Read_933 - The Secret to Vibe Coding

[00:00:00] Speaker A: The people who figure that out first won't just ship faster. They'll build things that the others can't even spec because the spec emerges last from a process that only exists in the doing of it. The best in Bitcoin made Audible. I am Guy Swan and this is Bitcoin Audible. [00:00:41] Speaker B: What is up, guys? Welcome back to Bitcoin Audible. I am Guy Swan, the guy who has read more about Bitcoin than anybody else. You know, we've got a really, really cool one today. So I have been vibe coding like a maniac and working with my open claw agent and it has truly, it is without a doubt, there's a lot of, oh, this is like revolutionary. And I see the path to a one man billion dollar company and all of this stuff. And there's good stuff. There's. It's absolutely a powerful tool and it does feel like a real unlock that I can do things. And this feels very different. This feels like the first AI tool that's truly different from just chatting with AI in a chat window. And so I don't want to discount that because it's insanely powerful. And I went out and just immediately bought a Mac mini to put OpenClaw on to start testing this out. And I already wouldn't go back. I'm not like, this is an incredibly valuable tool. I feel like I have a secretary to finally take it, take care of and manage a bunch of different things about my setup that I have never [00:01:50] Speaker A: really had under control. [00:01:52] Speaker B: And, and for the first time, it actually feels like I largely do have it under control. And I just hope that this continues [00:01:57] Speaker A: to scale and stays in line with things. [00:02:01] Speaker B: But it's already doing a heck of a lot better than I can by myself. So I'm just gonna be figuring out how to make this better and better. And that is a very, very powerful tool. But there are a lot of things that I think people are saying. They're like, oh, it just, you know, went out of its way randomly and wrote a newsletter for me that I kind of think is BS because my agent has never done this and it's never acted in the way that some of these people are claiming that it acted. And what's funny is that, like, I don't know why you would have to, like, why you would make that up, because it's still, it's amazing the things that it can do and the things that you can do with it without it, without needing to make those things up. So I don't know what the context is or what they've done with their [00:02:38] Speaker A: setup that might have led to that. [00:02:40] Speaker B: Or what about their conversation? Because I watched a video that some guy was saying this and it was like he'd had it for two days and I don't know how I would have led that to happen within two days of me actually having my bot up and running. But regardless, I thought this article specifically was just really powerful and in framing what the unlock is like, what it is that's changing and what role the [00:03:05] Speaker A: human has when you can just spin [00:03:08] Speaker B: up agents and you can put an AI to a particular task or to a particular hierarchy in a stack of [00:03:15] Speaker A: tasks and rules and purposes and this [00:03:19] Speaker B: was just a really, really powerful framing. It's really short. An average bitcoiner on Primal actually reposted it so you don't have to go to X to read it if you don't have an X account or you know you're trying to give it to your bot to see and you the API doesn't let you access. So I'll have the link to that in the show notes as well. Shout out to him for sharing that in the Bitcoin and AI signal group. Actually a huge thank you to the HRF and their amazing work as well as for sponsoring and supporting the show. They have the Oslo Freedom Forum this year from June 1st to 3rd. This is one of the best places, if not the only real place, to really meet with the people fighting for freedom around the world and to be inspired in a way that few places and few conversations will allow. Tickets on sale now? [00:04:11] Speaker A: There's a link right down in the [00:04:12] Speaker B: show notes and don't forget to check out their Financial Freedom newsletter if you want to stay up to date on the news around all of these things and the tools for defending freedom around the world. With that, let's get into today's read [00:04:24] Speaker A: and it's titled the Secret to Vibe Coding and the Only Skill that Matters in the Age of AI Written by Jesse Posner the most important new skill in software engineering isn't prompt engineering. It's the ability to extract process from practice in real time while the practice is happening. I've been building with AI since the day ChatGPT was released to the public. But something changed in the last week. Across nine repos and 500 commits, that feels like a permanent unlock. A methodology emerged. Concurrent agent work streams managed by a meta layer. Specific scaffolding patterns, verification protocols, prompt templates. None of it was planned. All of it was discovered by doing the work and then naming what happened the act of naming is the skill, the shape. You build something, it fails or succeeds in a way you didn't expect. You name what happened, not a vague retrospective, not lessons learned. You name the specific thing. The agent skipped the design loops because the prompt framed them as suggestions that naming produces a pattern. Prompts must make critical steps into hard requirements with explicit checkpoints. The pattern becomes a choice you make going forward, the choice becomes