[00:00:00] Austin: the thing that you have to funnelly understand about all these LMS is they're like really, really smart and good at executing with a limited amount of context, but they can't handle too much context. [00:00:09] like some of the core structure that Claude has given the world around how to work with agents is really important. the idea of memory, memory files, agent files, the idea of skills and skill files, commands, these objects basically allow you to reconstruct the human brain in a way, People often think of memory and Claude as just like a big file, but it actually, should be a series of files. [00:00:29] like, go back to the nineties when you have really intense folder systems. That foldering structure is essential for Claude. Now, [00:00:36] You kind of like, you start to organize, your own personal machine around a human brain, and what that does for you is it means you're now not having to think about context anymore. You just put things in folders and you point it at it and we'll bring it in. [00:00:49] ​ [00:01:16] In This Episode --- [00:01:16] Phil: What's up everyone? Today we have the pleasure of sitting down with Austin Hay MarTech, rev Tech and GTM Systems advisor and AI builder, writer, podcaster, and ex founder. You'll be hard pressed to find someone that understands MarTech and is more advanced in their CLO code journey. And there's conversation. [00:01:32] We cover the low friction path, the cloud code. Important concepts non-developers need to build in cloud code. Why meeting heavy schedules are a compounding career liability. Why good taste beats technical skills, and Austin makes the case for a new class of professionals. He's calling white collar super sains, all that and a bunch more stuff after a quick word from two of our awesome partners. [00:01:54] Sponsor: Revenue Hero (Inbound lead conversion and scheduling) [00:02:48] Sponsor: Mammoth Growth (Customer data experts) --- [00:03:51] Phil: Awesome. Thank you so much for your time today, sir. Really excited to chat. [00:03:55] Austin: I'm so glad to be here, Phil, this second time. Lenny [00:04:00] hasn't invited me back twice, but you have. [00:04:03] Phil: Yeah. Yeah. Missed opportunity there. Thanks for returning to the show. Only a handful of returning guests. [00:04:09] 1. How Code-Driven AI Workflows Outperform Chat-Based Prompting --- [00:04:09] Phil: I'm an avid reader. Austin of your Growth Stack Mafia newsletter have been for a long time. [00:04:14] Uh, on Substack. I'll link out to, to for, for folks that aren't already subscribed. Um, but I loved your post on, uh, white Collar super Sains. Uh, I had my head spinning. Uh, like I said before we started recording, I'm a huge Dragon Ball Z fan. I had grown up, that was my number one show. So, uh, initially caught my eye, but [00:04:31] let's talk about like the hidden AI chasm that you talk about in the post. A lot of folks are talking about the debate between using AI versus not using ai? You think that within this chasm, there's like a hidden chasm that's actually using AI as a simple chat interface, a UI, versus using AI as code and and wielding stuff there. So my first question for you is, when you look back at your own shift over the last year from copy paste prompting to [00:05:00] code driven workflows, what actually unlocked it for you? [00:05:03] Like, is it better models, better tools, finally getting a taste of CLO code or changing how you thought about delegating work? Maybe riff on that for a bit. [00:05:11] Austin: Yeah. Um, gosh, a couple things happened. First, I think, uh, the, uh, the kind of advent of cloud code changed everything in my mind, and I was not an early cloud code adopter. For what it's worth. I continued to like, kind of operate the old school way and just thought they were chatbots. Um, but you know, if you back up like six months ago, the, the dissociation that I had was you have the apps, you have Claude and um, you know, and chat GPT, and you can go in there and you can work and do stuff and get responses. [00:05:44] And then you have API keys, and you can use API keys to like, make requests and have threads and calls. But like there was this. Kind of gap between, oh, I'm using an application that's already been built and I'm writing code to use ai. And my first kind of foray into [00:06:00] using API keys at the time, and I think everybody did this was like NAN and Zapier. [00:06:04] Like you go and you connect your key, you have kind of like a model outputs. You can put random data into some type of zap or some type of NAA kind of node and then it'll produce a result. Um, and my own ability to like use the API key was somewhat limited because, you know, there was no terminal interface to like, just randomly construct API calls. [00:06:23] So I felt like I had to do a lot of copy pasting, which is what you talked about. So I could make something great and I could have it in GPT or clawed, and then I could paste it in the terminal and start building the terminal. And then when I needed help, I have to go back And, um, on a related note, like. [00:06:38] I remember actually I was using a tool called Runway, which is a company I used to work for. Amazing Tool. Um, it's an fp and a suite and it connects to QuickBooks. And QuickBooks is basically a database that has UI on top. And I experienced the same problem. Um, I was writing SQL to like query our financial data. [00:06:57] Um, this is when I was working, uh, at clarify. [00:07:00] And then I would have to make a SQL change. So I have to like copy the SQL back to Claude or to GPT and make changes. And one of the kind of first steps was I started a folder and I just had all of my SQL queries in that folder. And this was like in the summer of last year. [00:07:18] And whenever I wanted to update all of them, I'd update them all into Claude and then have Claude look at all of 'em. But I felt like the limitation was the UI for Claude Limited how many files you could upload. The model couldn't handle lots of files. Um, it was sometimes slow and because it couldn't handle all the context and it couldn't connect to my local repository. [00:07:36] I just felt like I was stuck in this world of kind of calling and answering. Then Claude Code arrives. Um, and I, and I think even before that there was like the advent of, hey, there's some type of coding interface, you know, in these apps. Cursor, arrives, cursor, basically an IDE with a text editor, and it can run any model you want. [00:07:57] So now you're model agnostic. Now you have, [00:08:00] um, you know, uh, the ability to run command line and also to write and type code, and then the agent is baked in. And I started using Cursor last summer. I think a lot of people did. Um, and, you know, uh, kind of kept doing that. But I think what changed is, I heard, heard a lot about Claude Code came up, a bunch friend told me it was so much better. [00:08:19] So I, I spent a weekend. I trying to set it up and understand for myself, like why it was so spectacular. And I, I think like for anybody who is used to the development process where you kind of like create a repository in command line, you like create a repo and GitHub, you start to code and then you push to Maine and you start pulling these things down. [00:08:38] These are like, just like really common, you know, somewhat technical, I would argue somewhat technical, um, workflows. Cloud code just makes that so accessible. 