Dax's Lack of Stress, Confusing AI Models, and Organizing Twitter
The and and this is supposed to be the first episode we put on the terminal YouTube.
Adam:What what's the deal with the white face thing?
Dax:What white face? What are talking about?
Adam:Why are you trying to look like a white guy?
Dax:I'm trying look like a white guy?
Adam:Look you look whiter than me.
Dax:You're right. You know what, I do. That's really messed
Adam:I you were a white guy. Yeah. My eye is getting pretty good.
Dax:You do look more tan than me right now in this video. Right?
Adam:Yeah. And this is the week before I go on vacation.
Dax:That's true.
Adam:You're gonna come
Dax:back with a sunburn.
Adam:Mhmm.
Dax:I don't know. My my my camera's all messed up. Not capturing reality.
Adam:Yeah. When you start getting natural light into the picture and then you haven't adjusted your camera, I know that feeling. I know those days.
Dax:I have adjusted my camera but the natural light keeps changing because you know, the sun moves around.
Adam:It does move around. I've heard
Dax:that. Probably just adjust it exactly this time. This is the only time where I think actually on camera. Yeah. Yeah.
Dax:Maybe I'll just do that.
Adam:That's good idea. It's Friday. It is the day before vacation. You know that feeling when you have so much to do the day before vacation and we're recording a podcast. This is like dedication right here.
Dax:Yeah. Wow. We gotta like pack. I mean, you have the whole kids thing which I I don't know what that's like but
Adam:Like we're we're not it's We're not leaving till Sunday, but this is my last day of work before vacation.
Dax:Oh, I see.
Adam:I see. So it's like wrapping up all the work stuff before you go out and you're gone for a week. When you really wanna be gone, know, you're like, you're gonna really like not do work.
Dax:You have nothing hanging over your head.
Adam:Yeah, exactly. So it's like lots to do always. Do you Are you stressed? Like do you generally feel like in life, with work, with whatever, are you stressed often?
Dax:I probably would say technically yes, but it's just been a constant thing for years now. So I don't know if I really like register it. Think I have moments where they'll be I think it's probably how everyone feels. There's like random sets of weeks where it feels like nine different things are crashing together at the same time and I get pushed like to the edge of what I can possibly have the capacity for. But most weeks I'm like right under that.
Dax:So I think it's one of those things where
Adam:Okay.
Dax:I'm pretty much maxed out. So if anything like coincides in a certain way, I like will feel like it's overwhelming. But generally, I don't think I'm like a stressed person.
Adam:You don't seem very stressed. Do do you feel like that stress, like when you do have stressful weeks or above average stressful weeks, does it leak into your daily life like when you're not working, like with Liz?
Dax:I mean, I think in those moments there's like not really many minutes where I'm not working.
Adam:Yeah, okay.
Dax:Yeah, that's kind how I define it. It's like there's literally no room for anything, even a little thing
Adam:outside of work. So Okay.
Dax:Yeah, my memories of that is just all I can remember is just doing a lot of
Adam:work. Yeah, think I think I'm stressed way too much. I think it's a there's some health signals that I just spend most of my life, I think stressed. It affects my sleep. It affects my relationships.
Adam:I'm told I haven't fact checked fact checked this, but I'm told by someone that I trust that when you're stressed, you can't bond. Like when you have like a high level of stress, you can't really like build relationship or bond with people.
Dax:I would say I can think of the perfect counter to that. So I don't know, like maybe that's true, but there's also the classic thing of like going through some traumatic stressful situation with someone that like permanently bonds you with them. Like war. Yeah. Like really intensely.
Dax:Yeah. So maybe they mean something a little bit more nuanced but
Adam:Yeah, maybe.
Dax:Yeah, the distressing is funny. It's me and Liz always talk about it. I actually posted this yesterday. Like whenever you see someone that's like, I'm a 115 and then like you hear about their life and it just seems like very low stress like their whole life was like pretty low stress and that's how they get there. Or the other thing which I was saying is they're really evil.
Dax:I feel like that's like another aspect. Know, like there's always just like that evil dude family that just is just horrible but like will never die. Yeah. I think that's like another factor. So, if you can't lower your stress, I think you got to up the evil is what I'm thinking.
Adam:Okay. Well, I'm not sure which of those is easier. I'm really trying to like do the math in my head like, could I become evil? Am I already evil? Do I just need to lean into it?
Adam:I think
Dax:it will be hard for you, I think it's a little bit easier for me. Yeah.
Adam:You've got that you've got that evil streak. Yeah. You did some evil things, just like thoughts, things you post on Twitter out loud.
Dax:Are you referring to my post yesterday about slavery?
Adam:Yeah, I saw that. That was that was a tweet.
Dax:Someone someone so I posted, man, if slavery was legal, I would get so much more done. And a bunch of the replies were like, you're like, this is people were like, this is why people think you're white. I was like, no, you dummy. If I was white, I could never post this.
Adam:You'd never repeat that.
Dax:This is actually proof that I'm I'm not white. It's like the most definitive proof.
Adam:I actually had the thought like how confused the people who think you're white and the people who think you're black, like they all have very different interpretations of this tweet. Just like people on both sides.
Dax:I got some messed up replies. Like, I got a reply being like, aren't you black? You would be the slave. And I'm just like, that's like wrong in so many dimensions.
Adam:Twitter's fun. What what do want talk about today? What that felt there's a lot of directions we could go.
Dax:I don't know. What do you have?
Adam:What do I have?
Dax:Like, do have on your mind?
Adam:Like, I wrote something down on a little notepad. Let me go through my list.
Dax:You just said we have a lot of directions we can go in, and then I asked you what are they, and then you're like, I don't know. Why are asking me?
Adam:Maybe I didn't have a lot of directions in my head. There's one I mean, there's open code. We could talk about open code. Now that you've tweeted publicly. That's kind of a tough one to talk through, I don't know.
Adam:But we should I
Dax:we haven't talked about it. Yeah. Yeah. So we ended up having this split from the other person working with us on OpenCode. We just he wanted to pursue selling it to another company.
Dax:Just didn't make sense for me or Adam for that to be the outcome. So we ended up forking it. We're doing our own thing now. It's going pretty well. We've got like a bunch of nice functionality that you've added.
Dax:I started digging into the code base yesterday and I just see all these places where we can just make the experience nicer. It's just funny, it's like, again, so much easier when you are a user. I just identified all the places where I was like, I'm avoiding this part of the product because I don't get it or it's confusing.
Adam:Just gonna
Dax:go fix those now. Mhmm. Yeah. So I was kind of working through that.
Adam:Yep. I'm excited for you to do some things to the code base. I've just been like pushing the main like lots and lots of little commits and I'm just I'm moving at a weird like, I don't know what I'm focused on pace, like I'm bouncing on different stuff to have you come in with your like nice chill, unstressed vibe and just be like, I'm just gonna rework all this and you'll just do it like next week while I'm out. It's gonna be amazing. Very excited.
Dax:Yeah. I hope so. I was just sitting there staring and just trying to absorb Mhmm. The whole codebase into my head. I think I need
Adam:to I think I didn't do that. I just started slinging. I just started like making changes and that was my problem.
Dax:I'm surprised you're capable of doing that because I literally can't. I'm just like, I just get really like, what is going on? Like, where is this coming from? Like, what? And then I'm like, can technically make this change here but And I just get really lost in in those things and I can't do do anything to clear I my
Adam:think what I was trying to communicate that I did a poor job of it. Like, I just start from like the feedback loop of I'm running it, I don't like that thing, I fix that thing. I And I just keep doing that and then eventually it goes all over the place. Because some of the stuff's just like not important, but I'm like, I hate that. I wanna fix it.