process. And then the process itself is subject to the same cycle. This is not reflection. Reflection happens after. This happens during you're building and watching yourself build at the same time. And the watching changes the building. The Greeks had separate words for this epistemi knowledge you can state techni knowledge that only exists in the act of making the practical wisdom to know which situation calls for which response. What I'm describing is all three at once, collapsing into a single recursive act. You are writing software, designing the process for writing software, and evolving the process for designing processes all in the same afternoon. The scale the meta process is too [00:06:58] Speaker B: complex for one person to hold. [00:07:01] Speaker A: Right now, I'm working across five TMUX windows, each with multiple panes. An automated agentic benchmarking lab runs for weeks. In one, a kernel fix feeds into a brainstorm for the next feature stream. In another, a research agent is reading documentation. In a third, a planning agent is structuring the next phase. In a fourth, a flex pane sits [00:07:22] Speaker B: ready for code reviews as the other streams produce PRs. [00:07:26] Speaker A: The meta agent in each window helps me track what's happening in that space and when I need to see the whole board. The strategic AI visualizes the layout, maps the dependencies, identifies the critical path, and tells me what to spin up next. This is not just writing code. Every one of those pains might be brainstorming, researching, planning, or meta planning the full cycle of building software. What should we build? What does the landscape look like? How should we structure it? [00:07:56] Speaker B: How should we structure? [00:07:57] Speaker A: The process of structuring it can each be delegated to their own agents in their own spaces. The meta agent helps me oversee the entire workflow, not just the implementation. The cognitive load isn't coding, it's orchestration, context management at a scale that exceeds human working memory. I can name patterns as they emerge. What I cannot do is hold the state of every active work stream while crafting the next agent prompt, while remembering that a different stream shipped code. I haven't reviewed yet. The Partnership so the meta lane has a collaborator, a strategic AI that holds the context. I can't. When I need to send an agent into a new work stream. I don't write the prompt alone. I describe the intent. The AI drafts the prompt structure, tells me where to place instructions in the document, explains why. When that agent is running and I have a gap in my attention, the AI offers options. Think through the strategic layer or check on a different work stream. When I say let's check on that other lane, it reads the code, runs the tests, summarizes the architecture and tells me whether the lane is done. I never loaded that context. The meta lane carried it. The meta agent also runs check ins on the meta process itself. What's working, what isn't, where do we course correct? Because its context is kept purely strategic, never polluted with implementation detail, it can manage complex projects over weeks and months without losing the thread. And when context does run thin, we handle it explicitly. Clean handoffs before compaction, markdown logs of progress and decisions, memory files that survive session boundaries. And the meta process has its own meta process for persistence and cohesion. The recursion. This is the recursion that matters. You are using an AI to manage your process for managing AIs. And that AI maintains its own continuity through deliberate context engineering. The human provides direction, taste, and the judgment calls that only come from using the product. The Metalane AI provides working memory, context management and a structured surface for decisions to happen on. Neither side is complete without the other. The human without the metalane drowns in cognitive load. The metalane without the human optimizes for the wrong thing. The system is the relationship. [00:10:34] Speaker B: Every concrete pattern I've found is an [00:10:37] Speaker A: instance of this relationship at work. A spike proved the technology worked and simultaneously proved we didn't want it. The insight isn't do spikes. It's that contact with a working prototype produces knowledge that specifications cannot. An agent said all tests pass when it was 90% done. The insight isn't verify your agents. It's that most people building with AI cannot distinguish a real result from a plausible echo of one because they're looking at the self report instead of the artifact. A single context window couldn't hold the why and the how without one compressing the other. I watched intent erode into implementation token by token. Named it Vision Compression. And the name produced the architecture. Separate them into different context windows. Each pattern emerged the same way. Build, encounter, surprise. Name it. Let the name produce a principle. Not designed, discovered and discovered in partnership with an AI that remembers what I said three hours ago and asks the right question when my attention is free. The process is a product. It evolves through Use, not through design. You do the thing. You name the thing. You look at the pattern. You name the pattern, you choose the pattern and then you hold it loosely, because the next round of contact with reality will reshape it. Named problems become workable. Unnamed problems become technical debt with philosophical implications. The difference