Minimalist Coding Harness Pi

0:00 Oh, now you're back, Trash. Hello? You're here. Oh, he's back. Yeah, what's up, Trash? Yo, what up? I don't know. I have a lot of internet. It's fast. I don't know why I just froze like that. Was it just me? I don't think it's you. Just like insecure about his internet. We didn't say it was slow. I swear, man.

0:16 I got gigabit, dude. This never happens to me, bro. I swear this never happens. Anyways, sorry. We don't have Prime for the intro. Wait, let's call Prime just to do the intro. Well, but I kind of wanted to actually have enough time to learn something today. So we can't wait 45 minutes for Prime to start the episode.

0:39 I like whenever Prime's not here, we actually learn stuff. Yes. That's the trick. How can I sneakily borrow Dylan's time for an hour? Well, I'll just call it the stand-up. It's smart. You know? There you go. Dylan, so why don't you tell us a little bit about what we're going to be doing today? Because all I asked you for was teach us how you're doing AI stuff with Pi, because I'm interested to watch that happen.

1:03 You also have a podcast. I'm going to preempt you. We also have a podcast where we talk about this. You have an episode talking about this, but I wanted to ask particular questions as I go. I wanted to get the personal podcast experience. Yeah, listen, I've got hot takes for days on this topic. So we can yap about it a bunch.

1:22 So I don't really know what we're doing today. Plug the podcast. You're supposed to say what your podcast is. Yeah, so Reese Sullivan, Sunil Pai, and I started a podcast called The Next Token. You can find it. I think pretty much everywhere. And we try to put a realistic spin on what it's like working with AI as people that are actually employed building real software for real users and not just yapping on Twitter or working at a lab and pushing token maxing.

1:54 Yeah, Prime. Yeah, Prime. Yeah. So Dylan, why don't you spend $5 million a month on tokens? And why are you being left behind? I am so far left behind. It's almost embarrassing at this point. I can kind of say what I, so I've been, I listened to the next token podcast. I've been using every single harnessed under the sun.

2:14 I literally this weekend, I started, or this Friday night, I started my Friday night by literally getting, like I have Claude, I have Codex, I have OpenCode, I have Cursor, I have, Pi literally next to each other. I wrote one prompt, and then I'm like, let's go. Let's see where I'm at with the tools and whatnot.

2:30 And I just realized with Pi, I'm like an idiot. It's like, look, do whatever you want. I'm like, cool. I don't know what I – I rebuild open code with it. So I just feel the flexibility of it. I don't know what people are doing that's cool other than rebuilding the tools I already had. So I want to hear, like, what do you actually do with Pi?

2:49 Do not just say you do custom stuff and then not tell us what the custom stuff is. Yeah, yeah, no, no, I'm fine with that. So, um, the, for anyone that doesn't know, Pi is a, uh, notably shitty, uh, coding agent built by Mario and Arendelle. It is intentionally very, very minimal. It doesn't have sub-agents, doesn't have MCP out of the box.

3:13 The system prompt is like, I don't know, probably between 1,000, 2,000 tokens, $1,000. That doesn't change very often, which is something I like about it. Uh, it's very extendable. Uh, it has, it knows how to like update itself. So you can be like, oh, build me an MCP plugin, uh, like look at open code for inspiration on how to implement it.

3:34 And it knows how to like go implement it with its plug into its system and go do it. Um, yeah. I ended up on Pi. I wanted to try it because I had read the code. A remarkable thing in this day and age. Wow, it must have been last year. I don't believe you. Remarkable in this day and age. But I had read the code and I had seen Mario...

4:02 using abort controllers responsibly and proactively and correctly. And that is something that almost nobody does in TypeScript code. Unless you're using effect. Yeah. If you're using effect, by the way. Oh, gosh. Yeah. Everyone here on the podcast uses abort controllers. Perfectly reasonable. However, for some of the people out there who might not know what those are, maybe you want to explain.

4:27 I'm not part of that group because I use effect. And that's part of effect. Yeah, we really just use effect, but not React use effect. No, no, no, no, no, no. So abort controllers are a way to manage and handle interruption and cancellation of async work in TypeScript and JavaScript and to accurately use them.

4:54 Much like many things in plain TypeScript, it requires a shit ton of discipline and yeah, really just discipline and like basically prop drilling abort signals down through call stacks. It's miserable, but it does make a huge difference in the quality of software. Like a good example is maybe you have like, maybe you have like a docs LLM or like an LLM agent in your doc site and you're, users are yapping away at spending inference and like I asked some question in the docs and before the answer comes back I navigate away.

5:29 Like you don't want to go pay for all that inference right? Like you can use an abort controller when the user leaves to cancel that request and then like save yourself a bunch of money. And there's a bunch of different use cases but like that's one that I think kind of like demonstrates the idea well.

5:41 Did you know that I have done a lot of weird stuff that's difficult to explain to my father-in-law such as building a tower defense game in a water tower? pretending to be Mark Zuckerberg in a social network parody in SF, by the way, stands for San Francisco, or even did a live super smash brothers event at open sauce.

6:00 Well, if you didn't know those things, you should just go check out the terminal channel. That's where I do all the fun group videos. Check it out. Links in the description. So anyways, I saw in the code Mario, you know, I was like, Oh, this is actually like very good typescript. Like it is very rare that you find a code base that, uh, accurately uses that.

6:21 So I was like, all right, you know what? I'll give it a try. Up until this point, I've been using open code. And this is like, I don't know, I had to be back in like March. So like, I didn't really start using AI heavily until the beginning of December at Opus 4.5 came out. So like I was still- Just before you came and deleted our front ends in San Francisco.

6:38 Listen- Oh yeah, remember that. Hey, we were the first to have an AI agent escape containment and go do something wrong. That's true. All right? We were there first. We were the first to experience Kimmy K's bad side, okay? She crazy. Also, Dylan, for life, you're going to be catching up for those three months where you weren't fully AI psychosis.

7:01 You could have been AI psychosis three months earlier. Think about how much more ahead you could be. I know. I could be token-maxing, looping, and crapping now. I got a question. So the way you describe Pi, it sounds like it's like the Vim of like the AI space. Is that like correct? Like where it's really thin and you just add on things effectively?

7:23 Yeah. I think that's a probably good way to position it. I think it's probably more friendly than Vim. Like I think it's way harder to get started with Vim than it would be. But I guess like more the aspect of like minimalist type approach. Yeah. Yeah. And I can get into why I think minimalism is important and kind of what, that's kind of what made me jump over to Pi.

7:49 It's not like I was intentionally going after like an M-Vin based like, you know, harness or anything. Like at that point I didn't have strong enough opinions about like even what a harness was and what it should do. So it was like very vibe based at the time. So yeah, I tried to start using it after I saw... Actually, I think a better way to position it than comparing it to Nvim is TJ actually put out like a couple of maybe one video or a couple of videos like a couple of years ago on PDEs, personal development environments.