Adam:And it's just like, all of it's mixed in. So I don't really look at the code base as a whole. I'm just kind of like, as the thing, you I'm using, what would I want better about this?
Dax:Yeah. But I actually think this works out perfectly because you're working on fixing this stuff for people at the end of the funnel, like they're using it every day and then like here's the stuff that sucks. My immediate place where I'm going to is what does the initial thing look like, like getting
Adam:it set
Dax:up and like how do we make sure that high percentage of people are getting through that flow.
Adam:I wonder how many people like read the read me and like try to run it and it doesn't even run. I have no idea. Because like I've just been
Dax:It's funny because like it's even before then. Like I look at the read me and I'm like, this already the way it's structured right now, we're probably losing x percent of people just because Yeah yeah. It feels a little overwhelming in some places. And I think I ran into where I tried to start it, just to test it in my home directory, it just completely froze.
Adam:I was looking at that this morning, which by the way, if you comment out like the LSP discovery stuff, it still freezes. So I I don't know what it is. It's something else.
Dax:Yeah. Yeah. So I'll I'll kinda look into all that. So yeah, I think I just I really wanna focus on if someone hears about OpenCode, making sure they get to the moment of being like, woah, I was able to like, you know, build a feature with it. How can we get as many people do that as possible?
Adam:Yeah. I'm very excited for you to do that. Do that this week and then we'll tell everybody jump on board, use OpenCode. The people listening to podcast unfortunately, you're gonna come in before Dax has done all that and you're gonna have the rough like founders series
Dax:experience. Founders edition.
Adam:Where you have to dig in and find stuff out. It's amazing though the people who have, like there's already awesome PRs that have been Yeah.
Dax:It's crazy. Like Yeah. People understand that code base better than I do already.
Adam:I know. It's crazy. Like, they don't complain about tech choices or they just start jumping in and making stuff that's awesome.
Dax:Yeah. I love Open Yeah. So I'm gonna spend some time absorbing a lot more of it. I think the nice thing is it roughly is the same shape as the SSD CLI and I've gone through a few iterations on like kind of architecture for that. Mhmm.
Dax:I mean, like, giving you an API, like a real time API and all that stuff. That's gonna
Adam:be be really cool. Nice.
Dax:I looked into connect this this is like very obvious. I don't know why I thought it was possible. I looked into like, can OpenCode just connect to your running LSPs? Like if you have Versus Code up or any of them up and they're responding to LSPs, OpenCode just jump into that? Mhmm.
Dax:But all these LSPs use standard input output which is like
Adam:One client.
Dax:Yeah. So they can only Yeah. So it's like you have to spin up two, so now you're using twice as many resources but whatever. It's a it's not a huge deal.
Adam:I mean, every time you run NeoVim I guess, like in different projects and stuff, you're you're doing that, right? Like there's an LSB Well, they're usually for different
Dax:code bases, right?
Adam:What do you mean? Yeah. I'm saying like, if I run NeoVim in four different projects at once, which I do every single day, like Yeah. Different code bases, it's spawning repeat LSPs for any technology that overlaps, any language.
Dax:But the LSP scope to the project. So yeah, you can't run less than four. But if you're running OpenCode and NeoVim in the same project, I was hoping they could both use
Adam:Oh, in the same project. Got it.
Dax:Yeah. Yeah. But whatever. It's not huge. I mean, the reason I think about that is, I have one project that is like an LSP torture.
Dax:It like destroys the TS server and I've
Adam:Yeah.
Dax:Never fully been able to figure out why. So whenever I think about on this, spend up two of those, I'm like really sensitive. But most most most like 99% of people don't have that problem.
Adam:Well, are you sure about the the standard IO thing? Because well, never mind. I thought I I guess it varies based on the LSP. But I thought I read that they had a different architecture.
Dax:It's not it's not required. Like you can it can serve over HTTP and there's like different protocols. But it depends who spawned it. So if Oh. NeoVim spawns in and it spawned in standard input output mode then, you know, there's no other process can really tap into it without like a lot of really crazy hacking
Adam:Yeah. That think it would be nice if we just like copied over into like an open code directory, cop like install into Yeah. You could copy them if they already exist on the system or reinstall it or whatever. Yeah.
Dax:Again, very similar to the SSTCLI, like there's a whole thing where like we have binaries that we depend on that we just auto download for you and and all that so
Adam:Love it.
Dax:Yeah. There's a lot of familiarity there.
Adam:So stepping back a bit, there's so many people working on this. I mean, this problem, this like Mhmm. Actual software developers building stuff. So there's like the whole category of like lovable Bolt
Dax:Yeah.
Adam:Yeah. Replit. You put in words and it just builds you an app or a website or whatever. And then there's like Cursor and Windsurf and others probably that are like, I would call them more accessible. Mhmm.
Adam:It's like developer focused but more accessible. And then there's like this whole other category which is the only one I'm even thinking of. Just throw all those other ones out. I'm only thinking about like, you're a real engineer. You use the terminal a lot.
Adam:Cloud Code was kind of like the beachhead where it's like, hey, we're just gonna interact your file system and like you don't have to have another editor. We don't make any assumptions about how you edit code or run your code. We just we work on your file system. And that just kinda like created this whole category that now there's so many different teams. And I when I think about like, what am I like, what do I think I can't get out of those?
Adam:First of all, Cloud Code, once you've used one of these tools, one of these agents that has diagnostics in the loop, LSP diagnostics, Cloud Code feels so dumb. I don't know if you've used it.
Dax:Yeah. Yeah. We
Adam:Yeah. So you but yeah. I don't know if you use OpenCode with LSP diagnostics.
Dax:I got it working.
Adam:Okay. Yeah. The difference is night and day. I mean, it's like it's a it's a huge change and it it makes Cloud Code just untenable for me like until they add diagnostics, I don't even go back and play with it just to see what they're working on, like all the features they add. Just because it's like it just writes code that doesn't work and then you're you're kinda like having to jump in and make edits, so you just get stuff to compile.
Adam:When you put LSP diagnostics in the loop, that's just kinda like table stakes, I think. Yeah. So it doesn't make mistakes. Like, makes mistakes, immediately fixes them. And you just kinda need that, I feel like, as a bare minimum.
Adam:But there's like five or six of these now that do that. Yeah. And I feel like they're the other thing so then why OpenCode? I don't know. Why am I making a pitch for OpenCode?
Adam:I don't know. I just kinda jumped into this. Open source. So open source plus has LSP diagnostics, plus it's terminal based. I think that's like the intersection that is like the uniqueness of OpenCode.
Adam:Not that there won't be other people that do exactly that. It's like ours will just be better
Dax:because we have taste.
Adam:That's
Dax:Yeah. It's it's this funny thing where I I was I talked to you about this a little bit. So it's clear that the editor with the AI thing together is the thing that gets the most traction right away. Because we've seen that with with cursor, etcetera. Mhmm.
Dax:So I think there's a lot of projects trying to jump onto that form factor. Mhmm. But long term, this feels like it's already too late. Like, it should just be Versus Code. It shouldn't be a Versus Code fork.
Dax:Yeah. But that's there. Okay. If it's a Versus Code fork, there's already two, cursor and wind surf. So at this point, you're like so far down Yeah.
Dax:To do another one that's like in the same form factor. It feels like just accept that we've lost that opportunity.
Adam:Mhmm.
Dax:To me, yeah, that just leaves the terminal opportunity because no one's really done an amazing job there. So even if the terminal one ends up being not as popular as the the Versus code flavor of this It's at
Adam:least an unsettled space where there's not a clear
Dax:like, I think if you like 100% win that, it might be bigger than getting like fifth place on on the Versus Code flavor of things. Yeah. And I think the other thing is I don't like, I don't really understand so if you look at, to the announcement yesterday from Windsurf, they are, you know, they they have their quote unquote own model now.
