PodRocket - January Mini Panel - Audio Edit === Noel: [00:00:00] Hello and welcome back to Pod Rocket Web Development Podcast, brought to you by Log Rocket. Log. Rocket provides AI first session, replay, and analytics, which surface the UX and technical issues impacting user experiences. Start understanding where your users are struggling by trying it for free@logrocket.com. I'm Noel, and we're back with our panel episode here. ~Uh, ~we'll be talking about how AI is reshaping developer tooling in the web ecosystem,~ uh, from monetization.~ ~To traffic on docs, pages and support and how that's impacting things.~ We're gonna touch on some,~ uh,~ acquisition news and,~ uh,~ what coding even looks like in the next,~ uh, you know, ~six months a year going forward from there. Jack: Will still be coding, Noel: yeah. Right. Someone will be coding. Jack: Somebody may not be me. Noel: Will it be us? ~Um, so, uh, with that, ~with that, let's get into our panel. ~Uh, ~you just heard from Jack Harrington, Jack's the blue collar coder on YouTube. Also co-host of the Front End Fire podcast. How's it going, Jack? Jack: Awesome. Happy to be here. Noel: ~Good. Good. ~We've also got Paige needing House,~ uh,~ co-host of the Front End Fire podcast as well. Paige: ~Hey, ~hey, I am glad to be back. Noel: Nice. And, uh, Paul Mikulski's,~ uh,~ PO Rocket Host YouTuber. A [00:01:00] little bit of this, a little bit of that. How's it going, Paul? Paul: Hey. ~Yeah, sometimes I make YouTube videos. More recently though, focused on building versions, which is this new app we're working on to help artists license music and push it to Spotify.~ Jack: ~cool.~ Paul: ~Yeah. And like share on their royalties and stuff. So that's been like my day in and day out and hopefully we're gonna get it to work and that'll, that'll be sticking around for a while.~ Noel: ~It's an interesting space. There's a new, there's fun new problems there in IP and stuff going on,~ Paul: ~Right. Exactly.~ Noel: ~I'm into it. I'm into it. Very cool. Um, yeah. Well, uh, on that note, let's jump into,~ let's jump into recent news a little bit. ~Um, ~the tailwind stuff. So Tailwinds, CEO,~ uh,~ shared the tailwind had laid off three of its four engineers,~ uh, pretty big.~ Pretty big cut. ~Um, and the kind, the blame there,~ the reasoning was that the,~ uh,~ documentation which they use,~ uh,~ to kind of be the funnel into some of their paid products, I~ don't remember, it's like called plus or something like that, um,~ is like just completely tanked 'cause no one's going and looking at docs anymore. ~Um. ~So now there's this kind of whole conversation ~going around, around like~ going on, around what business models are viable, ~some of these, ~some of these things which made sense before may, not specifically around like documentation, and I think even more broadly around like where people would, ~you know, ~pay to have some nice out of the box like. Tool set and suite. That's like a really good implementation. ~I'll, ~I'll buy this like UI package that makes it easier. And now if I can just prompt that away, it's like I'm probably not doing that as much anymore. So this is kinda all going on. ~Um, ~do you guys think this is like a tailwind specific problem or is there a domain where you think this is hitting particularly hard or is this something that's gonna just ~kind of ~happen across the board? Like for dev tooling in general? Paul: [00:02:00] I like the phrase prompt the pain away. I'm taking that. Thanks Noel. Paige: ~I mean, ~honestly, it. I am, I'm a little bit surprised that it took this long for the open source software maintainers to start feeling the pain or at least sharing the pain, because we've been hearing for a couple of years now, since Google started doing their AI overviews. When you search for anything that all these sites that depended on traffic like Stack overflow. Wire cutter and all these other kind of recommendation or just blog type sites have seen ~their, ~their traffic reduced dramatically. Like their monetization is way down, their page views are down, everything is down. So it almost is surprising now when I look back at it and think about it, that it took this long for people who run documentation sites or run things like Tailwind or other, ~you know.~ Dev tools based sites that ~they're, ~they're just now starting to see or talk about the drop [00:03:00] off. ~Um, ~I don't think that this is a tailwind specific thing. I think that this is a, like ecosystem thing that we really need to be more aware of ~and, ~and think about because open source software, if it can't survive in any Monet in any monetized way,~ there's,~ there's gonna be none of it left. That's a problem. ~You know, ~companies like. Clawed and,~ uh, you know, ~open AI ~and, ~and companies like that, they can buy these things up, but there's only so much money that they can spend on and support. And there's way more open source software than there are companies that can afford to bring them in-house or support the teams fully. So that is, I think this is ~a, ~a problem that we need to start thinking about more seriously. Jack: Yeah, I agree. ~It, ~it, the open source funding has been ~a, ~a problem for a long time. ~Um, ~I'm a SAC Core contributor and it's constantly top of mind. You know, how do we get people who are contributing to Tans stack paid? ~Um, ~and ~it's, ~it's tough because, ~you know, ~the current model that SAC is on is basically out of ~the goodness, uh,~ the marketing,~ uh,~ budgets of all these companies, right? And essentially [00:04:00] outta the goodness of those companies hearts, right? ~Um. ~And ~that's not, ~that's not a sustainable business model. And yet at the same time, like what we want is we want like open source, absolutely neutral libraries. And so we have to figure out a way to make that economically viable. I mean, come on. These companies are, ~you know, ~multi gazillion dollar companies. They can just bank and they can't support the, ~you know, the, ~the web stack that they're running on. ~I mean, ~what the heck? It's crazy. Paul: Do you guys have faith that like other funnels can make up for? Ads on a docs page or like a funnel pushing to a pro product because there is a variety of YouTube personalities out there that rake in a good amount of cash. You can put sponsors on your YouTube videos, and then you immediately have a funnel that's pretty fat if we're talking about ~like ~paying three people, right? ~Um, ~maybe it's just shifting. The bottleneck is shifting to a different spot in the pipeline. It's not going away. Dock site's ~just ~just less popular. I'm not an open source maintainer. ~Uh, ~and you need to pay people. ~Right. ~But [00:05:00] I'm just like, there's gotta be a way that's not like doomsday. Jack: Or to get like, you know, bund, ~you know, ~essentially ~like ~get acquired Paige: Yeah, you get acquired. Jack: or ~you know, ~just now re ~uh, ~recently Astro Getting acquired by CloudFlare. And ~you know, ~both of those groups are basically saying, Hey, ~you know, ~we really wanted to concentrate on the tech and ~you know, ~so we got the payout and now, ~you know.