between building with AI and being built by AI is whether you can name what's happening while it's happening and. And that naming is only possible when something is holding the context. You can't. The distinction between writing software and thinking about writing software has collapsed into a single act. And that act requires a partnership between human judgment and machine memory that neither side can perform alone. The people who figure that out first won't just ship faster, they'll build things the others can't even spec, because the speck emerges last from a process that only exists in the doing of it. [00:12:54] Speaker B: Do not forget to get your tickets [00:12:56] Speaker A: for the Oslo Freedom Forum. This is the place. This is the center of the world [00:13:02] Speaker B: for talking to the people fighting for [00:13:04] Speaker A: freedom, for hearing the stories of success, [00:13:06] Speaker B: the stories of failure, the warnings for [00:13:09] Speaker A: the Western world to learn about the tools for fighting back and forth and honestly to be inspired to go out and do something for the fight. [00:13:17] Speaker B: It is June 1st to 3rd this year, 2026. The tickets are on sale now. [00:13:24] Speaker A: You can find a link down in [00:13:26] Speaker B: the show notes directly to it. Check it out if you haven't. [00:13:29] Speaker A: And a shout out to the Human [00:13:30] Speaker B: Rights foundation for the amazing work that [00:13:32] Speaker A: they do and for putting on this event. [00:13:35] Speaker B: All right, so I wanted to read this, and there's actually a piece by Max Hillebrand that I may actually just do as a follow up to this one, because it's very much in line with this same idea really well, it's kind of the more philosophical take on what's going on, rather than the actual nuts and bolts of describing the process and kind of like how work or how the creation process is evolving. But absolutely, this is. It's crazy how. How much things are changing and how quickly. And I've been trying to kind of stay on the front edge of this in a loose sense, like I'm. I haven't spun up a bunch of different agents, although as soon as I read this article, I pasted it into my agent, into Jarvis, and basically got a breakdown and assessment of exactly how this, how Jesse was doing this, and what tools we have at our disposal to implement this. Because one thing I didn't. I didn't quite realize with quote unquote agentic like working with like, like using this kind of like meta agent setup or model is that you can have completely different context windows. And that is not. I didn't recognize that that was really the value of this is you can have a context context window that's specific to a certain intent or a certain [00:15:08] Speaker A: model or a certain task in relation [00:15:13] Speaker B: to the overarching picture of what you're trying to accomplish and when, especially when you have like I have my quote unquote agent right now which does a ton of various tasks, right? I've got one that's just managing or excuse me, I have it managing my to do list. I have it managing reminders and taking. Keeping track of and reminding me of the major things on my calendar of reminding me to contact people back and, you know, which things are overdue, keeping me up to date on like what's coming up in the next two weeks, keeping track of my reading lists of my bookmarks list in relation to projects and what sources I copy down. One of the things that I did was I was constantly taking pictures of business cards and information or social media accounts for people that I met at [00:16:06] Speaker A: Plan B in El Salvador. [00:16:08] Speaker B: Because one of the things that I constantly do is I'll meet a ton of people or talk to a ton of people about all these different various projects or things that I'm interested in or business that they're running that might be relevant or people who want me to get in touch with them again about Pear Drive. [00:16:24] Speaker A: You name it. [00:16:25] Speaker B: There's just lists, all sorts of things, and I always try to mark it down and try to record it in some way in some note or something, but it's just gone. It's just gone. Every time I do it. I do this at conferences and I'll remember one or two good things and then there will be 30 other things that I recorded down or people I was going to talk to or things I was going to meet. I can't tell you how many times I've had a conversation with somebody that like, oh, we should do this project or this could be really exciting. We could get back together and we could work on this. And then it's just lost. And because I am just not organized enough to keep up with the cascade of various things. And importantly is that it's never in one place. You know, there's never one tool to fit all of the things. And that's fine. I don't want one tool to do all of the things, but I do [00:17:11] Speaker A: need to see it in one place, [00:17:14] Speaker B: you know, like I need to see [00:17:15] Speaker A: it all in context. And so while I'm doing this, I [00:17:18] Speaker B: might be like coding with Jarvis or working on some sort of a project. And then I'm like, oh also save this ready so that I can check into it later. Or check, right read this article and pull any useful information from it. Or you know, save this to my to do list, like what's up, what's coming up on the calendar, et cetera, et cetera. So it will. I will constantly mess up the context of the project we're working on with a bunch of these like side