RobbieTheWagner 
Adam Argyle Welcome to Sentat! Welcome to a brand new episode of the front-end happy hour podcast. Welcome to this week's JS Party! Live from ship-shaped studios, this is Whiskey Web and what not. With your hosts Robbie the Wagner and me, Charles William Carpenter the third. That's right Charles, we drink whiskey and talk about web development. I mean it's all in the name, it's not that deep. This is, Whiskey Web and what not, do not adjust your set. Whiskey Web and what not is brought to you by... Today's episode is brought to you by Warp and their rad new cloud agents. Check it out at oz.dev.oz.dev. You know we're already big fans of this app, I keep a long list of all the features that I love.
Now with oz and Warp's cloud agent setup, I can escape my machine and trigger cloud coding agents from anywhere. From Warp, the oz CLI, the web app, which I can do from my phone, hell's-yeah, a cron job, github actions or an SDK that they so graciously provide. Think about it, with just a docker file you can front-load a sandbox with context and have agents open PRs across multiple repos at once, all with a single prompt. Warp rocks, enough said, all right back to the show. Hey, what's up everybody? Welcome to Whiskey Web and what not, with your host Robbie the Wagner and Adam Thomas our guide, the nerd.
Howdy y'all. Happy to be here. Hey, happy to be here as well. How are you doing, Adam? Good. You know, same thing as always though, which is half the day. I'm like, oh, tech is awesome. I love AI and then the other half of the day is, what am I doing? This is terrible. We are ruining the world. And I just can't decide. And so between that feeling and then the feeling of politics and then just being a dad and, you know, a spouse, that's great dude. I'm having a subregular tearing day where I feel like I'm pulled into 10 pieces. How about you, man? Well, I didn't read any headlines today or do anything that made me have existential dread, but today was actually good because it was 80 degrees.
So it makes me feel like summertime. Yeah. And some other reason you just snow blowing, you had a machine that you like, it's going to snow again. I think next week. So we except never ends, but, but yeah, what always makes you feel better. Some nice whiskey. A little bit of whiskey. We've got redwood empire pipe dream. This is their flagship one. They were kind enough to send us this bottle for us to check out. We've had several redwoods on the show before big fans. Thanks to Redwood for sponsoring this episode and sending us some awesome stuff excited to try the flagship. It is 72% corn, 18% rye, 5% wheat, 5% multi barley. And it is a blend of bourbons from California, Indiana, Kentucky and Tennessee with an age range between four and 12 years.
So it's got a lot going on. It is named after the 14th largest redwood tree in the world. Apparently Adam wrote this so he could have written anything and I would read it. Oh, I should do that to you one day and that'd be great. I'm Ron Burgundy. I do love the sounds. But yeah, apparently the bottle has the latitude and longitude so you could hike to this redwood tree. That's pretty dope. They also plan on some of the trees. They do. And I was like, I hope others to let along. Okay. So we got the lat is 40.34728 in the long is 123.96152. If anybody wants to hike there, I hope there's not a whole bunch of bottles here. You know, I hope there's like respectful people hiking there, you know.
Yeah. Yeah. All right. Well, let's give it a give it a taste. Oh, before we taste, do we have a toast? I haven't looked that far down in the notes. We got a toast, dude. Okay. So here's to the west coast, the best coast where the roots run deep and the networks run fast to the ancient redwoods perfectly anchored in the fog, reminding us that the most resilient architectures take centuries to scale. Cheers mate. Cheers. I'm smelling spear mint. Ooh, or maybe more of a fish, maybe not spear mint, maybe the fresh. Yeah. Mm, peppers. Um, whatever root beer is made of roots. Oh, man, it smells good, but it tastes good. And it's got like a good ribeye for bourbon. It's not too sweet.
Oh, wow. That is much different than it smells hits you really hard with some sweetness in the beginning and then some rye finish, which is kind of nice. I always like a rye finish. I went to the store today and I was like, I need a rye in the cupboard. I just like, and I like having a scotch in the cupboard. Um, hey, this is going to hit those senses. I like it. It's a little cedar and maybe it's because I'm, you know, reading redwood on there, but there's a, they had like a delightful woodiness to it. Yeah. Yeah, it does feel like some perhaps interesting barrels or something. I'm not really sure where they get all these flavors from.
But it seems like they're, um, they have a lot in the mash bill. They have a lot of blended stuff together. So perhaps it's more in the mastering the craft of the blend, which the folks at barrel do really well as well. So, you know, I'm a fan of of blends and figuring that out. I think that takes as much work if not more than distilling just something from scratch. So that goes into some of the history of this one. So apparently it's aged anywhere from four to 12 years. And I don't think we got a young batch. I think this one tastes really good. But they also do. I can't remember the acronym, but it was like MGD or MGP.
Someone's going to be really upset that I didn't write that out. But there's a, um, a famous distillery that has a whole bunch of whiskey. That's MGP. MGP. So that's where they picked and sourced all of these and blended them. And they chose specific barrels. So it's like they have been very picky and they did a good initial selection for this flagship. And so people gave them a lot of props like, oh, you picked the right distillery, the right blends. Yeah, this is good. Yeah, I'm noticing that my lips are very off from what I'm saying. Are you noticing that? It looks good now, but when you first said your first word today, it was quite delayed.
I like the AI reference on the front, which is the power of imagination makes us infinite. Way to go, John Muir. Yeah. All right. So for anyone who has never listened to this show before, we have a very complex rating system. Zero to eight tentacles, zero being the worst eight being the best four being four or five being middle. We should really make it have a middle number. So I can stop saying that. Anyway, I don't know what you think. Dude, okay. So I really like the ribeye. I like that this is a this is a bourbon. I'm going to enjoy a lot more than I think. I would have like a, you know, something a little sweeter. So I'm really happy about that.
It's got a like a woodsy vibe. There's just something about it. I know it's from the west coast and they're up in the red woods, but there is something kind of piney. It's very reminiscent for me of the west coast, just the way that it smells when you go to the woods. I don't think I'm going to give it a seven because there was the high and wicked we had recently. I gave it seven. I was like super good. I was like really into that one. But definitely high sixes. I don't think I'm going to go like a six seven five on this. This is really good. Yeah, I agree. It's very tough. I want to give everything a seven because it's nothing's perfect.
But I don't have anything super bad to say about it. But that being said, compared to the high and wicked, which was one of my favorites that we've had. Honestly, this is not quite as good. So yeah, I'm going to say, I'll say six and a half. Sounds fine. Nice. Subsequent Sips. More amber, a little bit of honey. Ooh, this is good. It's a good adventure. Yeah, it must be the blends. The blends just kind of keep rolling flavors over my tongue. It's exciting. Yeah, very pleased. All right. So I put a little couple of little different things in here. I was trying to do like a little would you rather game and I only came up with two. So it's short.
But would you rather review one PR with 200 changed files or 10 PRs with 20 changed files? I might have done this to my co-workers recently. So I did one of each. I wonder which one they didn't like most. I can tell you which one was harder to get reviewed. And that was the 200 changed files and one PR. But you know, it's nice to roll back. One PR with 200 changed files. Yeah. Yeah, I don't know. I'm somewhere on the fence here. Yeah, maybe you'll convince me. What do you think? Well, so it depends. But I think I tend to agree with you. Like there's a lot of initial shock when you see that many changed files that makes your brain shut down and you don't want to review it.
