Adam Wolff from AnthropicAI
You know, like, for me, transferring my what I know and believe about software engineering and also transferring just some of the love I feel. I feel like, you know, with your product, with with ClawCode, with any good product, what makes it good is, like, it tastes good. You know? It's it's like a dish that you cook. Like, you can just taste the love in it.
Adam Wolff:You know? And that's something that was just made by the guy who's clocking in nine to five is not that good, and and those products are the ones that win. Absolutely.
Dax:So where are you joining us from, by the way?
Adam Wolff:I live on the West Side Of San Francisco out by the beach.
Dax:Okay. Nice. How is that?
Adam Wolff:Oh, it's great. Well, actually, this time of year is the worst. You know, it's like, can't see your hand in front of your face, soupy fog.
Dax:Oh, really?
Adam Wolff:It's nice day. Yeah. That never happens on July 4.
Dax:So it's like summer is not the best time?
Adam Wolff:No. September, October, that's the best time.
Dax:Oh, really? Okay.
Adam:Like, at the beach, are there are there beachgoers in San Francisco? What's that what's that like?
Adam Wolff:Sure. Yeah. You wouldn't really go in the water, unless you have a wet suit. But people come out here. Actually, there was this proposition that passed, in the last election to close the Great Highway and turn it into a park,
Dax:and it
Adam Wolff:was very controversial. You know, the people who live out here wanted it to be a road because of traffic and, you know, they just are worried about getting their groceries and stuff. But everyone else in San Francisco is like, yeah. We just wanna come and walk around there. So they shut the road down, and it's been pretty good, actually.
Adam Wolff:Even I feel like even the people who are super against it are like, yeah. Okay. This is kinda nice. There's often people out there. You know?
Adam Wolff:It's it's a good place to be.
Dax:Was that related to COVID at all?
Adam Wolff:Yeah. That was when they first shut it down. And then then there was this compromise where it's open during the week and closed on the weekends, which that was fine too. I don't know. I was, you know, I was with all the residents.
Adam Wolff:I was, like, kind of opposed to it. And then it passed, and I was like, you know what? I'm retroactively for this.
Adam:I think
Adam Wolff:it's gonna be good. And so I jumped on the bandwagon. It worked out.
Dax:It's funny because there's so many stories exactly like that all over the place. Because I lived in New York during COVID, and they shut down a bunch of streets. And a lot of them are, like, oh, permanent shutdown, which is nice. They've turned to really nice things. Now I live in Miami, and there's a street that we love that's just all restaurants, like, it's just and they fully settled in.
Dax:I didn't know until recently that was used to be a normal road, then COVID happened and they and they transformed it. So, it's been nice. It's been nice having more, like, walkable areas to hang out in. I don't know if anything happened in The Ozarks, Adam.
Adam:There are no walkable areas in The Ozarks. I mean, you can walk, you just don't get to anything useful. Just kind of walk arbitrarily.
Dax:Walking for walking's sake. Yeah.
Adam:Yeah. Yeah. Yeah. I do that a lot, actually. Adam, how long how long have you been at Anthropic?
Adam:I I know that you've got the React history. How how long since you joined Anthropic?
Adam Wolff:Yeah. It's been a little over a year. Really exciting times. It feels like ten years, but Oh, yeah.
Dax:That that's plenty. Did did you join to, like, directly work on Cloud Code, or, like, what was the situation?
Adam Wolff:Well, I joined you know, I when I saw GPT three and three five especially,
Adam:Mhmm.
Adam Wolff:I was like, the the this thing is gonna be awesome for coding. You know, I just I've gotta work on that. I gotta get closer to it. Like, I do not just wanna be a user. I wanna be I wanna know how it gets made.
Adam:Mhmm.
Adam Wolff:So I eventually I I joined this team at at, Anthropic called the Cloudification team. And the intention there was to accelerate the internal technical staff with AI. It was, like, part of the developer productivity team. Mhmm. So that's how I joined, and that's what I was originally working on.
Adam Wolff:It just the thing about, like, kinda putting an effort like that inside developer productivity is, like it's always like, hey. We can have all these neat, whizzy things, or we could just fix CI. You know? And, like, you end up just going to the bottom of these infrastructure things. Like, I I worked on, like, how to log the usage of Zed at Anthropic for, like, a month, you know, which, you know, like, you you actually really wanna know that if you're really trying to empower your staff, but it was also it also really cured my Rust fanboy, my boost of saturation with Rust.
Dax:Yeah. It it it's funny because I never thought about any of those things because I've mostly worked at like, on my own stuff for, like, really small small teams. But especially with a lot
Adam Wolff:of this
Dax:LM stuff coming into, like, developer workflows, we've been talking to bigger companies and they're all really obsessed with trying to understand how their team is using this stuff better and having, like, analytics and and metrics and better insight into those things. It makes total sense, but it's just something that I'd never been exposed to before.
Adam Wolff:Yeah. If you're gonna spend as much as it costs to empower every dev with AI, you kind of need to prove that it's worth it.
Adam:Yeah. Yeah. Okay. So I want to ask you. Could Dax and I talk about this cycle we have where it's like some days or some weeks, we're like, this is the future of everything and it's going to change my job forever.
Adam:And then there's like days where you just have a bad day with the LLM and you're like, f this. I'm never using it again. Do you from your perspective, being inside of Anthropic, do you have a different perspective on that as a developer because you kind of like you know something about the internal things going on in Anthropic, whether that's future roadmap or whatever? Do you have a different perspective or do you feel like just every other dev where you're like, good days and bad days, I want this to be better. I know it's the future.
Adam:Some days it's just not there. I'm just curious if you have any thoughts.
Adam Wolff:Yeah. I listen to you guys religiously. So of course, I know you guys you talk about that a lot and I just sit there shaking my head because like or nodding my head, should say. I feel the exact same way. And, you know, I think there are two there are a few different aspects to it.
Adam Wolff:You know, like, one thing is, like, I've gone through all these little mini eras with technique. So, like like, at one point, my editor broke, and I didn't really fix it. You know what I mean? Like, all the highlighting was broken, all that stuff.
Dax:Let me guess what editor you're using.
Adam Wolff:Okay. I'm I don't wanna throw anybody under the bus. I went back to Vim. I was like, okay. I know this works.
Adam Wolff:Yeah. But but, anyway, it broke. And I so I was like, oh, I'll just use Claude code to do everything. You know what I mean? I'm like, oh, let's call this variable x instead of y.
Adam Wolff:You know? And, like, when you try to use the LLM for that, you're doomed. Like, that is actually a terrible use of the LM. Like and I think hand finishing things is is that's what I do now at least. And, like, you know, and then you also just sort of realize, like, well, how big a chunk do you need?