'cause all you do is you type CLO and now you're in a terminal and you have the ability still to look at files. Um, so I think, I think cloud code just as, as an idea of you're working in terminal and you're interacting with an [00:09:00] API key and you're interacting with Claude, but you don't have to interact with the app was really fundamental. [00:09:05] The second thing that I think actually was really fundamental is the fact that CPS came out. MCPS got way better. Um, max came out. Max lets you run Opus. Opus came out. So like a lot of stuff happened all at the same time. Like fir first is this whole MCP thing didn't really make sense at first, and now it makes a ton of sense. [00:09:24] It's basically an API with documentation bundled into, uh, an object that Claude knows how to use. And so, and anybody using Claude code for example, can say like, Hey, I want to, I wanna be able to consult linear and like, pull down all my tickets and organize them. And Claude will be able to help you set up the MCP Connect to your account. [00:09:42] It has all the APIs for linear, so it can do anything in linear for you. And so just as a practical example, like almost anybody with a little bit of te, just a very little bit of technical knowledge could say like, I have too many tickets. I wanna organize them and update them. It could talk to Claude and it could go build a little module for you and do that. [00:09:59] And so I [00:10:00] think that that was one big piece of it was MCP. Second big piece was the fact that Opus as a model is just pH phenomenal. I think it's much better, at least in my experience, than even Codex now. And again, they have the studies and all the other stuff, but I, I just can only speak from experience like the, the upgrade from sonnet to Opex Opus was just amazing. [00:10:17] Um, and then I think the, the third thing that really changed is they, you know, whenever I wanted to use Cursor, it'd be like a hundred dollars a day for like a really expensive project. You're trying to build something really great, it's really expensive. [00:10:30] And I think Claude crushed it by saying Buy the Max plan or the Supermax, and you get unlimited opus credits, so you can build as much as you want. [00:10:39] Um, another thing that I think really changed thing is this idea, like some of the core structure that Claude has has given the world around how to work with agents is really important. And I don't think most people know this. And I wanna come back to like, you know, where we are on the, the chasm. But some of these objects we can talk about are like the idea of memory, the [00:11:00] idea of memory files, the idea of agents and agent files, the idea of skills and skill files, commands, um, like all of those I think were essential. [00:11:09] Oh, and then the last one too is the idea of federalized agent agents where you can spin up multiple teams of agents that worked with one another and their sub-agent processes that can run. Because like, you know, the thing that you have to funnelly understand about all these LMS is they're like really, really smart and good at executing with a limited amount of context, but they can't handle too much context. [00:11:29] And, um, and they're also not really good at storing big things of context. So, so these objects basically allow you to reconstruct the human brain in a way, you know, memory I think is the best example of that. People often think of memory and Claude as just like a big file, but it actually, it should be a series of files. [00:11:47] You should actually be thinking crazy as it sound like, go back to the nineties when you have really intense folder systems. I'm like a. I'm a, uh, you know, a, a person who came outta that era. If you look in my iCloud, like everything [00:12:00] is organized by folder with name, and there's lots of convention around it. [00:12:03] And that was just the way that you grew up, if you're working on an old machine, is that you have folders to organize everything. And I actually think like that, that foldering structure is essential for Claude. Now, um, you have an index file at the top of your folder that tells you where everything is, but then you have memories, you have skills, you have agents, you have Chrons. [00:12:21] You kind of like, you start to organize, you know, your own personal machine around a human brain, right? Um, and, and what that does for you is it means when you're running Claude or any other agent for that matter, you're now not having to like think about context anymore. You just put things in folders and you point it at it and we'll bring it in. [00:12:40] So I, I wouldn't say it was one thing, I think it was this culmination of many pieces together, but I, but I'd argue is like, I think you and I, and probably everybody listening to this call, we're still on the top oh, oh, oh, oh, oh 1% of. People understanding the power of Claude, um, you know, or, or open AI or any of these tools. [00:12:59] [00:13:00] And I also think there's like a world pretty soon where this super savvy like us that functionality will come, you know, more easily to people who are not that the model I like to show, and I show this to everybody, it's like you have Claude and I'm, I'm gonna use Claude and Philanthropic for today's example, but I, I wanna be somewhat fair and agnostic that I, I think Rock is okay and I think, you know, open AI is, you know, okay, so if you're using these other tools, it's largely the same. [00:13:25] You have the chat interface. So if you pull open Chappie or, or Claude, you have the chatbot, then you'll have like this idea of cowork, which is where you're working on, you know, a clo, a clot or GBT instance, and that has access to your local machine. And it's like doing things on your machine for you. So you can pick a folder and you can put stuff in it. [00:13:43] And you can imagine it's starting to teach people. Hey, Claude is not just a chat interface, it's actually like a work agent that can live in a folder on your computer. And then we have this thing called MCP that lets you connect data sources from elsewhere. So it's trying to like bridge that gap between chat interface, [00:14:00] project workspace, and then you have cloud code, which is to say, oh, actually your entire computer is the agent. [00:14:05] You know? And when I say agent, what I really mean is it's a single interface that is smart enough to take any task and boil it down into 15 to 20 tasks and deploy agents for every one of those. You know, figure them all out, restructure them, rebundle them, and then serve you the answer. And you know that that is like the way that the world has moved really quickly is now you can almost do anything. [00:14:25] As long as you can break it into small enough sizes, you can figure out how to give it the right context and the right index and give enough time. Um, so anyways, very long-winded answer, but I think the, the answer is there's not just one thing. It's this culmination of many things that I think really hit this critical mass in about January of this year. [00:14:44] Phil: Yeah, it's, it's really cool to hear you kind of unpack the, all, all the changes in, in such a, like you could listen to this and be like, oh, that's probably like the span of two, three years, like now this happened [00:14:53] Austin: Six months. [00:14:54] Phil: yeah. [00:14:55] 2. How to Start Building With Claude Code When You Have No Time --- [00:14:55] Phil: And I feel like for me, like the biggest unlock was. Like going from chat plus copy paste, like you send a command, I want to like summarize this blog post. [00:15:04] You copy paste the output and you kinda run with it, and then you wanna do it again. Like you're just repeating that same process manually over and over again. Cloud code is really a way to have AI be just a step in that repeatable workflow. Like, once you've, uh, summarized one thing, you know you're gonna do it again. [00:15:21] Like, how can I put this in a system, save it in a file somewhere, have contacts, have the saved prompt and just like save a command. Maybe every time I wanna like, summarize something, it's just like a shift in the way of thinking. And that's one of the ways, like, I, I've, I've told people recently like, oh, like clock code, like, I don't really code, and it's like, oh, it's not the right branding or whatever. [00:15:41] But, um, I, I wanted to like, like people listening right now. Like I, I'm convinced that not everyone listening is, you know, definitely not at your level. Maybe they're like, where I am at and they're just a couple of months into this journey, but I'm guessing that a lot of people are. Still on the fence or [00:16:00] not on the fence about doing it or not, but just like a time commitment. [00:16:03] Like they're not sure when they want to do it. Like they have a nine to five. They don't have like AI hackathons within work and they're trying to figure out like, I, I know I need to do it. Everyone talks about it. It keeps me up at night, but like, I haven't done it yet, so I wanna be mindful of, of those folks. [00:16:17] Austin: Yeah, I mean, first of all, I think like if you're one of those people who doesn't have the time, 'cause you have won a family, which is a lot of people, Yeah. um, hard weekends, a nine to five. Um, I think I, I mean, I, I don't know, philanthropic strategy, I don't know chat, GPT strategy, but I think they're intentionally rolling out in this way. [00:16:35] They have, they have the chat interface, which is like for beginners. Then they have the cowork interface, which