~ That, you know, we're now on them, but is that viable? Is that gonna keep them? ~Is ~is BUN just gonna become basically the Anthropic engine and we don't hear it from it again? I hope, certainly hope not. ~Uh, ~but I don't think there's any guarantee. Paige: Yeah, ~I mean, ~short of acquisition, it seems like there's a lot less ways to keep yourself going as an open source product than there ever, than there has been in the past. 'cause like you said, you're either trying to build a hosting site or you're trying to add deep databases into your platform as part of your offering, or you have a paid version. But if everybody's using AI to. ~Kind of ~[00:06:00] implement your thing and they don't need to be aware, or they're not even aware of those product offerings, then what do you do? And I don't know the answer to that. Like maybe we have to revisit or more strongly push back on ~like. ~Some sort of a monetary incentive or kickback to the original, where the original documentation came from that these ais are leaning on. But I don't know ~how, ~how we enforce that, how we make that, ~you know, ~the ais referenced the Tailwind docs 17,000 times during this one session. So Tailwind gets how much money for that? ~You know, ~it's, I don't know how you'd measure that ~or, ~or. Give them some sort of compensation for the value that they've provided to the Noel: this is Cloudflare's whole thing, right? ~Well, kind of. ~They're trying to do this ~like pay to ~pay to read, pay to scrape model stuff. So there's this like May,~ uh,~ maybe, but it is, it's tough. ~Like, you know, ~like everybody's ~like, ~yeah, that could work. But ~I think, ~I think you mentioned something interesting there, Paige, and that ~like ~there's, there, I think that there [00:07:00] is a distinction,~ uh,~ between like ~knowing, ~knowing about the paid offering and needing it. It seemed like a lot of the discourse here was around,~ like,~ knowing that it was there, but ~I, ~I speculate that it's more of a need thing. It's just ~like ~if people, you just, if you don't really need ~the, ~the templates and the con, the config and the off the shelf thing. If I just like, if you're V one, if you're agent just ~kind of ~gives you one, you're like, eh, it's good enough for my needs for right Jack: yeah, exactly. Noel: ~like, ~like ~the, ~the incentive to go change it later seems ~so much, ~so much lower. Jack: Hey, yeah, I think I bought the Tailwind thing literally just to support Tailwind. Like I think I maybe used the templates like once or twice, and I was like, oh, okay, cool. ~You know, well ~designed templates, but ~like, ~it was primarily because it's like I, I just like tailwind and Noel: Yeah. It's probably a little bit of both. It's hard to know, right? ~Like ~who, Jack: I, unlike Paige, Paige is not like tailwind, Noel: Yeah. Paige: I am not a Tailwind fan. I'm sorry. I don't like having my classes cluttered up with tailwind names and weird stuff. I prefer straight CSS, but even then there are ~So, ~[00:08:00] tailwind I think is in a very hard part of the market where there are so many good component libraries out there, and so many of them are free also. It's just a really hard sell that I'm gonna pay for your nice looking components as opposed to using somebody else Shad cn or somebody else's free, nice looking components also. ~Um, ~but maybe that's just, maybe that's just me and other people feel differently. Paul: It definitely feels Jack: the end ~is, ~is maintained by basically versa. ~So, you know, ~similar Paul: now It is. Yeah. It got eaten up as well. Jack: did. ~I, ~I thought it just started there actually. I thought the guy worked for Versa. Noel: I'm not sure. Jack: Yeah. Paul: Yeah, I, it also feels like we're talking about a few different things all in the same basket. ~Like ~there's, do people know about it, like Noel said, and then there's the just raw traffic on a dock site, raw traffic on an x, on, on platform, on pipe. Those are like two separate things. I'm wondering like. Does one overshadow the other [00:09:00] in importance? Typically, like it sounds like from what Noel said, it's, Hey, this exists, you should know about it. ~Um, ~I don't know if that's like the key problem here, but those are like two different things. And then there's the ~like, ~do people actually want to know about it? Is it actually useful? Is it something that actually has value in the market? And that's like a third thing that just, yeah, there's some things that are getting squeezed out. ~It's, ~it's, they're not useful to that degree anymore. ~Um, ~tailwind might hit all three of these. It sounds ~like, um, ~but I am curious, like from your guys' perspective, do you think it's more of a traffic to docs, traffic to funnel thing? Oh, we can fund the project versus we have another product and we want to tell people about it. And there is income that comes from this outreach and that outreach funnel is now dead because that feels like it's fixable. You can like have agents read docs. They can read source code. My agent reads the source code all the time. It goes into node modules or whatever. ' cause of my cloud. I like it to tell it. So it does get confused. But you can say, Hey, Jack: The Paul: yeah, get in there. ~Like ~if you read this, make sure you mention X, Y, Z. If you sniff [00:10:00] A, B, C, I'm sure there's ways to get, so that's why I'm curious if ~like ~one overshadows the other in importance. Noel: Yeah, I don't think we know yet. I think it's gonna depend on the, on what it is. Like I think ~that ~that's the trick here. And ~that, ~that was kind of what, like why I think that this question is interesting. ~Um, 'cause like I.~ I don't really have numbers or anything in front of me, but ~it, ~it seems like at the other end of the spectrum, if you look at like a similar kind of open source business model where there's like a premium,~ like, you know, ~templates or support ~or like a, it's, it's, it's not, ~it's not so much like we're giving you something you couldn't do on your own, but we're just helping you get there a lot more quickly and ~it, ~it feels like, it's like ~that that is the, ~that is the interesting inflection there. 