little tangent things. And I've gotten to. And I've gotten into a habit and [00:17:56] Speaker A: I believe it is working better or [00:18:00] Speaker B: it is more aware of maintaining the stream of thought for what I'm. What we're like the core thing that we're working on. [00:18:11] Speaker A: Because I will explicitly say short tangent, [00:18:14] Speaker B: do this, update this thing or whatever and then we will go back to the kind of the more the project that we were working on more directly. But now I'm realizing after reading this article and kind of like digging into it a little bit is the thing. The real way to manage this is to have separate contexts, context windows for each of those things, separate AIs and [00:18:36] Speaker A: have them specific to which one does the best job. [00:18:39] Speaker B: So that my meta agent, my main Jarvis can manage and update the information for the one that's managing my calendar and to dos and reminders and maybe my bookmarks lists and all that stuff and how it's related. And then I have one that's specifically built for a coder, one that is working on Viber coding projects and you know, is more geared around the engineering and the design, like holding to the design principles and things that we've laid out. And then JARVIS is you know, connecting them in the sense that like they're over basically organizing and orchestrating these two working together so that I am not constantly messing up my. The Vibe coding context or the specific project context window with a bunch of reminders or to DOS or calendar items that have nothing to do with it. So I hadn't. And I didn't even realize that OpenClaw was kind of built for that already, that just creating sub agents is just kind of part of the process. And JARVIS has already just been like, well why don't we just create agents for these things? Because this is kind of the map of the things that we are working with. And now I just really, really want to see. This is also like a huge thing for okay, maybe this does make sense to local run a lot of this stuff because I can get a lot of benefit by having my to do and reminders and calendars app be local, to have the one that saves all my bookmarks and contacts, information and stuff from like a conference be local and private, and for my orchestrator to be local and private, but then my vibe coder to be CLAUDE code, you know, to reach out to the much bigger models that are going to produce better results, specifically only when they actually matter or when they produce the best results, and I actually need them. And he's already actually proposed which models would be best for which thing, because, you know, you might have one for mission you want one to maintain, like, okay, are we building the right direction? Are we thinking about this in the right context and all the things that you actually want to accomplish, Is this the way to do it? Is this. Is this project in line with your. Your thinking to make something that's simple and modular and reproducible? Or are we over. Are we making this overly complex? And that actually brings me back to another point that I was just talking about with my brother yesterday with PEAR Drive. Paradrive is a really big project with a ton of moving pieces and a lot of things to work out. And I keep going back to, okay, I need to build a simpler and kind of one off, like just kind of proof of, okay, how can we use this? And what are some of the things we can do with it? And in doing so, I have been trying to boil down some of the benefits of PEAR Core with just the general pair stack into pair drop, essentially an airdrop replacement that doesn't require you to be like, right near the device that you're talking to and is also just super simple, right? It's just gonna work. You just share a link and you can download a file. And using Paracore, I built this basically in a day. [00:22:04] Speaker A: And it's actually pretty great. [00:22:06] Speaker B: And it's funny because I specifically told Jarvis, or excuse me, here's the thing. So I coded a lot of it with CLAUDE code in its own session. And part of me has been regularly going like, okay, am I actually getting the value out of Jarvis? Because I bought a Mac Mini for this to have it on its own machine. And there are definitely some things that are really great, but I will still go back and just have a session with CLAUDE code. And I sit there and I wonder, like, is it that much better? Like, what's the differentiator? [00:22:35] Speaker A: Because it does feel different. [00:22:37] Speaker B: But I wanted to make sure that I'M actually being honest with myself for do I actually need Jarvis or is CLAUDE code? Is a session with cloud code in the terminal going to give me basically the same results? And so I actually coded with CLAUDE code for some time, but then I realized that I was making Pair Drive again. I kept adding things that are, trying to add things that I really wanted or solving problems that I just shouldn't be solving at this stage. Like trying to create a share environment without needing to have a link and, you know, creating like a permanent relationship and creating spaces, all this stuff just, just trying to do too much at once. And like, I still haven't even just done the simple thing, but I've been working on this session and, you know, trying to get things to work and sorting out bugs and processes and trying to create new rules and in like, how to work going forward. [00:23:36] Speaker A: But I have a ton of this [00:23:37] Speaker B: context and I have done