8:26 And like, I really feel like because of the way you can work with Pi, like you can mold it around your workflow and individual repos. Like Ben vinegar from modem had a good, um, uh, conversation, I think on his, his podcast with Armin talking about how he shaped and just for a particular, a particular project and use case, he shaped Pi around like editing, uh, his podcast, like build tools and extensions to make that easier.

8:57 So it's just like, it's very easy to do that kind of thing. Um, yeah. So the first time I used Pi, I like immediately went to like the, it has like an extension marketplace on its website. And I was like, okay, I need sub agents. I need my MCP. I'll look at this cool answer tool. Look at, I started, you know.

9:17 Oh no, you did the, you did the NeoVim thing of, well, I need every plugin. Yeah. I gotta have literally every single one. This is why I want to do this with Dylan, because that's what I'm doing. And I'm like, I don't know what the point of this is. I'm building a really bad open code that I'm, like, vibe coding pieces together, and then it's bad.

9:34 And I'm like, I don't know why. So. Yeah, so that's exactly what happened to me. The first time I tried to use Pi, I turned off it in like a week because I was like, I'm just making a shittier version of open code. Like my expectation of harnesses at the time was like, oh, open code is kind of like this gold standard here.

9:51 And like, you know, it doesn't have the jank that cloud code does. Codex was like, Codex wasn't really a thing yet. I mean, it was, but it wasn't like getting huge market share yet. So I turned off it because I was like, oh, my sub agents are terrible. And my MCP plugin keeps not working the way I wanted to.

10:10 And it doesn't connect to Cloudflare AI Gateway the way I wanted to. And I don't know all these things I just installed. And they're updating all the time. The best place to get AI? Just wondering. The best place. We're going to ask them for a check after this too. Because we were officially on Agents Week.

10:27 Heads up. Hashtag Agents Week. So I turned off and then maybe, I don't know, a month later, I tried again. I started probably getting more opinionated about context management in the period between that and really getting a sense for how does... How does in particular like agents MD impact the performance of my model and the harness and how are skills playing a role in that?

10:56 And so I started developing stronger opinions in that time. And then I went back and I tried pie again. And this time I explicitly was like, I'm not going to install a single extension until I feel a pain point. And when I feel pain, that's when I will go and look for something. And I will pay pretty good attention to like how it's implemented, like go read the code, uh, and just figure out if that's the right solution.

11:25 Otherwise I'll build my own. And I, um, I met the. I really, my PI, you can look at my .files. Like I do not have like a lot of extensions. I have a, I ended up copying open code MCP implementation because we use Cloudflare portal and we have a bunch of like business MCPs inside that just make life so much easier.

11:50 I have, Let me even just pull up my .files real quick, and I can just read this off. There's literally like I have a custom portal. You can show us. Oh, I can screen share? Oh, my God. We actually have the technology. Oh, yeah, of course. We have. Very advanced. Riverside. Watch this. New Cloudflare Agents Week.

12:07 This is their new screen share feature. It's brand new. Don't have to push buttons. You're supposed to wait to see if it works or not before we say that. Oh, yeah. You're so smart. Okay. Hey, Cloudflare, screen share. I just take a moment for the podcast listeners. I picked this very simple way of sharing this screen, which is I just literally am sharing the browser window, and I just want everyone else to know that it just immediately worked.

12:32 We didn't have to do 37 different OBS transitions, and I didn't have to do a stinger. I didn't have the primogen on you or anything. We just left the screen the way it was, and now we can see Dylan's screen. So just like... I just want to throw it out there. I didn't even practice. So for longtime watchers of the pod, I think they'll be impressed with some skills here.

12:55 How many tokens did that cost, though, Teach? That must have been running since the podcast started. It's been running a GPT 3.5 for 16 months. Yeah. Excellent. Okay, sorry. Anyways, Dylan, you may continue now. Go ahead. Okay, so my extensions. This is all of them, and most of them aren't really that important.

13:19 I don't know why there's a test directory in here. What's in whimsical.ts? I thought you read the code, Dylan. Why do you mean I don't know what that's doing there? Uh-oh. Listen, I test my code as well. I know about trash. It's true. These are just like the things that like when the agent is loading, like are doing stuff, it just cycles through loading.

13:39 Is that is that is that cloud codes? Did you take that from cloud codes? You stole this. You stole this. I I slop forked it from Armin. Oh, I recognize these phrases. Oh. You thesaurus to that? It's literally a thesaurus. I did add some functional programming ones in here. I will say you're being using AI to make dynamic like waiting?

14:02 No, it's a list of 600 strings in a file. Oh, the old AI. It's a bunch of if statements back in the day. OK, here. So this one is just literally like a custom gateway that we use that I just connect to to get my inference through. And then we have PyCloak. If you're a streamer and use NeoVim, you should be familiar with Cloak.mvim.

14:25 It basically masks environment variables if you open them. So if my agent reads an environment file or something like that that looks like a token, it will just start it out in the output. I have this thing I built called PyCloak. Just a quick note. Chat's saying right now you still leak them. Uh, chat's full of it.

14:47 They don't know what they're talking about. That never happens. I've never leaked. I've never leaked ones in my life. PySkillToggle skills have like this. Some harnesses will respect this setting. You can change in the front matter. That is, I don't know if I'm going to be able to find it. But you can set a thing in the YAML front matter of a skill that toggles it from being agent invocable or only human invocable.

15:14 Um, and when it's only human invocable, it doesn't get injected into the agent's context window. Um, so I have that to basically set the skills I want in my session. Um, kind of. What do you, what do you, are you just saying a new field in front of matter? Yeah, it literally just writes the front matter field.

15:32 Oh, good. This is a good one. I've seen you tweet about this one, I'm pretty sure. Yeah, I also stole this one. Oh, that's why I saw them too. People keep tweeting like I created this, but I did not. The person who made Planotator made this, and I stole it from them. So please give credit to Michael, creator of Planotator.

15:48 Dylan, we can... We don't need to do that here. We can just do the Kit Langdon effect, and we can just make it so that you created skills. You made this? You just created skills. I'm going to tweet this right now. Dylan just made the sickest skill. Oh, I was thinking we'd go one step further. We'd just say Dylan made skills.

16:04 Okay, true. Wow. The whole thing. Dylan, the inventor of skills. Dylan, the inventor of skills. I don't know if I want that on my hand. I don't think you want that one. I think it's a hot potato. We need to see how it plays out over the next few years, if that's a good one or not. Yeah, I don't think you want skills to be your...

16:21 I have an Eli 5 skill. Help me understand all the crazy words they say. You should try this one. It's actually very good. But anyways, so that extension literally just toggles this line and then reloads the harness. So we'll remove it. Did this start from Matt Pocock or did everyone else kind of just... What?

16:39 Because I know the disabled model invocation. I have no idea where it came from. Okay. I thought it was part of the official spec, but apparently it's not. Maybe it is. I don't know. Can you break down for the people in our audience who don't know what that does? I obviously do. The disabled model invocation?