is. Kind of a step up. And then they have the coding interface, which is more advanced. The cowork space is a great place to learn effectively how it's gonna work. You know, I could even pull up cloud code and show you guys, you know, if you go to Cowork, it's a prompt, but you'll see there's just a couple different things that are different. [00:16:55] You can upload files, you can make it local or kind of [00:17:00] cloud-based, but it's trying to tell you like, Hey, this is a folder that's CLO is working in. And I think once people start to understand that like, oh, this, this is not something that it lives only in the cloud. It's something that lives in a folder. [00:17:11] And I could have this folder on my computer and I could actually just like throw a bunch of stuff in that folder and start to work with that folder to do something for me a lot. That to me is kind of the, the next step is you go from copy pasta, you know, back and forth to repeatable motion you've defined. [00:17:28] And the problem with that though, is like you still have to run it, right? You have a, this idea of a session. I think what you and I talked about before this call joined and I think where a lot of people wanna get to is like, how do I make this feel like NAN and Zapier where it's just doing things for me automatically? [00:17:44] And that's the bridge. Um, and I, and it requires two things. One is you either have to have, your computer has to be on and able to run Aron job and Aron job for people who don't know, is just like something that happens on a chronological time commitment. It's like every morning at 6:00 AM run [00:18:00] this script. [00:18:00] That's all Aron job is. The reason why these things live on servers is 'cause servers never go down. So, you know, this is the reason why Claw Bot, I think took off in December is 'cause you're basically creating a mini server in your basement that can run Claude. Do Aron job. So it, like, everybody thinks it's, you know, a IG it's not, it's, it's just Claude running on a continual time commitment. [00:18:24] So you can basically think now about workflows. You can think about, Hey, I have a, a reliable server that can receive a web hook or receive data and do something with it. I have a computer that can run every morning and send me a thing. Um, and then I think the next level from moving from computer to, you know, to your basement, you know, iMac to the cloud, and I think we're getting there pretty quickly where, you know, your local machine and Claude or GBT running in the cloud will become one. [00:18:54] So now you can do a cowork thing on your computer and Claude will take those files and [00:19:00] automatically version control them and pull them into the cloud. And so you could say, now I wanna run this thing on a sequence, right? We're, we're moving in that direction. And I think the only difference between, you know, when we get there now is that like, that just probably takes time for philanthropic to build. [00:19:13] But I, I imagine we're not too far away where you'll have an anthropic folder and uh, or a cloud folder and you can build in that folder, and all of those things can be deployed as applications in the cloud. And so now you can just have a single terminal interface on Claude that just you can talk to and it can build things for you and run automations. [00:19:32] The difference is you and I, and I think folks who are learning cloud code now, we can get to that actualization faster. Um, I think they'll always have cloud code, you know, as a separate entity for builders. But, you know, I think we'll get to the masses later. [00:19:45] Phil: And [00:19:45] 3. The Programming Concepts Non-Developers Need to Build With Claude Code --- [00:19:45] Phil: you say that, like in your article, um, one of the ways to get started here is that workers should learn object orientation and the command line as like one of the first basic things that people should get familiar with if they're not already. And I'm curious to get your take on, you know, like beyond those basics, what programming mental models do you think are critical for like bridging this gap from going with like chatting with AI to wielding it as code? [00:20:13] Curious your thoughts there. [00:20:15] Austin: So in that article that I, that we talked about, uh, there's this graph and it shows like everybody thinks, you know, we're here, but actually we're going to Claude code. [00:20:27] And, um, the reason why I think command line and object orientation are two things that are so important. Well, one, I would say, I've been saying this for the last decade, and actually most of the marketers I've hired that are great have gone to some type of bootcamp academy where they learned this. Um, and it's because I think if you understand the command line, you understand how computers work. [00:20:50] If you understand object orientation, you understand how APIs work. And if you understand one programming language, call it Python or JavaScript or even HTML, you now understand the [00:21:00] way that the web works. And I think when you have those three things, command line for computers, object orientation for how data works, and um, you know, one programming language for how the web works, you're now able to teach yourself anything that you don't know. [00:21:17] And when we just talked earlier about how we're moving to a world where, you know, actually these things run in the server, in order to understand what a server is, you have to know what a computer is and what command line is, and what object orientation is and like what's a server versus a database. So, so what, what, what I'm saying with this is there's like, there's this entry level knowledge you have to, like activation energy you have to overcome to then be able to teach yourself anything and even understand what to ask. [00:21:40] Something like Claude four. Um, and to me that is the thing that I keep trying to tell people, especially folks in marketing, is that it's not about becoming a programmer. I don't care if you know how to like SSH into a tunnel and do whatever. Uh, what I think matters is being able to like, figure things out to build. [00:21:58] Like if, if, if a marketer [00:22:00] wanted to build a micro application and deploy it to GCP, like these are not that complicated skills, but they, they have to know like, well what is GCP and how does it work and how do I long into it? And like, what are the services that are available? And so when I talk to Claude, I can ask it to do the right things. [00:22:15] Um, otherwise you won't be able to kind of audit what it's telling you. And you could do stuff that's like really dangerous, I think in Claude code or just like silly and wasteful. So, so that's, that's some of where I go with this is like, it's like teaching you to be able to point Claude in the right direction to do the thing you wanted to do. [00:22:34] But again, I think that's the reason why Cowork exists and why the chat exists is 'cause there's gonna be a lot of layman users who. We'll never reach that point, and that's totally fine. [00:22:43] Phil: Yeah, I think that's the beauty of like, I think a lot of people are kind of still on the fence because what you kind of talked about, like maybe some of the things they're, they're saying, you're saying there, it's like, uh, like object orientation, you know, like I'm building lifecycle sequences and Marketo or MO engaged. [00:22:59] Like I, that's [00:23:00] not really my thing, you know, maybe I should learn more about that and take this course and that course before I get into cloud code. Um, but like part of your advice is also like people should be tinkering. Like, it, it shouldn't necessarily limit you from just trying it, like going in there and doing it. [00:23:16] Like you're not super sure what object orientation is, or, um, you know, Claude is saying you need to like structure this in A-J-S-O-N format and you're like, what the heck is JSON? Like you can ask. Claude to teach you those things and like it can be your teaching partner while you're also running it in terminal. [00:23:33] And so that's the part that was kind of a, a bit of unlock for me. It was just like, you know, I'm not exactly sure what I want to build. So it was also like a teacher. Um, it helped me like grasp some of those, you know, um, things before kind of diving into it. And it's, I dunno, it's hit or miss because like, I feel like I, [00:23:49] 4. How to Turn Repeating Prompts Into