Adam:Mhmm.
Dax:SWE dash one or whatever. And I don't know what the history with this is. I know that both Cursor and Windsurf, like Cursor has hired a bunch of people to work on their own model. Maybe Windsurf was doing the same or this is just some collaboration with with OpenAI who's now their owner. Mhmm.
Dax:It's funny because like, I was posting about this yesterday. Let's say this is an OpenAI model, they developed it and they're gonna only deliver it through Windsurf.
Adam:Mhmm.
Dax:To me, that is a counter to what OpenAI claims to be. And this has been a question this whole time, right? Is OpenAI going to Like when they say we're working AGI, we're gonna like, you know, build this crazy future. They say that all the time or at least Sam Altman says it all the time. But if you look at their business actions, it's not It doesn't look like a business committed to to that or betting on that, right?
Dax:So there's been so many I think we thought of this before but like probably even like a year and a half ago, they announced like some really boring enterprise y type thing with Microsoft. And then now, them acquiring Windsurf and being like, oh, we have a model that's like vertically integrated, that's such a heavy bet into the application
Adam:Mhmm. Layer.
Dax:And they've said a bunch like, we're not gonna compete with you in the application layer and here's how we define what's application layer and what's not. But they're clearly just doing stuff more heavily than that. If they were really betting on being a platform, that SWE thing would just be another model in their API that any editor could use because they're not trying to compete on code editors, they're trying to compete on on this this bigger thing.
Adam:Mhmm.
Dax:And who knows, like, yeah, model might end up in the API. But I I am kinda wondering like, are are they now getting more defensive about this stuff and then they're trying to make sure their applications have kinda unfair advantages that that other ones don't. And if that's the case, then yeah, open code will have some theoretically some trouble if you assume that models are defensible, which I don't think they are.
Adam:Yeah. I don't either.
Dax:If they are then then yeah, we'd have some trouble.
Adam:This one that just launched. I mean, the the SWE one thing is not as good as Cloud three seven. So it's just kind of like, I want something better than Cloud three
Dax:Yeah. Too.
Adam:They're not better than that, it's just like noise. I don't care about the news. Yeah. And maybe like there are people I know that that wanna use these much cheaper faster models. I just don't know what they're doing with them.
Adam:Like, I feel like I have like a Super Maven type thing for line to line completion that's really good. And then I want like big brain help me do hard thing. I don't know what these middle ones like what the purpose like people who get really excited about whatever, running Quinn locally.
Dax:Yeah.
Adam:I just I don't I don't really understand like what are you accomplishing with that because I've seen how not helpful bad model output is. So like, is there Yeah. That much
Dax:use three seven, which is an expensive heavy model and it still kind of sucks.
Adam:It's still You still have to do a lot of like Yeah. Your own work to get it to give you what you need. Like, it can be really powerful when you do the right things, but there's still like this learning curve and like doing all the right things. And still, I end up going to Gemini 2.5, dumping my entire code base to ask it to hard problems because 3.7 is just stupid. I mean, it's just it's effective at calling tools and going really hard until it gets to a solution.
Adam:And if you give it a good plan, very effective, not like I use three seven constantly, I have hundreds of millions of tokens at this point. Yeah. Like three seven is amazing for what it is. But like it's still not very smart. And everyone that uses it heavily knows the big brain questions, you go to Gemini 2.5 or maybe o three and like bounce between those.
Adam:And they can do some amazing things, especially I love just with Gemini 2.5 being able to drop an entire code base. Because usually like, I work on pretty big code bases. They've they're under 500,000 tokens and that fits inside Gemini's 1,000,000 token context easy. So you just like, I use RepoMix to just like create a big static file that is my entire repo. I drop it into Gemini and I ask it hard questions.
Adam:It's amazing. Like it's really smart, but it can't call tools for shit. Like it won't do it because it wasn't trained on it. I don't know. It's just not it's not good in that role.
Dax:Yeah. We end up complaining about this every single week. And I imagine it's gonna be one week where it just gets fixed because I don't this doesn't feel to me like an
Adam:unsolvable It's a temporary problem. Yeah. I do think Google's gonna win this. I mean, if I'm looking at the trajectory, I just I think they're gonna come out with Gemini three or whatever. It's gonna be really good in these agentic tool calling scenarios because I mean, I assume.
Adam:I assume they listen to developers and they actually care about this use case being Google. Yeah. And then it's just gonna be it's gonna be like I would never use Endtopic again.
Dax:Yeah. Which again, if you are OpenAI or one of these other labs or whatever and you see that same outcome, that would also make sense why you're trying to shore up the application layer because that I is I think at the end of the day, like, ChatGPT as a b to c product is gonna be really hard to dislodge. Like, I think about people like far removed from our bubble that still use these products, they just are not gonna like switch to something else. They're they're not like trying all these different things, they're just like, oh, ChatGPT is like this magic thing that does stuff for me. Like this other thing being 20% better like, it's not gonna be a thing that they switch.
Adam:So I will say Google on that front. I saw my wife on her phone using AI mode on a Google search result. I hadn't even seen this. Mhmm. There's like
Dax:The first tab. A tab now. AI mode.
Adam:Yeah. So I could see Google winning over They a lot of normies on that
Dax:always have that. And I think they have this weird challenge of they need to So the benefit they have is obviously like billions of people probably use Google search every single day. They they can leverage that but it's like double It's a double edged thing because yeah, they can leverage it but two now they figure out how to like integrate people's assumptions of the product and push them into this new direction. And that's been hard, like you've seen them like just insert AI response at the top, which has been Yeah. Like, not great.
Dax:Now they have this new AI mode tab. Like, will people switch and use it? Like, do they get what the difference of these two things are? So it's a double edged thing whereas ChatGPT can it's just like a completely fresh Yeah. Focused thing.
Dax:That said, the product worked on ChatGPT is horrible. It just is look it just looks crazier. I took a screenshot of what ChatGPT looks like now and I'm just like, this is insane. Like there's so many toggles and buttons and drop
Adam:pay attention. Now, I need to
Dax:I specifically looked at the mobile app and I was just like, this is such a busy crazy screen. And I think it's oh, they actually cleaned it up a little bit now. Yeah. It does look a little bit better, but still, know, everything's just hitting under
Adam:more I mean, I just use the web app, I guess, mostly. It is still kind of like when you really step back and you think about this era of like AI that's really helpful being injected into our lives, like choosing between these just ridiculously named models Yeah. The like search versus deep research and then you've got like this little dot dot dot canvas, which is where you collaborate on writing and code. Google has a canvas thing. I don't know when I should use that.
Adam:Like, can I not ask you questions about code right here? I don't there's just like so many weird little like idiosyncrasies these things have developed that Yeah. You can just imagine in five years, all of this will look so stupid.
Dax:You know what drives me crazy about this? Okay. So I'm like, alright, I get it. There's like different directions you need go with the models and you're experimenting so there's there's a lot. I was like trying to make sense of it.
Dax:Okay. I get your own models do this like reasoning thing. Okay. So I I know when I see o, I the reasoning. But then what is four o?
Dax:Like, why is there o in gbt for like, that mean it's like also kind of doing that? And then I talked to these models now and they all seem to converge on the same behavior, so I don't even know what's doing what. Like, one of these initially, was like, oh, this like searches the web a lot. But then now they're all kind of searching the web Yeah. But then also there's like a toggle for me to turn on web searching.
Adam:Search What does that button do? I have no idea anymore.
Dax:Yeah. Gonna search anyway. So I'm just like, what? And then and then what And then I'm like, okay, deep research. I like deep deep research, it's a great feature.
Dax:But like, should I be using it with four point zero or o three or o
Adam:four minutei Which model when you want to do a deep research?