However, you're not saving yourself any work. You're still going to review them all at some point. So I would say like keep the big PR but just review it a little at a time. Like go through GitHub has a nice thing where you can just check each one you've looked at and like come back later and start where you left off and do the rest because when you have it all split up, it's like, like you said, if it's all related changes and you need to revert it, then you're reverting five PRs instead of one. But yeah, I've had this problem at work where like, yeah, I couldn't get anything reviewed because it was huge.
And they're like, can you break it up and like, absolutely, which AI is super good at. If you have the GitHub CLI and you tell Claude like, hey, I'd like this to be five PRs, please. It's like dope. I found these like domain areas that these files kind of fit in. We're going to split them into those and then like, I'm going to open five PRs for you with the GitHub CLI. And I was like, oh, sick. I don't have to do anything. This would have taken me hours to like figure out how to manually split it up. So props to props to anthropic today. Nice. Have you done any stacks yet? So this is a graphite feature, but it's kind of like, you know, branches that depend on each other.
And like normally you'd have to do it and like, just have pointers in the comments and be like, this is one of four. But they actually personify it. Well, they have UI for it that shows that they're all related. It shows them as a group. You can see the status of them. You can even see if they're all approved and ready to go that the one at the end is ready to. The only bummer. So here's the kind of follow-up I was going to say is that stacks are cool. I'm enjoying the stack workflows. And I think I talked about it last week with Git Butler as being a really great app for facilitating stacks and having branches of branches or even having three branches checked out at once and work tree.
Being able to like try on a combination of different people's work. But the bummer comes from like, there's an order that matters. And so when you have the one PR, which is a feature PR with all the changes in it. You get to see it all. There's no, there's no order of operations to deal with. And so we've had this issue where like I've got four PRs or whatever and I have to wait for them all to be ready because they all need to merge back to back. However, dude, hundreds of people are trying to merge code. So it's not like those four just go in together. So it's like, okay, so now do I do I squash the stack after we did all the work and then we re re approve or whatever.
I don't know, it just kind of gets weird. So there's no easy answer to this. Kind of like get read all those rebases the right answer, but I think I would rather review the if it's one feature and these are all related to the feature. That's not so bad. And plus, if it if something does happen, you really can revert, make a couple changes to that and resubmit and go through another approval process and kind of learn from your priors. Man, if you had 20 different PRs and one of the 20 kind of borked the rest, dude, you're a mess. But real quick, the way that I got to this scenario was over the weekend, I figured out how to take one of our internal agents and Shopify.
And course it into kind of a rough wiggle loop. And it made 48 PRs on Saturday. And I'm only I'm only through five of them right now. And other people are just like, what are all these? I'm like, I don't know yet. I've got to figure out everything it did. I just gave it vague instructions of right PRs that said, I got you. Well, yeah, it was like because I'm I've added a scroll component. And so I'm like, hey, anything that's got overflow needs migrated. So just go find overflow and swap it for the component. And it's like, boom, boom, just did it in loops. Anyway, so yeah, that's that's my mess. I've created, but they're all individual ones. They're not related. So it's a little easier to navigate.
But hey, that's a fun question. I like that one. Yeah. So the only other one is would you rather turn on Yolo mode for your AI and just review it all later. Don't watch it at all or like work with it step by step validating like it's doing the right things so that it can't just, you know, go off the rails all day. Yep, I'm I'm still torn on this. I all up until about last week, I never Yoloed except with pie pie Yolo's out the box. So in case everybody's curious, if you have pilots, just going to do stuff, it's mostly cloud and warp that want to sort of like ask you these questions along the way.
But then again, so it's I started Yoloing about a week and a half ago because I was getting really tired of approving because I was con but approving is good because it'll never do something catastrophic without me seeing it first. And so I could and I don't like to let it do get commands. And so that's constantly where I'm like, mm-hmm. You don't write that get command I do. I'm in charge of what actually gets sent to other people. But then just this morning I put warp on a task and it spun through all my credits until it goes, I'm out of credits and I'm like, you didn't solve the problem. I could do it. I went and solved the problem. One line of code. One line of CSS.
And I was like, holy shit, this thing just spun all of my goodness. So yeah, I think I'm somewhere in the middle. I think I'm mostly Yolo these days, but it bites me and I would almost rather go back to where I was with which was my sound bites where I got ogres and peons and Duke Newcombs. And you know, eight bit sounds kind of telling me the status of what AI is doing and then I go check in on it. But then it doesn't do shit for you overnight. So yeah, it's trade offs all the way all the way down. What are you doing this right now? I'm always like working iteratively like I tried to do. I forget which tool does what I feel like they all handle it a little differently for the auto approve where it's kind of like, hey, I'm using the GitHub CLI.
Would you like me to auto approve any GH command? Yeah, anything that I feel comfortable with a whole subset of, I'm like, do all of that. But yeah, I like for it to come back because like you just said, you burned through all your tokens and it didn't even give you the result you want. It kind of prefers doing that a lot of the time. Which whether that's programmed in it to make you spend more money or it's just a happy accident. I think I think they like that that you want to you get tired of hitting approve. So you let it yolo and it burns through all your money and then you have to either give them more money or do no work for the next month.
Which is probably not what you would use. Yeah, I didn't do any work over the weekend because I burned all of my open clause a hungry token bitch. And so I got a try and I've done all these things to try to pacify and make it take less token. But you remind me to I have an allow list that I was really nitpicky about for a long time. So I can cloud you can go say right like you're saying wild card all of LS. You know, if you're just listing files, I don't even care what comes after LS like LS, you're hard away. You know, I'm never going to deny you that, but I'm kind of giving up on micromanaging the allow list.
Oh, this kind of reminds me too. Have you heard of dark factories? I have not. Man, I should have put this on the notes. This is a cool concept. Well, is it cool here? You tell me is this cool or is this creepy as hell? Dark factory is a. An AI mantra where you don't care what the code is at all. So you write specs and you write expectations. This is what I want out of the software. I literally don't give a shit how it's done. I just want the feature. So do it. It's almost like when you do bolts are lovable and you just get something back. That was a dark factory. You could go peek at it if you want.
Yeah. And calling it a dark factory or hosting when it is you very intentionally saying we will not review. We just tell it expectations and the factory makes things for us. What do you think? Would you want one of these? Is this the future? Is it dark factory or like how do you feel about this? I think as long as software development teams exist in a similar way to how they do now, it will never be acceptable to just do that. Like that would be acceptable for a personal project or a team of a couple people or whatever. Because I think that it's like there's always going to be code review. Unless you've then, unless you've moved all of your code review,
you're so confident in AI code review that you don't like it will review itself. You're never going to look at the code on either side. Then maybe, but as long as I'm reviewing something, I can tell if you didn't look at the code at all. There's going to be a lot of signs. It's like when you tell it to write text and it uses m dashes and like all the stuff it shouldn't use. You're like, yeah, I see. I am absolutely right. That is definitely AI written. So I'm like, yeah, I don't know. I think we will get there. I think at some point, software teams will be drastically different than they are now
to where we're kind of all okay with that. But I think we're, I don't know, everything moves fast. But I feel like we're years away from completely not caring what the code looks like. Yes, I agree. I think it's important right now that we continue looking at the code. It makes way too many mistakes. Way too many mistakes. Just like today, like I'm saying, I don't even want to go into the problem. It literally was one line of CSS. Who cares what it was? It burnt a lot of tokens. And I used multiple models. So after warp right now, I was like, I'm just exhausted right now. I'm trying to do work and I want to do some personal hacking on my open cloud.