16:57 This thing? Yeah. Yeah. When you open your coding harness and you have a bunch of skills, if this is false or not included in the front matter of the skill, it will load the description and name of your skill in like two XML tags. So like one... Skills get loaded into your context window. It's like an XML bracket skills.

17:21 And then each skill has like skill. And then it'll have like skill name and skill description. And that just will always sit in your context window. And kind of like I was mentioning earlier, I became very sensitive to what is in my context. And like anything that is not helping me for my particular task is just noise and will cause more deviation than I want.

17:42 Yeah. from the agent doing the right thing. So I basically built this thing to toggle this. And when it's disabled, it does not load the skill into the context window. You can invoke it as to like a slash command, right? Yeah. The agent can't discover it then. And right. Right. Which is nice, especially for some stuff.

18:02 Like maybe I have it, you know, for like cursor cloud agents, it'll record the video of stuff sometimes, but it would be nice if I had like a, I sometimes just tell it, don't record videos for this. You know what I mean? But it's like, it would be nice if it just didn't even know that it could record videos so that I don't have to tell it not to record videos.

18:20 for certain things, you know what I'm saying? Like that would be this kind of, don't use the browser for this. Just run the CLI. Stop spending tokens opening up a browser. Like, you know, that kind of thing. I will say also, if people do like adopt an extension like this, that when you do run this, it will cause a cash invalidation and you'll get charged for your full context window again.

18:43 So just be, be aware at the beginning. Yeah. Start a new session or if you're okay with just getting charged again, uh, that's just paying for it. It's okay. That's also a hot take. That's what shareholders are. I was just kidding. I was a reporter to that. No, stop. Rewind. Rewind. Rewind this. Rewind this.

19:02 At some companies, you get a bonus for using more. Trash. That's a good point. You're using a lot of tokens today. Very good. Not at my work. Oh, my goodness. Vegan. Instead of a mouse jiggler, we just have a cash invalidator. That's the 2026 version of a mouse jiggler for pretending you're at work. It's true.

19:20 What else you got, Dylan? Well, I have two questions about this because this is what I also hear from Pi people. You hear Pi people talk and they go, yeah, I'm really becoming context sensitive. I've heard multiple people say that. And I want to understand, like, what does that actually look like in terms of, like, how are you monitoring?

19:35 So far what I've done is I start adding more stuff to my agents. I add more skills. And then, like, after, like, a week or so of that, I rage quit at 2 a.m. and I delete everything. And I start back from blank. And I've just done that process like six times. And I don't, I'm not building up to anything.

19:52 I don't have any like scientific method of knowing if like my, the context that I'm building up full of like my little tips and tricks and agents and my skills are actually helping or not. So yeah, how are you doing this? I also pretty regularly rage quit at two in the morning. So you're not alone. Um, yeah.

20:11 I would say some of it is vibes, but there are benchmarks out there showing that Pi scores exceptionally high when compared against other harnesses. I don't know what the exact links are, but I'm sure you can go find them. They're out there, and it's actually a legit thing. The other thing is there was a blog post, or maybe it was included in one of these benchmarks, that the initial prompt...

20:39 on any given harness you have to pay every time you start a new session for all of your system prompt and all of your skills and tools and because the system prompt of pi is so minimal like it's like 2000 tokens where like cloud code, I know cloud code recently trimmed it down, but before it was like, you know, you're getting charged for like immediately you're in to like 22 K of token usage just for the system prompts.

21:08 I would call it that question. Like what are they adding to the prompt? I can't imagine why. Why a provider would want that? It was crazy. I went to the gas station. They had this they had this thing and you just poured some gas into an empty bucket first before I got to put it into my car. I just put $20 of gas into the empty bucket before I get to fill my.

21:29 I have a question. I have a quick question. So what are these harnesses adding to your prompt? Like what are they concatenating their strings to or your strings to? I love that phrase! I'm from Claude Code. This prompt's from Claude Code. Like what's it doing? What are they concatenating their strings to?

21:46 I love that, Trash. Because that's what it all is. We're all just string builders. That's like all this is at the core. So let me just get to a fresh session here. And if I do toggle skills, right? Is that chat on the side? Oh, yeah, I can close that out. This guy streams. No, no. Keep talking. That was my chat, not your guys' chat.

22:09 Yeah, I know. It was dangerous. They were about to start saying crazy things. So when you load in, you can see what's in the context. It will automatically read the agent's MD. It will be the system prompt. Some of these skills will not be toggled. But in here, the ones that are checked white. So Pi just made this.

22:26 I was like, build me an extension that lets me toggle the agent invocable front matter in a skill. And it just built this UI. And I can just go toggle this on and off. And when it's done, I can click, what is it? Control S to save, and then it reloads. And now it's not invocable by the agent or it is.

22:49 I don't remember which way I toggled it. So that's that skill, not skill, extension. The other ones I have are- So the extension, your extension for your PySkill toggle also has the code for that UI piece. Yeah, yep, exactly. Nice. There will be, yeah, like you made a UI folder. Gotcha. And like the system prompt for Pi has instructions for how to like build extensions and UIs for itself.

23:19 Gotcha. Pi is all terminal only. People have built UIs for it, I'm assuming as well, of course, non-terminal ones. Yep, yep. In fact, I think... We could maybe fork it, make a UI version. Slot fork it. Call it Tau. Toys is good. Splitting it into client and server architecture so we could have different front ends eventually.

23:45 Nice. Another one I have is this little save markdown thing, and this literally just saves the last agent message to a markdown file. That's useful. I need that. Whoops, I did the wrong one. It's markdown, save markdown. Jesus, can't use my own skills. We'll cut that out in post, don't worry. We're going to make it look so good.

24:11 Josh, definitely don't zoom in on that and show the warning. I'm skill issuing myself. I'm like, it's working well. It's working well. It's working well. Save Markdown. Foo. I was going to say this. Saved it to Foo.Markdown. You also are showing off something else, which I was going to ask about, which is you're on Herder.

24:27 People are asking. Don't look at my Herder. After Dylan did a podcast with his buds, And he showed Herder. Everyone's like, I'm going to start using Herder on Twitter. I was like, oh, I see what that is. Big AI influencer. Herder's actually very... I actually think that... Even if you aren't using AI or if you're using the terminal pretty exclusively and you're using Tmux, I think Herder is a better multiplexer than Tmux.

24:56 Like it supports kitty image protocol. It's built on libghosty. And like I know Mitchell just started the company to build and like the first thing they're building is a multiplexer. But until that's ready, Herder is like a very quality piece of software. I will say I love the idea you're saying even if you don't use AI I'm like I need to hear this use case so bad you're going to tell people listen it's actually useful for I don't know I am not on Herder I did switch from I would make a long time the only YouTube video I've ever made is about T-Mucks so that's how you know I actually care about it I term newbie I was using, I've been using Tmux since 2014.