Dax:Yeah. I have the same That one is such an opinionated type of product. It should deep research should have its own tab where it's like
Adam:Yeah.
Dax:There's like no options. You just ask your question and they just pick whatever. It's it's a slow process anyway.
Adam:It's super slow. I So I've done the deep research thing with like multiple models on the same question just to like see it. Can I figure out like which output do I like better? I can't figure it out. It's too much output.
Adam:It's like it's very hard to read through both and be like comparing in your mind which is better. But think of how plugged in we are to this stuff and this is still so confusing to us. Like, why I would choose o four minutei versus o four minutei high. I still don't get the high thing. I get the mini finally.
Adam:Like, o four minutei, they've been working on o four and they just gave us a little like intermediate here's a yeah. Here's a little taste of what o four is gonna be like. But o three is like baked. It's out of the oven. But four is bigger than three, but this is the mini four, so it's actually cheaper and faster than the o three.
Adam:Like, it's just a it's a mess.
Dax:But then there's a mini high. Like, is a mini high? I don't get What is
Adam:mini high? I don't it. And here's another thing. I I had not seen this before. Under more models, they have the GPT four one, which I thought was only available to developers in the API.
Adam:I thought we were special. We got four dot one through the API only. It wasn't in ChadGPT when they first launched it. Now they've added it to ChadGPT. I'm like, what's the point?
Dax:What's what's what's four one good for?
Adam:It's actually really good at coding. Like, if I'm choosing
Dax:best coding model across the board on OpenAI?
Adam:It's okay. So this is so stupid how much knowledge we have. Yeah. So it's the best if you wanted to like swap three seven for an open AI model because you just love Sam Altman and his doofus face. Four one would be like your closest swap that in and call tools and stuff.
Adam:O three, super expensive, but it is smart. Like I use it like a Gemini 2.5 replacement for like hard problems. Does that make sense?
Dax:Yeah. Okay.
Adam:But it doesn't have the context window that Gemini has, so you're limited. You can't just drop your whole code base. I don't know. It's all a mess. I hate it.
Adam:Four one has really high context windows because it's kind of like made for programming. I don't know.
Dax:Man, this is
Adam:This stuff sucks.
Dax:It's funny because I think for me, I don't I don't even end up doing what you do because I my habits are just like open chat GBT and start typing right away and hit enter right away. And after I do that, I'm like, oh crap, like, I'm on the wrong model or like, I'm like, I guess this one's fine for this. I don't know. Cause I don't wanna start it over. So it's just like, I just never actually do this intentional thing of I'm doing this task, therefore I'm gonna use this model.
Dax:Yeah. Yeah.
Adam:I mostly like, if I go to chat GBT, I just use o three for most of my like questions.
Dax:I just hate how slow it is because sometimes I just have dumb questions where I'm just like, just tell me off of your memory.
Adam:Like, I don't even I think everyone would say just use o four minutei for that. Okay. I think that's what people like, people love o four minutei because it's like 10 a tenth of the cost of o three, but it's way faster and not really any worse. I don't know. I just like to use the best, so I use o three.
Dax:I think the search also is another thing that I feel it's funny. I think what I think the UI reflects this. They're realizing that every there's like so few true improvements, everything is just moving like the piece around on the board where it just gets better at some stuff and works other other stuff. Mhmm. The search thing is a good example where it so clearly solves one class of problems that felt like all Like that's the problem, like it's not So it's not up to date.
Dax:But now I realize, now that all these models are searching all the time, sometimes I just have questions where I actually don't want it to get biased by whatever it searches. I want it to like give me like the Wikipedia esque answer to that of like Mhmm. I'm asking something boring, just give me the most boring like historical like permanent ish Yeah. Information about this. Like, don't get confused by whatever's going on currently with this.
Dax:Because there's like something in the current there's something that's like a current event and I'm trying to understand like the historical Yeah. Context for So I don't want it searching and getting like biased by all that.
Adam:You need like a don't search button. Let's just add that to the Don't search.
Dax:But it's supposed to when it's not toggled, it should be don't search.
Adam:It should be, but it But searches anyway. It still searches. I forgot, we haven't talked about four five, which is like the goofy uncle that I actually do switch to 45 when I want to write something to someone else. Like, there
Dax:have been a couple of
Adam:things I've written, I'm not gonna talk about it too much, but like a letter to someone and I need help with the blank page problem and four five is the most like, who knows what you're gonna get, but it's like more human somehow. I don't know, maybe I've just been like convinced of that for marketing but It goes that way. That I do actually use four five even though it's so stupid how they named four five. Oh, man. And it's in research preview.
Adam:What does that mean? I don't know. That's all Everything is in research preview. Everything AI related is a research preview. Just gonna go ahead and put that
Dax:out Yeah. It's it's research previews of billions of users. It's it's crazy.
Adam:I do I do feel like this is uniquely OpenAI problem. Like, most of other companies have reasonable model names. They're not that hard. I feel like Gemini, it's like 2.5 or two. That's it.
Dax:Yeah. And for Anthropic, I like never think about their models either.
Adam:No. Yeah. I just use three seven.
Dax:I just always use 37. Mhmm. Yeah. And I'm sure like as frustrated as we are, I imagine all the people working at OpenAI, like, they feel this even more because they're
Adam:Mhmm. Like
Dax:they see users like running into all this and they're like, ah, I'm sure I'm sure they want to fix it. But I don't know, is it is it fixable?
Adam:Yeah. It's interesting like thinking about the future of OpenAI where they did like do this massive hit with ChatGPT like consumers like crazy growth, fastest ever whatever at the time. Like they they hit it big with this big app. But like from the model side and when you think about like the future of model development and research, do you really trust them to like beat these other people? I just don't.
Adam:Like I don't I don't trust them to beat Google. I feel like
Dax:Me neither.
Adam:Like Google in the end, it's just like Microsoft's gonna win the agentic Versus Code. I feel like Google's gonna win this and then what is OpenAI? I'm even more worried for Anthropic and I like like the CEO, all the things he says, seems great. But if they lose the coding thing to Google, what do they have? I don't know.
Dax:Yeah.
Adam:It's a lot of money, a lot of capital involved, right, keeping these things going?
Dax:Yeah. It's funny. I think we're so used to I guess, like if you if you're around our age, I think most of our memory is new things coming out disrupting old things and like winning. So I think we're very primed to that type of story. But the thing that is common historically is there'll be a new thing that came out and all the old players would be like, that thing's dumb.
Dax:Like that's not gonna go anywhere. The new thing ended up being right and it would like displace the old ones. And we can remember this with stuff like like when the iPhone came out, all the existing phone companies were like, that's just like a toy or whatever. Didn't look at it being like, wow, that's really good. We need to immediately squash it.
Adam:Right? Yeah.
Dax:But for AI, it's not going that way. It's it's the opposite of what we're used to. It came out and every single giant company fully over rotated into it and we absolutely cannot miss if this goes somewhere, we like cannot miss it.
Adam:Oh, you think about Amazon, the meme of everything becoming AI Yeah. For AWS. Like they just from a marketing perspective, from their focus perspective, it was just like completely shifted.
Dax:Yeah. So it's very different from them from the previous shifts. Like it it wasn't like and when the internet came out, there was a bunch of incumbents being like, internet's not a big deal. Like, whatever big like retail like Walmart or whatever, so late to doing anything selling on the internet and then Amazon got in there. Yeah, I just don't see that with this shift.
Dax:So I don't think it's gonna be the usual disruption flow that we're used to.
Adam:I just remembered Apple and that they kind of are doing trying to do I don't know. Has there ever been a more like crazy dichotomy between like amazing company that makes these the physical devices that are perfect and like they still have Siri on them? Like what what in the world?