'cause ~like. You know, like, ~does Red Hat have this problem? ~Like, like, ~you know what I'm saying? ~Like ~if ~you, ~you're looking for corporate support for your Linux,~ uh, you know, ~boxes or whatever distribution,~ like, is, ~is that impacted the same way? I can't, I maybe a little bit. I'm sure you can prompt through some of your problems there, but I'm, I, it's, I'm,~ I,~ I'm skeptical that it'd be, it's the same magnitude as something like, ~you know, ~nice front end templates. Jack: Yeah,~ I,~ Paul: not to think about things. That's what they pay for. Jack: I [00:11:00] think the Im impact of AI at this point though. ~I mean, I talk with Adam, sorry. ~I talked with Adam last year at React com 2024 and or two years ago, and he was like, oh my God, we are doing so great. We have so much money. I don't even know what to do with it all. He was, it was hilarious talking to him. And then ~the, ~the next literal next year. It was that much of a drop off where he is ~like, ~I, ~you know, ~this is unsustainable. ~We're ~we're out. And it's ~like, ~whoa, that's crazy. And ~I'm, ~I'm kind it like I, maybe it's the, yeah, the Opus four five effect or whatever, where it's ~like, ~oh, okay, this is really taking off. It's really good. And it just got wider, much wider adoption, but ~it, ~it was a dramatic decline. And companies have a hard time dealing with that. Within the space of a single year, you lose basically. Your primary revenue source, that's really hard to deal with. Noel: Yeah. Paige: Yeah,~ it,~ it might even be more of an impact for something like Tailwind that is a UI library, because what [00:12:00] used to be. Very difficult. ~You know, ~you needed to have a front end developer who knew how to work with the library, who knew what they were doing. Just in general, building these types of sites, you don't need to have necessarily anymore. You can prompt it as a decent engineer and just say, I need it to do this. ~I need, ~I need this to look nice. You might not even need to even say anything beyond that, and it can come up with a really nice layout just from all the. Ways that AI have been trained ~and, ~and the improvements that they've had up till this point. So even if you don't necessarily have a design eye, if you're a dev and you can say, and you can point towards Tailwind and say, these are the components I want you to use. You know what used to require having a good sense of it, or having a designer or using their prebuilt components and templates and things like that. You don't necessarily even need anymore. And that's probably another thing that's contributing to them having a hard time ~staying, ~staying monetized, ~I guess, ~is because the AI can [00:13:00] do it and do a really decent job, like you said, without paying for anything besides token. Jack: Yeah. Yeah. ~Mm.~ Noel: Yeah. Paige: I don't know what the answer is. Unfortunately, I don't have a solution. I feel bad, but I don't know what the, ~how we can, ~how we can fix it. Noel: yeah. Yeah. ~It's a, ~it's a tough one for sure. Like I think that there's, ~there there're like, yeah, that there's, there's. ~Alternate paths and things for some of 'em that feel more viable, but ~like, ~yeah,~ front,~ front end tools like this, just like ~it seems, it seems, ~it seems so tricky. ~Um, ~I guess from that point, like we, ~we can, ~we can ~kind of ~broaden this a little bit. ~Um, ~Ryan Dahl recently tweeted the era of humans writing code is over. We've been hearing, we've been hearing this for the past year from various,~ uh, you know, ~people involved in the AI business at,~ uh,~ at ~various, ~various stages. ~Um. ~Apple's kind of shifted its,~ uh,~ intent to~ plant, to, uh,~ power apple intelligence,~ um,~ to Google Gemini. I think that originally said it was gonna be,~ uh,~ open AI or, or something. Um, and it, ~kind of ~just raises ~this, ~this bigger question,~ um,~ if AI is, ~you know, ~doing all of this initial ~ground~ groundwork for everyone, like what [00:14:00] part of the ecosystem. Do devs continue to work on, ~like do, ~do you guys think that there's a piece here that is the part that is still like where the agents need direction ~or is it kind of~ Jack: That's exactly it. The ~agents, ~agents need direction, right? They don't, they're not gonna just going, Hey,~ I, I, ~I need a product, I gotta make some money. ~Like, ~let the agent go off and create a, ~you know, ~a granola, ~you know, ~engine or whatever, I don't know who know, whatever you, however you're gonna make money,~ um, you know, you need to provide, like,~ that's what the senior engineering or the engineering job now is basically. Pointing agents and izing the task to where you understand what the agent is doing and what it's outputting and you can test it and that kind of stuff. So that it's just changing kind of the nature of what we do ~and, ~and people have likened it to the whole compiler thing. I'm not sure I like that too much. Where it's ~like, you know, ~the switch from like, ~you know, c plus plus or~ assembler to c plus plus. ~Uh, ~I don't know if I'm there because ~I mean, ~at the end of the day,~ like, you know, it, ~it's much higher leverage code. But yeah,~ it,~ it, we obviously, you still need [00:15:00] engineers. The question is, what worries me is that ~I, ~I don't know if you need as many junior engineers and eventually, ~you know, ~the senior engineers are gonna be like, ah, I'm good. ~You know, ~I got my money. I'm outta here. ~You know,~ Noel: Yeah. Jack: who's gonna take up the reins? Then, ~you know.~ Paige: Yeah, I think that's exactly it. ~Um. ~The way that Ryan phrased it made a lot of sense, and I think it was the caveat that a lot of people were looking for. When we've had other people say, AI is going to take our jobs as engineers, which was it is going to do the grunt work of writing the correct syntax and writing the actual code. We as software engineers still need to be there to check its architecture, to catch the edge cases, to make sure that what it's writing is efficient and repeatable and ~you know, well, well ~architected for scale and thing and security and things like that. So we're kind of doing a lot more of the oversight of ~what it's, ~what the product or I guess the project is doing instead of the actual [00:16:00] implementation of. These divs or this programming language or something like that. And I think that's what was missing was that little bit of nuance because we've all, everybody who's worked with AI as a developer at this point has experienced when it just hallucinates things or it makes up a function that you already know exists in the code base. 'cause it doesn't, it's not aware of it. And you're like, no, you don't need to rewrite that. Go use the one that already exists. It's over here. Jack: Yeah. Paige: So ~it's, ~it's that kind of thing. And,~ but,~ but at the same time, ~you know, ~the past probably four to six months as I've been working re really regularly with Claude Code in my own day to day. Like I am not writing code any longer. I am watching Claude write it. I am checking the code base. I am trying the app out and making sure that it works the way that it's supposed to. But Claude is the one who is writing the actual syntax and adding the new functions and things like that. I'm just watching it every step of the way to make sure it stays on track and on [00:17:00] target. ~So. ~I don't think that he's wrong as much as some people might want it to go back to more of a craftsman style of, ~you know, ~artisans writing code by hand. I just don't think that's the, I don't think we're ever gonna be able to put the genie back in the bottle in that regard. Jack: I'm not sure I want to, to be honest, like I'm so much more productive with ai. In fact, I've been, you know, just had a, like a side channel with Tanner Linley talking about ~like, you know, ~how much,~ like,~ you just think it, it's like, oh, you know, I, he. Recently wanted to put tans stack.com builder up, which is this, like you can create a Tans stack site or a, an app on this, in this builder app and just kinda ~like ~click on it and do all that. And so it was sitting on top of ~the, ~the stuff that Ira had written by hand, ~you know, ~late, early last year. And. ~I mean, ~he just took it to the next level, ~like ~so easily with ai it was like, and adding features left and right, and I was like, whoa. And yeah,~ I,~ I, yeah,~ I,~ I'm experiencing the same thing, which is like, [00:18:00] you think it, and you can have it as long as you can express it well. Noel: ~So, yeah, I, I feel in the past we've always like dev tooling and has always been things that help us with this. The former, like writing lines of code, right? Like for the most part it's like, how, how can we, how can we improve the ergonomics of that and make it easier to do? I mean, not exclusively, but that's been a lot of, a lot of the focus.