Focused - 252: Donkey Work

Episode Date: March 24, 2026

Welcome to Focus to Productivity Podcast about more than just cranking widgets. I'm David Sparks and joined as always by your friend and mine, the esteemed Mr. Mike Schmitz. Hello, Mr. Mike Schmitz. How are you? I'm good. How are you, David? Good. For our listeners at home, we record the show for a second of the morning, at least California time, I guess it's later for you, Mike.

But we have, for vacation reasons and other things, not recorded in like a month because you're out of town. We got ahead. now we're back on schedule. Can I just say I jumped out of bed this morning and thought, I cannot wait to sit down and talk about focus stuff with Mike Schmitz. Aw. I miss you. Well, it's mutual, although as maybe we'll talk about in deep focus,

I've been going through some of your robot assistant stuff, and you've thrown me into an existential task management crisis. I love that. I love that any time I can do that to you, Mike. It's just a little reward for me. You got me. Yeah, just a little reward. So you raise the topic. I've got a new field guide out called Robot Assistant Field Guide,

and it's all about using Claude Co-work and Obsidian or No Plan or text files to help do the donkey work for you. And we thought today would be fun just to talk about the concept of donkey work from the focused perspective. But I'll put a link in the notes for the new field guide. You get 10% off with a code focus robot. So don't forget that. It's limited time, but I want the focused audience to get in on it. Yeah, this is a good one. I've been working my way through this one, and I'm learning a lot. So I always learn from your field guides, but this one in particular has made me feel like a complete newbie in a lot of different ways, which I think is a good thing because you also make it pretty approachable on how to actually implement this stuff.

But this is a good one. This is a good one. I really wanted to get this one out for normal people, not just programmers. And I feel like we're on the verge of just changing the way we use. computers. This is a watershed moment, although I think most of the world doesn't know it yet. And I think anybody that gets in on it early is going to benefit from that. So I just think it's important right now to understand what's going on and how to take advantage of it. And we're going to talk about that in the show today. But Mike, you've got something going on too. Yeah, we've got a life theme cohort that is kicking off April 6th. And I've kind of revamped the way that we do these. Basically,

you just need to be a library member, which is my paid community for serious sense makers. So you can head over to library.prightlpkam.com at any point to join the wait list. Then you'll get notified when it opens up. We're opening up the library basically in conjunction with these cohorts. So when people come in, we're all there for a specific reason. We go through this stuff together. I think it's a better onboarding experience for folks as they join the community. But I've also got a separate page set up with a promo video and things like that.

if you want to figure out if this is for you. Life theme is basically my term for a personal mission statement. So it's a process that my wife and I walk people through in terms of identifying your own personal core values, your ideal future, all those kind of like big picture things that are part of my personal retreat framework. That's basically what the life theme cohort is for. The last session is where we all share our life themes. That's always a lot of fun.

So you can find out more about it at life theme. practical pkm.com. And if that sounds interesting to you, we'd love to have you join us. I feel like the whole key to focus, which is the name of the show, is to figure out what's important and do that. And life theme is a great way to figure that out. So everybody go check it out. So AI, you know, it's come up on the show before. And the thing that I keep thinking about AI is like, how do you incorporate into your life?

Because so much of this is about human expression. right we feel threatened by AI in the sense that it makes art and wants to write poetry and when you see that you're like but that's what we do why would we want the robot to do it and i feel strongly that we do need to embrace our humanness and human expression if anything more than ever because for the first time there's another thing that can also do this type of stuff you know and we've used the quote before on the show of people thinking, oh, I'm so glad the AI can make my art. Now I have more time to do my laundry. And I think that's the enemy that we're facing with this growth of artificial intelligence. So something we all struggle with as it becomes more powerful

and more present as it starts replacing our work is how to, what's our relationship with this thing that is capable of not human expression, but a facsimile of it. I would say, They have rather poor facsimile of it, but eventually it will get better at that. And then on the flip side, you've got the artificial intelligence companies, and they have a big, they're driving the U.S. economy right now, their stock prices are through the roof. They're spending tons of money on these data centers. And so in order to keep that, you know, to keep that thing in the air, they have to tell you that it's going to cure cancer and going to solve every significant, you know, P.A.

HD problem that we're going to have a what's the warehouse full of geniuses, I think is the term somebody used, that we're going to have all of our problems, global warming, cancer, whatever, it's all going to get solved by this AI. And that's why you need a buyer stock, you know, which I think is, it's possible. I'm not really convinced an LOM's going to do all these things for us. But it's possible. But it's not here. And there's an underbell. of this that is actually pretty good for us right now. And that is, can AI help us with the stuff we do everything? That's grunt work, what I like to call the donkey work. And I think the thing that people aren't realizing because the AI companies don't really want to get into it too much because

they're more interested in getting the high value is that your donkey work can be done right now by AI. And that's the experiments I've been doing since the new year. And that's what led to this field guide. but I guess I should describe donkey work further. Yeah, well, real quick, before you do that, the AI rhetoric or propaganda that you were describing, I actually listened as we record this a couple days ago, I came across the Cal Newport podcast, he did a bonus episode. I heard him the episode before talking about, I need to address the AI elephant in the room, but I'm not sure I want to make that part of the vibe of the Deep Life podcast.