Dax:They've taken so many swings too. It's like something is like broken with that part of the organization. I have very little hope for it. It's but it's like so confusing because this stuff just needs to be baked into the OS.
Adam:I know. I want it I want them to be good at this because I use my watch to say, hey, Siri, add this to the grocery list. Like, I want all of my Apple devices that I will continue to own forever until they die. Like, I want those devices to all be connected with this brilliant AI. I want the Apple intelligence things to be good and it's just not gonna be good.
Dax:Yeah. And that sucks. It makes way too much sense. So yeah. So there's that version of it.
Dax:There's a version where we want this feature and we can kind of imagine what a good version of it would be like and it's missing from something like Apple or iOS and Mac OS. Then there's all these products that should not have added any AI that Mhmm. Now have like AI features in them. They're just kind of annoying and stuff that's like we just have to like click around or ignore. Mhmm.
Dax:So it's like this weird like It almost feels like this weird infection. It's infected all of these products and it has made them worse and very few products are good about removing failed features. They like Mhmm. Because internally politically, like, bigger organizations just cannot do that. Yeah.
Dax:So it just feels like a bunch of software just got worse overnight. And like, is there gonna be some kind of another shift in the future that just like, there's a bunch of new products that just remove AI from the products that never did it well, you know?
Adam:Like Chrome extensions that just rip out anything that makes it like an AI looking API call or something.
Dax:Or just like just fully like new competitors, you know? Oh, I see. Yeah. There's like the products get bloated and then like new competitors show up that are more minimal and then they get bloated and then new products, you know, that whole loop. Yeah.
Dax:But this AI thing feels like it affected everyone immediately.
Adam:So that reminds me of some Twitter stuff that happened. I don't know if that was last week. The DHH stuff, did you see this interaction? It worked extremely online, I'm sure you saw it.
Dax:Yes. I actually was so annoyed by that whole situation. I totally reassessed how I used Twitter and redesigned everything about it.
Adam:Because of that one. That was the straw that broke the camel's back, Yeah. Yeah. Mean, there's been a lot of annoying stuff on Twitter. It's interesting that that one what can yeah.
Adam:Tell me what was so annoying about it if you're willing to
Dax:talk about it. Yeah. So the situation was DHH was doing his usual thing where they're doing something not that impressive and he's spinning it like he's a genius. So that attracts people to argue with him about it and so the Zach Kanter was like
Adam:I don't actually remember what DHH started with. I remember Zach was a quote tweet, but I don't remember what DHH
Dax:talking about like how much money they're saving by moving off the cloud and how it's a scam and how everyone else is falling for it and and they're so smart because they're the ones that aren't flying for the scam. And then you have all of these idiots being like, yeah, I'm also someone that doesn't fall for scams and like, I Yeah, like, that that makes sense, like, everyone's falling for a scam but not me, me and DHH, like, we know, we know what's going on, we're smart. And it's all like people that just haven't done anything and are basically losers, they're all like, pieing on top of this thing. That's going on in one direction. And then peep another group of people can't help but engage with this.
Dax:So then Zach Kanter posted being like, well, like, know, there's opportunity cost with everything and it seems like Basecamp totally missed the AI thing. They have no mentions of any AI features, which annoyed me because I'm like, the underlying point is correct. DHH can't just blanket say everyone is falling for a scam because everyone has different opportunity costs and it's embarrassing that Basecamp does not have more opportunity cost. That's a very good point to point out. Saying that proof of it is they didn't jam AI features into their
Adam:was product the wrong
Dax:Totally negates it.
Adam:It's just a softball for the other side of the argument to just be like, you're an idiot.
Dax:Yeah, exactly, you're an idiot. Then there's all these people being like, well, like, you know, he has he has like expensive cars and stuff, so he's successful and it's like, and then people are like, what's wrong with just like being that successful? Like, why do you have to be even more successful than that? Like, why not just be happy with your mansion and a bunch of cars and then It's just it's just that whole stupid conversation where like everyone talking about it I want
Adam:you to answer I want you to answer that question that the idiots asked and I didn't ask, just the idiots. Why why is that not enough?
Dax:Because it has nothing that has nothing to do with discussion. It is and it's totally it's fine. Do whatever you want. Right? But the moment you step outside of that and be and start to say everyone else is stupid, everyone else is falling for a scam, everyone else is wrong for not doing what I'm doing, then you open yourself up to be criticized by that, right?
Dax:And the fact that peep don't everyone tries to take this angle, it's like, he's right because he's rich. There's people that are more rich that disagree with him. So if we're just gonna go down an argument, I'm like, well, Jeff Bezos is more rich and like disagrees, so DHH is wrong. It's like a stupid Yeah. Line of
Adam:thinking. Yeah. So, forget about the argument on Twitter that made you change your Twitter habits. And I'm just a a boy standing in front of a boy asking him, why do you hate lifestyle businesses? Because I know you do.
Adam:Well, maybe not hate them. You just would never choose them. I don't hate them.
Dax:I just I I just hate this whole culture around like, the the the problem I have is anytime anyone points out any kind of failure of Basecamp, and there's a lot of failures of Basecamp, they have they've like failed in lot a of ways and they're not very Yeah. Impressive in a lot of ways. They retreat to we're not successful on purpose. If you try to be more successful than this, all these bad things would happen. Your life would get worse, your company would be worse, you have your employees would be more unhappy.
Dax:And that's just not true. It's like a really stupid way of looking at things like a false dichotomy. Yeah. And if we really wanna get into that, let's look at a company that's crushing Basecamp. Let's look at Linear.
Dax:Linear's employee retention is like absurd. It's like 98 of people they've hired have never quit. It's like really outrageous. So to say like, we're building a really happy, nice, chill company and any company that's more successful than us, you know, is worse on those dimensions.
Adam:It's just totally their playbook like Yeah. That's just how they've they've beat that drum for so long. Like that's how they market or something.
Dax:I mean, it's not to me, like I don't think it's like, oh, this is how they market for their product. It's just it's like a personal thing. It's like an ego thing. Like I
Adam:mean, like being arrogant is like their their whole thing.
Dax:Yeah. And like, you know what I did? This is actually what I I go back and I read this all the time. I go back and read DHH's post he put out in 2010 about Facebook. It's a whole big article about how Facebook is a total scam, how every investor is falling for it, they're never gonna make any money, their IPO number is ridiculous, etcetera.
Dax:It's like, if you read the way he talks about it, it's so convincing. It's really convincing, it feels like he's the one smart person using common sense in a giant crowd of people just not thinking and following each other. Yeah. And he was so hilariously wrong about this because everything in business is counterintuitive. So the person showing you common sense of like, obviously, it's straightforward, like why is everyone over complicating it?
Dax:Obviously, you just do a straightforward thing. It's almost always wrong. So I go reread that and I'm like, damn, it's crazy how everything he talks about now, it's the exact same tone. Like, I'm not even talking about the content.
Adam:It's Mhmm.
Dax:This tone of everyone's falling for it, just use common sense, it's straightforward. Again, I don't care about what they're doing personally. They're not stupid. Basecamp founders are not stupid. They're smart, they understand their business more than anyone, they understand their goals more than anyone.
Dax:Whatever they do for their business, I'm sure is better than whatever commentary that I would have. The problem is when they now need to feel the obsessive need to inflate it beyond their own borders and be Yeah. Like everyone is stupid for not doing it like this and then everyone piles on. Again, the people piling on are people that just do not do anything. So I was watching all this as back and forth and I'm just like, I just need to never see this shit.
Dax:Like it just Yeah. Gets in my head and I get annoyed and I end up thinking about dumb stuff. I've reorganized my whole Twitter so I never see a post from anyone that I'm not actively following.