~ ~So what, what like changes now do you guys, do you guys see like big advancements coming in, this kind of new. Flow of, you know, kind of, um, prompt engineering agent babysitting. Like, what does that look like? What do you guys, do you think, are there things that you've been pining for? Is there, is there something you feel is lacking?~ Paul: ~So I have a flow of TMS that is pretty manual with agents that I think we could start to benefit from cross pollination of the standard in standard out and what's going on shoulder in context and isolates versus kind of a subagent vibe. Um, I also wanna like, uh, I forget who said it, but something about junior roles.~ ~I don't know if it was you, Noel, or, uh, I, I~ Jack: ~It was me. Yeah. I talked~ Paul: ~it was, it was you, Jack. Okay. Um, I don't, I, I kind of wanna push back on that a little bit. I~ Jack: ~Go for it.~ Paul: ~junior roles are more useful than ever. Right now. What's. Not as useful. Our software as junior roles, like that's going away. But like on my smaller team that's building like the music app as an example, um, I would much rather have somebody that doesn't have, I'll put in air quotes here, a Lobotomized software brain.~ ~Instead they have a master's from some school in music business administration. And I would consider them technical because given the tool they have, the agency and the self propulsion. To like seize the tool and now I have a brain, an emotion on the team that is just like so far beyond what any software engineer could do.~ ~Yes, they're junior. Yes, they just graduated, but they are like so much higher value and output for the team. And so it's just like the definition of what is technical is changing. The definition of who is technical is somebody who can break down work and like orchestrate that work and, and, and work on a human team at that like human level.~ ~That to me is technical and like. Junior technical people, bring 'em on. Bring 'em on. I just don't want software junior people. And so maybe that's what we're gonna start to see snuff away. It's just there's a weird cold start problem. 'cause like Paige, you pointed out, we're still in the soup of like. There still needs to be a senior engineer on the team there.~ ~Just there, there's gotta be. Um, but if you're scared to, like, if you have a team and you're scared to like, think that way and you're like, well, I, I give somebody that's non-software or only business this and they mess everything up, I would push back and say, that's your skill issue. Like you have a skill issue of like making the cloud code, tooling the cloud code hooks the controls work for where your team's at.~ ~That's like the challenging thing right now. That's to become the new senior engineering. Medium that we're trying to hone in on. But yeah, I love junior people because they have the non lobotomized pure brains of whatever their subject matter expertise was, uh, that they studied, and it's really valuable.~ Jack: ~I think that's absolutely fascinating. I mean, you're spot on and that's really exciting to think about. Right? You know. Obviously we wanna bring in all these non-technical people so they can help us build really cool apps that work in their space, right? But how do we do that in a way where we don't get, like vulnerability as a service, right?~ Paul: ~Right.~ Jack: ~I want, you know, it's like, it's like it in a way. You kind of want like a, a person who's guiding the, you know, who has that subject matter expertise, who's guiding the design, and then that's. Kind of connecting to an LLM that's also got a human in the loop. So the human's like, Ooh, I see where you're going with that.~ ~We wanna make, we gotta make sure it's secure. You know, that kind of thing. And that,~ Paul: ~that's a great example of what Noel just said, like what's like one of those things we'd reach for, like what you just said, Jack. That's exactly it. Like that feedback. Loop,~ Jack: ~Yeah. And I don't think we have that yet.~ Paul: ~Right, right. Yeah. Yeah. We're, we're working like, you have to build it manually. You have to be like, Hey, my Cloud md if, if this developer with this go GitHub handle, he's like the music guy.~ ~If this developer with this gov handle, he's like the front end guy, new to backend, like. Ask them these questions, ask them that questions. It's so manual right now. Um, but yeah, that's how I, that's how I do it. I'm like, every single person has their own markdown file, their history of like what they learned, what school they went to.~ ~And so like, Claude can prompt them with much more veracity and like for them. Um, but it's manual. It's manual. I want it to be automatic.~ Paige: ~That's interesting, but that the way that you're describing it, Paul, almost makes me think that maybe having a computer science degree will be less important than it might have been in the past, because you don't necessarily need to understand the underlying algorithms as much, or you don't need to have memorized the syntax.~ ~Maybe if you have a degree in something that is related to, like you say, the product that you're building or the industry that you're going into, that provides you with more benefit as a, a new, new C, early career person. Then necessarily being able to write a great for loop or. Traverse a binary tree, or you know, do something that has typically been almost like the only way entry level into coding jobs.~ Paul: ~and, and, and I guess in that way, Paige, I get the compiler thing, Jack, I feel you. I feel you though, like the whole, like we are at the new level, not, no, we're not, but in these niche ways of like, I'm never writing bubble sword again. And if I need to, I'm gonna go like, what? Okay. Why A, B, or C? Gimme the options.~ ~Tell me why. Been there, done that. We don't need to do it again. Compiled. LLM. But other than, yeah, I think that's the only one smidgen where that makes sense. That was quite the face jack. I was, I was, I was~ Jack: ~Oh, sorry. No, that's literally my nose is going off, but, um, no, no, no, no. I, I totally get it. I, I'm just thinking like when we talked about like. The OSS and the licensing stuff we'd mentioned, like, how does Red Hat do it? And it's like, well, red Hat's not in any danger of people going like, yeah, I'm gonna go and like LLM prompt myself a red hat.~ Noel: ~Right.~ Paige: ~I'm done with implementation.