Adam Wolff:You know? Like, I think I so part of it, think, is, like, we are changing
Adam:It's skill issues. You're just being polite. But it is.
Adam Wolff:Skill issues is really true. And the skill is not, like, oh, I know how to write a prompt. It's just, like, I'm able to come to this with a fresh mind
Dax:Mhmm.
Adam Wolff:When a new model comes out, when a new feature comes out. You know, even like plan mode has kind of changed how I use quad code. So you really have to adapt quickly. And I think I you know, as someone who's always spent, like, and hours tweaking my VIM config, it's sort of a dream. You know?
Adam Wolff:Like, Yeah. I would prefer to only do the meta and not actually have to push anything to production.
Dax:But Yeah. It it is fun. On one hand, it's been a lot because things change so quickly and yeah, you have good days and bad days. But it is fun just because we're all just effectively just messing with our editors, which we all love to do and that's effectively what all of this is. Just seeing how workflows can change.
Dax:I've been I've been into it, and I think the part that I've enjoyed is whenever there's something new that's different than how I normally do things, I, like everyone, resist it to some degree because any kind of change is painful. But then something flips in my brain where I tried to overuse it on purpose to see, like, what the extreme version of this is like. And in that process, you usually gain a good amount of insight. So I also did try to use LMs to even do, like, simple edits, things that I could do in a few keystrokes with with my keyboard. And surprise, I think I was surprised.
Dax:I think it was a it definitely wasn't great. Like, obviously, using an actual editor makes is is the right choice there. But I found a different kind of workflow where I was just queuing stuff up. Like I was like going through a file and like every little thing I would just queue up and then I would go to the next thing instead of like getting it done. And I think in net that probably wasn't my favorite, but it was a new workflow that I wouldn't have thought of before.
Adam Wolff:Yeah. One big thing that's like kind of the fad at Anthropic right now is like not to correct the LLM. So, you know, because the LLM doesn't see interruption from the user in training, really.
Adam:Mhmm. Mhmm.
Adam Wolff:And, actually, you'll notice that, like, sometimes if you interrupt Claude enough, it'll actually hallucinate interruptions Mhmm. Which is super annoying. You're like, no. Keep going. You interrupt yourself, buddy.
Adam Wolff:But, like, you know, like, that's a good example of one where, like, then I sort of feel like I over rotated on that. I'm, like, afraid to update it. You know? So, like, I don't know. It it's funny just how much presence of mind it takes to, like, really make sure that, like, what you're what you're trying is helping you or, you know, to to as you say, to, like, learn something from it and not just, like, fall into
Dax:it. So when you say when you the fad is to not interrupt it, do you mean you started over or like what what like what exactly do you mean by that?
Adam Wolff:I mean, like, you give it an ex you give it a detailed prompt, you give it as much tooling as you can, you give it you shouted it like, must pass all the tests before you keep going.
Adam:Yeah. Do not add comments. That one doesn't work. That one doesn't That one work. Not effective.
Adam Wolff:Doesn't help. No. It doesn't help. No. Help.
Adam Wolff:You know, you give it all your instructions up front and then you just like kinda let it cook for like, kind of, you know, you're you're trying to get it to go on its own for as long as you can. Mhmm. And then you come back to it, and you, like, review the whole diff it's created, and you sort of give your feedback on it. And then you kind of, you know, move on to the next terminal window. The one thing I think about a lot is, like, there's the the difference between it being awesome and it sucking is the difference between you sitting there watching the LLM do stuff while you do nothing and the LLM doing stuff while you're doing something else.
Adam Wolff:And then there's also this question of, like, if the other thing you're doing is scrolling Twitter, is that good or bad?
Dax:Yeah. Yeah. Think we talked about it. Yeah. It's a I think for me, I my journey with this stuff was never gonna leave NeoVim.
Dax:Like, I just refused no matter what tooling came out. I was just like, no, I can't I can't leave my baby. And it eventually felt like, how long can I resist this because some of these tools look pretty powerful? And I got some of it inside NeoVim, like the Tap Complete through Super Maven, things like that. But I was seeing Cursor and I was like, man, am I just gonna be forced to use it at some point because the productivity is so huge?
Dax:And then Claude Code came out and I think it was a good Trojan horse and that I was able to use it without like shifting anything about my editor, like it was a separate thing that lived outside of it. But it also introduced a totally new workflow to me where whenever I saw people using cursor, was like very active and it was trying to augment you as the person writing code, which, you know, some of the stuff is greater than tab complete. I think it's great for that. But that never clicked for me properly. When the Cloud Code came out, it showed me that there are there's this completely different thing where you can kind of fork your work in certain moments and hand something off to Cloud Code while you continue to do other parts of it and then check back in on that later.
Dax:That was a new workflow that I'd never considered at all. And that was kind of like the breakthrough for me being like, oh, okay, now I see how I can incorporate this stuff in a in a much bigger way.
Adam Wolff:Same. I you know, the very first thing I did, this is GPT three five. I built an integration in NeoVim. It would open a buffer on the right, by the way. That is where the that is where the window belongs.
Adam Wolff:You would open a buffer. And, you know, just because VIM is so easy to move text back and forth, and this is you know, three five did not really have tool calling. People started doing that thing of, like, respond only with the new program or whatever, but it didn't work that well. So I would just use it for code understanding. You know, I would just, like, paste in a big function and be like, hey.
Adam Wolff:How could we refactor this? And, honestly, there's something I kinda miss about that. Like, that was, like, a golden like, I learned how to do Kubernetes
Dax:from that. Mhmm.
Adam Wolff:And I feel like that was, like, my golden era of, like, actually learning things from the LLM. I've noticed, you know, we talk about, like, these different fads. Lately, I've noticed I've gotten really lazy. I really trust, especially Opus. Like, I over trust it.
Adam Wolff:You know what I mean? Like, I'll go back into the code. I'll be like, who wrote this? What is he doing? You know?
Adam Wolff:It's like, oh, I just I committed that PR yesterday.
Dax:You know? Yeah. Yeah. Yeah. I I think I also had a phase where I try like, tried some of the NeoVim tools that kinda brought some of the functionality in editor.
Dax:And they like, I kinda used them. Like, it was nice to be able to select some code and be like, oh, you know, refactor this or add this functionality to it. But ultimately, it just felt a little too clunky to really stick as something that I that I that I did consistently. Kind of like a nice to have here and there. But yeah, like I I'm really into this whole thing.
Dax:I think that we've talked about this before, but this mental model where you have kind of like a dumb idiot sitting next to you with their own laptop. Can just give them stuff. They're not messing with what you're doing, you can still be doing other stuff. And that separation is nice because, you know, they are kind of dumb and you want to be able do this other stuff while they're doing their thing.
Adam Wolff:Yeah. But like, oh, you know, like, I'll just be like, hey, can you just write a doc on how the installer works?