Adam:Okay. I wanna ask you about how you reorganized your Twitter but just not yet. Hang on. Yeah. Wait.
Adam:Two things. One, we should have DHH back on the podcast. He could talk I for like two
Dax:don't wanna to him. Don't wanna I feel so bad. Just don't wanna interact with him ever.
Adam:It's a joke. I was gonna say we could talk he could just talk two hours and we could just watch him.
Dax:Yeah. Yeah. Exactly. And just listen to him again.
Adam:I saw he was at he went to Austin to be on Yeah. Lex Friedman's podcast.
Dax:So I really hope my Twitter is set up
Adam:so that whatever
Dax:dumbass clip goes viral that people are debating just does not Shut the hours of him recording.
Adam:Are you so excited? Okay. The other thing I wanted to say, I was still I was trying to get away from DHH and this whole this whole recent experience. And I'm really trying to ask you a question that I have like a perspective of you from kind of the internal view that the Twitter people and our podcast listeners don't see.
Dax:Yeah.
Adam:Where I know like a lot of ideas flow your way and most of them are not big enough. Like you have a filter and I just wonder like how do you do you stay so like locked in on I'm only gonna go for the biggest possible opportunities. Like what is that like constitution in your mind that's able to do that? Because it does seem very unique. I don't know other people who wouldn't be like just like most of these opportunities, like chase them.
Adam:But you're so like, no, it's gotta be the biggest thing I could possibly do on this earth.
Dax:I think it's really just having been through the loop already. I remember being someone that would hear a lot of this stuff and be very compelled by it. Like, everyone's trying to do these VC things and they never succeed and like the founders end up with nothing, like that whole narrative. Mhmm. I remember being like, yeah, I want to live a life where I build these small, calm products and like they get they make some amount of money, maybe not enough money for more than just me, but I like build a foundation and I'll build new things and they'll eventually give me the resource I need to do whatever I want.
Dax:Right? So the common narrative that everyone hears. And I've pursued that approach and I saw and I've deeply experienced what how wrong that is in like most cases. Yeah. It is just so hard to make something small work and the counterintuitive thing and here's the matter one that that the linear founder actually, he replied to one my tweets saying it's the exact same thing.
Dax:Weirdly, the bigger crazier ideas end up being easier to work on. So I just have the scars of thinking one way and then shift and realizing like, it goes nowhere. Deeply understanding that my time is limited and I can only work on so many things Mhmm. In my life. And being like, gotta do things that have a real shot at working.
Dax:So that hasn't changed. I'm trying to do stuff that has a real shot at working, but there's like a naive understanding
Adam:of
Dax:that and now there's like a more sophisticated understanding of that that I have now. And I posted just like one example of that the other day where I was like, everyone tells you, oh, like find like a small niche where there's nobody else in and then like build something there and you can probably like, you know, get traction. And there's just like a pure mathematical thing with that. Like if you're trying to build something where your audience is 5,000 people, you find one person that wants it, they love your product, they think it's the best thing ever, it's awesome, they're paying you. They have no one to tell about it.
Dax:They're like in some niche and yeah, you can argue like, oh, maybe people in the niche all know each other in practice, it doesn't really work that way. They can't go to their friends and tell them about it. They can't go to their husband or wife and tell them about it. They can't go to their parents and tell them about it because it's like too small of a product. Yeah.
Dax:If you build something that's meant for a lot of people, any user that's happy with your product is going to tell 10 other people. This like thing, even if your end goal is the same, like you're gonna size of business is the same, the one that can work for a lot of people is just so much easier, right? So there's like there's like that that thing, I learned that lesson in like a dozen different places.
Adam:Mhmm.
Dax:So now when I see stuff that seems like a good idea but doesn't fit those, I just see it as a trap. It just looks like a trap to me, like looks radioactive, so Yeah. That's why I kind of I'm able to focus on the other thing, just the scars of it Yeah. Not working before.
Adam:Before I ask you about your Twitter habits, I think my perspective is so jaded by I don't know if it's a perspective, I'm just gonna tell you my raw thoughts and feelings. It's like it feels like it's been a long career. Like I feel like I'm just tired and I've sold like my wife on enterprise value for so long and I've made so much less money being a founder and I have so much equity and on paper everything is so great. But like, it just kind of feels like what if the things don't work out and none of them end up like, I've had some small wins like some things that have you know, if you have a startup for a long time, like, there are ways that you could receive capital along the way or money along the way. But like, it could but all my startups that I'm involved in could just end up dying.
Adam:And then I spent my whole career like twenty years dying and pouring myself into making less like a fifth of what I could have made every year just like doing consulting. And like, I don't enjoy that work but like before I did the whole entrepreneur thing, it was a much better lifestyle and like it just is very tiring to feel like everything is deferring for the future and what if that future never comes and I'm like, why didn't I just do the race car thing? The DHH thing. Like lifestyle business and had a good lifestyle. I don't know.
Dax:Well, the DHH thing isn't people think that it's like a it's not it's not it wasn't a small thing. Like they that's a that's a tangent, like I don't I Yeah, don't really categorize
Adam:got you onto that. Yeah. Yeah.
Dax:Yeah. So I wouldn't categorize him in in this even though he claims to be.
Adam:Brushy just got a high paying job at Google and just been like, f it. I don't care about my work.
Dax:I I think what's funny is this is actually another example of the thing that seems to make sense, but is actually the opposite of the thing is true. So you described the failure case of your current path. None of these things work out and you just end up making a fifth of whatever.
Adam:Obviously, don't believe that's gonna happen or I wouldn't have kept going for
Dax:15 like, let's say, like, it just doesn't What work out, is the actual failure case, right? The failure case, when you work on these things, ends up being like pretty crazy opportunities that most people would kill for. Right? Like, take mean, I can't talk about this publicly, but like, the thing that for that I was telling you guys about yesterday.
Adam:Yeah. Yeah.
Dax:That would be a failure case for SST.
Adam:Sure.
Dax:Quite literally like a impossible to get dream situation for 99% of people. Yeah. So I think that's the other flip side. It's a
Adam:lot of the
Dax:opportunities that you have are because you did these you went for these things Yeah. That were
Adam:I guess, yeah.
Dax:Bigger, more visible, like And it's it's like so hard to tie because it's it's a whole playing a whole like what if game. I can talk about the filler case in the other way that sounds also really depressing, where you go work at Google and you do a good job and you make a bunch of money and then like ten years in, they just let you off because you're just a number. It does not matter. It does not matter. Yeah.
Dax:And anything you did doesn't matter. You're just Yeah. Laid off because there's they're cutting x percent of their workforce.
Adam:That's true.
Dax:And then it happened to be a time where the economy wasn't very good and you're stuck trying to find a job. No one's gonna pay you the same amount as Google because, you know, you're making way more than you probably should have.
Adam:Mhmm.
Dax:You don't necessarily have the skills that make too much sense outside of a company like Google and you have no idea what to do because your lifestyle is now inflated to something that only Google level position can fill, right? But anyone could describe the failure case in a way that seems like fatal?
Adam:Okay. Well, let me tell you a failure case that's worse than all of those, that I sometimes feel like I'm on this path and maybe this is irrational fear. But what if like, they don't I don't fail. These startups don't die, they just go on forever.
Dax:Yes. That is bad. That actually is the death thing. Yeah.
Adam:And then I'm 55 and I don't have any kind of like cool prospects in the job market and it's just like, I just spent my whole life making less money than I could have for a dream that like never really didn't happen but also didn't ever you know what I'm saying?
Dax:And this is a function of experience as well. And this is exactly why I think the small things, you should not do these like small lifestyle businesses.