So this is like, go nuts with cloud this time. And I just had to stop it. I was like, it's just going in circles. Especially when it starts taking pictures, it's like, I'm going to take pictures. And I'm like, no, please don't do that. You're just going to eat up tokens even faster. You can't even see you, Dork. I mean, what are you going to learn from some pixels? The thing that always gets me is like, all the things that says makes it sounds so much like it knows more than you. Of like, you're the ask us to do a thing. And it's like, oh, here's like 10 pages on the exact way all of this works.
And I know exactly the changes we need to do what you want. And I'm like, oh, my God, you know. Like, yeah, can you implement it? It's like, no. So I feel like we're still falling flat. And like, I'm having way more magical moments than terrible moments these days. But it still lets me down a lot of the time. And like, I was doing some stuff with ember data, which arguably is not used to tons. It doesn't have a lot of reference material. But I was like, OK, if you have a relationship, the other side of the relationship has like they have an inverse key. And you have to set it like this inverse, like the other side of the relationship on this other model is called this is like what is for.
So I was like, this you can programmatically know definitively whether there's one on the other side because you go find that model. And if it's there, I want you to fill that in for me. And it's like, cool. I looked at all of them. Here's a table of all of them. I'll show you all the values on every side. I spent days trying to get it to do it right. And it just would not. And I'm like, what are you doing? Like, it's right there. You're going to make me look through all of these each one manually and tell you which ones you did wrong. Like because I can't say the right words to make you look at it again and like realize your table is wrong.
Oh, yeah, it's, it's frustrating. So I mentioned 48 PRs got opened up on Saturday. But let me be a little bit more real. And I hope we are more real on this show than other people who are like hype training and stuff. But it opened up 48 PRs. And I was like, okay, give me a link to all of the PRs. I want to see them in a flat list and categorize them by the type, you know, like if they're mobile or just give them some general categories. Oh, which also reminds me every time pie spins up a sub agent. It's a video game name. I might have said that last week. Anyway, it's really fun. It's a fun site.
It's just so fun. Okay, yeah, pie is just so fun. You can customize the sheet out of it. I'm just having a blast. So like every sub agent has a name from a video game that matches the tax they're doing. And it's just so fun. It's like, I spent up three agents here. Snake and Pac-Man and something else. And I'm like, hell, yeah. That's awesome. Okay. So anyway. The translation opens up 48 peers. List them. 27 of them or whatever. Say overflow suppression. Right? Remember I was like migrating overflow and I was like overflow suppression. I was like, uh, what the hell is that? And it's like, well, I just turned, I turned overflow off on those 28 PRs.
So they don't scroll anymore. And I'm like, well, that is completely unacceptable. I was like, that's not a migration, dude. That's like, that's just you ignoring the work. And I was like, why did I get, where did I get the lazy agent? Come on. Why do these agents, they're either lazy or they're liars right now. They're like, hey, I did the work. And you're like, you did. So prove it. And it goes, oh, you caught me. I didn't know. Like, what do you mean I caught you? They should never tell me I caught it. Yeah. What the hell? So I had to rework it. And I did it. But dude. Yeah. I got out of trouble with OpenClaw doing that.
Where like, every single time I'm like, hey, here's a clip of the podcast. You have a podcast clip hosting skill. It connects to post bridge. It knows that there are platform specific captions. It's part of the like API call that they should be of a certain format. And every single time I do it, it'll like forget that there's platform captions and make eight separate posts for each platform. Or like, it'll, it forgot that it was called Twitter in the platform captions and called it X, which I could see a, you know, a person making that mistake or whatever. But like, you have it document. Like, I told you, remember in the skill, write it down that this was the format.
And I guess it's just not writing it down. I haven't opened up the skill to look at what it's written down. But like, it just does everything like, it'll do it perfect. And I'm like, that was great. Remember that. And it never, ever remembers that. Like, I thought that was supposed to be the magic of open claw. And I don't, what do I need another thing? Like a lot of people have been releasing stuff. Open claw memory messed up, like install this thing. I haven't tried any of them yet to see if it fixes it. But like, why? If I tell you explicitly, then you should be reading that skill and you should see it in there.
And you should do that every time. I don't understand what's happening. I've got the same, I've got the same issue. The one that keeps getting me is, I'm like, I'm in discord. And I'm like, never reply in the general channel. The general channel is where I spin up a new thing. And then you sub thread. Always. And never, ever, ever post there. It post there all the time. It post there all the time. And then the weird thing is it went and it was solid for like a week, dude. And I was like, finally, I nipped it. And I can, I haven't, I did the same thing you did though where I haven't gone and looked.
But it keeps saying it wrote it to memory. I wrote it to my, my files. I'll never mess it up again. And the other thing it does that I always tell it to stop doing is it stop reporting errors to me. I'm like, I don't want to see errors in GitHub. Just, I have one thread. It's called the errors thread. Put them all in one thread every time you catch an error. Sticker out there. Hey, it's don't care. And it's constantly posting them everywhere. So yes, I totally agree. There's like a couple of really weird quirks where I'm like, I know you're a puppy and shit. And I'm supposed to like train you.
And I, and I'm trying to be patient. And I'm never mad at it or whatever. But I'm just like, some of this stuff is really annoying. And it's so apologetic too. You know, it's just like, oh, I'm so sorry. You're totally right. You've told me 20 times now. And he can even count it. And you told me 19 times. Did I not do that? I'm like, well, how many more? You're like my children. And I don't want that. I would like a computer, not a child, please. Yeah. Yeah, I mean, the strategy that I think we both talked about that does work better is like anything that can be a script, just make it a script.
And then it's like, it can never get it wrong because it's just invoking the script. But like, yeah, is it just, do I have too much context? Do you like reset before you do anything every single time? Or like, I don't understand when it decides to clear the context if ever. I don't know. I haven't gotten that in the weeds about it. I have a little bit. So I had it make a token dashboard. Oh, here, I'll share that blog post here. I'll go to nerdy.dev and find my, I talked a little bit about some of my open-class stuff. But I haven't make a dashboard of its own usage because I want to go look through the usage
and be like, what's going on? What was expensive? Oh, here's a blog post. I should talk about it at some point today. Swoop, swoop, there it is. Oh, yeah. But I had to make it through that one. I didn't read it, but it was cute. Nice. Thank you. It was very cup head inspired. But the idea was make a dashboard. Track all your usage. Show me which github or not github. Show me which discord threads are consuming the most. Which I assume is going to be the store I'm building. I'm building a Shopify store and I'm like, hey, that's probably going to be, and it was. It was one of the top ones. But it was general channel because every post that goes to the general channel,
it would read to the beginning to find context and pack all that shit into its brain and then charge me. Right? And then I was like, stop. As soon as I saw it, I was like, new rule. What I make a post in the general channel. It is a completely fresh context. I want you to only reference your little internal memory files and your little things that I've touched you. And start fresh. Furthermore, open-cloth punk-ass as I call punk-ass. Furthermore, if I post to a thread, I want you to only read up to past messages to get a sense of what it is I'm talking about. If you don't have enough and you don't feel confident, you can read a couple more and a couple more.