Adam:Yes.
Adam Wolff:And, like, that's just incredible. I love it for
Dax:that. Yeah.
Adam:Like, we use we use a lot of mono repos and just having, like, a web project with, like, MDX files for the docs in the same mono repo and being like, hey, document this packages whatever feature in the web project and it just writes the MDX file. It's amazing. I love it for that kind of I agree. Yeah. I guess it is like a continuum that like we Dak's talking about this workflow and it's the Cloud Code kind of workflow that you guys pioneered where the human is very in the loop, but there are like so many bets being made too on the completely asynchronous, the jewels, the whatever.
Adam:I don't remember all the names of Devin. Yeah. They're kind of like give it a problem and then it comes back with the PR. I guess Anthropic's not really doing that today. Right?
Adam:Is that do you view like some point in time where the models just get so good that we don't need to be in the loop? That's one of my fears is like not like I'm gonna lose my job thing, just like I really enjoy this whole Claude Cody kind of workflow, and I hope that it's here to stay. But I wondered, does Anthropic do you have a different perspective being inside Anthropic? Like, is this temporary?
Adam Wolff:It feels transitional. I mean, just the fact that just that I feel like every week I come to my work and it feels a little different. You know? So, obviously, things are changing really fast. I would say, you know, like, one measure that we talk about at Anthropic is how long does the agent work autonomously.
Adam Wolff:And I think everyone's goal is to increase that number. So that is where we're headed. It conflicts with something I've learned though as, like, having been a software developer for a pretty long time, which is that, like, the hard part is always figuring out what you're trying to do and and why.
Adam:Yeah. You
Adam Wolff:know? Like, it it's it's always been pretty easy to, like, poop out a PR. The question is, like, you know, is it is this actually a feature that is gonna help? And, you know, it's amazing in retrospect when you look back at, like, a year of what a team did, how how little of it has actually helped at all. You know?
Adam Wolff:So I like, I don't there's part of me that really is dubious that, like, a three sentence prompt will be enough to make anything better. I I'm I'm I'm really curious to see how that plays out, honestly.
Adam:That's how I feel when I see these people. Don't know if you've seen the Twitter posts where somebody's got like 16 Cloud Code instances running in parallel and like like how do you have that much stuff that needs done that you actually can keep it going? I I just like to me, they're just spitting out garbage and it's like, look how fast it can run. I I don't know. That whole thing just kinda
Adam Wolff:Oh my god. That whole world is I I you guys rant about that a lot, but still I find it kind of amusing because it's pretty clear like, well, obviously, you're not doing anything you said posting on Twitter.
Adam:Yeah. The main one I can think of, he's like, ignore my mom in the background. And that's just so perfect for me. It's like perfect. Perfect explanation.
Adam:Exactly. Yeah.
Dax:I don't know if you saw my post the other day, but I was like, 95% of my energy goes to figuring out what the hell we should be doing and then 5% goes to coding. And now with AI, it's like 96% and then 4%. So Yeah. Yeah. It's definitely it's definitely better, but like the agony of work is always, what should we be doing?
Dax:Why should we be doing it? Should we be doing this thing or the other thing? And it's cool that I can spend more time on that stuff. And I think the part that's a little bit unclear for me is a lot of the process of figuring that out is trying stuff and is doing stuff. And I I wouldn't say that the LMs take away from that because I can think of discrete examples where it's exactly the opposite.
Dax:I had a situation the other day where I had two totally different ideas for how, like, a certain, like, UX flow could work. And in the past, I would sit and think really hard about which one makes more sense and I'll commit to that and we'll build that. But then now I just had both built, And then I tried both, and one was obviously worse than the other once it was implemented. And in the past, I would have never been able to do that. So that process of exploring still exists, but I will say some of that is literally doing the coding as well when you're trying to figure out like architecturally how should something structured, what patterns are good, what patterns are annoying.
Dax:But yeah, have a hard time imagining how I could be effective solely from like opening GitHub issues and watching them get completed.
Adam Wolff:Yeah. Yeah. Well, I I totally agree. I have this theory actually that, like, if we use these things right, we can make the code a lot better. You know, everyone talks about how, like, AI makes code worse, but one big difference I've noticed is, like, I'm much less attached to any given approach because I didn't do that much work to to get it going.
Adam Wolff:You know? So I I totally agree. And, also, if you try two approaches, you know, you just sort of look at the PR and you just look at, like, how many files were changed or, you know what I mean? Like like, macro things. And you can kinda tell one of these is gonna go in more easily than the other one will.
Adam Wolff:And then often what I'll do is I'll, like, try a few different ways, and then I'll be like, oh, this is a way to do it. And then I'll go back. I'll reset everything and I'll have Claude be like, hey, write an MD file about how we're gonna do this one. And then then we'll kinda do it step by step together.
Dax:Yeah. The step by step together is also interesting. I've also found that I haven't used this for programming as much, but for generating other things, I've had it Instead of giving it a prompt and being like, generate the thing I need, I will give it the prompt and I'll say, ask me questions step by step to figure out what to generate. And it does such a good job of going back and forth with me and then pulling out the information it needs. Yeah.
Dax:So some of those workflows are are super interesting. And then the thing you're saying about improving code, this not totally related, but Adam, did you see that issue that somebody opened today about how the they were like the the issue title was Improve Code. Did you see that one?
Adam:Yeah. They pasted like an LLM gave them a review of our code. They just pasted as an issue.
Dax:There's a there's a file in our code base that's really bad, but it's bad because it's the most complex part where I've where I understand the least. So I'm like letting it become a mess as I understand what we need to do before we refactor it. Someone clearly prompted the LLM with something like provide an argument as to why this code is horrible because it uses that phrasing a lot. It's like, it would be justified to say this code is horrible because and then it like breaks it down and then it like and this person has opened this issue with this like several paragraphs of why this code sucks and it was just like, should improve it. Was like, are you even a user of this app?
Dax:Like, what you went and did this? Like, why? So I was annoyed. Then I and I replied being like I replied calmly explaining why it's in that state. And then I threw in a once you get more experience, like, this will make more sense to
Adam:you. Nice.
Dax:And I closed the issue, like and but, yeah.
Adam:The good and the bad. We get the good and the bad with LLMs.
Dax:And source. Open source. Yeah. Great combination.
Adam:I'm seeing a lot more of the AI content stuff or just like search results and like Twitter interactions, just all kinds of random places that I I so clearly, is not a real person. That's getting more and more annoying. I hope that we figure out a way to solve that. I don't know. That's not your problem, Adam.
Adam:I don't know. You
Adam Wolff:know, I worked at Facebook, so like I am incredibly paranoid and suspicious of basically everyone. I just assume everyone else is an LLM. That's a good starting point.