all of [00:23:39] Speaker A: these rules with Jarvis, but CLAUDE code [00:23:41] Speaker B: doesn't have it right. CLAUDE code is fresh session every single time. And so I've been doing this with clog code and I'm trying to figure out if the direction I'm going and the way I'm building this is really the right way. And I was like, okay, this is the perfect example to take where I am right now and what I'm working on and bring that context into Jarvis and see what Jarvis says about it, see how my agent treats it differently from CLAUDE code. And I was beginning to think that I had taken a wrong path and this was leading me to bloat and complexity that didn't need to be there and that I needed to simplify something. But I didn't know exactly how or what I should focus on. And I did not include this context in the conversation with Jarvis. I just kind of like took, okay, here's the map of everything that we have and I pasted it in. But importantly, Jarvis has the context from our dev room. So I have my open claw bot in the room with the developer group, with the people that are working with Pear Drive. We are building the core and the desktop app and the mobile app, everything. So it is just kind of copying stuff down. And one of the rules and process [00:24:57] Speaker A: routines that I gave it is to [00:25:00] Speaker B: basically note down anything significant or decisions we make or, you know, new releasing new tools that we discuss as options for accomplishing this, that or the other, and record this down in checkpoints every two weeks, like, basically summarize the past two weeks. And whenever we have a meeting, I always record it and then transcribe it with Whisper and then I give it to, to my agent as well. And then again it pulls the to do lists, the, the process for the next week, the decisions we made, the, you know, what worked, what didn't, what the progress was, what we had to backtrack on, etc. Etc. So it's basically creating this map of everything that we've done and keeping that in context. [00:25:46] Speaker A: So Jarvis has this. [00:25:48] Speaker B: And when I took what I had coded with CLAUDE code and put it into Jarvis, Jarvis's response was the exact same thing that I was beginning to think says, it looks like you were trying to do too much. You were trying to build PEAR Drive, when the whole discussion and everything that you talked about wanting to do was [00:26:06] Speaker A: to make PEAR Drop, which would be [00:26:08] Speaker B: much, much simpler than Pair Drive, and to basically get a foot in the door and to get people to see that you can do this and to make a product, get a product out there that basically leads the fact that you're going to have Pear Drive. And that does. The one thing that's really useful is that it just delivers files from A to B reliably, easily, with interruptible connections, with a simple link and without a platform, without needing a website or anything like that. It just works device to device. And it made a suggestion of, based on your conversation in the last meeting that y' all had, when you talked about this at length. And then I said I should essentially remove of the seven or so things I was trying to do, I should remove these four or five and boil it down to just two or three and accomplish those tasks reliably and understand. [00:27:03] Speaker A: This was the first response. [00:27:05] Speaker B: This was the first output of Jarvis after a day's worth of coding with CLAUDE code. CLAUDE code didn't give me any of those suggestions. It didn't really care or have any context about my mission or why I'm doing things, whereas Jarvis explicitly has that context and is applying that context during conversation or during design decisions. And it completely changed the course of the afternoon. And by that afternoon, within, within an hour or two, we're using. In fact, I just sent a file to Jarvis this morning and I did it with a PAIR Drop link. Like, it's working and it's working great and simply. And I'm still gonna add like one or two little things just to. Not just because, you know, you might wanna share a link with one person and a different link with another person. And you need like a history, you need a list of those things. Cause right now it is Very, very simple. But I also had Jarvis very quickly create like a little command line side tool for it so that Jarvis can very easily. It doesn't have to like, look at the interface, right? It can just very easily send and receive stuff to me. And now literally all of the new versions of Pair Drop are pair dropped to me. And then I am pair dropping back any edits or changes or things that I'm doing. And we're literally. It's just Pair drop link after Pair drop link. And because of that, we're not blowing out the context of the Telegram chat. And the simpler this app is, the easier it's going to be to really quickly develop an iOS app and an Android app and test all of those things and make sure everything's working together [00:28:43] Speaker A: and then just get it out there for people to test, to play with. [00:28:47] Speaker B: While I continue to build out things for Pair Drive, which will be compatible with pairdrop, I will be able to do the exact same thing. I will be able to give a [00:28:56] Speaker A: Pear Drop link to Pear Drive and [00:28:58] Speaker B: Pear Drive will be able to download something very easily and vice versa. And that was kind of the thing that like, that really set it in stone. I have felt like this thing was different. I have felt like obviously this is useful in