But do not read, flat read, the entire thread to catch up to this one comment. It's very likely it's within the last two chats and I don't need you packing up context for that. And that has saved me a lot of tokens. But it's nonsense. Like, how does it remember that instruction? I mean, it says it writes it to the same spot. It writes the other instructions that it doesn't actually listen to. Yeah. Exactly. I don't know. Luck of the draw, I guess. Luck of the draw. Yeah. We could be talking about this for the whole episode. So I want to steer us away from this because we always get trapped. I want to talk a little bit about, did you see, I feel like this has maybe been a week or two now,
because we had guests last week and I think maybe it happened around, I don't know, where CloudFlare rewrote all of next JS in a week. Oh, yes. I definitely saw this. Yes. Okay. Yep. Yep. Did you watch everything unfold after they posted that? No. I heard a couple of things about it. Like someone's like, it's going to be packed. I mean, so there's been security vulnerabilities in next forever. So it's like, how many did this introduce it? I'm sure it's like the first vector they attacked. But yeah, it's going to be more. Yeah. So I didn't see everything. Everyone was talking about it. And everyone was like, oh, my God, AI is so powerful.
You can just rewrite a tool as powerful as next JS in one week. Like, what could we do next? And then Guillermo got really salty. Of course. I think the first thing he posted was like, no context, just like a link to like some Versel docs on how to migrate from CloudFlare to Versel. And then I think within the next six hours, like I think he had his whole team up at arms like doing stuff on it. He was like, oh, we found three critical vulnerabilities. We fixed them all by the way, because we think the web should be like a safe place to be here. I don't know. And then, yeah, it was, and then like everyone was like, oh, there's no vendor lock-in
in the next JS, no vendor lock-in. And then they're like, real people who have used it are like, there's some vendor lock-in. It's like, you can technically deploy it elsewhere, but you lose all of the goodness. Versel is a great platform. We use it for a lot, like no shade. But yeah, if you're doing next JS, you're kind of committing to using Versel. So, you know, I think this was a interesting experiment, but I just love the drama. The drama was well worth it. And I like the drama because every weekend somebody rewrites some shit in a day. So, yeah, you just had this with Andre Kaperski. I'm going to screw up his name so badly.
He's like one of the famous AI folks right now. And he's like, overnight, I had my new research loop, which is like a Ralph Wiggum incremental research. Have you seen this actually, the research loop, this person wrote? Is this the auto research thing? I saw something called auto research that was like supposed to do something. I don't know if that's what you're talking about. But it's like another one of those scenarios where someone overnight had AI working and who knows how many tokens are racked up. And it solved some hefty problems. And so, this particular one was a, they wrote a research loop that researches the topic,
tries to make an incremental improvement and measurable improvement. And if it works, it commits it and then it runs the loop again. So, it's like Wiggum making just nudging something. Just nudging towards victory. And sure enough, they woke up in the morning. It was a really tough problem. And they had incremented the progress by like 13% and made these, you know, wild improvements overnight by just, I call this brute force. I don't know what other people call this because it's not elegant. It's, you know what I'm saying? Yeah, it's Wiggum is not elegant. This is a really rudimentary, this is like hacking when all you did was bang on the door with one.
One didn't get me in. Two, did that get me in? No. Three, I'm like, oh my god, how long is this going to take? I think it's a computer. It's going to sit here and do it. It's a dumbass. It's just, it's just a loop. So anyway, but the research loop was interesting. But what I just like every weekend that happens is somebody recreates something that took other people years to do. Like, what was it two weeks ago? It was like, we rebuilt a browser. It cost us $400 in tokens and we rebuilt X. Hey, we rebuilt X with this and it's just like all of that to say, I can't rebuild the scroll implementation on my site with any LLMs.
Yet you can rebuild an entire rust backed server. We're in a weird space right now and these like wild sweeping things that people are building are kind of pissing me off. Because they just give this false impression. And even for me, it's like, I've kind of one-shot it a few things. But the one-shots aren't the cool feature of AI, but they're the things that get you the likes and the, I don't know, man. Yeah, one-shotting doesn't really work. And if it is going to work, I think the like, the pre-work really matters. Like, you need a really fleshed out spec or at least more details than like, I want you to build a social media platform.
Like, you can't do that and it's like, hey, cool, I built X for you or I built Instagram for you or whatever. Because there's so many layers to that too of like, well, are you going to use SaaS products or everything like you roll in your own off? Like, are you just building visually the same thing? Because that's not as hard. You can fake it with like, you know, CSS and make it look like the thing without having the functionality. So I think it all comes down to, is it for fun and for it to display similar? Because if so, you can do that sometimes. If it is supposed to have anywhere near the same functionality where it talks to some sort of API,
and that whole API is all fleshed out and your authorization and authentication and like all the things are done, like that stuff is not getting one-shotted. I don't think, at least that I've seen. I mean, I feel like the backend, you know, rust things, these are just like the contracts are so tight that they're more primo for rewrites or optimizations than the front end. So that's kind of the, it's kind of the issue that I had. It's like, I'm giving it a front end task that's just wildly wiggly, especially I've got what it worked. It's like only in Safari in this one weird little case. And AI doesn't know how to solve that.
AI can't compute the combinatorial explosion of like multiple browser versions plus the way that heights are calculated with the fucking 40 million boxes that it took to get down to that one little box that had some text in it that's broken. I'm like, that's just, I'm eager though for to take some of those jobs. Like those are some of my least favorite ones to do. It's like, I don't want to go over there. Have you tried installing like a combinatorics.ai skill? I have not. Yeah, I haven't either. I wonder if there's enough fancy algorithms to brute force, you know, all the possible. Like more or informed brute force, you know, like, oh, we're using the, what's the solve thing?
I don't know. We're using Dijkstra's algorithm to like find the shortest path to this thing. It's a process of elimination. So yeah, you could brute force your way to understanding what it's not until you finally discover. I mean, that's honestly what I do. That's what you said. That's what any of us who have ever coded for a long time, that's what we do is we've now acquired so many wrong things that we know it's right. And when we see enough of wrong things and we see a bad behavior, like a lot of people are like, dude, how do you know that was doing that? I was like, I have smelled that pile of shit before.
And I know I've smell, that's my 20th time smelling that and I know exactly what it is. And it's just like, I'm not proud. Yeah, it's hard to explain that too because people will be like, like, I don't know, I'll be pairing or whatever. And we're debugging our way to find the bug and we get there in this super roundabout way. And like, wait, how did you know like, that's where it could be? And I'm like, honestly, I couldn't tell you, like we're just getting there because I like, and I can feel my way towards like this being the right thing or whatever. So yeah, it just comes down to having done it a bunch of times.