Adam:I love
Dax:it. One of the most frequent comments we get in any video we do is everyone asks if Adam is AI generated. They they say I'm
Adam:a skin walker. I've literally I never heard of this stuff until until I got on the Internet. Yeah.
Dax:Yeah. It's funny. I can't I have this thought all the time where I'm like, you know, a year ago, if you told me how much time I'd be spending thinking about whether a reply was AI gender or not, I wouldn't have believed you. But it's crazy how much of it exists and it's in a format that isn't immediately detectable sometimes, or like there's something in it that triggers something else in your brain so you don't think to think that's an LLM. Like, it'll compliment you or it'll ask you a question, and immediately your brain is like in the mode to wanna engage.
Dax:And there's people that have been repeat quote unquote people that have been consistently replying to my post for a while now where I only put it together that was an LLM because they were replying so consistently. They were replying to every single post and very quickly after I posted it. But other than that, I maybe would have thought that they were real. So, yeah, it's crazy how much my energy is going into figuring that out.
Adam Wolff:Well, yeah, I mean, again, starting with Facebook, I sort of noticed how the technological limitations and implementations
Adam:Mhmm.
Adam Wolff:Really influence the culture. You know, like like obvious examples, like eight bit. Right? Like, that that's a whole style even when you have plenty more than eight bit. You know, you're rendering four k HD, but it looks like eight bit.
Adam Wolff:Right. And I feel like AI now has really suffused the culture, like, that kind of hallucinatory, dreamy. You know, there is, like, that little, like, retro futuristic time we had, you know, when mid journey was really good at that. Yeah. It's interesting.
Adam Wolff:I guess I'm what I'm saying is like the people start to imitate the AI, I think, too, is like another aspect of that.
Dax:Yeah. Yeah. I I had this crazy experience once where I somehow ended up on some weird Twitter space, and I was I'm pretty sure it was a person acting like an LLM, where they would respond with weird delays, but the responses were like very good. And the person was claiming that they were like an unreleased model or something, but ultimately was an actor at the end of the day. Like, this is so circular and crazy.
Adam:Can't believe So confusing.
Dax:This type of stuff is happening.
Adam Wolff:Then you go try to explain that to like your normal friend and they're
Adam:like, what are
Adam Wolff:you talking about? Like a person pretending to be an AI, what are you doing?
Dax:Yeah. So that's another thing. So as much as every single conversation we have, everything we see, everything we read is all about, like, coding and coding with LLMs. And obviously, people that are programmers by trade are, like, more plugged into this stuff than the average person. Despite that, there's still so many people that have not tried using LLMs for programming at all.
Dax:Still, it's still
Adam:like Mhmm.
Dax:Such a small percentage of people that have tried it or incorporate it into their workflow. Do you guys feel the same way? I mean, you're like more in the heart of everyone around you thinking about it. But do you get glimpses of how big the world actually is?
Adam Wolff:Oh, yeah. Well, you know, set aside like all the regular people who aren't programmers who haven't touched this stuff. Although I feel like that's actually flipping. Like Mhmm. Two of my best friends are like a doctor and lawyer.
Adam Wolff:And they a year ago, they were like, these things are not useful. And now they're like they send me their, like, deep research on, you know, our backpacking weekend. But but for programmers, yeah, I mean, you know, we with Cloudco, we end up talking to a lot of enterprises.
Dax:Mhmm.
Adam Wolff:And, like, there there, you know, there are people who are like, well, our our staff is really skeptical of AI. And I honestly, I just have to suppress my feeling of like, are you kidding me? Like, why how could you be skeptical? Like, you've gotta try this thing or, you know, I mean Yeah. It's got problems, but, like, how could you not be curious about this?
Adam:It's always amazing. I mean, the divide between like our little circles on the hyper internet Twitter crowd and like normal devs. Because I work with a lot of normal devs, like my normal job and it's just not on the radar. Like or it's just like not good enough or they don't they have some perception of it that it's not good enough. And yeah, it's there's a disconnect.
Dax:The thing that's always weird for me, and I've observed this with every tech and also building stuff for developers, I often get this question of, oh, like, why should I try it? Or like it it it to me, it feels like people want a mountain of explanation and research and argument before they even try it. And in my head, I'm just like, you can just try it, you know? Like, why do have to be skeptical of LMs? You can try it and you'll probably land somewhere in between, like most people do.
Dax:But that that always confused me. Maybe they're really busy with other stuff and they wanna like disrupt any of their workflows, but I can never make an argument for any new tech because, you know, you should just try it and see how it applies applies to you.
Adam Wolff:Yeah. And, you know, having worked again, like with a lot of different engineers over time, I just try to only work with people who write code because they love to. Mhmm. And those people love these tools. Mhmm.
Adam Wolff:You know? Because, like, again, it's just like it's fascinating. Like, I I program computers for twenty years, but now I can talk to them. Are you kidding me? Like, that's that's insane.
Adam Wolff:You've gotta use that. I think there are plenty of people who just they go to their work. They get their tickets done, they go home, they just don't even think about it, you know.
Dax:But it's funny, even for for those people, like it's even better, you know. They can't earn totally. But I get what you're getting at. There's like, there's a level of reflecting on like the meta that not everyone has the habit of doing, which I guess is alien to me, but I've seen it enough where people do need to be given like, again, I just always imagine like a larger book report. Like, feel they're asking for like a big book report explaining all the reasons and all the pros and cons before they'll even dip a toe.
Dax:But again, just weird for me because they're like, it just seems to cost nothing.
Adam:Okay. Do you think these devs that don't that aren't really interested in the meta, that aren't on Twitter and that aren't all these things, the ones that just like check into their job, do their work and then go home, do you think they're happier? Not in like the I know that was a dumb question. Do you think like they're more present with their families? Do you think they're more like, I have this issue where I'm very like entrepreneurial, whatever, ambitious in life and like I go very hard at programming and everything I do and I'm very online and all these things.
Adam:And I worry that it's maybe like, I don't know, taking me away from the people I love. I'm not as present as I could be. I'm sorry, Adam.
Adam Wolff:No. You'll be sorry. I think that's a tough question. I love that question. That's great.
Adam Wolff:My first answer is like, no way. How could they be happier? Because I my guess would be they approach their family and their hobby and everything else the same way they approach their job, which is like they walk through it. Like, I I think Yeah. To to to be engaged, you know, is like, it's also pain.
Adam Wolff:Right? Like, it it's it's quite painful to sort of realize, like, man, I spent a whole day arguing with the LLM. I should not have done that. Or I spent the whole day, like, you know, kinda being mad at my little kid even though they couldn't help it. Mhmm.
Adam Wolff:Those are like the same thing, I think.
Adam:Wow. Okay. You didn't know you were coming to give me therapy, but that was amazing. I actually You actually helped me in my real life. So thank you.