a lot of other ways. Like I said, I don't need, like, this is basically my secretary. But the really cool thing, the really powerful thing is literally the processes that are pulled out of this. Because I. One of my big things is I'm trying to manage memory so that every time I reset the context window, which I try to do a lot because after it gets over, like 50%, really, really more like 65, 70% is the time when I really notice it, it will start to forget things, it will start to mess up. It just gets. It gets strained. It's like trying to, you know, open a bunch of stuff when you have 80% of your RAM already occupied. It just. The context window starts to really mess up. Things that are important just get kind of buried under the mountain of other random instructions. And so what I do is I've tried to create this kind of hierarchy of things to remember or to things to have in context. And whenever I feel like we need to reset the context window, I will, you know, record down those important things and specifically not record them down in memory, but record the place to find them and what they are related to in memory. But then put the actual, the actionable files and the description or the mission [00:30:30] Speaker A: or the project manifest into its own [00:30:34] Speaker B: file and and it's the reminders and to dos in their own file so that we are not bringing them up every single time I boot it. And so when I reset my context window, I started off and it's like 7% of the context window. And I try to keep it to exactly what is most necessary so it can reach out and find the context for the other stuff when we are engaging in some other project or going down some additional path. [00:31:04] Speaker A: But that for general purpose and general [00:31:07] Speaker B: items and saving and everything is like we just reference and we go to the things that we need and then come back. And because of that, whenever I start fresh, I don't really lose much at all of the important stuff that we are working on. But what I hadn't realized and what makes me wonder about like there's. There's an element of what Jesse Posner broke down here that just makes me wonder like, is this really. Is he really working like this? But Jarvis suggested like, like I said, a series or a stack, a hierarchy of a few different models or excuse me, agents and which ones may actually be optimized to use different models and then how to run them together. And honestly it's actually it, it seems like it would be pretty powerful because you know, in the context you're man double the use of these words in the nature of having like the context window blow out that causes the degradation of the operation and you know, having reminders and TO DOS and calendar and stuff messy up the waters of the context of our Vibe coding or Paradrop or paradrive projects is it actually makes, if I can't get an entirely different context window, it makes it a very, very powerful thing to be able to just have to literally have my main agent manage. The fact that the reminders and the TO DOS and the bookmarks in the [00:32:38] Speaker A: calendar never mess up or muddy the [00:32:40] Speaker B: context of the Vibe coder. And that the intent and the reasons and the philosophy for building the thing and the decision making about which path to take is its own agent with its own context that controls or manages the Vibe coder to not get lost [00:32:59] Speaker A: in bloat and bad decision. [00:33:02] Speaker B: And just that on its face, because a lot of my kind of micromanagement of context window and stuff has been about that. It's like, okay, how do I try to separate these jobs? It's like, oh well, I can just spin up a new agent. And it just hadn't registered with me. But there's something that I really wanted to point out and that I really loved in this article was the fact that what you were doing now as kind of the human coordinator is the one managing this is that you are providing judgment and feedback in the real world and a connection to the reason for doing these things. And this is something that the AI specifically cannot do, and that our role goes from explicitly building and more toward orchestrating. And I hadn't. Yeah, I've been thinking about this for a while. Is like, okay, what does AI do? Like, what. How is my workflow and the things that I've done changed with AI and what am I able to accomplish now? And why. Why does that. Why does it feel differently? What am I doing differently? Because so many people are like, oh, you're not going to have a job. And, you know, this is totally wasted. I learned how to do all of this, and now this is completely worthless. I'm about to be replaced. And I was trying to come up with a good analogy, and it seems like. Or a metaphor. And it seems like most people didn't get it when I posted it. I posted it on X and Primal or Nostr. But the idea was that this was supposed to be an analogy or a metaphor about the perspective shift versus the reality. So a lot of people fear AI, And I like the idea of music as a comparison because I think there's a very easy picture, a mental picture to take when it comes to music because so many people are focused on the actual doing of the thing and the fact that, like, oh, I learned how to do a thing and now AI can do this for me. And that's a terrible thing because this is my only value. But when kind of reflecting on my own situation and how I'm using it and what I'm able to do, I've been thinking about it very differently. And what's funny is that this is very, very, very in