Automations That Run Themselves --- [00:23:49] Phil: I've chatted with some developers that, like engineering buddies who are just like, you know, like months and months advanced in, in building some of the things in, in cloud code. [00:23:58] And they're like teaching me a lot of [00:24:00] the learnings along the way. Like, you need to have your context file instead of like calling the API here, like you should actually be like prompting Claude directly for like non-deterministic things, things like that. But then I also chat with people that. Are non-engineers. [00:24:14] Non-technical, and they've built some crazy things. And like they would be the first ones to admit to you, like they don't know a thing about object orientation or like what is gonna be a deterministic or probabilistic task. Like some of those like basic mental models, they're just like, I opened clock code in my terminal. [00:24:30] I'd never opened terminal before. And I was just like, I'm trying to build this. Like how would I plan out doing that? This is why I'm trying to do it. And they told me to do this and that, and I told 'em I don't know how to do this. And then they walked me through it. So it's like, yeah, there's like the basic stuff you want to get down that would help you, like you said, save on tokens specifically. [00:24:48] Um, but yeah, it's, it's, it's such a wild time because like the learning curve doesn't have to be as deep as, you know. Some people look, uh, or think it is from the outside. [00:24:57] Austin: Well, I mean, you're, you're hitting at important [00:25:00] points here. One is, um, you know, there's a lot of value to be gained for individuals learning these tools. Um, if you're an individual practitioner at a company and a lot of your job is administrative bullshit, like you have a lot to gain by understanding how to use these tools. [00:25:19] Um, if you're an engineer whose job it is is to proce production, grade security, safe, highly available, amazing applications, of course you're gonna get on the rabbit hole that you probably went down around context files and memory files and commands, and. You know, use the tokens on probabilistic stuff that you don't know and then like, make everything else in code. [00:25:40] And for what it's worth, like I've learned a little bit of that over the last month of like, if you want something to be consistent and repeatable, you have to kind of push it from a prompt into code. And so, for example, as I'm building things, anything that I wanna be repeated with high precision and fidelity, I'm prompting, hey, build this into a shell script. [00:25:57] Build this into a Python script, right? Like, [00:26:00] take the probabilistic thing that you're doing and make it a script that runs. And g guess what? At the end, if you want to add an agent that like, looks at the results and then determines if it's good. Um, a great example of this would be like, I have, um, a news briefing that comes to my inbox every day. I have a big list of all the sources that I care about. You know, your podcast Lenny's, uh, Elena Verna, Brian Valour, like just, uh, you got the Wall Street Journal, got the New Yorker. 'cause I was tired of getting like a million emails and then. Having to figure out what to read. Um, and I just wanted one thing that, you know, I could flip to throughout the day. [00:26:36] So I built a giant file of all the things I care about. I had Claude get their RSS feeds. Um, and as an example, in the beginning it was like every iteration it was going out and getting all the files and then determining what to do. So like a huge context block. And it was really random. It was inconsistent on what I was getting. [00:26:53] But then I was, you know, I worked with Claude to say, no, actually go get all the data and have [00:27:00] each kind of source go through its own agent, its own Claude session, to take maybe five to 10 articles and compress them to three. And then each of those three goes into another agent that then says, which of the three are we gonna keep? [00:27:11] And then each of those goes into another agent to say, okay, well these are the ones that make the final cut. Right? So it's like, but that's an optimization problem that engineers know about, not something that a first time builder really cares about building a simple script to help them do their job. [00:27:25] The things that I, I like, like to encourage people with examples are, are like, how much it can make your life easier and better. Um, one of the things that I have been using a lot is actually the linear MCP. I'm a linear guy. I like to project manage and linear, but everybody knows, you know, when you see a giant board, if it's. [00:27:42] Overwhelming. It's hard to do anything with it. With the linear MCP, you can say, Hey, gimme all the tickets on the board and categorize them by based on priority or themes, and tell me like, let's walk through them together and I'll write the updates for you, and then you can post 'em. All right, so that's an [00:28:00] example. [00:28:00] You have 10, 15 things to go through, especially as like a manager managing a team with 2030 tickets, you could have, you know, Claude, pull them all down, look at the highest priority ones for you to review personally, and then say, well, like any ticket that hasn't been updated in three days, can you write a generic response back to the requester and be like, Hey, what's the status update? [00:28:20] And then we'll just use the MCP with linear to go do the work for you. Um, another example that I I think just recently came unlocked, two that are pretty amazing is one, um, you know, you can brainstorm in a conversation and, and then you can take that context and do whatever you want with it. And I think people are really sleeping on the idea that. [00:28:41] You can use every human conversation to drive the outcome with AI now, and that is actually what I think should be happening. Most people think about AI as a content creator or like an adjudicator, but I actually think, like I wrote about this, I think the, the flywheel for work is changing very rapidly and we're getting to the point where you could just show up for calls [00:29:00] all day and have conversations and the work comes out of it similar to, to, to this conversation. [00:29:04] You know, we're having a conversation, how do you get this to like a pristine podcast at the end? Well, there's editing involved, there's creating the socials for it. There's going to the tools and uploading and all that is just busy work. The real work is, do I think about the questions? Do I plan what the person do? [00:29:19] I have the call and, and like, yeah, maybe I want to come up with the right socials for it. But like the work to be done is in the thinking of this pod. The same thing is happening across so many surface areas in work. It's like when you plan OKRs, how do you plan? You have a conversation, you have a doc. [00:29:35] Then the hard work is like we gotta get everybody's opinions into that doc and formalize it. Um, so I ran a process with a company that I'm working for where I went and had conversations with every team member. All these had granola recordings and clarify recordings. I took all that context. I had it all ported into a series of agents that divided into groups and themes. [00:29:56] I took all that OK R context and I came up with their OKRs and then I [00:30:00] sat down again with the team and we did a brainstorming session with their, with those OKRs on a Figma file. And then afterwards people dump their, like emojis, their hearts, their thumbs, their comments on this giant board. And the Figma MCP lets you take that into text and then put it back in Figma. [00:30:17] So, you know, in an old world, like just two years ago I did this at ramp and like even at ramp, one of the best companies in the world, it was like a months long process and now we're talking like a week, you know? And honestly, it wasn't even a week. It was like. Probably five hours. And if I really had no other time in the day, I could have done it faster. [00:30:36] But it's like, you know, life still continues when you have work. So. [00:30:40] Phil: It is such a cool use case. Austin, I, I love that. I feel like, you know, folks sometimes are a bit on the fence because there is like a lack of inspiration or, or use case. Like I'm stuck in the mundane of my nine to five. I don't know if I need to like. Code anything and, and, and Claude code. But I think like hearing examples, just