Adam:Yeah. I'm gonna take this one home.
Dax:Okay. The other point I'll make is, I think it's easy to forget the bad parts about a life lived that way. So I think the upside is you can check out. There's a good ability to check out. But the problem is you can't check out a 100% of time.
Dax:There's times of your life that you have to be checked in, you know, for eight hours a day, whatever it is. And most people, when you talk to them about their job, every minute is just misery. It's just my boss is being annoying. I have to do this thing. I don't think it's useful, but I have to do it anyway.
Dax:You know, it's like they they feel like they're working on stuff that has no point. So it's like enduring misery for eight hours so you can escape for the others. And I think most people unfortunately are in situations like that. And I think the thing I easily take for granted is every minute of my day now, I feel like I'm working on, I'm like spending it the best I could possibly be spending it. And parts of it are hard and painful, of course, but there's no chance I would ever get burned out because I'm not doing stuff I don't believe in.
Dax:But, you know, there was a lar like, there was a point in my life where that wasn't the case. And I remember what that was like and I remember, like, always being annoyed and so many cycles in my brain just thinking about how stupid my, like, company was for a certain reason or, like, my team was for a certain reason. And that all faded and I haven't been in that situation for years now. And that that like shows up in other parts of your life, right? Doesn't just stay at work.
Adam:Yeah. What Adam helped me realize is, man, this conversation is going all over the place. Just helped me realize that like, I've always had the work part checked where I'm happy with my work and I'm excited about my work. I can't hardly sleep because I'm excited to wake up and work. I've always said that part.
Adam:It's the interpersonal connectedness relationships that I've not been so good at. Like, I'm a present dad, but I'm not so emotionally available. Anyway, you've made me realize that I can go hard at that too. I just haven't learned how to do it yet. Like, I go hard at everything in life, and I just don't know how to do that part yet.
Adam:So I'm gonna figure that part out too. And like if I were somebody who just checked in, maybe I'm a just kind of a slothful person and I wouldn't do those things to better myself on that area. So anyway, I'm better. Yeah. Thank you.
Adam:You fixed me.
Adam Wolff:That's awesome. That's so real for me. I you know, I'm I'm my kids are older, but I I struggled with that when my kids were younger for sure. It's tough. I've gotten better, though.
Adam Wolff:As you say, you can apply yourself, and even figuring out how to apply yourself to that is one of the greatest challenges.
Dax:Mhmm. Mhmm. Yeah. I'm about to have my first kid, so I'm gonna I'm thinking about some of that stuff.
Adam Wolff:Mhmm. I'm excited for you, Dax. I'm expecting it to humble you a little bit.
Dax:Yes. I will. Everyone literally, there's like hundreds of people hoping for that and
Adam Wolff:I we're rooting for that, Dex.
Dax:Yeah. I've already told everyone. No matter what happens, I'm just gonna be more annoying. You guys you guys need to, yeah, give up hope.
Adam:I gotta say, this is just like for anyone who notices. I'm noticing I've talked to Dax a 150 times on this podcast. Dax's voice is even different, and he's like lounging because his neck is hurt. You may not have wanted to say, like, you're this crippled person right now, but he yeah. I can tell you're like, your voice is different.
Dax:Yeah. Your voice is quavering. It's weird. So this morning, after my run, I literally bent down to pick something up, and I turned my neck at the same time and something just like popped in the back. And it's been really painful and like, it's affecting the way I speak.
Adam:It's Yeah. Pretty crazy. And you're like You're lay you're laying back like this.
Dax:I'm, like, trying to find guys? Positions that are that are not painful.
Adam:Uh-oh. Yeah.
Dax:So I just ruined my Oh,
Adam Wolff:we're rooting for that at all. That sucks.
Dax:Well, I was thinking. I'm like, I'm glad I don't have a kid yet because, like, I have to, like, hold a baby and do all that stuff right now. That'd be pretty hard.
Adam:Oh, jeez. Yeah. So Cloud Code's, like, kind of taking off. I don't know. We haven't really talked about, like, the rise of Cloud Code.
Adam:I feel like Yeah. I I don't know what chart it was I saw, but, like, it it definitely exploding. Especially on Twitter, like, feel like a lot of people is it is it just does it just come down to like, I don't want to use Versus Code? Is that it? Is it just like people that don't want to use an editor baked into their AI thing?
Adam Wolff:I'm, you know, I'm not sure. I think, like, the the growth has been absolutely insane. Like, it's like 300% usage increase in the last four weeks. Wow.
Adam:This kind
Adam Wolff:of stuff that you just don't really see. 90% of the code like, within Anthropic, it's even really taking off. Like, you know, our team, like, almost all of the code is written with the help of the LLM. Like, a couple things. First of all, the new models, the four models
Adam:Mhmm.
Adam Wolff:Are much better. Mhmm. You know? And it's funny because, like, you use them at first, and you're like, this isn't this isn't that much better. It's not like it's so much smarter, but just being a little more steerable and a little more you know, I had this amazing experience with Opus where I kinda gave it an ambiguous prompt, and it came back and it was like, wait a sec.
Adam Wolff:Do you mean a or b? And I was like, oh, I've never seen that from the model before. That's like a new behavior. So I I think the new models really unlock something new. And in some ways, I think quad code is just the easiest way to get at those things.
Adam Wolff:You know, you can you, like, bring your own subscription and you especially if you sign up for the max plan, it sort of feels like all one thing. Whereas with these other things, you have to, like, plug your keys in, and it's just, like, one extra step. Yeah. I I, you know, I I can't quite explain, though, why why Cloud Code has been so popular, especially there are other good options out there for sure. But I think like a lot of it is the model, the new model.
Adam:Do you just default to Opus? This is a question I have for an Anthropic engineer. Do you default to Opus? Ever use Sonnet? Like, is there a reason?
Adam Wolff:No. Actually, I'll tell you. I joined Anthropic in part so I could have unfettered access to the top model. I was like, I do not I'll pay anything, but I'm not getting rate limited. Yeah.
Adam Wolff:So now I get my own inference tier. It's awesome.
Adam:That's awesome. But there's not like I I haven't quite figured out with programming. Dax uses Sonnet. I mostly use Opus. And I don't know if there's material difference.
Adam:Like, don't know if either one of us has got a sense for if we should use one or the other.
Dax:I'm very happy. Like, I'm a very happy user with just Sonnet. I think I will switch to Opus occasionally, but yeah. I think I'm sure Opus is better, but I'm surprised at how satisfied I am just using Sonnet.
Adam Wolff:You know, Sonnet is just as good as at Opus for coding tasks, I would say. And also, when we do we we actually have a horrible problem trying to tell internally if a model if one model is better than another.
Dax:Mhmm.