And maybe that's what AI is missing. Maybe when AI has done it as many times, if we can figure out the like, when you learn it, refer to what you learned. How do you know when it's right? Yeah. Which I feel like it's like what we're doing with Anthropic. This is kind of why I feel like Anthropic is what a scenario to be in. They are definitely using analytics on everything that's going in and out. So we're training their models and paying them what I'm like, how do I buy stock? Yeah. How do I buy? I've never heard of anything so amazing. Yeah, I mean, so no shade on Anthropic because they're giving me free cloud right now.
I want it, dude. I have a, I have a repo with 5,000 stars. I'm like, give me the thing. Did you apply? They'll give it to you. I hope so. I don't have it yet, but yeah. Yeah. So I, I mean, it was easy for me. I was like, I maintain Shepard. I would like free LLM, please. And they were like, you got it. Like I didn't put that much explanation in there. Am I in the dark or am I still in the window where it could be a yes? How many days until you got a yes, dude? It was like three days, I think. Oh, I probably am not going to get it then dang it. Well, I also, I think I saw the tweet like minutes after it was like, I was one of the
first people to apply, I think. Oh, so that might have been good for me, but. Okay. So there were some layoffs. We got some notes here. Let's, let's count these. Yeah. Yeah. So I wanted to see if you had any, any takes or things to say about. Block laid off 40% of their workforce, which has been maybe a couple weeks now. So it's not fresh. And by the time you all hear this, unless you're listening live right now, it'll be super not fresh. But I still think it's an interesting thing to address because like they lay off 40% of their employees and then their stock goes up 25% that day. Classic. So.
Yep. Like one that signals to other companies. If you do a similar thing, you will be rewarded by investors. But then. So I saw a video of, I think this was on some other podcast or something, zero context. I saw it in passing in my feed. So sorry to whoever I'm not giving you any credit, but they were like, you know, this video Brian, I don't think so. I have. Okay. I don't know. This is dumb because, you know, why lay off your peep, if your core product is doing really well. And like this is what, you know, Jack Dorsey at block the CEO had said, he's like, we're doing really well. We actually didn't have to lay everyone off.
We have plenty of money, but it like, you know, trims the fat, gives us more profits, whatever. They're like, why would you do that when you've spent all the time hiring all these people vetting them, having good developers. If you think you have too many, then just take a bunch of wild ideas that can make you tons of money and throw everyone at them with all the AI, like you should be trying to get to the next level and like blow everyone away, not try to maintain your current product and cut costs. Yeah. And I'm like, I kind of agree with that. Like that does sound like that. Like as soon as you fire everyone and you're just trying to have a lower cost to maintain
the product that makes you money now, you've already lost because like everything moves so fast that I feel like if you're doing that, you're never going to be at that next level where like someone's going to vibe code their way past you tomorrow and you've lost all of your advantage now. I love that. But I kind of want to maybe go a little further instead of just giving them wild stuff to do. To attack the SaaS products that that company's using to save money and move it in house. So all your competitors, kill them all your competitors, anything that you're paying for monthly that's charging you a just a stupid amount of money vibe coded vibe coded internally
and cultivate it. So put a team on it and just harness it until that thing is now your own your own tool doing your cause this is the malleable software dream that the dream is is you don't have to bend your workflows to someone else's SaaS product, you build your own product internally and it bends to your will and anytime you want something new, you just vibe it and then you wait till it hardens and then you ship it. I think that's the way I do it. If I was a SaaS company right now, I'd be crappin' my pants. I think we talked about that a couple times but like Shopify's already ditched a couple
SaaS products. I wish I could tell you what they were and I don't know what I can or can't tell you but like yeah, well without saying specifics because we're ditching a ton of SaaS products as well. Are you really too old? Which is hard of moving to IBM because like IBM has a lot of internal products that do the same things but then other things are just like well, why would we spend I don't know, $200,000 on launch darkly when like we could not and probably just make something that does similar like depends on what you're using, you know what features you're using of the SaaS product and how long it would take you to rebuild something similar of course.
But yeah, it's very compelling, you're like, you know, if you gave a team of 10 unlimited AI usage and you said you have six months to replace launch darkly, do you think that they have that much proprietary stuff that like a team of 10 good engineers with as much AI as they want couldn't replace it for less than $200,000? I feel like I feel like everyone should be shaken in their boots. It's like everyone should be shaken in their boots, man. So I've reflected on this a lot because I work for SaaS companies. However, I think Shopify is in a unique scenario that some will ditch the SaaS to run a store
but a store is more than a database and products with purchase buttons, there's just a lot going on and especially like even like Shopify has has this feature now where if you put products into their database, they'll actually surface you as results to chat GPT. So now like that, you as right. So imagine your products are now getting bought through brand new modalities that you didn't even know existed just because you're in their product book. And here's the second place. If you're shopping for another product that's similar to you, you will get surfaced as a similar company to the one that you're shopping at.
That is really cool stuff. You can't do that if you're running your own solo store on the internet. Plus I've never, I've used Shopify super duper lightly or some iteration of this podcast store was on Shopify for a couple of months. And it was not expensive. Like I'm sure if you're, I don't know, I don't know if it was a month. Yeah, I don't know if a big brands pay like like if there's an enterprise plan where you're going to pay a shitload of money or how it works, but like, you know, I think for the like, yeah, 30 bucks a month, whatever, chump change, like it, all the features you get, there's, it's dumb to build your own at that now, $200,000, then you should
build your own, you know, like it starts to be more compelling when the price goes up. So I think that like the volume game is very smart. If you're a SaaS product and you charge a reasonable fee per month and it's not like something that you can vibe code on your own in a day and be confident on, why would people not continue to pay you those reasonable fees? I think that that volume game will continue to be strong. You might have a few people that drop off, whatever. But the things that will get killed are the ones that have that bespoke enterprise. The only way we make money is if this one company signs for $500,000, $1 million, whatever,
otherwise we can't, that people are going to replace that big time. Yeah. Oh, man. So I called it like, we could go two weeks ago. I was like, how long until we replace Slack, because I know Slack is a fortune, like it is so much, so much wildly expensive for a chat app. But every, so I worked for Amazon, we had Slack, I worked for IBM, now we have Slack, like these big organizations who probably hate using Slack, like we have their own internal chat apps at like every big company because they've built them and they use them for whatever, like Amazon had a chimes, I think it was chimes. Like everyone has a thing and it's like, but we can't because there's like this built
in, you know, we need the integrations where you can like slash my bot, do the thing and like we need all these different things that you don't quite have and it just feels nicer and you get like, they're backing up all the history, so we've got 10 years of history of all the conversations you've had or whatever. And like, yeah, I think there's a lot of little things to solve, like whereas, you know, a chat app at the surface is super easy to build, but like those things that they've, they've got a big moat because they've been the, the choice for like, maybe a decade over a decade, it's been a long time.