Adam:Because
Dax:Yeah. The worst place to be is in a business that makes money but is not growing. Mhmm. Because mentally impossible to drop it because it seems like it's working, but guaranteed to waste years and years and years and years and years of your life. Lifestyle businesses tend to hit that pretty quickly.
Dax:These like bigger things I'm talking about tend to be very binary. Like they very quickly don't seem to work Mhmm. Or they they like you're definitely on like, you know, a a really solid path where you have a lot of options no matter how things change. So yeah, I do think what you're describing, that's actually the real the real risk.
Adam:And I don't really like dwell on that or I wouldn't have kept going this long. Like I I'm optimistic, I'm a founder. I think things are going to the moon. But like after a hard week when I just haven't had vacation in three years and I really need a week off like next week, that's when it all just starts to kind of like you play out those worst possible case scenarios and you get really depressed and you're like, man, I might I might like a lifestyle business where my lifestyle mattered. The
Dax:thing is that the point is that lifestyle business situation is a fantasy. It's not real. It's Yeah.
Adam:Yeah. Yeah.
Dax:It's like not like a real option. It's extremely rare and a lot of the success cases that people point to could not exist today. They kind of only existed in a moment
Adam:It's in like levels saying you should be an indie hacker when he's like the only one that's ever Yeah.
Dax:Made a lot of money. The great the best time to have been an indie hacker is when it wasn't really a term.
Adam:Yeah. Right.
Dax:All the ones that are successful started then. There's no room for new indie hackers. There's no room for new base camps. Like it's Mhmm. You gotta look at the world the way it is today, not how it was when these companies were started or these people started.
Adam:I'm feeling better. I needed that pep talk.
Dax:I confirmed your fear. I was like Yeah.
Adam:Was too dumb to pick up on it. I got confused. I started thinking about OpenAI model names again and I just got very confused and it worked out
Dax:for you.
Adam:So keep going. Yeah. What did you do to fix your Twitter? Because I I hate the For You tab and I think you said you hid the For tab?
Dax:So on on on my desktop, I switched to Typefulli Chrome extension. There's a Typefulli Chrome extension Okay. That lets you Chrome toggle
Adam:extension. Much everything.
Dax:And you can go to the settings, but my Twitter now only shows the following tab. It just hit a bunch of other like unnecessary stuff too. I actually don't even see tabs anymore. I just see my following feed.
Adam:Oh, you don't even see tabs.
Dax:And it's like it's just the people I follow and then I read up to the point where it's a tweet I've already read from the previous time I opened it and then there's nothing else for me to do. There's really nothing else for me to do on the app. I can't like
Adam:Wow, I did it. Wait, no, I saw for you, Ted.
Dax:Why did I say have
Adam:to like
Dax:go into the settings and and configure a bunch. There's a lot of options, but they're all just toggles. So easy easy to like toggle stuff and see what you
Adam:wanna I'm just doing this on the fly real time.
Dax:Also unfollowed pretty much everyone and started re following incrementally to like build up this feed. And then on my phone, now that I have this habit of like only going to the the following tab, I just never go to the for you tab on my phone. I like I got into a good habit where like I just don't have that swipe habit where I go into the other one.
Adam:Okay. Are there is there like a way to export these settings or something? There's so many toggle can you just tell me what you did? Like what buttons you hit? There's so many.
Adam:Zen rider mode? Do I want this?
Dax:No. Okay.
Adam:Trends on home timeline, recent media.
Dax:I just got a DM from Rahul Ligma because Wait,
Adam:the guy that was fired at Twitter or fake fired? Well,
Dax:follow him because I like his company. Like I the way he works
Adam:at What's his company?
Dax:Julius. Julius AI.
Adam:What is that? He was making like a stat meze competitor for a while. Like a basketball search thing.
Dax:Yeah. It's like a it's like a nice AI powered tool to do a bunch of data analysis stuff. It's like if if you build ChatGPT but entirely around like Mhmm. Processing Excel sheets and like creating visual data visualizations and stuff. Mhmm.
Dax:Interesting. But yeah, he was like cause I just re followed him because I unfollowed everyone and re followed them.
Adam:Mhmm. Because I
Dax:was resetting my list. And he was like, what did I do? And I've gotten a lot of messages like this and I'm like, yeah, I bet I bet I like because I unfollowed everyone. So I think people just saw people saw that I unfollowed them and they didn't realize that I unfollowed everyone.
Adam:Got So
Dax:it felt like really personal. I got a few messages like take
Adam:it very personally. Yeah. I mean the unfollowing.
Dax:Yeah. It feel it feels bad. Now I have to explain to everyone it wasn't personal. Well, for a lot of people it was personal, but if I repalled you, it's
Adam:You had to hide the ones that were personal under this like guise of I just unfollowed everybody. Don't understand That
Dax:was actually the solution. You know how we always you would never unfollow anyone? Yeah. This is a solution.
Adam:Unfollow everybody. Yeah. Did you like use like a tool to do that? Did you like hit a button? Think I
Dax:just wrote a like, I just wrote some JavaScript in the Nice. Console and just let it run and then obviously rate limits and stuff. So I just like left the tab open for a day and unfollowed everyone.
Adam:Okay. This this whole minimal Twitter thing, I just went through every single setting and there is no way to turn off the four u tab.
Dax:No. Yes, there is. I I turned it off. What? I don't know which one it is, but I turned it off.
Dax:I can show you a screenshot on my screen. It's literally just Well, I removed the tabs in general. There's a way to remove the tabs.
Adam:Oh, remove the whole tab thing, Okay. Whatever. I'll do this later. This is this is not our podcast listeners problem.
Dax:Anyway, it's a typefully minimalist Chrome extension.
Adam:Yeah. I'm I'm not gonna endorse it yet. Jury is still out for me, but Dax says it's great. It's great.
Dax:It's great.
Adam:Oh, timeline tabs. Okay. I can x them. Remove distracting elements. See, I thought these were checked okay.
Adam:Never mind. That's just It's inverted. It's inverted. Yeah.
Dax:Yeah. It's still confusing. Anyway,
Adam:there's something else I want to talk about. Oh, man. I just saw it now. I'm on Twitter and I just saw tweets. There's like I'm I'm developing this list.
Adam:It's not actually a Twitter list. Just like this small group of people that I follow that I wish I could just unfollow everybody but them and only see their tweets. I think that's what lists are for. But the people who are like super smart about evaluating all these models, I've talked about it before. But now that I just have the following tab, I'm just seeing lots of those tweets and it feels good.
Adam:I want all of those tweets.
Dax:It's great. Yeah. And so now I'm like, wanna follow more people because I wanna see more stuff in there. It's
Adam:Mhmm.
Dax:It's good. But it takes commitment because I'm like, on my phone, you can still just use the other tab. But like I said, that last round of stuff I found so annoying. Man, it's crazy how deeply it hit me. I was literally like, I I'm like, I just give up on people.
Dax:Like, why am I trying to help people? Like, all everyone is just so stupid and just falls to these same stupid cycles. Like, I was like seeing myself totally like retreating from talking about anything publicly. That's how Hey, we got that's how deeply it hit me.
Adam:It hit you so deep. We got breaking news on the podcast. Feel like our podcast is half an AI podcast at this point.
Dax:Yeah.
Adam:It's hard not to talk about AI constantly. OpenAI, introducing Codex. Oh, I thought they already introduced Codex twice, you might say, because there used to be like an old model called Codex and then they released a CLI tool called Codex. No. This is an actual new model called Codex.
Dax:ChatGP Codex one. OpenAI Codex?
Adam:It's an OpenAI new model. It supposedly beats three seven in the SWE bench, you know, writing Python.
Dax:Oh, yeah. Because OpenAI did say they were announcing something to eleven and they it's 11:10. Yeah. Okay.
Adam:They just did.