Maybe I spin it off as its own separate thing. Anyways, he ended up doing a bonus episode. And one of the things that he talked about, which I completely agree with, is that a lot of the narrative around AI is going to cure cancer and do all these things is being driven by the people who are building the AI tools. And of course they want to describe it as the next big thing.

And he kind of made the point that they're all kind of struggling to figure out how this is going to work. And I just appreciate his approach because he's a computer scientist. He talks about focus and the deep life and a lot of the same themes that we do here. And so if you're listening to this podcast and you're interested in trying to understand this stuff, I feel like that's a good resource to do so. I'll try to find the link to that bonus episode and put it in the show notes. But the sky is not necessarily falling.

Just take what you hear with a grain of salt. And then like you were at the main point you were making was how do we use this to do more? of what really matters. That's the focus angle to this. It's not like, well, AI can do everything or AI can do nothing, but how can AI do the things that I don't want to do so I can do the things that I do exactly. That's refreshing to hear that Cal Newport agrees because I feel like he's a lot more knowledgeable on computer science than I am. He's a PhD. So I'm going to go listen to that myself. But I do think that what's happening is let's flip that earlier quote on its head. Let's get the AI to do our laundry so we can do our creation. And the thing can do it now. We're going to talk about that

a minute, but it really can. And there's reasons why. But donkey work is really insidious for a lot of us in the sense that it causes an disruption in our focus. So donkey work can be, we all have jobs that where, you know, part of our work isn't actually the work. It is updating the spreadsheet, it's setting the meeting. It's all the nonsense that we do every day. And one of the insights sites I've had, as I've been building this robot out, to help me is I just did not appreciate how much of a focus block this is. Because I always hated doing this stuff. Like somebody sends me an email and says, oh, my password, stop working. That's a fairly easy process for me. I go to the website where I sell the course. I find the customer account. I go to the

go there, I push a button. Then I, you know, I draft a thing to them saying, you know, it's been reset. But the problem is whenever I do that, that completely stops me in my tracks in terms of creative work. So imagine me sitting here working, take a break, check email, and I've got to go do this stuff. And then when I want to go back to the work, it's really hard to get rolling again. And maybe because I'm getting older, I don't know.

I think this has always been a problem for me. But the reason I hate it so much is not that it's hard work. It's that it blocks me and it's like it stops the focus train. And then it's like really hard to get back on the train. And the same thing with the blog posts. Like when I write a blog post, I get the post done. I write it, you know, my editor, my text editor. But then there's a whole process I've got to go through to get that onto the website.

And so if I'm writing posts and I stop to put them on the website, it's really hard again. It's this jerky stop, stop. motion that really gets in my way. And now that I've got this robot kind of doing for me, I really kind of appreciate how much of a barrier that was for me. Yeah. And the term donkey work is something that I've heard you say for quite a while. And I'm glad that we're devoting a whole episode to this because I think it's one of those things that you can kind of generally grok what it is. But I know I'm going to take a little bit of credit for this episode because because I sent you a message a while back.

You used the term, and then both Nick, Nick Milo and myself quoted it in a newsletter. And I was like, you really should have a signature piece of content you can point to. It's like, this is what I mean by doggy work. So I'm glad that we're diving into this specifically here. And the whole robot assistant field guide that you built on top of this concept, you know, I didn't really know you were going to go that far with it. that however has opened up my eyes to a whole new dimension of this that even I did not consider when I made that comment you know I thought maybe you'd have a little bit of uh of uh some additional

context for like these are the sorts of things that you can chat with an LLM and and get some feedback on and make it a little bit easier maybe you know cloud co-work you can automate some of this stuff a little bit, but you've really kind of gone deep into the code aspect of this and providing it the context and really just like treating it. I've heard you say, you know, treat it as an intern. You've built the intern. Yeah. And I think that's a brilliant approach to this. Yeah. And we all have different donkey work in our life, right? No matter what your job is, whether you're retired, a student or in the middle of your career, there's things in your life that you have to do that involve computers in the internet that is not the work. And that's what donkey work is. In the old days,

if you had to carry something up a mountain, you gave your donkey three carrots and a pat in the head and your stuff got carried to the top. Now you've got to babysit a little robot and develop one and you get the same benefit. And I'm just kind of surprised it's not a bigger deal for people to understand. Part of this was the open claw kind of explosion where that was a system people were using with like a Mac Mini where you run an independent computer as an independent agent.