And so that, that is a problem, like they're, it would take so much to replace that, not the people aren't trying. I'm sure people will try because I'm sure, like, you know, I know as ship shape, a tiny company, Chuck and I are running a consultancy, we were paying, I don't know, $700 a year or something, which is not huge, but not small. And like, you know, that, so that was for like 10 to 20 people or something. Imagine companies of 5 million people or whatever, like it's like, they've got to be paying so much money and yeah, so, but, but then again, I never hear anyone talking about replacing
it. There are so many other tools that are like, replace that. Let's get out, but I've never heard anyone be like, we've got to get rid of slack. So, all right, let's, let's change shop a little bit. I saw that Amazon was like, hey, we've had a lot of downtime from vibe coding instance. We need to stop a little bit or like, they were calling like for like a reform or like, more reviews or something like that. So it's like, I loved the tension, the tension. I'm sure was at Amazon at one point, it's like, AI fucking rules. You better start, you better have PRs burgeon like wildfire, y'all like, because this is
where we're at. And then all of a sudden they're like, slow down, you know, like, and I'm like, ah, the pendulum, it's swinging. So, you know, you cannot resist the way that it swings back at you. What do you know about the Amazon, uh, uh, pendulum swinging here on this? I have not talked to my, my bros at Amazon to get first hand accounts, but, um, this, yeah, I think this is going to be a thing at all companies where like, they're, we're all kind of forced to use AI and like, I, I'll take as much AI as you'll give me because I'm enjoying having it do these minute stupid tests that I don't want to do.
However, when you yolo mode something and it changes 50 files and all the test pass, do you push it up? Probably. Like, um, that, that is just going to slowly erode your, you know, decades of enterprise level safety. And there'll be some little bug that like, AI didn't think about that like a human who had managed that and got paid way too much money for a decade, uh, would know. And that you've laid off all those people. So fuck it. Like, it's all going to get screwed because like, yeah, I mean, there's, there's a problem there. The reverse has happened to Shopify where, um, human, human reviews were too short or
whatever. Like, humans reviewed this, humans wrote it, humans reviewed it and it caused an outage. And then, uh, we have people saying like, look, there should have been an a, here's the new mandate. If it, if AI wasn't part of the review, there's a problem. It's not, was there a human in the review process is it's required now for AI to have seen the PR and say something. So now we're actually at this point, now where we have corums. So it's not one AI also, also, oh, who was it? Oh, there was a service. There was a service that was charging 10 to 15 dollars per PR review. And they were trying to pitch you like, can I get $10 to $15 per PR view?
He is a service. Me is a service. Um, but right, I was like, oh, it's a kind of, is that cheap? Is that, that's actually sounds expensive. Um, but they're like, what we're doing a, a wild suite of tests for you, right? It's not like one AI with one opinion. Um, and so we're, we're testing things. It's like a corum. You have a multiple different personalities that kind of come at this and try to, uh, establish something. So it's a really interesting like landscape, they're having to deal with right now. The tensions between human reviewers and AI reviewers is just, this is why I brought up dark factories as you know, it's like, if you're in the opinion that AI would have
caught that and a human didn't, then you're of the opinion that you should just be running a dark factory. Um, so I think this, um, less about what would the AI, would the AI catch in a human wouldn't and more about like the, you know, you have to have an AI review it. It's just like, why wouldn't you, right? Like it's an easy check box. Why wouldn't you? I agree. I could find something. Yeah. Like code rabbit almost every time I get code rabbit to do a review, find something might not be critical, but it's like, you know, this, this is a little weird. Have you like looked at this and like, damn, that is so thorough.
Like no human would have ever told me that. So why would you not? Like I think that, uh, I think it, it's definitely a good idea to have that requirement. I don't think that it's going to save you necessarily. But, you know, why not? Uh, the more reviews, the better. So if you can get a free one from AI before any human looks at it, yeah, I'm cool with that. Totally agree. Yep. I'm not sad that it's there. I think it's kind of smart. Um, I do appreciate what it's doing. And they, uh, they ask good questions. Like it's just another way. And I, I run the same thing locally, like question this PR.
I'm about to put this PR up. Please be critical. I want you to like look at this, fresh eyes, you've never seen it before. Judge me for everything that's in there. All right. We have way too many topics for the rest of the 10 minutes. Uh, we need to take all our topics. We don't get to and putting them in a list for next time because we say that. And then we forget about them next time. But anyway, um, the stacks one, dude, let's do it. Okay. Yeah. So, you know, similar to everything else we've talked about, AI, AI, AI, um, Dax had tweeted a thing that I think is super important. And I think, uh, Dax being, I don't know what his title is, but, you know,
basically being in charge of open code, whatever he does there, um, is thinking a lot about this because it's like, we're growing really fast. Have a lot of developers. We're trying to, you know, think about what's actually useful. So he's like, everything great comes from being able to delay gratification as long as possible. So he's saying basically like, don't just go, oh, cool. New feature, uh, I described it in like one sentence, AI wrote a ton of code. Let's ship it like maybe that's, maybe don't do that because, you know, you should, you should actually think about which features people want.
Not just like, can I build it real quick, but do people actually want it? Are we bloating our application doing stupid stuff? And then like, um, similarly, while you're like, if you, if you spend the time to build like a really good spec and you're, you know, tackling a problem, and throughout the iteration of, uh, implementing that problem, you end up with some hacks like you need to work around something. Then AI makes it really easy to work around that. But maybe what you should do instead is say, we should rethink our design. You shouldn't need to be hacky. You should just like, maybe your design was flawed to begin with.
And then the third point is like also prior, prior to ties cleaning up and doing tech debt remediation because like, that's not as sexy as shipping new features, but like, you know, he's just saying everything should be high quality. And you should think about features, be really selective with it because, you know, we're going to hit a world very, very soon where everyone is shipping things so fast that they're all sloped. Like they might work fine, but it's like a thing you didn't need makes the app take another second to boot. It's like, you know, you really, like that bespoke crafted software is going
to do so much better than like just the throw everything at the wall software. This reminds me of a, uh, if you heard of the golden mean by Simon Sinek or the golden circle, I am not. This is, it's got to be 20 years old at this point. And it's, it's what made, what's what made Simon Sinek popular as like a thought leader. Um, he says, it's not what you do. It's why you do it. He goes, when you buy an Apple product, uh, at first, you might be thinking about what it is you're buying, but he's like, that's not why you buy it. You buy it because why Apple exists. Um, and so what you just pitched, and this is why I shared it is you just
shared exactly as he describes Apple. Apple will, um, create the best, most bespoke solution for the problem. They'll take the time and they'll do it right. They're going to give you hardware that's tight. And this is how, uh, you wish it always was by Apple, you know, and it's like, holy shit. Okay. So yeah, what is it? What are they selling? Yeah. I love the way that sounds like I don't really care. It sounds like they really care about scratch. Um, I have a follow up, though, which is, oh, go ahead. Yeah, actually, I was going to say it kind of comes back to what you've said a bunch of times is like, it's about taste and it's about like curation of like, you know,
it's, everyone can build everything now. So it's no longer about, can you build it? It's about, do you understand what should be built? Yep, it's not what you built. It's why you built it, which is also, I heard, um, why didn't hear, um, but there's a, a undertone of design is now we can going to become more important than it ever has before, which is similar to saying that taste and vocabulary are the way that you get what you want, which is saying that design is now going to distinguish everything because if, if what you can build is just a prompt fart away, it's what did you, it's, it's like, and it's like, what are the attributes of what you built?