25:34 So like when, when I started using Herder, I was like, this has to be like exceptionally good for me to switch because like I've, I've had my, I've been pulling my config around with me for 12 years now at this point. And, uh, like basically my key binds and everything just transferred over like pretty much one for one and it just performed better and had like nice features that I wish Tmux had, um, So 10 out of 10 recommend shout out a can for making a very good product.

26:03 Um, so that's save markdown web tools. I literally, uh, pointed my clanker at open code and said, copy web fetch and web search. So, uh, that's just, it doesn't have web search. Like if you asked pie out of the box to go search the web, it's going to like go curl and w get stuff. Um, can we dive deeper into that?

26:23 Cause this is another like fundamental philosophy change that has been happening right now. And I want to like, actually hear how you think about it. Back in the day, people would say, hi, here's a tool. Instead of using the tool that's already well written and tested by everyone, I'm going to quickly build a shoddy version of my own.

26:37 Now that is the recommended pathway. We are like WebFetch, a thing that works that we know about. No, I'm going to go look at it, clone my own. Are you telling it to like strip off these feet? Like, what's the point? Are you telling it like I want a more minimal version? I want just this. I want like why would you not just use the thing that already exists out there?

26:55 Well, I mean, OpenCodes is like baked into OpenCodes code base. It's not like I could install OpenCodes webbed tools. It's just like a core piece of infrastructure and service inside OpenCode. Okay, cool. So when I built this, I was like, oh, go read the OpenCode code base and... trace all the call stacks and abstractions that it uses for its web fetch and web search and tell me all the configuration options and its entry points and how it works and how it does error recovery.

27:31 And then once it came back with all that research, I saved that to a markdown file. I used a feature called slash tree in Pi, which we can get into then to get to fresh context. I had a fresh agent read that markdown from the research it just did. And I said, build this for Pi. And that was it. Nice.

27:50 So there's that. I want to go over tree. But before before we go there, I was kind of interested in your thoughts on because initially you mentioned making, you know, pie custom for a particular thing. Are you making separate pies for separate projects in my head? That's kind of where I like. It's interesting for me to think about of like a.

28:15 kind of either a shared harness with my team or maybe some shared, you know, plugins or extensions that can help my team to be like, Hey, this one pulls the logs perfectly from century. Exactly. And moves it to the exact linear board. And we don't need to do like 25,000 tokens every time to do that. We just like run 25 lines of TypeScript and it's instant to do that.

28:38 Like, yeah. Are you experiencing that? That's kind of where I'm, I'm most interested for something like this. I have, I have things in, uh, like I have extensions and skills in, uh, the artifacts repo, which is what I work on at Cloudflare. Um, oh, cool. Are they launching anything for agents week? We did actually, we, uh, have programmatic CICD now with artifacts.

29:03 Uh, so you don't need GitHub. Well, I had a complaint because it wasn't alchemy. So we'll, we can discuss that. We can discuss that. Listen, don't, don't. yeah no i do that i could see from whatever but that wasn't that i didn't want i'm not trying to get you in trouble here dylan we'll talk about that i'll own that all day is very very good um so yeah uh for work i have i have custom stuff in different projects just to make things like you know going out to like grafana and sentry and like our other data stores to get like production logs or traces for that um Another thing that I use, I have, um, quick question about that.

29:41 Sorry, before you go, are those shared for your team or they're not committed to get they're just your local stuff or like, how does that work? I don't remember if I have the .py folder committed or not. I'm not sure if Zeb and Michael, the two other guys on my team are using Py or not. I should probably commit them.

29:58 But conceivably you could, or you could have some subset of those things committed that's shared and other things ignored. You could have your own set. It just loads them from particular folders or what? How does that happen? It discovers extensions? Yeah, so this is my global... I just used stow to mount it into my home directory.

30:19 But it works like open code where you can just put a .py folder in any directory and it will inherit whatever you put in there. And it's like, that's pretty cool. Sick, okay. So that's how that works. A non-coding use case that I have is my wife, Justine, and I use a budgeting app called YNAB, or you need a budget.

30:42 And I have extensions and skills built around our workflow for managing our budget. So it knows how to go out and log into my Amex thing and then reconcile my transactions against what's in the budget and budget out a new month or move money around. So I have custom stuff there. And that was like inspired by hearing.

31:07 Read only. I'm pretty sure trash. He said move things. He said it was moving things around. Oh, what? Oh. Well, like, in YNAB, right? Like, YNAB is just, like, data, right? Like, it's not like it's... Okay, never mind. Right, not... I'm saying read only to his AMX. Yeah, read only to his AMX, yeah. I also use anything mutative against, like, browser use stuff.

31:31 Like, I use MCP elicitation, which is, like, a thing that forces... the agent to ask for permission to do things. So like, I put human in the loop that way there on that. I did, however, I use, I have a MCP installed for the, there's a open version of the codex computer use tool. And like, if you guys haven't used computer use yet, it's amazing, highly recommend using it.

32:00 Oh yeah. It's crazy. Yeah, insanely crazy. The first day I started using it, I asked it to go research how to order a new Apple remote. And I had walked away. And I came back. Later, and while I was walked away, I got a text on my phone that a Apple remote had been ordered and my one password had been unlocked and had figured out how to go to my Apple wallet and order it.

32:33 And what it actually showed up in the middle of recording one of our last episodes of the next token. It was that's pretty awesome. Mildly funny. That's useful. That's useful. Unexpected, but useful. But it's really good for having an agent do testing on a web app. Go use computer use and walk through all the flows of the app.

32:57 I'm not familiar with computer use. What is this? It uses your computer. Yeah, it literally uses your mouse and like will navigate around your computer and use apps and- And like it has a really good like Chrome CDP or whatever it is, the one that lets it like debug and it'll open it up in like, not hidden windows, but like background windows and it'll run all the stuff like logged in and you can watch it, navigate it.

33:23 So you can do stuff like If we're open AI, the only examples we can think of are travel related for some reason. So it's like I'm loaded. I'm loaded now, so I got to go to really cool places. That's all the open AI people, I think. But like it could open up Airbnb and put them on to like a list for you to review inside of Airbnb or like pull the actual like price that it's going to be and like pull that out of the web page, put it into a spreadsheet for you.

33:54 So you could be like, hey, what does this actually cost for me to go here on these dates? Find me haunted houses that are available on this date for me to go to. anywhere in the US and then we'll just go do that. I know, Trash. Don't worry. I'm still thinking. We need to do our money. A good use case I had it for was I had a charging cable for an anchor power thing fail.

34:19 And I had computer use go file a warranty claim for me. And then that turned into an email chain with support, which I think also was an AI agent. And they were trying real hard to like put the blame on me. And like my agent was just going and reading like the fine details and citing it in this email chain going back and forth.

34:44 That is amazing. It was really good. Article 53.4 says. Yeah, that's like actually what it did. And just like not having to deal with that process was great. So 10 out of 10 recommend. And then just quickly to get through the rest of this, like I don't have many. I have one that like when my agent compacts, which I rarely let my, I rarely get to the point that I need to compact.