~ Jack: ~Yeah. You know, but they are gonna go and LLM prompt themselves another CSSF. You know, potentially, I, I've literal, literally I've had the experience where I asked, uh, an LLM to go and use, I can't remember. It is some new JS framework that came out like a couple months ago. And I said, oh, I wanna try this out.~ ~You know, go, go find out. And, and it literally just, it totally BSed me. It created its own framework on the fly. Then use that instead. Ripple. That's what it was. It was one, it created no Ripple files. I'm like, this can't be right, by the way. Um, but yeah. And like, okay, cool, but where's the level at which this stops where? We, we, we say, okay, well, from here down, like we actually need, like, like full on CS degrees. Like these are the, you know, like if you're gonna be doing, I don't know, embedded work or, you know, clearly os work. That's, that's, I don't, you know, or, and or maybe like our oss now get really amazing because now we have AI we can throw at them and start really doing next level stuff that we weren't doing before.~ Noel: ~Yeah, I mean, I think, I think it's an interesting question. It's like at ~at what point. At what point do we feel comfortable getting ~on a, ~on a train or into a car or whatever that's ~like, you~ Jack: ~Oh,~ Noel: ~some someone's written the, like, you know, the, the software that's on the boards that's controlling these things.~ ~That was, it was like, you know, like literally like in hardware before, but now there's like some abstraction. Someone's prompted it and done the thing like at like, maybe we get there at some point, but I'm not there right now. Like, no way in hell, like.~ Jack: ~where the, the flight control system is, is vibe goated. Nah. That, that's not happening. Right? Yeah.~ Noel: ~um, and you know, like I, it just, it just feels to me like I still, and like~ I was ~having, ~having another interesting discussion. We, we were talking about like errors and hallucination rates and ~all this, ~all this stuff before, It's like ~at at what point.~ At what point do we feel comfortable getting ~on a, ~on a train or into a car or whatever that's like, you Jack: Oh, Noel: some someone's written ~the, like, you know, the,~ the software that's on the boards that's controlling these things. That was, ~it was like, you know, like~ literally like in hardware before, but now there's like some abstraction. Someone's prompted it and done the thing ~like at~ like, maybe we get there at some point, but I'm not there right now. Like, no way in hell, like. Jack: where the, the flight control system is, is vibe goated. Nah. That, that's not happening. Right? Yeah. Noel: ~And we kind of like, we kind of found ourselves in this. Um, I was having a conversation with a coworker and~ we brought up this thing that had always been the argument against like self-driving cars forever. It's like, well. Humans make mistakes too, but we just expect it to be the norm when a human does it. So it's ~like, ~maybe we get there eventually, like with Vibe coding, where it's ~like, ~it's good enough most of the time that we just say ~like, ~it's fine. But ~like, I, ~I had to bring up the point. It's like, yeah, but ~like ~there are lots of [00:19:00] places where the like code bait always, like there's, it's, there's proofs. Like we've,~ it's,~ it's solid. ~Like there's, ~there's no. Yeah. There, there's no way it can go wrong. Exactly. Jack: Yeah, you put in X, you always get out Y Noel: ~yeah,~ Jack: Whereas with an ai, it's like you put in X, eh, you know, you kinda get a little something like this one time, something like that. The other time, Noel: So ~the, ~the shape kind of keeps changing. It'll be interesting to see, I think like my outlook on in these like different modes ~kind of ~keeps. ~You know, ~waxing and waning as to like ~where, ~where we'll end up. Like I'm always, I, ~you know, ~I think Opus four five's really good and then ~like, ~just like I have some stupid thing where I'm like, Hey, tell me if there are any breaking changes between these two versions of the api. And it's like, no. And then I go look, and it's like, yeah, there are like, okay. ~Uh, like,~ Paige: I didn't break those unit tests that are now failing, Jack: Yeah. Paige: failing before. Like I'm Jack: ~Well, ~that's what keeps you up at night, right? Is that, I don't know. 'cause I don't have full control over the code really? Come on anymore. Paul: that happens, like my [00:20:00] takeaway, I'm trying to reframe it as ~like ~that's my skill issue, where like I didn't set it up correctly that I allowed it for that to happen. It's not like Claude, it's me. like, yeah. So it's like maybe I should throw a hook in there or a script or like the cloud code hooks great validation tool and you can run some crazy stuff like there. There's this one set of hooks I have where Claude runs a subagent and when the subagent completes, it runs Codex. 'cause Codex is a bit more fine grain, but it's not good at planning. Then Codex reviews it, then hands it back to the main agent and then it hands it back to the subagent to complete. So this sort of like. I never had the issue ever again. So every time this happens, I'm just like, it's not Claude, it's me. I need to learn how to use this better and it'll be better next time. It just ~kind of ~compounds over time. Noel: yeah. I think you're right. Like you can ~kind of, ~you can ~kind of build these, ~build these pipelines. But I think this then brings us back to like, what, what does tooling look like? Like back to the original question, because ~then it, ~then it feels like dev tooling is making that thing you just described. Paul easier. ~ how, how do I, ~how do I make it so I can tell agents what to do and ~like, ~[00:21:00] get, figure out how I can make them catch their own mistakes and stuff more easily? Because now it's clunky, it's still not super Paul: Yeah, totally. Noel: ~Um.~ Jack: I don't even think we know what the problems are. ~Right. ~We're still coming up with the solutions, like we just like wig loops have started becoming a thing and yeah, because people ran into an issue where the LLM just ~kind of. ~Stops like just halfway through. You're like, okay, I'm just done for the moment. You're like, no, you need to finish this. ~You know, like, ~oh, okay, ~well ~now he's got this thing, this wig, basically just kick it in the butt until it gets it done. Noel: Yeah, but ~it does, it does, ~it does ~kind of ~lead to this interesting question then,~ or,~ or like ~what, what, ~what do the, what are the frameworks of the future look like? ~Do we stop, ~do we stop worrying about like frameworks as this abstraction that, like how we think of them right now? Front end frameworks or, ~you know, ~like your web server frameworks or even, ~you know,~ Paige: Or even coding, like coding languages. It's ~like, ~if. If we're using agents to do it, do I really care what coding language it's written in? No. Whatever's the most efficient or is gonna solve my problem the best ~if, ~if I don't need to know the syntax. Jack: I ~kind of ~disagree with that because it's [00:22:00] like at the end of the day, a really good line in the pragmatic programmer or just. Idea that like never ship code, you don't understand, right? If you get paged at ~like ~two o'clock in the morning and you're just, ~you know, what, ~what is this code? Why, you know, trying to figure out what, why this code's blowing up. Like you, you don't want it written in, ~you know, ~cold wall 62 or whatever it is. Zag, ~you know? ~Yeah. Paul: I, ~I, uh,~ do think frameworks have immense value, though? 