line with what computers and software originally did as well. It's just clearly an exponential extension of that because it's a layer that's able to utilize all of these things. So the analogy I had is that, like, imagined you learned to do one thing right. You, you spent this time getting this skill. You learned how to play the guitar, you learned how to play an instrument. And so the fear of AI is I've spent all of this time learning how to play the guitar, and now AI can do this very easily, or it can just produce a guitar riff or music, and now my entire skill is worthless. While I think the reality of this, and for the people who recognize this as a tool and want to use it to Produce and do better things and have a vision for something that they want to see in the world is that to the contrary, I just spent years internalizing the soul of music. Like, what is music. Learning to play, learning to feel the actual music. And even though I can only play one instrument, AI just gave me the ability to conduct an entire orchestra. And that's what I think is happening, is we're all getting our own, the equivalent or the metaphorical equivalent of our own orchestra and all of the things that we want to accomplish. And rather than thinking, oh, my one skill just got replaced by AI, what you should be thinking is that now I can accomplish things and have something else do my skill plus have access [00:36:55] Speaker A: to 30 other skills that I do not know. But that would be incredibly valuable to [00:37:00] Speaker B: be able to make use of. And honestly, any other skill like you think about, this is a great example of kind of the concretation. The. That's not a word. The. The sublimation out of a pattern or process. Like that's what software does, right? Is okay, I built a cool pattern or I built a cool tool in my computer. Now, boom. I can immediately replicate it a million different times to a million different people. And now everybody has access to this tool. That is what software did. That is what computers did. And that's what made them so disruptive. It was the ability to expand and extract or extract and then expand the use of one particular process, one particular tool, or one particular way of doing things such that you didn't have to rebuild it yourself. You didn't even have to understand how it worked, but you could utilize it, you could take advantage of it to build a new thing. Well, this is exactly what AI is doing, except for patterns in general. So you can literally create a ephemeral like skills that just evolve and work. So rather than taking one piece of [00:38:10] Speaker A: hardened and static software, you can modify [00:38:13] Speaker B: the software, you can modify your process, you can work with any and all apps. It literally is just that layered one up. [00:38:20] Speaker A: It is an exponent of that. [00:38:22] Speaker B: Rather than having to map this thing out in a concrete way, you can literally just train on all of the various maps, on all of the various pieces of software and code of people trying to utilize and build out these sorts of tools and relationships. And then you can have a generic way to build tools that you can just utilize directly through conversation. Like in hindsight, this is kind of a perfect extrapolation of how these tools expand on themselves, of the exponential trend of taking what software and computers did to all of our Tooling and then [00:39:00] Speaker A: doing that to software and computers again [00:39:04] Speaker B: by creating a pattern compression, a literal model, a one giant or not even one, but just like a series of models that extract patterns of images, of language, of code, of app, skills of things, processes, all of these things. [00:39:25] Speaker A: And then to be able to pull [00:39:27] Speaker B: those things back out and then when [00:39:29] Speaker A: your process is actually evolving through the [00:39:32] Speaker B: act of doing that's one of the craziest things about this, is that in just figuring out what does and doesn't work and then explicitly naming what occurred explicitly, like detailing out the rule of or the finding, and then continuing to move forward on top of that new foundation. Like, it's like building blocks. As you're trying to build the first, the first, the first floor of whatever your project is or your concept is, you're realizing when the first floor doesn't work or when something is out of place and you need to put another [00:40:05] Speaker A: block into the foundation. [00:40:07] Speaker B: And then you start the first floor, start to work, starts to work, and you build on the second floor. And then you change the foundation again because you realize that it only scaled up to this size and you had to make a modification. And then you're learning the rules of like, okay, this part of the foundation always needs to be like this. If I want to get to three stories high, this part of the foundation needs to always be this wide. And specifically, you're. [00:40:28] Speaker A: You're doing it in the process of, [00:40:29] Speaker B: like, what do I want out of this building? What's the purpose of this thing for me to actually construct all of this? And is this actually adhering to the vision that I have? [00:40:39] Speaker A: And all of it is molding and modifying it. [00:40:42] Speaker B: It's this, this, like, ongoing process of learning and reflecting and doing that creates not just a structure, not just the building, but the practice of creating the structure, structure as a repeatable process. So it's