like your, your [00:31:00] newsletter parser, uh, like content aggregator of just like, you know, I, I, I'm sick of looking at like 17 new newsletters in my inbox. [00:31:07] That's an easy use case that a lot of people could benefit from. [00:31:11] Sponsor: MoEngage (Customer engagement platform) --- [00:31:11] [00:32:07] Sponsor: Knak (No-code campaign creation) --- [00:33:15] Phil: Maybe you can give me a couple of other use cases that are a bit more on like the tinkering side. Like that's one of your mandate from, uh, the, the blog post that you wrote was that like, there's kind of an urgent need for workers to build and tinker daily with AI and professionals that are just like clicking buttons and third party tools and doing copy pasta to perform manual tax or are kind of cooked. [00:33:37] 5. Why Spending All Your Time in Meetings Is a Career Liability --- [00:33:37] Phil: So my question for you is like. For non-technical execs that, you know, their schedule is dominated by meetings. The, their we're, you know, maybe solopreneurs that are lost in the grind like you and I. Um, what does like a high ROI 50 minute daily tinkering habit or use case examples kinda look like that prevent some of these folks that are on the fence from, from being cooked? [00:33:59] Austin: One. [00:34:00] You know, I feel very strongly about this. I think whether you're a founder, an executive, a vc, it doesn't matter who you are. If you're spending all your time now in meetings like you are, you are missing out. I just have to stay to the record like you are. I think you are cooked personally. Like I would not want to be any role where I'm in a nine to five in meetings all day because I think you are going to get left behind. [00:34:25] That is just my honest to God personal opinion at, in the best case, like, you know, you might survive, but you're gonna be really inefficient. I think in the worst case is like your job is totally eliminated. Um, and so even me, like as a founder, five time startup guy, like I'm spending all my time thinking about if I were to join a company right now and start executing a role, how would I do that job with as little human kind of inter intervention as possible? [00:34:50] And like, so, so the, the, maybe to answer the second question around what are like a couple, couple tasks you can do. My whole career has been defined [00:35:00] by finding boring tasks that I do in the day and trying to automate them. And I think if you're not doing that, if you're not spending an hour today to say, what are the things that I spend all my time on and how could I automate it, then you're missing out. [00:35:13] So even if you're somebody that has 5, 6, 7 calls a day, ask the question more. Like, what do you do with after those calls? Do you get action items? Do you follow up with people? Do you send Slack messages? What could you build that would make that process five to 10 minutes at most, and not copy and pasting, but a fully automated workflow. [00:35:32] Because if you start to think this way, then what hap happens is this is you say, well, I'm in meetings and calls, and that's where all the context is. So I gotta get a call recorder, and if I have a call recorder, I have to have transcription data and I gotta put that somewhere. And once I have the transcription data, I gotta do something with it. [00:35:45] Well, what do I wanna do with it? Am I summarizing it? Am I sending it to a tool? Am I putting it in a Slack board? Am am I putting it on linear? Like, you know, it's, it's a little bit of just like working backwards from where you are. Um, [00:36:00] so depending on the role and, and what you're doing, I think honestly just starting with like what, like what takes up your time and, and what's boring and, you know, can be automated is the, is the way to go. [00:36:09] Um. And then, yeah, I mean, I, I think the, the more ambitious thing is to say like, what are cool ideas that you've always wanted to build and you've never had the opportunity to. Um, this was an example that I talked about back in September, which is like, I was, I bought a, um, have you heard of, uh, eight sleep before? [00:36:27] Phil: Yeah. [00:36:28] Austin: Okay. [00:36:28] 6. Why the Best First Claude Code Project Is the Task That Already Annoys You --- [00:36:28] Austin: So I bought eight sleep back in the summertime, and I know it's expensive. I'm, I'm very privileged. I am, I'm a huge fan of eight sleep and uh, this is like a first world problem, uh, you know, issue. But I bought this eight sleep and I remember, uh, looking in my credit card offers that there was like 20% backer, $400 off or something. [00:36:51] I was so stoked and that's why I bought it. But I didn't click the offer. So I spent all this money and then I like didn't get $400 back [00:37:00] 'cause I didn't click the button and I was so enraged, like, so enraged. This was back in October. I like stayed up till 3:00 AM. With Claude and GBT to build a Chrome extension that every time I logged in, it would just click all the offers. [00:37:16] I was like, never again. Every time I pay my bill, it's gonna click all the offers. Um, but like, I think that's, that encapsulates what I wish people would think about and do is like, if you have something annoying or interesting, like go build it. It's, it's actually, you know, and, and, and maybe in the beginning, maybe the first rev will take a lot of time and effort, but you'll find pretty soon that actually you can build ideas very, very quickly and you can go, like, you'll start to understand how I think an LM needs data structured. [00:37:45] This is maybe the, the one of the fifth gaps that we didn't talk about, which is people starting to understand that like context is what drives the outcome. So if you give. Ai, nothing. You're gonna get a shit result. If you give it really structured, [00:38:00] detailed, thoughtful context, it's gonna give you a great result. [00:38:04] And so that's the reason why Whisper Flow is taking off is 'cause if you think about if I wanna build a big application with a lot of detail and there's all these things I gotta think about typing that is really hard. Um, you know, unless you're, you can, you have the coffee and you're in the zone and you structure it. [00:38:19] But like what you actually need to do to get a really good result is have like line by line the exact things you want and you be specific about where it is and what the context is and how to find it. Like you have to be very detailed if Claude is gonna navigate APIs and some type of application. So that's why talk to text is so valuable. [00:38:37] That's why Whisper Flow is so valuable. That's why the big announcement from Applaud that. You'll have talk inside of Claude Code is so valuable. 'cause now when you're reading something that's wrong, you say, no, no, no, here's what I want you to do. And it will restructure it in a format that Claude can interpret better. [00:38:52] Um, and I think once you start to understand that you now can start to get the results that you want from Claude running. And, but again, that's maybe like, [00:39:00] I think that that for me was a big unlock and I think maybe January, February, where I started to understand that I could use Whisper Flow to do these things. [00:39:07] I'd heard of Whisper Flow and I'd used it when I was talking to the Claude Bot app, but I was like, this is not that useful. 