my work days looked like practically where I'd spend six hours on kind of monotonous stuff that I knew I needed to bang out and I could listen to music. I'm realizing I just completely forgot your question. I'm just kind of like rambling about what I miss. Yeah, it's, I mean, it's it's something that I don't think the genies going back in the bottle at this point for me, at least not in the near term.

there is a part of me that kind of like hopes the whole thing just crashes and burns and we all go back to writing code and like they can't afford the data centers or something and like all of it just goes away. But yeah, as long as I'm, you know, focused on open code and working on this stuff, I don't think there's a world where I'm going to spend a lot of time not using these tools. Was that your question? I don't even remember. He was kind of asking more about whether you think there's actually a productivity multiplier that you can say is significant. Yeah. Yeah. Especially with your deep familiarity, like, you know, you were a very, very good.

programmer. You keep saying that I think I'm just so distracted by the flattering comments. I think I work really hard. I think my thing is I go very hard. I don't think I'm a very good programmer. It's such a funny way to say like good programmer. Well, see, if you worked with me, I would work really hard on the problem and do a good job. So you wouldn't really call me a good program. Right, right. Exactly. It's like, no, I'm sorry. Someone doesn't really work very hard. Right. We're like, do a bad job. That's crazy. I mean, like, I would do terrible in, like, an interview or something, like, on a whiteboard.

But, like, I can throw effort. Don't worry about that. Yeah, probably. Do I think it's a productivity increase? I think, like, the way I'm seeing, and Dax has talked about this a lot publicly, I am seeing how you can justify putting a lot of effort into stuff that before you couldn't, because this thing does, it eliminates a lot of the tedium. So, like, you can actually try to get to better outcomes with your code in ways that you

wouldn't justify, you know, in the past, maybe you'd just say, like, we're going to shelf this. We know it's not ideal, but we're going to come back and take another pass later. Now you can kind of do all the passes at once and kind of hone in on making something what you really want to see, like the shape of it or whatever,

make the code how you like it in an ideal state. So I do think it's an increase in that. I don't know if that's productivity. I don't even know how to define productivity at this point. We know it's not lines of code. It's not lines of code removed. It's not diff size. I don't know what it is.

It doesn't feel like, like we're necessarily shipping a lot more features. But I do think it could allow us to ship better, better systems with the same amount of effort. That's my current view of it. That might change in three months. It might change in a year. But that's kind of how I view right now.

Yeah, I've found for me personally, just like there's stuff even like if you say, oh, it's like dumb, simple thing or whatever. But it's like, oh, this function, like the shape of it is wrong. but it's not something you can just like right click refactor and it like it's like it's shaped wrong in some way that can't be mechanically done by like a like a language server or something right and maybe you have like 200 uses of it or something like that now i feel very easy to be like switch this to this way it should look kind of like this if you run into something that's weird send me the list of them that were weird go and then like yeah which before i would just say like sorry guys guys like we don't have time for this. Yeah. Yeah. There's tasks like that that just weren't really on the table that now it's so good at those

types of things. I would say it's like certain things sped up a lot. But at the end of the day, the bottleneck is still kind of the same, which is like most of us don't have good things to work on or we don't know what we should work on. And that's like the product decision making and all the hard things that come into it.

I think that stuff's end up being the bottleneck anyway. And you can kind of outpace that stuff with just like, what if I just refacted it again? Or what if I had it. look for performance improvements and did that overnight. And we did it like every night. There's like a lot of pseudo productivity. Yeah, exactly.

Just stop sleeping and keep rewriting the code. Well, that's kind of an awkward ending. No, no, it's not an awkward ending. It's actually a great ending. We're all just considering, we're all bathing in what you're saying because, you know, all of us has varying levels of,

of AI psychosis at this place. And some of us, maybe we have anti-AI psychosis. People have been saying that's a real thing here. You know, maybe half of me. I have two wolves in me. One half battling it, one half going forward. I don't know. Very confused right now.

An AISIS. An AIS. An AIS. You're an AIS. I'm an AIS. It's very difficult. But yeah, no, we're just all really appreciating what you're saying because it is, I mean, whether, no matter what someone's opinion is on AI, it is single-handedly the biggest thing that

has happened in the average developer's life, at least, in the last quite a few years. you know, there has not been such a big change of possibility for a long, long time. So, I don't know, I find it very interesting. I think, so are you going to start the Adam Foundation, which is going to be like a place for for people who have AI psychosis? Like, I could imagine like an actual like thing for treating this or I don't know what you would call it. Kind of not really joking.

Like I feel like there's enough people now who've said they've had this. that there should be a thing that's like give generously to the Adam Foundation for you know Flying a ice like Ozarks right and like step two you got to give him the cocaine and then they oh that was that was I'm mixing metaphors. No wrong. Yep. We backslid poorly. I thought that was in the movie. Isn't that what the show?

Isn't that what he's watching? Anyway, give generously to the Adam Foundation everyone out there once we found it, of course. And once we figure out how to actually curious. Right, that too. But if they could give us the money first, they'll give us the money first, and then we'll figure out how to cure it later. I mean, that's how that, you know, it's kind of like, it worked for, what was it, the, like, Lance Armstrong thing where he's constantly like, give us money to cure cancer.

And you're like, have you cured it yet? You're like, well, you know, we're working on it. I do love the idea that, like, a doctor, you can have, like, a really serious doctor writing down a note. Yeah. And you look at the prescriptions, like, take a hike. Yeah, yeah, yeah. Go golfing.

some grass and go golfing. You're like, damn. I actually took a hike this morning. I took an hour hike in the woods this morning. I've been doing that every week, and that's helped me a lot, just generally with my mental health. So getting outside.

You're supposed to call me when you do those so we can talk. You keep saying, I'll call you and I'm walking so we can catch up. And now you just outed yourself, Adam. You just told me I spent a whole hour walking by myself and I didn't even call you. I'm so sorry.

Imagine how much more soothing it would have been to take a walk in the woods with Teage talking out the entire time instead of the beautiful sounds of nature which are terrible for AI psychosis. Exactly. I could have started reading him the Vim manual Think about all the positive effects.

There's so many things I could read to him. All right folks. Yeah. We got a couple. Do you want to do a closer here? You want to close us out? Honestly, I'm not sure if we really need a closer at this point. I think this episode. We just wave goodbye. Well, thanks Adam for coming on. Bye, thanks for having.

By, my way, it's Adam. dot dev on Twitter on Twitch You also did some YouTube videos in the past You've been around You work at open code Try out open code today At opencode.AI That's right

That's right. There we go All right bye bye Bye everyone Bye Bye