'cause they sell guarantees to us. ~It's, ~it's in the compiled vibe. ~You know, ~this is a compiled piece of human thought. We're guaranteed that it does X, y, z in these ways. ~Um, ~so those have value. ~I, ~I have trouble foreseeing how those don't have value just because it opens up more fray on the, on, on the end of the development planning. ~Um, ~in response to like, what is development, like that skill Noel that you're getting at, what are those frameworks looks like? ~I mean, ~we had one come out at the end of December that I've been finding really interesting, which is Gastown and I forget. Yeah. Gastown. It's the guy who made the beads orchestration,~ the,~ the [00:23:00] beads. ~Uh, ~project, I forget his name. ~Um, ~I should know his name. He's a pretty famous dev, but ~it, ~it, you allow, you define the tasks, it sets agents off. They have review loops. They have ~like ~a mission control like. It looks clunky. I agree with you, Noel. I've, I haven't used it to do any serious work, but it's that first flavor of ~like, ~here's a highly opinionated way that you're gonna have a team of like 20 people. And it's done by somebody who has been at Amazon, they've been at Facebook, they've done the rounds,~ uh,~ they've managed teams. So it's like they're really serious about this. It's not like some YouTuber who threw together some like agent network thing that has wasted my weekend away once again. ~Um, ~so these things are starting to get a little more. Formalize and it's interesting to see how people ~are, ~are approaching them. Gastown, who knows where it'll be in a couple months. ~Uh, ~but it's definitely gaining some gas Jack: Okay, wait a second. So ~I, ~I'm looking at, ~you know, ~there's a medium article. Welcome to Gastown. And is this the one,~ uh,~ what the heck is Gastown? ~Uh, ~Gastown helps you with the TD of running lots of Claude code instances. Stuff gets lost. It's hard to track what you're doing. Gastown helps you with all that yak, shaving [00:24:00] and lets you focus on what your Claude codes are doing. Y got, man, that yak shaving. I was getting bummed by all that. Yak, shaving. I gotta say Paul: like four is my limit. Like I Jack: I, I, Paul: at four, but Jack: I'm like Paul: need something. Jack: Yeah. Yeah. Paige: I don't know. I'm maybe I'm too much of a control freak, but I still like one. I'm just gonna watch Claude as Claude does it because so often and so quickly it gets off topic or. Goes strays from what it was originally supposed to be doing. So I don't feel like I can leave it alone at all, but maybe that's just me not relinquishing control. Paul: Yeah, go strangle it Paige, and tell it. Tell it you're bad, Claude. Paige: No, I don't like punishing Claude. That's the other thing. It's like I don't try and threaten Claude, like some people have talked about. They get better results with that. ~I just, ~I just tell it it's wrong ~and, ~and then redirect it. Jack: ~Well, ~apparently then you're at stage five according to this medium article in your journey,~ uh,~ where stage [00:25:00] five is you use the, you use ACL I, I guess I Paige: Yes. Jack: a single agent, yolo,~ uh,~ and then disc scroll by it, which you may or may not look at. Stage six is progressing onto CLI, multi-agent yolo, and you pretend you know you regularly use three to five parallels. Then stage seven is 10 plus agents, and stage eight is building your own orchestrator. Okay. Noel: ~The article, it kind of, ~the article ~kind of ~reads like a meme. It's hard to tell. Paul: Yeah, that's Noel: ~the, ~where the membrane is. Paige: Yeah. You've achieved mastery when you're at stage seven or eight. Noel: ~Mm-hmm.~ Jack: ~still, I'm still, I get, I,~ I don't understand why, ~you know, ~like the CLI is thing is inherently better. ~I, ~I ~kind of ~jump back and forth between Claude and CLI in versus in the IDE. It's mostly where I am ~with my, ~with my monthly quote. it's like, because, I mean, I like it in the IDE, I like to see the diffs in the IDEI. I like the experience better in the IDE. I don't understand why ~the, ~the pixels being reordered in a certain way is like just inherently better ~if it, ~if it's in the CLI like, like, okay. But ~I guess, ~I guess that means I'm just stuck [00:26:00] in stage four. Paige: ~it's been, ~it's been a few months since I tried the Claude. I-D-E-N-V-S code and maybe it's improved since then, but originally, and I think it was pretty early days when it had come out that it just didn't feel like it was as accurate or that the code suggestions it had were as good and it also ~kind of ~crashed my ID sometimes. So that's kind of why I reverted back to the CLI end. I've just gotten so used to the workflow where the CLI is open in a terminal and I've got VS code open. So I'm watching the changes as they're getting added and then I'm just, ~you know, ~running it locally so that I can ~check, ~check and actually see if the, ~you know, ~if it's doing what it's supposed to do. But I've just ~kind of ~gotten used to it. So it's almost like it's not, it might not necessarily be better at this point or worse, it's just ~kind of ~the flow that I've gotten comfortable with. So I'm happy. I'm happy with how it works. Jack: doing cursor and just recently, I guess what, probably a month ago, they added like a really [00:27:00] good browser mode Paul: ~Hmm.~ Jack: ~you know, ~so nowadays it's a, it's, I can just ~like ~say, Hey, I. Go make, ~you know, ~fix this, run it until the page works. I'm gonna go get a coffee and ~you know, ~I go off, I get a coffee, I look at my 3D printer or whatever, yada, yada. Come back. It's done. ~You know, ~and it's, it actually, you could, having you, you will have seen it automate the whole thing, click around the whole deal, and then I, ~you know, ~talk to these CLI folks and they're like, oh man, I got my MCP playwright stuff going. I'm like. That was just an integrated feature. Man, I just got that. Like I, is there something special about it? ~Like,~ Paul: Claude caught up a little bit. ~They, ~they got Claude and Chrome,~ uh,~ now, which is a plugin. So you just install the plugin and Jack: oh, that's Paul: kinda wor it works. So they, Jack, great point. I think they heard that point and they're trying to catch up with it, but it's still not at the cursor level. The cursor browser experience is just unrivaled. Can't beat it right now. Also have you guys ~um, ~checked out? Happy Coder. That's another tool ~I, ~I definitely have enjoying, enjoyed telling people about. ~Um, it, ~it's a wraparound Claude Codes. You just write Happy instead of [00:28:00] Claude Jack: Okay. Paul: up a tunnel Jack: Code, mobile Cloud. Paul: and it's on your phone. Yeah. So I'm in the steam room,~ the,~ the YMCA and it has a voice assistant power through 11 labs. So you can just say ~like, ~Hey, gimme a rundown. It'll talk to you, wait for you to respond. No buttons, no pressing, 'cause you can't press buttons in a steam room. Jack: Okay. Yeah. This is dope. ~This, ~this is, yeah. Okay. ~I, ~I have this like fantasy of ~like the, ~the movie theater experience where it's ~like, ~I'm just like, ~you know, ~at the movie theater, I'm watching, ~you know, ~some movie and I have ~like ~a brainstorm and then I'm like, I'm all just gonna step out for a second, ~you know, ~go over and ~like, you know, ~kinda walkie talkie to my, ~you know.