'cause I can just type pretty fast. But now, now I have come kind of like six to seven terminal windows open. And even to my time, those are running CLO agents. And so I spend a lot of my day now just going from terminal to terminal and running tasks and then walking away and coming back. [00:39:32] Um, and, and again, if there's any advanced users out there, I would say the big thing that changed for me is I actually started using Warp with Cloud. Code. Warp is a command line, or kind of like a, a terminal substitute, kind of like I term. But what's cool about it is it itself is running it's own version of ai. [00:39:49] So like if you are trying to learn command line and you're bad at it, it will just help you, but then it can also run cloud code. So there's virtually, there's like not a world in which you don't have AI helping you do [00:40:00] things. Um, so yeah, I mean, I, I just, uh, but, but I would argue like now everything is in clo. [00:40:06] Like all the, the manual scripts that I used to run myself, I've migrated fully to a scripts folder and I just tell Claude I've made them all commands. Um, so it's, it's really, really wild. [00:40:18] Phil: Yeah, no kidding. I feel like the analogy there is like a AI inception, but the, [00:40:22] 7. Why T-Shaped Marketers With Claude Code Will Cover the Work of Entire Teams --- [00:40:22] Phil: the analogy I did wanna unpack with you and, and. Just selfishly get a chance to do some mid journey images for it, is that you essentially coined the term like white collar super sands in, in your blog posts. Like I said at the top, they're massive, uh, DBZ fan, uh, growing up and you kind of describe them as this like new class of elite workers that have kind of mastered AI coding agents and can essentially build anything that you can kinda dream up, um, but also be really good at strategy and the comm side of things. [00:40:52] Um, you, you set yourself in the post and, and today that like you've built more high quality software in the last three weeks than you did in like 10 years of [00:41:00] bad engineering, quote unquote. Um, I myself have built, uh, some like no JS scripts, like I told you before, we recorded like a lot of the stuff I'm automating for the podcast production. [00:41:10] Um, so the painter, you're the paint, the picture you're kind of painting is, you know, one person managing an entire lifecycle marketing stack. Telemetry, attribution alone, but. Like the counter argument there. I guess like the, the devil's advocate question for you is like, doesn't the super sane model create both like a single point of failure and also this like filter bubble problem within companies? [00:41:35] Like is it less about replacing teams with like a lone wizard or a lone super sane human and more about like, dramatically increasing the surface area that like one person can actually cover? Can, can you chat about that? [00:41:49] Austin: Yeah, but before, I think that there's an important thing you gotta explain to anybody listening. What is a super saying, Phil? [00:41:57] Phil: Uh, Goku essentially is, uh, [00:42:00] the role model of all super sane. [00:42:03] Austin: Yeah. And look, I, I don't wanna embarrass you with my knowledge of, uh, dragon Ball Z, but, uh, y you know, the reason I picked that term is 'cause like at, if you're like a nineties guy like I am, um, there were a bunch of these episodes where, you know, a say in becomes a super saying and their hair glows gold, and they basically get like super power and super strength. [00:42:22] So the idea was to say, Hey, like using, using Claude Code, I think like, makes you one of these super powerful beings. And if you're ignoring this, you're gonna get left behind. Um, and, uh, and look, I, I think there's two parts to what you described. One is how, how do you actually use cloud code and this newfound power in a way that's, um, kind of distributed to a team? [00:42:47] And I have some ideas on that. And then two is like, you know, your question around are you creating kind of a filter or like a, a central node that could collapse? And, and to be honest, like I, I think like this is [00:43:00] happening whether I have an opinion about it or not, I, I, I think like the reality is, is people that are really savvy with these tools can do the job of five to 10 people. [00:43:08] Um, there are parts of jobs that are not replaceable. And, and that's where I think like the oversimplification comes in. Like, like, let me give you a great example. I think this is highly relevant for technical marketing. Right? Um, you know, I'm, I'm doing a little bit of consulting work to this company called A to B out in San Francisco, and we have four or five marketers. [00:43:29] We're gonna hire a couple more in the, in the next few months. Um, there's a team member, uh, who does event planning and community kind of management, and she's awesome. And like that work is gonna be really, really hard to replace with AI because AI can't navigate the web, it can't build relationships, it just can't see venues. [00:43:47] So, you know, I think her job is really, really safe. But the part of her job that I think could get way better with AI is sending lists, lead lists, inviting people to the conference, tracking and [00:44:00] measurement around her events. So there's lots of stuff like, even though that core part of the job is not gonna change, a lot of stuff around her job will change. [00:44:06] And the question is, is she gonna be expected to do it or are there gonna be other kind of operators around her that are doing it for her? And where I think this leads is like. I personally think that the people who can do a lot of these jobs all in one, are going to command much higher salaries than the people who can't, which is like the unfortunate reality is like everybody's claiming about jobs getting lost. [00:44:29] I don't think jobs are going to go away as much. I just think that they're gonna get concentrated into people who are paid extremely well to do a lot of really high quality work. Um, and this is what I'm seeing in MarTech right now. You know, the job that I used to do was just one small piece of marketing engineering. [00:44:45] It was like telemetry and attribution, and I was kind of a product manager of engineers and data managers and paid marketing. And now I find myself doing all those jobs and also like, why would I ever ask to be [00:45:00] paid what my old salary was? If I'm doing three people's jobs, especially if I have the knowledge to do all these things and I can, I can guide a paid marketing agent on paid marketing, I can get a MarTech agent on MarTech, a telemetry agent. [00:45:14] An attribution agent. And so the other part of this is I think like people with really broad experience, kind of T-shaped experience, where they might be really deep in one vertical, but they have a lot of cross-functional domain expertise. They know statistics, they know data modeling, they know attribution, they know telemetry. [00:45:31] Like those are the people who are gonna be so powerful in the future because you don't need to be an expert anymore. You don't have to know exactly how to send an email in Marketo. You don't have to know how to use SQL to run a query. You have to know what SQL is. You have to know object orientation for how to do it. [00:45:49] You have to, you have like good taste for like, was this campaign good or not? So I actually think we're moving towards a world where like being thoughtful, highly intellectual and intellectually curious [00:46:00] across a lot of domains, those people are gonna win. And I think right now, if you're an expert in one thing, I would really, really encourage you to go learn Claude code as soon as you can and like use that expertise to anchor yourself. [00:46:13] But then be able to spread out amongst a lot of different surface areas. So you can go deep in one vertical, but you can be an expert a lot. [00:46:19] Phil: I love it. It's such a cool perspective there. Like you mentioned good taste and that pops up a lot in a lot of people's articles about, you know, like [00:46:27] 8. Why Marketing Taste Matters More Than Technical Skill in the AI Era --- [00:46:27] Phil: what is the most future proof thing with, with AI right now and good taste comes up with, with a lot of folks. Um, when it comes to like, hiring, you just said like, um, you're working with a client and, and you're hiring a team right now, or let's say, you know, promotions in this area. [00:46:41] What specific tests do you, or like signals do you look for that can help you identify a marketer or an engineer that can handle creative stuff and, and has good taste? Like how, how do you assess that? [00:46:57] Austin: Oh man, I, I've been thinking about this a lot [00:47:00] and, um, the background context, all this as I've been watching, uh, mad Men Recently, so I feel like a lot of Mad Men vibes are falling into my podcasts and my, uh, my substack. [00:47:10] Phil: A little Dragon Ball Zian madman on the side. [00:47:12] Austin: I know, I know. Well, I mean, it's just, I, I feel like, um, in many ways we're like reverting towards the mean when it comes to marketing in the sense that at the same time that like all this technical marketing is getting easier to do, it's also like less effective. [00:47:27] And what's more effective is just being creative and thoughtful and, you know, for example, RAMP is killing it right now. Not be just 'cause they're making a lot of money, but because they're doing really high quality campaigns that emotionally attract people. Um, you know, for what it's