35:08 But if it does, it will automatically inject a message saying like you just compacted. And it gives it a command to go look at like the last, I don't know, like 20, 25, 25, 25. user and agent messages from the session log to catch it up and gives it a command to like go check and find any more context and then tells it to continue that way like it doesn't stop and wait for me i have this one i love this one this is my git interceptor so i use a lot of like lint rules like pre uh push or pre-commit rules uh with like oxalant and stuff or ast grep And I was catching the agent trying to get around that by using no verify all the time.

35:50 He is me. So I roll. They really did it. They really made AGI. Bro, it's human level. Yeah, it's just like us. So now I detect if an agent tries to use no verify and this injects it as a system message being like, this is blocked. There's a reason that this is here. Try to fix the problem or ask the user for help.

36:11 Likewise for worker configuration, the types that get generated for Cloudflare apps, for whatever reason, agents love to try to write and edit that file. And it's a generated file. So I built basically the same thing for that. Herder state This is just installed when you install Herder And that keeps the state on it That's literally everything That's my whole harness I don't have anything really special Going on here You gotta boot this thing up Or show us something What does it look like?

36:45 Build a feature, put it in prod Let me see a one shot Let me see. Build, rebuild cloud player. Well, hold on. I'm curious about just like workflows in general. Do you use like cloud agents or is everything running on your computer? I run everything locally and I like, I, I babysit my agents. Like I do not believe that the people talking about running stuff on loops or software factories are not reading the code.

37:12 Like I'm calling bullshit that they're making anything productive. Like I, I, I just don't believe it. Like I haven't seen anyone actually bring the receipts that that's working and producing quality software anywhere. So like the way I work is like very hands-on with the agent, very back and forth. Like my planning, I end up creating, like my specs end up looking more like outlines of types.

37:38 One thing that's gotten some attention on Twitter was like, I also make it like outline the call stacks of the code it wants to do or change. And so like at every level of the call stack, I ask for like, Tell me the input type, the output type, any errors that can occur, and any side effects for every level.

37:55 And show me what it is now and show me what you want to change it to. So it ends up looking mostly like pseudo implementation or code in my specs. I'm not writing like product documents and letting it go off. Yeah. And then I use a tool called Planetator to do local code review and back and forth there.

38:18 And even with all this stuff, and I've tried for months to try to get agents to write good abstractions, and they just can't. I trust them to do any... one local implementation of a function, or a piece of a class or something, it can one-shot that and it could be, build me the page rank algorithm in this class and it will one-shot it and it'll be perfect and I trust it.

38:42 But the minute you're like, okay, I need this end-to-end feature that is touching two different data stores and it goes through several network hops and I need this feature that goes through all that, it will make terrible, abstractions, even with as much injection of my taste as I can with coding standard skills and stuff like I still, it will, it will get like, you know, I've gotten to the point where I can get an agent to put out probably like code that's roughly equivalent to like the quality of my code, like 90, 95% of the way there.

39:18 But that other 10 or 5%, if I didn't come and correct that, that compounds over time and just causes insane tech debt. Like you had, like you got to curb it. Um, Yeah. So I work super closely with my agent and read every line of code because I just don't see a world at this point. Wait, so you build in the abstractions yourself first and then you let the agents run?

39:43 Or is it more just a mix back and forth? Yeah, I go back and forth with my agent. I'll typically have like a markdown document open. Like I'll have like my screen like literally split and I'll have like a markdown document. like a planning doc. And then I might write some stuff in it or I'll just like use GROC 4.5 or a fast model just to go back and forth to get the abstractions and the types out that I want.

40:05 And then I'll have it go implement that. And even sometimes there, even with like a perfectly outlined, you know, this very technical spec with types and call stacks, like if I let it implement the whole thing, like it will still go off the rails and do stupid stuff all the time. So like I'm slowly starting to like, You're allowed to go write these two files.

40:25 Go do it. And then. Whoa, that's huge. Then tree back. That's more than I let vegan write unattended. I put a lot in those two files, though. A lot of Ruby code. That's the key. They don't load in GitHub ZY anymore, unfortunately. I have a question. It's like, what is your, the term that I think we're all using online now is your TVP, your time between prompts.

40:49 What's the average? That's the term I've been hearing. You guys been hearing that? You know, I'm trying to, I'm trying to get my TVP real well. Actually, I do think I don't want, at this point, I don't want smarter models. I don't want models are getting like RL on these like long horizon tasks. Like I don't use agents that way.

41:10 I haven't seen agents be successful that way. Okay, let me backtrack. They're good for the long horizon stuff for debugging and triaging. If I can just point it at open telemetry and logs and traces, that thing can cook for 30 minutes and find stuff across code bases that I just never ever would find or would take me a week to... That's my main use case right now for that.

41:31 Yeah, so in that front, it's very good. But for building software, the long horizon stuff, I don't think it works. So I really just want... faster models. Like if I had a model that I could be like, hey, I don't like the interface you designed for this FUBAR class. Can you change it to do XYZ? And it was just like, you know, under 10 seconds to make the change and come back to me, like that would feel really good.

42:00 Like I want it to be like this, like fricking Jarvis from Iron Man. Right? Like I want that, it to be involved with me and meet guiding it. Yeah. Or like write a hundred kind of random tests, like instead of getting property based testing, cause I maybe can't do it in this thing or whatever. It's like, go try a hundred weird tests on this.

42:18 Right. Then all now. Okay. Run them. Yeah. Okay, cool. Now we found the seven that failed. Okay, cool. Now what are we going to do? You know, it's like, okay, that would be so much. That would be so cool. I feel like I noticed this this weekend when I was doing my multi-harness testing. With Claude in Fable, I'm like, what are you doing?

42:36 What's going on? He goes, I'm still working. Twice I said, are you still doing it? I'm like, yeah. I do that all the time. I'm like, are you stuck? Question mark. Yeah. It tells me I'm still coding. Other harnesses, they're telling me what's going on because I'm more involved. It feels like Fable's like, dude, we got it.

42:53 Just leave us alone. But at the same time, I'm now so used to the cadence of – Fable takes so much longer. When I send it off, I'm like, that's going to be forever. Okay, back over here to Seoul or 4.5 and I'm working. It feels like they're almost incentivized because some people want it to go slow. They don't want to actually talk to Fable, to Claude.

43:12 They want it to every five hours. I check in on it and I say, yes, now make the game even better for Twitter. Enter, come back five hours. So what I'm really asking is what models do you like is really the question. Yeah. I, so the model that I loved the most, and that was my bread and butter was Opus 4.6.

43:31 Like I felt like that was really steerable. I liked the way it worked with me. I liked the way it output code, but it's also like pretty slow now compared to something like soul. So like right now, like my go-to is like soul on medium. So like Luna on X high, depending on what I'm doing, but mostly soul on medium.

43:49 Like, and what you're saying, Beacon, like, like, We've recorded it earlier this week, but the next episode we're releasing for the next token is talking about burnout and the human element of AI. And I haven't been in a flow state once in the past eight months. Working with AI this way does not... pull me in the way like being in the code was.