Adam Wolff:Like, that's actually much harder than it sounds. The users will generally pick the shorter answer that was delivered faster. Right. And especially because we can't really do, like, long trajectories with one model versus the other. It's, like, kinda one tool prompt usually, and it's really hard to tell often what the difference you know, they're very similar responses.
Adam Wolff:So I just stick to Opus. I you know, one flavor of Twitter maxing that I really hate is like, I use this model for planning, that model to write Java, and this model for It's like, dude. You
Dax:should just
Adam Wolff:get a life, dude. Yeah. So I can just keep it pinned on Opus. That's fine with me. But, you know, Sonnet's really good too.
Adam Wolff:They're both better than Sonnet three seven. I you know?
Adam:Yeah. Sure.
Adam Wolff:That guy was a jerk. It was just like break your test and write stupid comments.
Dax:So let let's talk a bit about the form factor of Cloud Code, which is all in the terminal. Yeah. Like, what like, what are your thoughts on that? Is this you see this as like just like a like nice angle you guys found? Or is this like a form factor you kind of are are betting on?
Adam Wolff:I think it's fundamental. And, actually, I think the fact that, like, it's it's in the terminal interactively is almost like a stepping stone to it being a much more composable Linux utility.
Dax:Mhmm.
Adam Wolff:You know, like, the thing I've always wanted is kubectl list pipe clod. What's what's wrong here or describe or whatever? And, actually, it's surprising how we've we've all leaned hard into coding, but, like, it's been harder to crack, like, some of these SRE things. There there's some great new products coming, but it hasn't quite happened yet. Ultimately, I think the fact that you end up typing into the terminal and doing diffs in the terminal is it's kinda neither here nor there.
Adam Wolff:You know? It it it's it's fine. But the real the real power is that it this can run anywhere you have a prompt. And for me, you know, the fact that I can run this inside a Docker container or I can easily like, I run it on a Pi Zero, you know, that I connect to using Mosh. That's where the like, I think that that's what I'm really excited about with these going back to the terminal.
Dax:That that's how we saw it too. We're like, this is just so flexible. There's no excuse someone would have to not use this no matter what their workflow, what their environment, on your computer, on a server, in a GitHub CI thing. I think that's that's the brilliant part about packaging it in this way. The other funny thing I've seen is there's, especially with Cloud Club blowing up as it has, there's been a few, like, big posts where people were like, I don't get it.
Dax:Like, why do people like doing stuff in their terminal? Like, it's a like, how how would you ever prefer that over a UI? And I think if Cloud Code wasn't growing as much as it was, I would like, look at those posts and be like, that's a good point because maybe this isn't as mass market as it is. But in the face of people saying that and it growing so much, there's like a more interesting thing that happens, which is clearly something counterintuitive is happening that people aren't getting, but it's still clicking with a bunch of people. So I love seeing that.
Dax:It's like a good sign that something is really working when people are confused as to why it is working.
Adam Wolff:I you know, a part of me thinks too, like, when I open Versus Code, you know, almost always by accident. Like, I just I just sort of feel like I feel like this thing has gone over some kind of critical threshold. Yeah. Like, like, the fact that we find ourselves in this place where there are no viable IDEs, I wonder if that fuels this a little bit too. Mhmm.
Adam Wolff:Like, it it just everything inside Versus Code to me, it just it turns me into this, like, consumer, you know, which is, the opposite of how I wanna feel when I'm programming the computer.
Adam:Right.
Dax:Yeah. It's in one of these posts someone took a screenshot of, I think it was cursed, they're like, like a terminal could never recreate all this. And I was looking at the screenshot, I'm like, that looks crazy and overwhelming and there's so much stuff going on. Well, one, the terminal could recreate it, but like, I don't think I want to.
Adam Wolff:Yeah,
Dax:absolutely. So I mean, going going more on this format, on this form factor, so you you probably obviously seen this stuff we're doing with open code. We're taking in a like a more extreme direction of like how far can we push all the tricks you can do in the terminal given like the years we've spent doing tricks in the terminal with with terminal.
Adam:Too too many years.
Dax:How did
Adam Wolff:you do that thing when you figure out the theme? I was so impressed. I haven't haven't yet looked at that how you did that.
Dax:How how it like matches your system theme? Yeah. Oh, so it's like a system theme is just your terminal is just remapping as the standard set of colors to other colors.
Adam:Yeah. And so
Dax:as long as we render using the standard colors, it'll it'll show up right. Oh. The
Adam:ands is zero to 15 basically. Most I guess now up to two fifty five you can customize in most terminals. So if you just use those colors,
Dax:yeah.
Adam Wolff:It tends to work. Okay.
Dax:Yeah. So this has been fun for us because we're trying to like see how much we can do. We already know what the the ceiling is. It's like NeoVim exists and it can do quite a lot. But a lot of this is just like kind of inching towards that direction.
Dax:I get the sense with Cloud Code that's not like your interest at all. But yeah, I'd kinda love to hear your thoughts on that.
Adam Wolff:Yeah. Sure. There are couple things that I think are really interesting about the terminal form factor that I do lean into. And then what you guys do is amazing, but it's, like, not really what we're after. Mhmm.
Adam Wolff:So I think the first thing that's interesting about the terminal is the concurrency model. Because on the web, you can have these, like, sort of threads of processes that run completely unsynchronized. You know what I mean? Like, a modal over or or, like, a a window over here has an has an async process, and then the another, like, divided window has its own async process. And those things never have to come together.
Adam Wolff:You know? And and they often don't, and it's confusing as hell. You know? Because, like, they're event based, you know, they they actually they they actually don't really make sense that they're running completely not concurrently. So one thing that's really nice about the terminal is, like, you do all this concurrent stuff, but then you ultimately have to come back to one prompt.
Adam Wolff:So you get this, like, really thick braiding of, like, concurrent tool calls and, you know, all this stuff where you're trying to pack it all in, and then it all has to stop when you take input from the user. So I think you can build you actually end up building a lot more sophisticated concurrency that way because you have these, like, stop points that are easy to reason about. And then there's a related thing that we've noticed, which is, like, just simplifying the UI enough so that we can show it in a few lines of the terminal has forced us to do so many other good things. You know? Like like design constraints free good design Yeah.
Adam Wolff:And the terminal is such a constraint that way. It it that actually really makes me happy.
Adam:Yeah. I lose sleep over vertical space in a terminal. Like, there's just so it feels like there's never enough vertical space. It's so constrained.
Adam Wolff:Yeah. That's right. And also with ink, if you try to rerender too much of it, you get all this flickering and stuff. It's it it's you you you have like like five lines effectively you can update.
Dax:Yeah. It's like a footer that that's that's live and everything else is static. Yeah. I mean, I I also enjoy the constraints. I think there's such basic things like you can't have different font sizes.