This is what's going to distinguish things. Same thing with like a remix. You can remix anything anybody's built at this point, but you need to remix with purpose, uh, not just remix to remix, although those are both going to exist. This is like what ship posts are. Ship posts are basically like, I remix it because I could. I put boobs on it. That's big. Why? I don't know. That has to be, but people laugh. I did it for the walls, you know, or whatever for that ad revenue. But yeah, okay. Wait, uh, I want to get into T shape versus X shape. Yeah, yeah. Um, so okay. So T shape people are, you have, um, a couple of skills, but you have one skill that
goes deep. That's the T in the X shape. So you're wide. You're like, the top of the T is shallow because it's like just the top of the T. And then your, your middle is deep. Okay. Got it deep. Yeah. So like for me, if someone was to describe Adam, they'd be like his deep skill is CSS. Obviously, his T though goes in both directions for front and development and back and development and it's got some product and some design. I don't know. He's got a few different T shapes that poke out the top. But like the reason we're bringing him in is he has an expertise and CSS. He is a T shaped individual with a T stem that goes deeper than the rest of us.
And that makes him valuable next to other T's. Okay. The X shape is what is being pitched as now more important work is now valuable in the age of AI because you are cross referencing. So the X meaning that you are now taking your skills and crossing into other domains that you did not know before using agentic flows, using your taste and your vocabulary to reach and to extend beyond your T into domains, causing positive change across the organizations and slices of your code base. And I think this is cool. This is also plays into like you've heard people talk about like AI is now enabling you to
have agency and I hate that agency and agentic like where flows are sort of like mixed up. But like agency meaning like, look, I see a problem. I'm using our tool and it kind of sucks over here. And I used to go, hey, you and that team, you should go enhance that feature. Oh, maybe I lock a bug, you know, I can take a video with sucks. But you know, like now you're like, dude, I can actually just go change it. I'll just go ask Cloud and be like, Hey, Cloud, that little feature over there. It sucks. Well, like, let's make it do X, Y and C and all of a sudden now you're X shaped. You're not a T that's going deep.
You're an X that is reaching in a crud. Your opinions and so anyway, I liked this concept where like your skill sets and what's valuable to the company is your ability to not be asked to do this. You didn't get asked to go into the finance department and change the way that totals are done. You just did it because Cloud enabled you. You had you saw a problem and you fixed it. So we might become more and more generalists. I guess an X is kind of like a generalist and a T is a specialist. And EI is promoting the transforming across section. I agree that I think like, you know, it's similar to like I've talked about a few times where I
built this whole color picker, magnifier thing and rust, just five coding it because I didn't know any rust, but I know programming. So I know like, you know, the right things to ask for. I need to make sure that we're writing the right sort of unit tests for all of these little functions. Like all of the things to have quality software. Even if I'm a, you know, I only do front-end most of the time. I haven't done back-end work in like over a decade now. I feel like so, but I know enough about programming that I can inform AI to make good back-end decisions. So I think that's like the thing there is like if you have a little bit of skill
that goes pretty deep on one end or the other, you can use those best practices to then get a better result on the other side. So it's like we're still all collaborating, but you might not know a lot about the thing, but you know, then theoretically someone on that team is going to review it or AI, I guess whoever's reviewing it. So yeah, I think there's definitely something to be said there of like if we all kind of are very experienced in a certain domain, which is like, you know, generalize as programming here. You can get better results on either side, not knowing how to do that thing. So I buy it. Yeah, I liked it. It's say the reward system says, if you see it and it
sucks, go fix it and quit being a douche. Wait, hold on the last part, I added. Oh, it doesn't say that in the in the box. It doesn't. All right, we do we need to speedrun these last five. What's important? We can save some other stuff. What pick one more from from we haven't go to what not? There's two tools. Oh, wait, I'm just going to speedrun a couple of these. Actually, wait, I can speedrun all four. So anthropic is like 24,000 accounts hacked us to try to figure out how we do things. And they're stealing. They're stealing from our models. And I'm like, that's funny. You're sad people are stealing from you. You stole from everyone. And I'm like, there's
got to be a joke about the thief stealing from the thief who stole from the thief. And there's just like a thief circle of like, is there a circle jerk of thieves? Anyway, that's my joke. That does make me this will take an extra minute to think about this, but I do wonder. Okay. So in real life, right? You steal, let's say you go and steal a gold bar. It's worth like $20,000 or whatever. I don't know how much a gold bar is worth. And then I steal it from you. And then someone steals it from me. Who do you arrest? Like, are we all thieves? Or is just the original who stole it from the person who actually owned it? The thief. Someone tell us in the comments
because I have no idea. Yeah, because I'm like, I wonder if this is the sentiment that you'll get from AI, right? It's like the one who made the egregious theft was the first one. Anthropics stole everyone's code so that we could all use it. And then if I then use your product to make a thing that steals, it wasn't my fault because you gave me the stolen content. So I think like, there's some weird levels, but yeah, I think we've talked about this before. I think everyone that is like sea level at all these AI companies should be criminally liable for the amount of theft they've done. Yeah, you know, they approved it at some point, totally agree. Okay, speed run,
next one is cursor cloud agents. So cursorize cloud agents by the way now. And they'll take videos of the work that they do. So it's no longer screenshots that are expensive. It's an MP4. And I'm like, 4K. Probably, dude, you get to, oh no, you pay more. But anyway, I just like, I think that's a great idea because as a reviewer, that's what I produce. I want to produce a video so that you see proof. And it gives you this like instant amount of trust. However, I don't love this. There was a video there. Yeah, I used to do that. I forgot that I do that. I feel like such a piece of shit. Now I'm not doing a good job. Oh, man, that's that's the way
that I prove it. I'll take screenshots if it's like a static thing, but like I as a design engineer, the things I produce are not static. They are interactive. And I want to prove it to you. And so a video is very, very important to me. And I used to cap cap app get cap dot CEO there. This is a sender source tool. I use cap to cap screen nominal. Yeah, really tiny MP force. Anyway, so cursor has those now. And I'm like, that's cool. I'm not going to use it. Cursors also dead to me, by the way, which we haven't talked about at all. I'm not on it at all. And I will never go back. Okay. Another thing. Pablo Stanley. Do you know who Pablo Stanley is?