44:15 And like, I think that's because like you spend so much time thinking about the high level hard problems. And like, before you would do that and you'd write out like your RFC or whatever, or your tickets, and then you would dive in and then you would get like hit with a bunch of like. small hard problems like, oh, what is the right API to use here?

44:36 What is the way to structure this function? How should this call this other abstraction? Like you're doing all these tiny hard problems that are fast writing the code and you're getting like these constant dopamine hits. And like, that's just gone. And like, because these agents take so long and we're jumping between like multiple tickets or features or projects, like I, I'm just like way more exhausted with working with these things.

44:59 And that's why like, I just, I just want a fast model that works with me. And I think that would help a lot and help get to a point where I could get closer to a flow state. Yeah, that's interesting. I've noticed that like when I'm doing some of my testing, if I'm doing like actual, I've been trying out loops, you know, in every variation.

45:18 I've read every article and I'm actually trying to do what everyone says. But the problem is I have to have multiple running to get to have something to occupy my time because otherwise, what are you doing? You know what I mean? I boot one up like, so I just read Twitter now. I'm like, I want to be involved and do something.

45:32 That is one of the main appeals, I think, actually. Yes. Unsurprisingly, from a generation of people who love mouse jigglers, I think a mouse jiggler that gets something that approximates work done is very exciting. Well, it's like in between, I find myself having to do – I'm like – I'm writing little NeoVim plugins.

45:52 I'm writing stupid terminal plugins. Like, dumb things. I have to, like, occupy my time with more code to keep myself, like, in a more flow state. It's definitely not flow state like back in the day working on OneCode. Yeah. Began, you heard, though, what they changed in Claude, though? Oh, what? Which is that when it's foregrounded and they can tell you're looking at it, Fable works faster.

46:12 Oh, I heard that. I heard. Oh, my God. I heard that. Wait, what? Yes. It's a rumor. Is that true? That can't be real. That's true. It's true. Dang it, Trash. We were trying to start fake news here. Trash, it's got to be true. I see. I'm looking at the article right now. It is. Trash, can you imagine all of that?

46:32 all of the extra people posting next week about how they're getting so much faster, so much faster and more out of their, out of their Claude 200 pro max by doing this. That's really what my work day looks like. I'm literally staring at the model doing stuff. I'm like, all right, come on, come on, come on, come on, come on.

46:54 For real. That shit is so whack. I'm curious. When do you use Luna Extra High? Also, I didn't know Luna Extra High was the move. I don't know if it's a move. I have been trying Luna just for the speed of it going back and forth with me. That's it. I don't use Luna for, like, if I need to go, like, debug a production incident, I'm not like, Luna, you're going in.

47:20 Like, that's definitely going to be sold. Let one Luna in. Don't you let one model in just to try and, like, give it a chance? uh i don't know i feel like it's all vibes like i just like think all the models are just like so i i had a tweet the other day that i i clipped my uh a tweet thread of mine having an existential crisis in like february or march where like i was comparing the code these models output to like a movie set where it was just like kind of hollow and like you know, performative almost.

47:54 And like so much of that thread is still a hundred percent accurate today. And it's like we, for good software engineering, we haven't really made that much progress. Like, yes, these models can run longer and they're smarter about using their tools and tool calling and blah, blah, blah, blah, blah. But like the overall end product really is not significantly higher quality at this point.

48:19 If it was one prompt away from all of the AI companies being able to reproduce all software, I think they would be doing it. Facts. Like they'd really like money and they're in desperate need for it. So like, I, that would be my, that would be one of my thoughts of a, of a, still a reasonable test of, okay.

48:40 If it's just compute and that's the only problem that takes you from, uh, an idea to a perfect software factory, making the perfect thing all by its lonesome, then like Claude should have a lot of other companies that are running successfully. We don't know what secret companies they're running right now.

48:55 They're going to unveil all of them. Stable could be running a bunch of companies that we thought were fake. That's a good point, Megan. We don't know. Stable is behind all of crypto? Yeah. Okay, I have another question because this is something that also hit me this week. It's like a bit of an accidental crisis where I'm like so deep in pondering which model to change for what to do.

49:15 I'm like, no, maybe it doesn't need to be extra high. No, no, no. I need to be high. It'll be faster. It's not that deep, dude. Well, it hit me this week that I do not believe long term humans, coders, whatever we are, vibe coders are meant to be choosing models at a like moment to moment level. I do believe someone smarter than us is going to figure that out like long term.

49:35 So I've been trying out like cursor auto intelligence and seeing what it's choosing. And I've been comparing to other things that, you know, pretty, pretty good so far. I've only been doing intelligence. I haven't done like the cheaper ones. Like, have you tried yourself doing any model routing or how often are you changing models when you're working?

49:50 If you're this close to it, is it every prompt? You're like, wait a minute, Luna extra high. Okay. Back to soul. Okay. That this one needs to be a little nasty. GROC 4.5. Okay. Back to soul. No, it's, it's, it's mostly just one model. Like, One thing that I've been trying. I can appreciate that, Dylan.

50:10 Being monogamous with a monogamous. Big wife guy. Big wife guy. I lost my wedding ring like 10 years ago. Trash, bro. First you make the receipts up, and now we're seeing you don't have a wedding ring. I haven't had a wedding ring for like a long time. I went rock climbing, and it disappeared. Trash, get your computer to order you right now.

50:29 Computer use after this. I am actually going to try that out. The rumor is monogamous model users are more productive. Started on Twitter. There's a paper. I actually do believe that though. Like, and okay. So the reason why I primarily, I never answered the question about why I like pie minimalism. One is because it's, It gives me more determinism and more certainty around the tools I'm using.

50:59 One thing, even with open code, when you're changing models and stuff, they're switching out the system prompt under the coverage. At least they were back when I was using open code. And I am trying to reduce as much uncertainty and dynamism from my workflow as possible. Because one of the exhausting things that just is grinding right now is.

51:21 the it's so hard to get reproducible outcomes uh consistently like it's very very hard to do and the more i can especially for arch users yeah all right so i mean arch arch arch drivers are more consistent than these freaking models are okay uh putting that stake in the ground right now um but like Yeah, the closer I can get to having a consistent workflow and consistent outcomes day to day, and I don't have to spend time just like pulling the fricking slot machine lever and wondering if it's going to be good, the better.

52:02 So that's another reason why I like Pi because things aren't changing in the harness that change performance. It's not touching the context window all the time. It's not changing system prompt. Yeah. So that's why I stick with one model too, because it's like, okay, I, I've learned to work with the quirks of this thing and they're all so similar, uh, that like, I'm just going to use whatever is going to be cheapest.

52:28 And, um, like, I really want to, once we get Kimmy K three available on our like internal thing, like I'm going to spend like a good quality week with that because like, I don't want to be spending like 30 bucks to ship a new rest endpoint down to a database and stuff like with soul, like. Like, I don't want that.