Dax:So if you wanna show emphasis or highlight things, you can only do that through colors or bolding. And at first it feels like that this is too limited, but then you realize you can actually do everything you wanna do. You're forced to do it in a better way. And now that makes me feel like, man, I think web apps should all just be one font size.
Adam:Yeah. Honestly, now Yes. So when we did the terminal website, David, our amazing designer at Terminal Yeah. Just said like, let's make it like the terminal, all the fonts the same size. So the entire terminal website, the pages on the terminal website, it's all one font size.
Adam:And it just feels so good to build as a developer where you don't have to think about like, what size is this one? Like, I know it's a muted color, but like how much smaller and like Yeah. You don't think about it at all. You just literally use the same. It's it's the best.
Adam:I wish every website could be designed like that.
Adam Wolff:It looks great too. Like it's actually like the perfect aesthetic for
Adam:right now.
Dax:Yeah. Yeah, so we do go a little bit crazier, even despite that, there's still so many constraints that we're to think about. I mean, we just had a situation this week where Adam worked on a feature and it wasn't it didn't feel exactly right. And you have this moment of feeling like there's no way we can make this work, you know, like it's we're just screwed. But then inevitably, something does emerge and you're like, okay, there's a better way to go about this and something clicks, oh, we're missing this.
Dax:This thing is important and we can restructure it in this way. It always shows up.
Adam:So There's so fewer patterns to draw from, so it's like discovering all the time. Like Mhmm. There's so many web patterns that you just there's an answer to every problem and like you do that thing and it's just boring. With the terminal, it feels so much more uncharted. I know people have done some crazy stuff in terminals, but you run into a situation where there's this UX you need to provide and there's no cookie cutter answer.
Adam:You have to like make it up. It's just so fun.
Dax:I I think also balancing because we are still trying to make this these tools something that a lot of people can use. Obviously, a Neovam user like is all excited about all the different hotkeys and they're gonna like buy a hockey for every little action. But it's very easy to make features hidden by accident because they're under these complex things that most people aren't gonna discover. So yeah, balancing both of that and being like we care about power users but we also care about easy adoption. It's definitely like a very challenging product task to take on but yeah, it's really fun.
Dax:It's like the right level of challenge for us.
Adam Wolff:Yeah. Yeah. I think it's fun too. I mean, the one thing I would say, like, I we found a way to deal with everything except discoverability. Mhmm.
Adam Wolff:That is the thing that just kills us. You know what I mean? Like, see all these projects where it's like, oh, I made neo I made Cloud Code do this. And it's like, no. We we already have that feature.
Dax:Know? Yeah. Sorry.
Adam Wolff:I built something to hit the enter key and all this. It's like, actually, we have a dash p flag. It does exactly.
Dax:Yeah. It's it's funny because someone asked us the other day for a tutorial and we're like, hey, what do you mean you want a tutorial? Like, what like, what does that mean? But then I thought about it more. I'm like, yeah, you know, we have a lot of features and I can see how if you don't read through all the help menus in the docs, like, you're not gonna you're not gonna see that.
Dax:Which I guess the Neovim world, that's how it all works, right? There's all these people that make great content explaining things that aren't obvious. But yeah, as like someone working on a product you hate when stuff is hidden.
Adam:Is it any different than when you worked on React? Like did people build versions of built in hooks and stuff that like you guys have these features already, but people build their big machine on top not even knowing? Because they don't read the docs. Just a lot of people just don't read the docs. It's the same devs that just clock in and they don't love their families.
Adam:Those ones. Yeah.
Adam Wolff:Yeah. That's funny. No. You know, no. React was, like, completely different.
Adam Wolff:Like, actually, the whole the whole game with React was getting people to build stuff on top of React. So and it was open source, you know, and and, like, we really wanted 90 different combo boxes because that's how you end up with the best combo box.
Adam:Yeah. It's the library versus app thing. Cloud Code is an app.
Dax:Yeah.
Adam:Yeah. I was just gonna say, it is like I wanted to bring up OpenCode because I've been obsessed with this whole idea, this whole concept of terminal based coding assistance. It was very awkward to bring it up. Don't know. I'd be curious, Adam, like you've seen we're working on this.
Adam:How do you feel about us working on this? Does it I don't know. I don't know what I'm asking.
Adam Wolff:Guys, I love it. I think it's awesome. You know, we talked about like, when we were shipping Cloud Code for the first time, we talked about making it open source. And, you know, I love open source. I did it for a long time, but I really felt like it was not the right thing for
Dax:Yeah.
Adam Wolff:For Cloud Code because I feel like the first and top voted PR or issue for the open source version of Cloud Code would be like, give us access to other models.
Dax:You know? Yeah.
Adam Wolff:We are never gonna do that. So so the fact, like, having an open version of this with an open router, you know, kind of model selector makes perfect sense to me. And it's just so obvious. Like, this is huge. Like, what
Adam:Yeah.
Adam Wolff:What we're all doing is so big that we all need to work on it. There's, like, plenty of value to capture, and there's plenty of, like, totally slightly different takes that actually totally change the inflection of how you go after this. Even the fact that, like, vibes wise, we're, like, a lot more in line. You know? Like, you can jump in and out of Cloud Code.
Adam Wolff:You guys use alternate screen buffer. And that actually like, that is a completely different feel even though they're both AI terminal agents.
Adam:Yeah. Yeah. Yeah. I I feel like there's not I don't are there other people doing what we're doing with the alt screen thing? Because I feel like there's, eight exact cloud code copies.
Adam:Like, even the big companies that are rolling out their
Dax:thing Awkwardly similar.
Adam:Just like they look so much like cloud code.
Dax:Does the to a point where I'm like, this like, I feel it even has the same like same, like, behaviors where I'm like, is it literally just, ripped from Cloud Code?
Adam Wolff:I'm with you on that. It's sort of like, did you just, like, have, like, write a prompt after Yeah. Have Cloud write a prompt after using it or doing that? Like like, why? Create a pod the same technology decisions.
Adam Wolff:It's crazy.
Dax:It's the exact same libraries, like, It's like it wasn't thought from scratch. So I feel I mean, that just that signals to me, like, they did it because they felt like they had to have a product in this shape, but it's not something they're like passionately pursuing. And I've found the anthropic products in general to be very good. I feel you guys are like a very good product company where I think you guys ship a little slower than some some I mean, Cloud Code was was before everyone else, but I've noticed this with like using Cloud on web. It tends stuff tends to come out like a little more baked and I appreciate that.
Dax:I feel like, like I stopped using ChatGPT because their interface just became a crazy mess of like toggles and selectors and I don't even know what's going on anymore. So I've always liked like, I can can tell there's like a good product sense at the company. Yeah, but some of these other these other tools like, I don't it just feels like someone was forced to do it, you know? It's like the immediate vibe I get from them.