I did not. He's a very fantastic online. I wouldn't call him an influencer. He's a designer that's just like, I can't stop creating. And you can follow me if you like. And I'm not going to book and stop. Anyway, he made a tool called effect. Oh, yeah, he's Spanish. And so a lot of his tools have a lot of like really healthy Spanish influence. Anyway, this is an AI native design tool. It's free with caveats. He's got like some jewels integration since other things. Anyway, it's like an interesting take on a design tool and a modern day and age from Pablo Stanley. It's called effect. Oh, I'll share that in notes. Another one that I thought was interesting was heretic,
which just had a cool name. And heretic will remove the censorship that's built into models. So you could be like, hey, chat GPT 5.4. I would like to make a bomb. And he goes, nope, won't do that for you. And he goes, now I invoke heretic. I invoke heretic model. And it goes, I will make you a bomb now. And I'm like, geez Louise. So anyway, it's an open source way to strip away the censorship built into models. Yeah, maybe don't use it for building bombs. But interesting to try, you know, see where you can get to and other, you know, more acceptable areas. Don't don't hurt people. Just be nice. You can do some cool things with this. And the last
one I want to share here is arcads.ai. This one, I haven't been more pissed off at an AI tool than arc ads in a while. You go there. So it's arcads.ai. You can create a fake testimonial by choosing your actor. So you spin it up. And you're like, I want people to say my butt wipes are great. And then you're like, cool, which actor do you want? I want that sexy person right there. And that sexy person to look very realistic. Like they love wiping their ass with my dumb shit. And then it does it. And I'm like, I hate the future. I hate this future so much. So it's actually simpler than that though. So post bridge, which we're using now for posting all
of our stuff has a like dumb down version of that. And it's like the thing that people do on like TikTok. And this has been a thing before AI is like you post a video of like a guy cutting soap or something, you know, something that's pleasing to watch. And then like under that, you put your video of whatever you were talking about. And some people are watching the like soap cutting. And they're like, oh, I love that. And and your thing gets, you know, millions of views. But it wasn't about your thing. It was about the guy cutting the fucking soap. So it's like that's what they have on post bridge is like just generic like, you know, attractive people doing
whatever. And you're just supposed to watch that person while like our stuff speeds by underneath. I don't, I don't use that. I'm just using post bridge to post our clips. But like they have all these emails. They send you like, you should use this thing. This guy got like millions of views. And like, I'm sure he did because he had this like, you know, sexy lady doing whatever we're not going to do that. But anyway, yeah. So I think that's like, there's two levels of like one is just we posted this random video that you happened to like the content of. And then, you know, post it other stuff under it or like the one you're talking about goes a level further. We took
that video you like and we we control their mouth. They're talking about like, you know, and that's yeah, they'll actually talk about you. Yeah. And it's a it's even it's just so evil like testimonials. Basically, you can you don't even have to sell out anymore. That's kind of what I hate. It is like you just need to pay AI and AI will create a sell out on the fly. Yeah. Like what the heck? There's people making tons of money that like have completely fake influencers that post stuff for them. Just because you can now like because why wouldn't you take your shot if I'm like, I need some money, open claw, make me an influencer on Instagram. Like I want them to
say whatever and I want to get good ad revenue. What like, I can't fault them for doing it because everyone needs money. But it's yeah, I don't love that world. It's it's just the side of like AI where I'm just like, I just wish we hadn't. But there it is. Um, so I want to I want to regress back a couple points to heretic. Did you play heretic the game? Do you know what I'm talking about? Well, yes. But dude, so when were you born? 90. Okay. I'm 86, baby. That was like too old for me. I remember playing it. So like, I don't even know what to do. I was like, I was like, I'm a kid and this game feels five. Maybe. So you were five playing heretic. I played it
a little later. Let's see. Okay. Yeah. Heretic. When did it come out? That's I'm so glad you're bringing this up because I very specifically remember this game. It says 2000 or out of 2024. No, 94. Yeah. It was like a first person RPG that was like just over my head. I also didn't have a good computer. I had no older brothers or anybody to show me. So so this was this goes back to this is like this was my computer that I think was an IBM computer. This game was on floppy disks and you would hit a certain point in the game where you would be walking towards a thing and it would go put in another disk and you would have to like take it out. I put another
one in and like, people don't have to deal with that anymore. But it was so fun. But yeah, heretic I think is my in my top five favorite games. It was really good. There were remastered ones you can play for like Xbox and PC now. You should check it out now and see see what you think. Heretic two was awful. Don't play that one. But heretic one was very good. And then like after that, there was hexin hexin one and two, which were like the same sort of like first person. But it's like magic like like you're not shooting guns. You're like you have like magic abilities and like shit like that. Yeah. Pretty pretty fun. I like them all.
Man, I knew you liked RPGs, but I kind of thought you were more of a first person shooter RPG player. But you've got roots. They run deeper in the fantasy world. Yeah. My number one game is Diablo. I played Diablo one and two and three and four. Oh man. Do you remember when they asked if you were legit when you when you logged in? In Diablo. In Diablo. Yeah. You're legit. Yeah, because it was so easy to hack and I hacked my character. But like you could run all sorts of apps that modded your tune. I gave you super abilities or whatever. So I used to crawl the dungeons. You didn't get your ass kicked. And so when you joined, I never did that with other people.
Oh, you did. No, maybe I did. I think I know what you're talking about now. I think I did play with that because it yeah, I wouldn't let you do it online. But like, oh, that's the thing is see if you if you played online is you would join a group and people would be like, are you legit? And like the whole question is like, is your character modded? And sometimes and you never knew if they were like, if if legit was going to be good or bad, like, are you legit? And you're like, yes. And they're like, good, because I'm modded too. Let's go kick some ass. Like sweet. I'm a child. So let's go play. Do you remember this happened to me? I feel like this happened to everyone.
It's like a ride a passage. You joined a game and they were like, Hey, if you let me kill you 15 times, all of your items will be duplicated. Do you remember this? No, I would have said no so fast. Okay. So I was like, this is weirdly specific. So I did it. Got killed 15 times. And there's like an overflow, like a stack overflow of some point. When you get killed 15 times, it can't store your items in your bodies anymore. And so they all just fly out into the world. And then the people take them all and leave the game. Ganked before it was called getting ganked. Yeah. I was like, I'm not even mad because that's
like such a specific, ridiculous thing that like, like, you know, that's like telling to run pseudo RMRF, you know, like, Hey, you'll get the world. Like I'm going to give you some gifts. You just got to run pseudo RMRF. Yeah. Let's do it. Just like the things that like in games, also used to be based on like your processor speeds. Like if you try to play an original install of like Warcraft 2 on a modern PC and you try to like scroll, it goes all because it's like based on the like the processor speed and now the process is really fast. So that was their request animation frame was based. So yeah, if the RAF is faster, now the actual gestures faster.
Oh, I love it. Yeah. So a lot, a lot of crazy stuff. Yeah, we won't we'll save the the other what not for next time or way over time. Yeah. Anything you want to mention before we end? No, man. This Redwood Empire was quite good. It was. It's nice seeing you. I know it's every week, but I feel like it could be more frequent. I just some days feel so long and then the weekends always feel like a million things happened. It's crazy. Yeah. Whenever like like when the the whole next JS rewrite happened, I was like, oh my god, I like we should record today. Like I feel like we could do more like maybe we could do shorter more frequent episodes. I don't know, but yeah,
we definitely could do more. Oh, man. Yeah. Toby released QMD 2.0 today. His query markdown data format database system he has. I don't know if you've seen QMD, but that's pretty interesting for next time. All right. Cool. All right. Thanks everyone for listening. If you'd like to please subscribe. Leave us some comments and reviews. We appreciate it. We will catch you next time. You've been watching Whiskey Web and what not recorded in front of a live studio audience. What the fuck are you talking about, Chuck? Enjoyed the show? Subscribe. You know people don't pay attention to these, right? Head to whiskey.fun for merchant to join
our discord server. I'm serious. Like two percent of people who actually click these links. And don't forget to leave us a five-star review and tell your friends about the show. All right, dude. I'm out of here. Still got it.