worth, like the campaign with, um, Brian from the Office, the actor, um, you know, we thought about doing that a runway. [00:47:47] It was like, and there's some creativity here. Like, oh, let's find an accountant that's really well and beloved, well, who's the most famous accountant of the early two thousands. It's Kevin from the office. But what that's striking at is like, [00:48:00] does this person have interesting ideas? Are they tapped into culture? [00:48:04] Are they tapped into the heartstrings of people? Are they willing to pull on nostalgia? And for what it's worth, I know this sounds like very ageist, but like, I think a lot of that is developed with experience. So taste can be developed by having a certain skillset and a, a certain professional bend, but taste can also be developed through experience and time. And so I think that, you know, we'll probably move to a world pretty soon where there's like operators and creatives in the same way. In Mad Men there were like the creatives, and then there was the art directors, and then there were the kind of account heads. And, and that might be okay, that might be the world we live in. [00:48:37] But if you're thinking about trying to like, develop taste, or if you're looking for taste, I often just go back to like asking basic questions around what was like the most creative idea of, you had recently tell me about something you've built. You, you know, like draw me something even. Or like, if you were to build something, what would you think? [00:48:52] How would you think about giving to people? Um, and then I, you know, I, I think particularly for marketing, asking for a portfolio of campaigns, people have run. [00:49:00] And also asking the question of, Hey, let's create together a prompt of some design and then critique it together and ask for their feedback on how they would make it better. [00:49:08] How would they feel? How would they prompt? Um, 'cause some of this is like, do you have good taste in the moment when you see something? But also how can you create good taste? Like how can you take something that's already been created and make it better? Um, yeah. So I don't think there's any hard and fast rules. [00:49:23] And for what it's worth though, I think like lots of design and creative people have often been hired and hired under the premise of like, you know, when you see it. So, I don't know, like, I don't dunno if we have to deviate too more far from that. [00:49:36] Phil: Yeah, man, I, I, I think about like the, the ageism aspect of this a lot myself also, um, like [00:49:43] 9. How Early-Career Professionals Build Judgment When Entry-Level Work Gets Automated --- [00:49:43] Phil: having two young kids, like my wife and I als always talk about like, what, what does the marketplace look like for our kids? And like 15, 20 years when they hit the workplace and like, who knows? But these things like that, like you, you just noted like experience and judgment. [00:49:59] Like it [00:50:00] definitely favors more experienced folks and, and mature workers. And like the question of like early career professionals that are just coming out of their MBAs or their bachelor's degrees, like they develop, how do they develop like worldly attributes in if like entry level technical tasks traditionally used to learn the business are now. [00:50:20] A lot automated and they're just like, yeah, I don't really know what taste is be like, they lean in on, on the pulp culture side of things for, for creativity. But yeah, it's, it's, it's, I don't know. I like, I don't have the answer for like, young folks out there, but it's, it's a world war weird world for everyone. [00:50:38] But I feel like for entry level folks, like it's, it's especially weird. Like do you have thoughts there? [00:50:43] Austin: Yeah. Well, I mean, look, it's funny. I, um. I used to be kind of like anti secondary education, but now looking back in hindsight, I would say it's like the reason why I know a lot of things about a lot of things is because I spent a lot of time learning about stuff [00:50:58] Phil: Yeah. You did A-U-M-B-A last year, [00:51:00] right? [00:51:00] Austin: yeah, did an MBA last year I went to Stanford for grad school. [00:51:03] I taught myself how to code. I did a programming academy. And then the last time we talked, I told you I had a lot of hobbies and interests and I, I think like this is a big part of just being a well-rounded person. Um, and, and so like maybe part of that is taking Maven courses. Maybe part of that is secondary education. [00:51:17] But I was surprised to feel at the end of my MBA that actually like. Even though I didn't go very deep in a lot of topics, it gave me access and knowledge to a lot of topics, which I now rely upon very, very frequently. Like a great example is like, do you know what Venetian modeling is? You know, if you don't, you have to go look it up and like you'll read a Wikipedia about it. [00:51:39] But like to actually go through a course on how these models work and how the conceptual model for them was extremely informative in how I think about modeling and marketing. So, you know, not all secondary education is bullshit in my opinion. Um, the other thing that I really, I think is happening in, in today's society is that like there are lots of ways to meet people, but there's [00:52:00] often not a lot of ways to meet the right people. [00:52:02] Um, you know, lots of courses, lots of communities, lots of events. But I often found before I went to my MBA program that they, they weren't always the right mixture of people. So I'd show up, there'd be a hundred people and there might be one person there that I really. Thought was like the right person to connect with, you know, so communities have a, like a, a, a wide variability of who you're gonna meet. [00:52:23] And I think like, you know, some of these old traditional programs like secondary education or or stuff like that, they just have more control over what the audience is. So that's just something I think for people to tra to think about is, you know, I'm not saying that online courses are bad, I'm not also, not saying that they're not great places to meet people, but everything has a time and place and purpose. [00:52:44] And you have to consider for the fact that like, they all have filtering mechanisms. And so what's the filtering mechanism on your community? Um, you know, and are you gonna be introduced to the people that are, that are gonna like, sustain you? That increasingly as I've gotten older, that's just been like a theme of my life is finding communities of people who I [00:53:00] relate with and care about and, and enjoy talking to and interacting with. [00:53:04] And, um, and you know, it's really easy when you're in college because it's a lot harder when you're older and people around the world and everybody's busy having kids and dogs and families. [00:53:13] Phil: I love it. [00:53:14] 10. How Austin Hay Runs His Career as a Flywheel --- [00:53:14] Phil: Austin, it's a perfect transition into, uh, the last question that we ask everyone, as you know, um, actually just recently dropped like a big round table of all the answers for the happiness question for this year. We're kind of tweaking it just a little bit, but I'm curious to get your take on it If this changed, uh, at all. [00:53:31] Like I know you left clarify, you had this like founder persona and now you're back to like the, the solo consultant, um, maybe even busier than, than you were when you were the founder, but um, you're like now a fractional technical CMO. You're still teaching, you're a MarTech practitioner, MarTech engineer now you're also a husband, dog, dad. [00:53:51] I still a water skiing fanatic, avid runner. Bunch of hobbies you have on the side. One, one question we ask everyone on the show is like, how do you decide what deserves your energy at [00:54:00] any given moment and like, what's your personal system for staying aligned with what actually makes Austin happy? [00:54:06] Austin: Hmm. Hmm. Yeah, so clarify was a wonderful journey, and I think to anybody, whoever asked the question, should you try to found something, I'd say, yes. It teaches you so much about the experience. Um, you know, it gives you a lot of empathy for the folks who do start companies. Um, so I, I think like, you know, that was an amazing experience that I would never trade for anything in the world. [00:54:32] Um, but I think when I emerged from being a founder, I found a couple of things. One