~ Paul: Oh yeah. Jack: Quad instance and be like, Hey, can you go do this? And by the time I get home, it's ~like ~done. Paul: Happy coders that you, or at least the way they're building it right Jack: ~Yes, yes, yes, ~yes. Paul: ~Like, ~I have a home directory,~ um,~ that I use with Claude Cowork that connects to my Gmail, my Todoist, my like granola recording. So I can also ~like ~be on a walk and just say ~like, ~Hey, what's like this dev and that dev doing what happened in standup this morning, create some Todoist items and it's in the central, like kernel of a repo that has a bunch of servers and I can just talk to it as I'm walking, getting coffee or whatever. Jack: Oh God. [00:29:00] You've got amazing setup, man. Geez. Paul: ~it's, ~it's great. Jack: Yeah. Okay. Noel: Nice. Nice. Okay, cool. ~Well, ~I did mention in our intro that we would discuss acquisition stuff briefly, ~so we've gotta,~ we've gotta ~at least, uh,~ at least touch on Astro CloudFlare. ~Um. ~Basically, ~as~ Astro announced is joining CloudFlare will remain open source. ~Kind of the, the typical, ~the typical spiel here. ~Um, you know, ~CloudFlare said the justification is that they get a lot of use from it, et cetera, et cetera. ~Um, ~there's a new version of Astro. I don't use astro day to day. I think it's six, something like Jack: Six now. Yeah. Noel: ~Um, ~ Jack: Live collections I think is the big, Noel: Yeah. I mean, and,~ I mean, and, and there is like, ~Astor does feel like very ~kind of ~in line ~with the, ~with the cloud flare worker, runtime edge thing that, that's busy is their whole kind of moat and shtick. ~Um, ~do you guys think that this is like ~a, ~a play that makes sense? Is this,~ uh,~ this ~all, ~all happy vibes all around? Jack: I think this is a play that makes sense for clout. ~Uh, ~it makes sense for both, to be honest. Like Fred wants to do work and Fred's the guy behind the astro. Yeah,~ they,~ they want to just do technical work. ~They're, ~they're geeks at heart ~and, ~and ~you know, ~this is whenever they try to get into the business world to make a thing go, it was a lot more, [00:30:00] basically, more hassle than it was worth. And so that, ~you know, ~this works out for them. They're, you know, they let CloudFlare handle that part and I'm sure they got a nice payout. The, on the CloudFlare side, ~you know, they.~ I think this is more than just we love these guys. ~It is, you know, it, ~it plays into a full like virtuous cycle strategy for them where they can go and have essentially like a next Js s Versal style offering where it's just all sort of one ecosystem. And they also actually picked up on an AI content play at the same time as this, which, ~you know, ~if you think about that, now you're starting to think about ~like ~a full on AI content. Deploy AI content. Noel: Yeah. ~Yeah, yeah, ~yeah. ~Mm-hmm.~ Jack: of deal. Noel: Yeah. Paige: Yeah, I think. Noel: go ahead. Paige: ~Well, kind of ~echoing what Jack said, I think that ~this is, ~this is really good for CloudFlare and for Astro. ~I mean, ~Astro, we use it for our own front end fire website because it's great for static sites where you don't need a whole lot of functionality. You just need a nice looking, easy to set up fast website, and that is. Still a [00:31:00] lot of the internet to be completely honest. ~Um, ~and Astro is easy to, it's easy to learn, it's easy to get going with. ~Um, ~it's very easy to use. So I think that this is a great thing if CloudFlare can say, Hey, I. You could try this framework that we support, it's open source. It's not gonna go away because we're supporting it. So you're not, you don't need to worry about it just disappearing or having funding issues in the future. It plays well with our network or it plays well with other ones. It's easy to take wherever and host. ~Um, ~and yeah, it's, it seems to be right now the only kind of guarantee you can get as an open source creator of software. To be around in a few years time with all the changes that we've just been talking about. BUN got acquired by Anthropic. CloudFlare has acquired Astro, ~you know, ~Al has bought Next Js and a bunch of other stuff that I can't even keep track of all the stuff they've acquired,~ but,~ but it also ~is a, ~is a better guarantee that they'll be around in a few years time and [00:32:00] that they're a good bet. ~So, ~yeah, I think ~it's, ~it's unfortunate that this seems to really be one of the only ways to. ~Kind of ~guarantee your success for ~a ~a while, but at the same time, it's great for ~the, ~the people who are, who can stick it out and get to that point of being really beloved or really in or very central to some bigger product that does generate revenue. ~Uh, ~and then just get bought up by them. Noel: Yeah, I agree. Jack: Yeah, I love how Astro is. It's really great in that it ~kind of ~does this awesome scaling curve where it's ~like, ~ooh, you could start off with something. You can start with HT ML pages. You could start off with, ~you know, ~markdown, like however you wanna play at the low end content level. And then, oh, cool. All right, we wanna start building out our app. You can literally just start ~like, ~oh,~ I,~ I could bring, felt, I could bring in React, I could bring in Preact, I could bring in tens, act start, like I, ~you know, it's, ~it's really, it's got a great beginning curve there. ~Um, ~whereas I think. Next Js it tends to ~like ~not be as good [00:33:00] for those kind of low end content sites, but be, ~you know, ~more amenable to ~like ~the kind of complex apps of the top end. So yeah, it's an interesting, ~you know, ~position positioned framework, but ~it, ~it is unique in that space. Noel: Yeah, I agree. I agree. ~Um. It did, ~it did feel like this one had a little bit less, there was a little bit less ~kind of, ~of that inherent,~ uh, you know, ~versal conflict of interest animosity here that I think we see a lot of the time with. ~I guess. ~I guess what, as we've seen with some of these acquisitions,~ like,~ it seemed like the environment, the community was a little bit more kinda this, ~you know, like, ~ah, this makes sense. Like ~no one was, no one was, ~no one was,~ uh,~ pushing back quite as hard. Jack: I think the community likes CloudFlare more than they like Elle. Paul: ~Or, ~or so we think Noel: yeah. Currently. Paul: seen Paige: has done a better job of positioning itself as a good steward of the web and an open steward of the web than other companies. Jack: yeah. Noel: Yeah. Paul: entertain the,~ uh,~ age old Reddit meme. I've seen time to time of everybody likes CloudFlare because the NSA made it so.