Adam Wolff:Well, thanks thanks for that. Boy, I gotta say it's pretty hard to know how to drive a product org attached to an AI research lab because it always feels like, man, we're not doing enough, you know, to, like, to, like, elicit the frontier capabilities with new products. You know? Like like, they're just there are things right at the frontier that is like, how do you unlock that for product? You know?
Adam Wolff:And, like like, memory is a good example. That's an area where I feel like OpenAI is maybe ahead of Anthropic a little bit in in their product.
Dax:Well, I turned that feature off very quickly.
Adam Wolff:Yeah. I'm it's also a little freaky. I some people are like, I love that it knows me. I'm like, that is so weird, though. It was like you know?
Adam Wolff:But but, yeah, like, I it's just a huge challenge, you know, and it all moves so quickly that, like, some of the things you think even just like the the sort of product development cycles that we're used to have to change because the the underlying technology moves so fast.
Dax:Yeah. The thing you were saying earlier about Cloud Code not being open source, I think a lot of people try to like form a they kind of say some kind of criticism around that. I think people so like our stuff is open source, but we don't really run it like a pure open source thing. We have a very strong opinionated way of how we wanna build this product. And there's a lot of features that are going to be popular, like popular and requested that we're just not going to implement.
Dax:And the only reason we're open source is because we have this like long tail of supporting all these different providers and plugins and stuff like that, that we do need the community support for that. I think people underestimate it's very difficult to once like once like a product, like an app like this, running it like a pure open source project, it's like a lot to commit to and still expect the results to be good. I I think people really underestimate that. So I I think Cloud Code not being open source or not run like an open source thing is that makes total sense to me.
Adam Wolff:Yeah. I you know, we could talk about this a lot, I'm sure, but, like, working on React for as long as I did, you know, I think I think there's a lot of open source projects where it's like, yeah, you can look at the source. But, like, committing to React at this point, like, it's very hard to do. Like, even if you wanted to, it's, like, extremely abstract. And, you know, you you basically need to know someone on the team to get your PR through.
Adam Wolff:Yeah. I, you know, I think I I think there's something to be said for, like, hey. You can look at our source. You know? That that's, like, certain kind of open source that I I would say you guys are mostly interested in.
Adam Wolff:You're right, though. The the great thing about open source is, oh, someone wants to run it on Arch on a, you know, like ARM processor
Dax:or whatever. We we just got the Knicks PR. Someone kind of working as a Knicks package.
Adam Wolff:Knicks. Yeah. Those Knicks people yeah. They they have lots of complaints about how it could work.
Adam:Yeah. But I mean, yeah, to to your guys' credit, you look at OpenAI who they I mean, they open source I guess Gemini's open source too, but they open source the the codex thing, codex CLI, like immediate PR to rewrite it in Rust. Have they even emerged from that? Like, they're still in the weeds rewriting it in Rust, think. I don't think it even I don't know if anyone uses it.
Dax:Yeah. I looked at their usage numbers too. They're like, they're below hours and we launched like three weeks ago. So it's Yeah. I don't think it's really sticking with anyone.
Adam:Exciting times. And I do, I love this space. It's just like, I we're all programmers. Like, a programmer, there's there's something so invigorating about working on something that feels like you're helping shape the future of your field that you've worked for fifteen, twenty years in. Like, it's just such a I don't know.
Adam:It's not like like I'm solving climate change, but it's like it's a tier below that in terms of purpose. You know what I mean? Like, there's like things that would be purpose that like the whole world benefits from. Then there's like, selfishly, my career could change completely over the next five years, and I wanna be part of that. I don't wanna just sit back and watch it happen to me.
Adam:So it's very invigorating. I don't if you guys feel that way.
Adam Wolff:I feel the same way, Adam. And, you know, one thing I think about is, like, we talk about dis distillation from one model to another, but there's this other thing happening, which is, like, the distillation of all of this human knowledge and wisdom into the models and the tools that use them. And, you know, like, for me, transferring my what I know and believe about software engineering and also transferring just some of the love I feel. I feel like, you know, with your product, with with ClawCode, with any good product, what makes it good is, like, it tastes good. You know?
Adam Wolff:It's it's like a cook a dish that you cook. I know you guys both cook. Like, you can just taste the love in it. You know? Yeah.
Adam Wolff:Something that was just made by the guy who's clocking in nine to five is just not not that good. And and those products are the ones that win. Absolutely.
Dax:Yeah. I feel like the timing is also so great. My personal timing, I feel like I finally know enough about my field, how the world works, how things work, and then like something exciting is happening where I've been alive for things like that before, but I was just like too stupid to really work on the right thing. So now I'm really excited because I feel like, okay, everything is finally lining up. I can actually, you know, take part in this in the way that, you know, I've seen other people be able to take part in in previous shifts in the past.
Dax:So, really fun times.
Adam Wolff:I feel like this is kinda I feel like this is my second go around. You know, like, I really thought bringing functional programming to UI was gonna work, and it really did. And now I get to do this AI thing all over again. It's I I feel very blessed.
Adam:Oh, wow. Yeah. You the React thing. Like, you've basically been part of the last shift. It's like there's the broad tech shifts like the Internet and mobile.
Adam:But like within programming, you would say AI changing everything right now. The last thing that really changed everything was probably React. So you've you've kinda like rode both those trains. Right? I mean, like
Adam Wolff:It feels that way. I mean, you know That's a very good developer. AI is way bigger.
Dax:You
Adam Wolff:know? Like this thing is a tidal wave, and we just are just starting to feel the water go out. Yeah.
Dax:Yeah. It's I mean, it's it's I mean, just having done this type of stuff for years, you can immediately feel that because when we put stuff out in this part of the world, the response is insane. Like, we've had success before, but this is like so different. It's like hitting so many nooks and crannies that, we never were able to touch before and it just is reminding me how massive the world is. And when you build something that potentially every single person could use, it just feels totally totally totally different.
Dax:And very fun. We're we're all very overstimulated right now. Think it's Yeah. It's having this funny effect where I personally have been working less on the weekends because I'm, like, so overstimulated during the week that I need to actually intentionally, like, feel like I need to not do anything on Saturday. But yeah, it's it's it's fun and exciting.
Dax:We're all just really amped up.
Adam Wolff:Yeah. It's awesome. It comes through in your product. It really does.
Dax:I appreciate that. Again, thank you again for for joining. Glad you glad you reached out and it was fun. It was cool to hear the other perspective that, you know, we've we've only been experiencing the product so far.
Adam Wolff:It's great talking to you guys. I really enjoyed it. We'll have to get together sometime.
Adam:Yeah. Thanks for coming on. Yeah.
Dax:Well, I will definitely have you on again. Yeah. This is this is cool.
Creators and Guests