like imagine if rather than software being the one tool that we're replicating to a bunch of different people. [00:41:08] Speaker A: It's. [00:41:09] Speaker B: It's this structure of how to think [00:41:10] Speaker A: about software, how to keep it in [00:41:13] Speaker B: line with your goals, how to ensure [00:41:15] Speaker A: it doesn't mistake and it doesn't bloat [00:41:17] Speaker B: and it doesn't get off track, and [00:41:19] Speaker A: that you can, you can then take [00:41:21] Speaker B: this thing and recreate it in a new environment. And then where that new environment is different, we continue to reflect and modify the process and the structure to make it work everywhere it needs to work and to build anything that we envision [00:41:34] Speaker A: in any possible environment or with any set of tools. [00:41:38] Speaker B: That we can have access to. That is a wild. That's such a. Like the exponential unlock is just crazy [00:41:47] Speaker A: when you start to think about philosophically [00:41:49] Speaker B: what is different, what is actually occurring [00:41:52] Speaker A: with these new systems. [00:41:55] Speaker B: It's absolutely changed the way that I've been doing things and I will not go back, I will have a hard time. In fact, I kind of feel like I want to just continue to upgrade and figure out open claw even more advanced. And I'm probably going to spend a [00:42:10] Speaker A: little bit of time tomorrow. [00:42:11] Speaker B: I don't have time today, I got too much work to do. But I'm going to be spending time tomorrow on figuring out how to do this multi agent thing and doing a Vibe coder. I might just have an agent for paradrive, you know, in fact I might have multiple agents for Peer Drive just because I want context Windows to be very specific and for context not to get bloated at all with any other project. Because not just like a project like coder, not just a Vibe coder, but a vibe coder only for paradrive and then a more broad vibe coder because there is a lot going on in that project and there is a very big vision that has to scale incredibly well. And it's. I'm starting to see, I'm finally starting to see the fruit of all of the things that all the groundwork that we have laid. Because I've built so many kind of like single purpose one off things, I'm [00:43:00] Speaker A: starting to reuse them. [00:43:02] Speaker B: And what's great is that I can test and improve and modify them with new things that come out or new [00:43:09] Speaker A: tools or underlying models or whatever that [00:43:11] Speaker B: come out as I continue to go. And it doesn't break anything because I'm not actually changing code. That everything works with like a basic set of functions and APIs so that they're isolated and the overall, the, the larger applications that use these things like if one of the internal tools crashes or something, it doesn't do anything to the broader application. The broader application actually has the ability [00:43:31] Speaker A: to restart the thing and try again [00:43:33] Speaker B: or to just note that, you know, there is no output from this one piece and let's go ahead and continue to the next piece and we'll figure out or maybe do a backup option for you know, not having, not having this one piece of data or this one part of the process working properly. But because of this kind of like collection of stuff that I built up, it's all just getting much easier and much faster. And I finally have a tool, like I said Like I'm using pairdrop to communicate back and forth and this literally just happened yesterday and this is the easiest and cleanest that I think I've had it built to the point that I think this is worth releasing. When I have iOS and Android app and it's like nothing, it's nothing magic. You know, there's a handful of other P2P tools that you can just. Okay, yeah, or just file sharing tools out there where you can just directly [00:44:24] Speaker A: send stuff back and forth. [00:44:26] Speaker B: But I think they've all missed just a massive opportunity with where you can take it. And I think I've got one feature and this is really kind of the key is that I have one additional feature in pairdrop that I'm going to be adding that's specific to how we have built the core to work that will allow you to do something or allow me to do something that I definitely can't do with really anything else. And that will be the beginnings of like, oh, this is what paradrive could be. And I'm just not going to stop building. I'm just going to keep just going to go as hard as I can and to try to make everything that I actually want out of this. So we'll see. I'm going to try to stay at the edge of how to use these tools. I hope you enjoyed that article and this episode and I will catch you guys. Don't forget to check out the hrf. A huge shout out to them for supporting the show. To the audionauts. Always for supporting the show. Show. I love you guys. I'll keep you updated actually on this. You will be my first testers for Pair Drop and it might be any today or tomorrow. I don't know when this will come out but I'm gonna have something very, very soon. So stay tuned if you're in the Audionauts. I got some fun stuff for you with that. Thank you guys so much for listening and I will catch you on the next episode of Bitcoin Audible. And until then, everybody, that is my two sets. [00:45:57] Speaker A: Nature is the source of all true knowledge. She has her own logic, her own laws. She has no effect without cause, nor invention without necessity. Leonardo da Vinci.