[00:34:00] Jack: Okay. Noel: Mm.~ Mm.~ Paul: Yeah. Noel: big vacuum. Cool. ~Um, ~I guess on that we can,~ uh,~ we can take a quick break and go to hot takes. ~We'll have to be fast though. Is everyone ready? With a truly primed, uh, hot take? We can rattle through quickly. Okay. You can think of one here while we do a quick, uh, ad break. Um, ~we'll be right back. This episode is brought to you by Log Rocket. Log. Rocket provides AI first session, replay, and analytics, which surface the UX and technical issues impacting user experiences. Start understanding where your users are struggling by trying it free@logrocket.com. Okay, you guys have had a minute. Does everyone have their hot take? I'm just gonna go around the circle. I've got here. Jack, what's your hot take? Jack: ~Uh, ~I'm gonna go non-technical ~as, ~as is my, my,~ uh,~ want here. ~Um, ~I'm gonna say Borderlands is a good movie. That's my hot take. Noel: I like it. ~the, the mental energy to, uh,~ Jack: ~To fight that one.~ Noel: ~Yeah, yeah,~ Jack: ~no problem.~ Noel: ~I'm on board. Sure.~ Jack: ~reason. No reason to fight it. It's~ Noel: ~Yeah.~ Jack: It's a fun movie. What do you want? Do you wanna be entertained? Be entertained. Noel: Nice. That's funny. No. On Paige, what's your hot take? Paige: ~Uh, ~my hot take is that light mode is gonna make a comeback this year. I think that there are so many SaaS companies and AI companies right now who have dark mode websites to all look the same, that I think light mode. Is gonna help you stand out in a [00:35:00] way that it hasn't in the near recent past. So ~let's, ~let's see if that happens. Noel: I've been on this team forever. I think part of it was just like, ~you know, I like, ~I like to be my millennial special snowflake and everyone was like doing dark. I'm like, no, I'm gonna have, I'm gonna have my weird custom off beige, ~you know, ~fungi color scheme. It's my thing ~and I, ~and I'm, I still maintain, it's better for our eyes. Everybody thinks it's not, but if you're in a consistently lit environment, you want ~like, you know, ~like a, ~not ~not pure, full white, but like a kind of off white, a little bit darkened with ~like ~a dark gray black text and ~like ~that's avoid, avoids Jack: aren't we supposed to just give you both? ~You know, ~you can just toggle whatever way you Noel: There is accessibility stuff Jack: as a can, as a content creator. I can tell you, please make a dark mode because, oh my God, like I'll show, ~you know, ~documentation site and it's only in light mode. ~Go, ~Go, go. Oh my God. Oh my God. Noel: The flash bang is real.~ Yeah. Yeah. Mm-hmm. Yep. Okay. Paul,~ Paul: ~Uh,~ definitely some of my favorite websites I've used recently, Paige, I've been in light mode. I've just been like, it's so tasty. It's a very tasty website. ~Like, ~I don't know why, but it's [00:36:00] delicious. ~Um,~ Noel: Nice. Paul: my hot take would be, I think it's very similar to something I said earlier, which is ~I've, ~I've worked with ~like ~a bunch of small teams in the past few months. ~Uh. ~Finding my way in the world and being non full-time W2 and ~you know, ~just doing this and doing that, working with this team, getting a little product out here or there. And it's just like a lot of times people wanna set up Claude code and then they get to this frustration wall pretty quickly and it's like, oh, okay, ~well ~we're only gonna let it do these things. And like my hot take is. You probably have, and it's not like against you as an individual. It's just ~like, ~I challenge to rethink of it as like a skill issue. How I'm like, oh, it's me. I think nine times outta 10. It's always that. It's never that Claude can't do it or do it correctly and people just, I don't know. ~It, it's, ~it's always like pushing ~a ~a, a rock up a hill sort of feeling where it's just ~like, ~no, it really is ~you, ~you're it. You know that song from Red and link the breakup song. It's not you, it's me. It's you. Okay, that's the end of my hot take. Jack: No,~ I,~ I completely agree. Like the, I've been doing more with Claude than just [00:37:00] coding, and it's been amazing. Like, I had a project that involved soldering just recently, and I hate soldering. It's like this, Ugh, God, I hate soldering so bad. And I was like, you know what? It's a skills issue. I'm just gonna have Claude,~ like,~ walk me through soldering. Like I was a 5-year-old with a hot iron and. I now know how to solder. 'cause like I, you know, it took ~like ~30, 45 minutes of just ~like, ~practicing and now it's ~like ~done. I like,~ I,~ I know kung fu and it's like my world just opens up.~ It's like, oh, I, I need to know, learn how to do blah, blah, blah, blah. Just, ~just head up. Go on man. Sum, total your knowledge like right at your fingertips. Go. Noel: No. No. Yeah, ~I feel like, I feel like you're, it's kinda like that. It's kinda like the old, it's like any, any like. ~Any project, there's ~kind of ~that, like, how much prep do I wanna do here? And I think we tend to fall too far on the side of ~like, ~I don't want to have to do that much prep. 'cause that's how engineering has always been. It's just ~like, ~eh, just ~like ~jump in and, but now I feel like we're in a weird world where it's ~like, you know, ~it's like painting a room or something. Or it's like if you spend three quarters of the time doing the prep work, you'll get done faster because it'll go so smoothly once you're configured and set up correctly. Yeah. Paige: that is, yes. Making [00:38:00] plans with Claude before executing has been like a game changer in terms of the ~end, ~end product for me. Noel: Totally. Jack: Yep. Absolutely. Have it. Have it go through and ask you questions. In fact, actually there was a one called ~like ~Next Flow, which is an addition to Claude, like a set of Claude extensions, and one is ~like, you know, ~maximal introspection or like interview or something like that, where like really interviews the heck out of you when it comes to ~like, ~what do you want, ~you~ Noel: ~Mm-hmm. Yeah. That's a, it, it, yeah. There's like, I think like ~tuning the interactivity there is like a, is a challenging problem. And I think, again, like back to our prior point, that will be, I think, and ~it's, ~it's already happening like as right now, it's mainly happening ~kind of ~in the editors and stuff. Like, oh, we'll add another tool and makes it a little bit better. But I think we'll see more and more. Sorry, I shouldn't, uh, the tool, the term tool is overloaded. They're adding like a new way to interact with the agents that like, makes it marginally better. ~Um, ~but, ~you know, I, ~I think we'll continue to see that. I'm just, it still feels to me like there's not a lot of kind of external activity going on. ~Um, you know, for, ~for this kind of, ~uh. ~Like tooling in this space. It'll be interesting to see. ~Um, ~anyway, we're up on time. I don't have ~a super, ~a super [00:39:00] interesting hot take anyway, so we'll,~ uh,~ we'll leave that ~for next, ~for next month. ~Um, ~but thank you for joining us. ~Uh, ~thanks for tuning in. Have a good one, guys. Paige: See you next time. Jack: See you next time.