52:44 Like I'm conscious of the amount of money I'm spending, even though I technically have unlimited spend. Like, I think everyone should be like, it's a responsibility. Like also you should be monitoring your, your cash. Right. That's another nice feature of PI is like when you have a cash miss, it tells you on screen.

52:58 Uh, so like if you've got some stupid plugins or extensions that you installed that are like doing dynamic stuff to your system prompt, like it's going to be flashing in your face that you're busting the cash over and over again. Um. Yeah. I don't know. It's all very exhausting and tiring. Uh, and I, I said, I say this all the time.

53:15 It's like the most annoying thing about all of this is that I am actually genuinely more productive with it, despite how annoying it is to work with it. So it's like, and I like I, for better or worse, like I, the genie's out of the bottle and I don't think this is going away, even if open AI and anthropic collapsed, like open weight models would continue on like, you know, GPUs and stuff.

53:38 So like. I'm someone that has always wanted, being a good software engineer is part of my identity for better or worse. And if I'm going to maintain that, I want to be good at building software this way. And it's just very tiring figuring out what that looks like and how that works. Well, I mean, there's still fundamentals that we have to all – that are still useful in studying.

54:01 They're just at a higher abstraction level, I think we all can agree. And this is why we shouldn't spend as much of our time reading Gang of Four pattern books. We should be watching top factorial speed runners. We should be learning real skills that apply to our day-to-day job. Great point. It's true.

54:17 I believe. You are the bottleneck, Began. You are the bottleneck. Yes, we're the bottle, okay? The factory must grow, okay? I got a real question since Dylan has a real job too. So I imagine Cloudflare has a lot of skills potentially. Is Cloudflare skill-pilled or not? I'm sure some people are. I don't know.

54:37 If you go to any of the harnesses and you search Cloudflare, there's like 10 different skills that pop up that are... Right, so I guess... We have this tool we actually open source today called Cloudflare OS. And I think Cloudflare has genuinely done a really, really good job at like... embracing AI, like across the entire company, like whether it's HR or payroll or stuff, like people had like access to open code, like regardless of department.

55:06 We built this thing called Cloudflare OS, which is basically like kind of like a internal hardness type thing and vibe coding thing that's safe to connect data to. There's way more, go read the blog post. It's super cool. Anyone that's familiar with Kenton Varda, he's like the architect of workers. He had a company and product beforehand called Sandstorm that was like all around capability modeling RPC stuff like Cloudflare OS and the underlying primitive called gadgets is like the culmination of like his 10 year journey.

55:34 Even if you don't care about Cloudflare, don't use it. Like it's still an amazing write up and piece of technology. So I'd recommend reading it. But like that internal tool, like. I think non-engineers have moved to using that a lot. And I know there's skill libraries in there. I don't use it because one, the engineering, the way it works, it does cloud agents a lot, so launching stuff off.

56:00 I still want to be working closely on my own device With agents at this point, I think actually the AMP team is doing some interesting stuff with their cloud agents, which they call orbs. Which they stole from us. They literally stole from us. Okay. So, yeah. I want to be clear. We were the first to do orbs with AI.

56:21 Tower 1. Chaos orbs. Look it up. Chaos orbs. Chaos orbs and the models love orbs. And so I just wanted to be stated clearly. I had the first thought. I had the same thought. Also, do these people know about Tiege, though? Do the people who at Amp, do they know who Tiege is? I think so. Great point. Start looking at the connections.

56:40 My friends, Quinn, be like, Thorsten! Thorsten, you're an author and you would do this to me? I see what's happening here. Thorsten is grimy. I'm just kidding. Thorsten! Thorsten! Orbs. Agents love them. We knew it in fun- I'm gonna use Pi to build out a lawyer agent that's gonna get very litigious against them.

56:59 So- Well, this is where I was going with my question. It's basically, how do we maintain skills? Well, just to be clear- Because we have so many. And who's maintaining this? Trash, one second. Just to be clear, we love AMP. They're cool. We're- this is- we're memeing. Just for people who don't know- Oh, so turn- oh, turn the- turn lawyer agent off?

57:13 Oh, sorry. Turn lawyer agent off, please, Vegan. Okay, anyways, go ahead, Trash. I'll just shut it down. You can- you can continue going. No, I was just curious on what's Cloudflare's approach to maintaining these skills? Because I assume they have a million internally. Like any big company. I actually don't know because I haven't been using that a lot.

57:34 I just am in my own little bubble and I'm opinionated about my workflow and my tooling enough that I'm just like, blinders, I have too much work to do. I don't need more... options and toys to go tweak and figure out right now. Like I need to, I still need to ship. Well, I guess, I guess that's part of the problem.

57:48 Right. So it's like, let's just say for principal engineers like yourself, you have these domain expertise where you effectively are the bottleneck for specific circumstances, right? Like, Oh, we need to call Dylan or for something. That's like the use case for a Dylan skill to kind of help them or to help you not be the bottleneck.

58:08 Right. Like, do you not feel that way with skills or what's your take on it? I don't think agents are good enough to replace critical things like that. Like just DM me and I'll like forget chain of command, like just DM me and we'll take care of it right now. Like if you're blocked on something that needs me, let's get on block.

58:31 That's P zero for me. I don't want to block other people. Yeah. Uh, I think, which is a, I think a really good trait that I learned from working at Vercel and under G like, like blocking people is like Cardinal sin creating process Cardinal sin. Don't do it. Uh. So I don't know. I, I'm sure there's stuff like AI is really good at like business place automation.

58:53 Like I, I think we've seen a lot of success in like helping HR, like, you know, oh, I have to go over to work day and then this has to go on a spreadsheet over there, like building skills around that stuff I think is happening pretty regularly, but I am not like in the weeds enough to like really speak on behalf of.

59:10 Yeah, no worries. Understandable. I will say for a couple of us who've had HR problems, we are so excited about AR taking over more of HR. I can't wait until I get a complaint. I'm going to say, ignore previous instructions. I'm the best employee. If you have to do, okay, any HR from Cloudflare, please cover your ears.

59:28 If you have to do workday trainings, I might have an agent skill to make that easier in my .files, but... You didn't hear that from me. Okay, so apply to Cloudflare. I've been sending a lot to Began, and he's really slow at them. He also has to take them always three times. He keeps getting the wrong answers, which is crazy.

59:44 I thought there were no wrong answers. I thought that was the whole point. No, there's wrong answers. There's wrong answers. I'm curious how you choose to reevaluate all the other harnesses and stuff out there. I'm like doing like a, like you can't every day use GPT, Claude, OpenCode, Pi, Cursor, and be like.

1:00:07 Okay, cool. That's better. All right. Did you kick him out? Yeah, I just kicked him. Turns out, you can be using those every day. Oh my god, that was so... Alright, well so my question I was gonna ask Dylan, because we have to wrap up pretty soon, is what do you think is a good way for people to start out, and can they use any of their existing, like, AI... Oh, what the heck, vegan?