is that I get a lot of energy out of having time in my day to build and having my time in the day to right, and having time in my day to learn. Everything that we talked about today has come from me sitting and reading and tinkering and playing and, you know. [00:54:55] I personally believe that that's just like part of my DNA. So that [00:55:00] was, it was hard to be actually to be a founder 'cause I felt like I didn't have much at that time. So maybe the first thing I'd say is just like, know yourself and know like what you actually wanna do. And for anybody asking the question of like, what you should be doing, I just think we're in such a generational wave that spending a little bit of time building is not the worst city in the world. [00:55:19] Um, the, the second piece is, I'd say that I kind of got to a place in my career where I transitioned from caring a lot about, you know, outcomes and progression in my personal life and professional life to outcomes and progression against like a couple of core metrics. One is how do I, being very selfish, how do I make the most amount of money working the least amount of time? [00:55:43] Um, and that might mean a dollar a day or maybe it means more, but there's this, um, there's this optimization que question that people usually run in their heads around like, how much money I'm gonna make. But they don't consider the time element. And so I think for me, the leaving a founder, I really started to consider [00:56:00] like money is not everything. [00:56:01] It's the money to time ratio that matters. So how do you maximize outcomes financially while minimizing inputs? And again, when it gets back to the earlier part of our conversation, that's been a guiding force in how I think about automation. 'cause if you can employ automation to do things for yourself, you can drive down the number of hours spent while increasing or keeping same the amount of, you know, financial income you make. Um, second thing for me is like, I'm in my thirties now and I felt like I spent all my twenties working really hard and building a career and reputation. And I, I got to my point in my thirties where I'm like, actually what I care about is having community. And there are a lot of factors behind that. You know, my wife and I went through some tough family times, clarify was hard. [00:56:45] MBA and Philly traveling every other weekend for two years. There's a lot of things preventing. Me from prioritizing that. And so that became like a core focus. Um, the second thing that happened in the last couple years is, uh, just like fixated on, um, on [00:57:00] health goals, on like, physical fitness. Not because I feel like I'm gonna be a championship runner or a triathlete, but because like, doing hard things proves to you that you're capable of more. [00:57:10] Um, I, you know, I, I've, I have a friend who I've been running with for years now, and maybe two or three years ago I told him like, I'll never run a marathon. Like, there's no way. I can never possibly imagine it. And I started working out with him and running, and that led to like faster runs, which led to, um, a triathlon, just like a basic Olympic triathlon, which then led to a faster half marathon time, which then led to. [00:57:36] You know, an, uh, Olympic, uh, iron like Olympic race, which is half of a half iron. Then I did a half iron marathon, then I did an ultra, then I, now I'm doing an Iron Man. And it's not like I'm a crazy type a like obsessed athlete. Like I'm, you know, I'm not Nick Bear on Instagram. Uh, but it's just, it's nice to do hard things and [00:58:00] to prove that you're capable of more. [00:58:02] And you know, what he says, and other athletes always say is like, you know, if, you know, progress isn't made overnight, right? Like, this takes a long time. Um, and, and so that, that's also been part of my journey is just figuring out how I can balance physical motivation and, and physical goals. Um, and then the, the, the last thing I'd say, which I, I think probably you're very familiar with, is there's a lot of peace that comes out of creating a flywheel for your life. [00:58:28] And so the flywheel that I have right now is. You know, learning is maybe the crux of it, you know, and, and learning comes from two things, like having time and having clients to work with or teams to work with, whether it's clients or companies. But you have time and you learn that then populates into enjoyment. [00:58:46] And that enjoyment becomes like reading and podcasting and sharing ideas, which then manifests in like a substack or a podcast, which then kind of creates more opportunity for you, which then creates more financial outcome for less time. [00:59:00] And so I think a lot of this, the North Star that I kind of got in my head is, you know, what it really comes down to, at least for me, is I prioritize learning and education and writing and thinking. [00:59:10] Every single day before noon. And I got this advice actually from Lenny, I can't remember which podcast, he talked about it, but like he, he told me he never takes a call before 1:00 PM and at first I was like, wow, must be nice, you know? Must be nice. No calls before one. And uh, what I realized though is like, le he is in a position to do it because he doesn't work for a company that forces meetings on him. [00:59:31] But two, he prioritizes it. And I think like that's a huge shift from being a founder to being an operator again, is that like, I have complete control over my schedule and I can prioritize before 12 working and tinkering and reading and learning, which then means I'm more prepared and I'm able to apply myself with teams in a, in a functional way. [00:59:52] And so, look, not everybody has that opportunity, which I, I think is tough. But like, if you can try to create that opportunity on even the margins, I think you're [01:00:00] gonna be more successful. [01:00:01] Phil: Awesome answer, Austin. Uh, it's actually a bit different than, uh, than the one that you shared earlier. Some, some truths that like stuck around there, but yeah, not, not everyone has the, the dream of no meetings before 1:00 PM but, uh, like when you talked about earlier, like if, if you're a solo founder or, um, like, uh, a, a manager and your full day is just meetings on meetings, on meetings, there's still like automation that you can do with the outcome of those meetings. [01:00:28] But I was thinking more towards the, like I, I had a short in at automatic and they're like fully remote, fully distributed, and they're super async, like all in on async, one of the pioneers sort of async and they're like super EET culture. And it was actually possible being a full-time employee at that company and not have a meeting until 1:00 PM or just have like a couple of one-on-ones, like throughout the day. [01:00:50] And it was like a very written culture, all the like. Updates and project statuses. They were all in, like project management files. We had this like [01:01:00] internal WordPress system that like people use to like share updates. And so part of the job was really just like being like siphoning information from all these different sources to big company also. [01:01:11] And I'm just like, when you're talking about that, I was like, I can't imagine working at that company today would be so much easier to siphon that information, distill it to like what actually matters for you. And just like automating status updates for people. But um, yeah, that, that's the dream. Like no meetings before 1:00 PM I also have the ability to, to do that today and not when you have consulting gigs on the side that I, I feel like it's very similar to just being a founder there, but su such great advice. [01:01:38] Austin, I had a blast chatting with you. I feel like we, we could keep doing this. Um, we'll, we'll have to get you on, be our first, uh, threepeat guest that, uh, thanks for your [01:01:47] Austin: To repeat. Love that. [01:01:50] Phil: Let's do it. Yeah. Hopefully, um, folks after listening to this are like, all right, that's it. This is, this was the last thing I needed before jumping into cloud code. [01:01:59] I'm stopping [01:02:00] what I'm doing right now and I'm just giving it a try. I appreciate the, the use cases and uh, like the openness to let me poke some, some holes at, uh, some of the claims you made. But yeah, excited to make this a super sane episode, man. [01:02:12] Austin: I, I'm stoked too. I hope this one for folks. [01:02:16] Phil: Awesome. Boom. I.