Dax:Wait. So wait. What is this? It's a
Adam:It's a model specifically for code, I think. It Yeah. It beat three seven. It's number one now on the sweep bench verified.
Dax:Wait. But is this what SWE one is?
Adam:No. This is Codex one.
Dax:But like
Adam:Forget the Windsurf thing. They've probably been working on that for eight months and it had nothing to do with OpenAI and it sucks and we don't care.
Dax:Okay. Okay. This beats It's
Adam:not a frontier model.
Dax:This beats
Adam:Oh my god. Okay. I'm so pissed now. I'm reading the stupid the blog post. Codecs can be guided by agents dot md files placed within your repository.
Adam:These are text files that came to read me. So now we have another one.
Dax:Agents dot md.
Adam:Now there's agents md, cursor rules, windsurf rules.
Dax:Agents md seems generic, so it seems like maybe that's what we have
Adam:to It does. It's better. I just I was trying to make OpenCode use context md. Context.
Dax:That was
Adam:gonna be the one. And then like, AMP just launched. They use agent, not with an s, agent dot MD. So now they're gonna have to change like, everybody just needs to converge on one thing. Let's go with agents.
Adam:I feel like OpenAI has the weight, the billions of dollars.
Dax:I can't even find this blog post.
Adam:Introducing Codecs. A cloud based software engineering agent that can work on many tasks in parallel powered by Codecs one, available to ChatGPT Pro. Oh, it's cloud based agent. Oh, I'm sorry. This is not so the model is separate.
Adam:Codex One is the model. Codex? Oh my god. Why did they release a CLI tool and a cloud based agent? Name the same thing within I like
Dax:literally cannot find this. I'm like going to their news news page
Adam:They have an info. Where do I give this to you? I'm gonna give it to you in Discord. Not Discord. The other one we use, Slack.
Dax:Introducing Codecs.
Adam:Uh-huh. Okay.
Dax:It's cloud based software engineering agent. It's
Adam:cloud based. It's cloud based. It's it's AMP. No. It's not.
Adam:AMP is a Versus Code extension. I'm sorry.
Dax:Is it Devon?
Adam:What are there any other cloud it's Devon? Yeah. It's Devon. It's cloud based. So you like it connects to your repos, I'm assuming.
Adam:But the more interesting thing to me from an open code standpoint is Codex One.
Dax:As a model. Yeah. So I'm looking I'm
Adam:assuming it's available to the API.
Dax:It's better than it's the best one on so
Adam:It's percent sweet link.
Dax:Time they're like, this model beats is the best benchmark and I go use it and I'm just like
Adam:Well, yeah. There's the benchmark and then there's actually good. Like, if it doesn't call tools effectively, then it's worthless. Like I mean, it surely does their
Dax:agent thing.
Adam:Yeah. This is like the purpose of it. So I I would trust I'm I'm still trying to find
Dax:And the API is Codex Mini Latest.
Adam:Oh, it is? Nice. Codex mini latest. Let's go. Oh, that's a smaller version of Codex one.
Adam:Okay. So that's specifically for using Codex CLI.
Dax:But but this is what yeah. Codex CLI is using exactly. Alright. I guess we gotta try this this product.
Adam:Damn it. This is the problem. With these stupid companies not knowing that they're platforms.
Dax:Like, it's just not another thing. Like, they're this is another bet into the application layer. Like, this is how you
Adam:They have like a models page. Right? Like, OpenAI just has like a I wonder if it's already in there. OpenAI models. Model reference docs?
Adam:No. Where is it?
Dax:Of course, there are examples all Python.
Adam:I know. What is it with the resurgence of Python? There's been so much stuff launched around Python. Like, I feel like we let the data scientists win or something, and it's so stupid to me that, like, because data scientists, all these people at these frontier labs all love Python because NumPy or whatever, I feel like that like became this thing that now all of us application developers are gonna start writing Python because I don't know.
Dax:Feels like it's an invasive like species or something. And like we like
Adam:Yes.
Dax:We kinda push it down and but we didn't kill it and now Like for you to crash. Resurging.
Adam:Uh-huh. Is ridiculous. Please, let's just squash the Python. It's not good, people.
Dax:Superhuman uses codex to speed up small but repetitive tasks, improving test coverage and fixing integration failures. Temporal feature development. So I guess some people had preview to this. Okay. So but like if this is in the app, I can use this now?
Dax:I think I have ChatGPT Pro. Where is Codex?
Adam:Is Codex a separate try Codex? It's there's a button. Where? Oh, It opened ChadGPT. I don't see codex.
Dax:I don't see it either. And I have pro.
Adam:Could be oh, wait. Available to pro team and enterprise. I think I only have plus. Plus I
Dax:have pro.
Adam:But it doesn't work when you do it?
Dax:Well, what does
Adam:Is that what Canvas is? No. It's such a cluster.
Dax:Oh, man.
Adam:So don't use o four mini high for your coding anymore or four one. There's codex one for that.
Dax:But it's only in the API. Only yeah. Or only in this Codex now with open code.
Adam:Yeah. I don't know if the mini one's gonna be it's not gonna be as good as the cloud based Codex one. You know what I'm saying? Like, they're reserving the best model for coding their little cloud tool now.
Dax:Yeah. Which this is just not what they claim to be doing.
Adam:That's what I thought. Do I have to get ChadGBT Pro? I'm gonna have to get Pro, I guess.
Dax:Is Faster Workflows optimized for low latency q and a and editing while retaining the same strengths in instruction following and style. A nice snapshot will be regularly updated as we continue to improve the Codex Mini model. Okay. I mean, I it could every time I've been optimistic, I've been wrong, but it could maybe it maybe it could be the new default.
Adam:Oh, it could be. I I'm like, they've definitely been pivoting towards we don't want Anthropic to win the developer market. And it started with 4.1, then Codex, the CLI, and now they've got this, which clearly they've been working on for a while. So the naming is intentional, I guess, that they have Codec CLI and they have Codecs, the web version, cloud based. I I haven't read the article.
Adam:I I'm curious. Did you pick up is it like what we would imagine? You you just like connected to your GitHub repo and it can just go to town.
Dax:That's what it seems like, connects with GitHub. But again, it like takes me to slash codecs and then it just redirects back to ChatGPT home. Maybe it's not like properly released yet.
Adam:Yeah, maybe not.
Dax:I'll probably have to watch a live stream so I can see what the deal is. I really hate the thing they do where they like crowd around the table together.
Adam:I know. I don't like their videos.
Dax:It it feels AI generated to me.
Adam:Yeah. We recommend assigning well scoped tasks to multiple agents simultaneously and experimenting with different types of tasks and prompts to explore the model's capability
Dax:This is a Devon killer. This looks Yeah. Like
Adam:It sounds like it's less like tool for developer, more like does stuff in your GitHub repo for you, which is also awesome. I mean, I'll take that. I mean,
Dax:I'm I definitely I mean, I liked the UX of Devon. I just I just thought it sucked and was so slow.
Adam:Yeah. It was just a bad implementation of it. Well, turns out breaking news is kind of boring for our podcast listeners because we're just reading the blog article.
Dax:Yeah. Well, that's what Prime does for a profession.
Adam:Oh, that's true. Yeah. Yeah. You should actually read it out loud.
Dax:Also, this isn't breaking news because this is like, you know, they're gonna hit a podcast next week.
Adam:Oh, yeah. Breaking
Dax:news for us, not for you.
Adam:I don't know. We're on the bleeding edge of internet online to online, whatever. Maybe people are just learning about this on Monday even because they don't use Twitter incessantly. Those lucky ducks.
Dax:Yeah.
Adam:Alright. We should probably get off here. This one's gotten long.
Dax:Okay. Alright. See you. Alright. See you.