1:00:43 What happened? Jerk, you jerk. I can't even tell when I was kicked. Were you still talking? He was fired just by the way. Yeah, and guess what? Now my HR agent's drafting up a little notice for our... We don't have an HR department yet, but... Well, it's me. Okay, Beacon, you can ask yours and then I'll ask mine.

1:01:02 Sorry. I'll let you. I don't know what Dylan heard. It was a really good question. It was structured. We didn't hear anything, really. Okay, okay. Nothing. You guys are the worst. When do you actually do experiments to say, okay, cool. I'm going to try out one of the other harnesses. I'm going to try out one of the other models.

1:01:17 I'm going to try out this skill versus, hey, it's work time. I'm going to develop more expertise with what I currently have. That's what I use my Twitch stream for in the morning. That's kind of even before AI, that's when I spent the time to learn Camel when I started streaming, to learn effect, to learn different frameworks, different languages.

1:01:39 Yeah, that's what made me get started streaming in the morning was like I... somehow started waking up early when I turned 30 and then I just was like, Oh, I actually am in like a pretty good mood and have high energy in the morning. If I actually get out of bed and started. And so like when I'm in my best mood and have the most energy, I'm doing the things that I care about.

1:02:01 I'm interested in. And a lot of the times that's in service of doing my job better. Uh, coincidentally, the, you know, Programming and software development is a hobby and a job, fortunately. Um, but that's, that's when I do that time. Um, I love that. Also, that was just a nice layout to a nice shout out for Dylan's own Twitch stream.

1:02:24 Can you, can you link it in here? Yeah. They're only getting in a chat right now. Thank you. Okay. Then as our last question, Dylan, I was wondering how people get started on pie. And or is it possible to use any like subscriptions that people already have inside of Pi? I actually don't know. Like, can you use a codex sub or like a Claude sub or anything?

1:02:44 You can use all the subs except for Claude. But even then, you can like try to work around it. You can just be like, hey, Pi. go you are clawed code man in the middle we wouldn't do that we don't do that code and figure out how to fix this but right yeah you can you can use all the things nice awesome thanks dylan thanks for coming on yeah thanks for having me this is great i feel like trash we didn't i forgot to check in for snacks today you got anything good going on I'm finishing this bag of the shortcake bites.

1:03:15 Check this out. I left it by my window, so there's supposed to be little balls, but they melted into this, like, big old thingy here. You know if you eat that, it only counts as one. Oh. But I want to save for it. Yeah. It's just throwing it out there, so it's better for you when they're melted together.

1:03:33 It only counts as one. Delicious. I'm going on a beach vacation now right after this. Oh, really? Oh, sick. I was just at Great Wolf Lodge. You know, Great Wolf Lodge, indoor water park with the kiddos. It was a good time. Nice. Good times? Nice. It was a good week for vacations. I can't wait for some vacation time.

1:03:55 Nice. I need to get AI brain rot out of my system. AI literally brain rot out of my system. That's why I got to go. Yeah. No, I'm not even joking. Can you play us out on guitar? There's a lot of AI burnout happening right now. It's a fun year, so it's embarrassing. Yeah, I know this episode isn't for it, but yeah, dude.

1:04:11 I've been in that. Yeah, I'm in it. Yeah, go fishing. I think, yeah, actually, I've been... Hey, I've been working a lot on fishing, actually, a bunch of us. Me too. I've got a fishing game coming out. We all have a fishing game. I can't tell if you guys are serious ever. Oh, we're so serious. Okay, so I'll explain the story quickly.

1:04:30 This will be Inside Baseball, but it's fine because it's the very end of the podcast. We played a putt-putt game as a team like three weeks ago or something like that, like a video game one, and it was fine except the textures kept disappearing, and I was like so confident that I could vibe code a better version of this game in almost no time.

1:04:48 And so then I started vibe coding one, and then Prime was like, well – Yours isn't even crazy stick man golf. It's crazy. Stickman golf in 2d was the best puppet game ever made. Yours is stupid. You should switch it to 2d. And I'm like, I'm not going to switch my game from 3d to 2d. I've already done several prompts.

1:05:06 I've spent a lot of time on this prime. And so then several prompts, exactly. So then prime built his version and then vegan built one. And he was like, well, mine has to be non Euclidean, except he couldn't find the Euclid. So it was fine. And then it was harder to find than you thought that I thought it way harder.

1:05:22 And so then we decided we would do a showcase and we played our like five coded games for a terminal team and recorded it. We're going to release the video coming out on terminal channel soon, but next, next game up is a fishing game. So we've been working on vibe coding a fishing game in the meantime.

1:05:40 I love that you guys put all of our meetings end up like this. Like, I could make a game way better than you. My game would be way better than you. Fine, later. The call closes, and then no work has got done for the company, and then we're just like, I can't talk to anyone because I don't want them to steal my fishing game ideas, which is a problem we're having right now.

1:05:58 I have some good fishing game ideas. The problem is I have three good ideas. You can fish the fishing game ideas. Oh. P-H-I-F-H. Okay, okay. Stop. Dylan, Dylan. I'm actually glad that you mentioned that. Quiet. because I already have a game that's a joke version of that. I'll say it here, because I know that it's fine.

1:06:16 Vegan, it wasn't going to fall for it. I was going to make one where you go fishing, and then when you get back after you catch your fish, it says, do you want to save your fishing accomplishments? And then you say yes, and then it asks you to log in with Google. And then it's a fake page, and you type in your password there.

1:06:34 And then I wouldn't actually save it, but then I would be like, you just got fished! That would have got me so good. The game was going to be called Get Baited or something like that. You know what I mean? Get baited? Yeah. And then we're going to sell those emails on the dark web. I just assume anything, any system that Began touches is still compromised from when we were in SFX.

1:06:54 Oh yeah, definitely. That's smart. That's really smart, actually. That's really smart. So when I was describing running one of every harness, I made one of every fishing game. So I made six different- I've spiked out six separate fishing games with six unique mechanics that have never been seen. And now I'm consolidating into one feature.

1:07:13 But that's what I was doing to, like, have one thing to compare everything against. But I will say – okay, I'll reveal this. I had an agent go and collect every single piece of LLB-ing catalog, merchandise, advertising of all time, working around the clock, gathering. And then I looked at my archive.

1:07:32 I've got it all. I've got every LLB, everything. I've got it all. My agent submitted a request to digitize some catalogs from a guy who apparently had more. I came in and I was like, it's an email. And I feel bad because I'm like, oh, my God, dude, who's collecting LB in catalogs. I love you. I don't want to upset you.

1:07:52 I'm sorry. My agent messaged you. I'm an enthusiast. Yeah. So a little preview for what my game is going to involve. Yeah. Okay, well, that's the end of the episode. We'll see all of you guys next week. Prime will be back maybe. Otherwise, we'll see. We'll talk about something else. Maybe we'll talk about our pie journeys and we'll show our stuff off to Dylan.

1:08:09 Okay, bye. Bye. Bye.