But we'll get into the mechanics of it more. Another thing I was thinking about donkey work is the way that I feel like it's an excellent off-ramp for procrastination when you've got that kind of busy work in your life. You know, you're really, you're getting paid or you're living your life to do the hard stuff. You know, that's the reason you go to the cohort with Mike or you go through the productivity field guide. You figure out what's important to you and you want to pursue that.

But there's a resistance inside you that doesn't want to do that hard work. We all face that. And donkey work is such an excellent excuse to not do the hard stuff in your life. So not only is it a block to you being focused. It's also a prime excuse for that terrible. of your brain that doesn't want you to live your best life. So, I mean, there's actually a lot more to this than just answering email. Right.

And one of the things that I want to get into with that is where those lines are. I think that's probably the next main point on the outline about the threshold. Because just because you don't want to do this thing, I don't think that means you should automatically automate it. There are some things maybe that you feel like I need to do this. And if you take a step back from it, you realize I really. really don't need to do this. That's a distraction.

So don't automate that stuff, but the stuff that you do need to do that you don't really want to do that's getting in the way of doing the stuff that's the high leverage, moving the needle type stuff, that's what you're talking about with donkey work, right? Yeah, exactly. I mean, again, building this robot can be another prime source of procrastination if you don't, if you're not careful. But I think you go into this as an adult and say, look, there's a new technology available that allows me to offload a bunch of this donkey.

work that I used to have to do or have to hire somebody to do. And now I can do it. And so you take an adult approach to it. You know, you take it one step at a time. And over a course of months, you've built out a reliable assistant that does a bunch of work for you that used to do yourself. I mean, the other piece of that is if you get it working for you, that doesn't mean you automatically take on more work, you know. Let's not make the same mistakes of the past here. And I'm not teaching this with hopes that everybody's going to try and increase their output by 50%. What I'd like you to do is have more time to read a book and take a nap. But it's totally within our reach now.

And you don't need to be a programmer to do this. It's all plain text. And we'll get to it. But the thing is built on Claude. You want to start there? Yeah. Yeah. So when you're talking about a robot assistant, there's lots of different.

tools that are out there. And I think chat GPT was for a long time considered the one that you should use and you could do the custom GPTs and stuff like that. That was great. But Claude has definitely emerged, in my opinion, as the one to pay attention to. I think you would agree. They just are building all the pipes to connect this stuff in useful ways. I believe they're doing it the right way.

And when you're talking about Claude, there's three different pieces of Claude. There's, as we record this anyways, who knows what they add tomorrow. There's chat, which is the standard, you know, you're having a back and forth conversation with an LLM. There's code, which is the next piece that was developed, and that was really for developers. And then there's co-work, which is kind of in between. And I have to admit that co-work is the one that I haven't really dipped my toes too far into the water. I'm using it a very, very, little bit, and I got a whole bunch of ideas after watching some of the videos from your robot assistant field guide. But there's basically three branches of this, and you can use probably

any of them in specific ways to alleviate some of the donkey work, but the way that you've approached it using code, I think, is probably the one that provides the biggest bang for your buck. Yeah. So the big piece is, so they make cloud code, which allows you to program and and improve your applications. Mike has used it to develop plugins for Obsidian for himself. It's really powerful. But what they found was that a lot of these programmers were also using it to manage their calendar and stuff.

Because if you're on a Mac, it has a terminal and you can program the terminal. So they were writing code to kind of do productivity-related stuff. So then they, at Anthropic, they use Claude Code to generate Cloud Co-work, which is kind of like ClaudeCode. for normal people. And it's really, it's really impressive.

And we'll get into what it does in a minute, but it is kind of the next iteration of this. But it solves so many problems that you, you couldn't solve with chat. It gives it memory, it gives it access, it gives it ways to do things. And just anthropic in general,

if you look at kind of their company direction for a long time has been make useful things for business and for people, I feel like, of all the big models out there, they're the one most focused on people like you and me and the listeners trying to just get the stuff to work for them. And as a result, they've got all the pipes built,

like you said. And I don't know this is the end of the game here. I think people are going to pay attention to this. I feel like this is a great opportunity for Apple. Like if they would build this type of functionality into MacOS where it was private and local, that would be even better. And I don't know if they've got the foresight to do that.

guess we'll find out. But the fact is you don't have to wait for Apple because Anthropic has done it and you can start running this stuff today. Yeah. And the key to making it work, I think, is in the perspective that you have in trying to work with the robot assistant. I was listening to a Nathan Berry podcast where he was interviewing someone who helps small businesses and solopreneurs specifically incorporate AI into their businesses. And he used the analogy, which I thought was brilliant, of AI as a souschef. So you can assign it specific things. And you can, it can do useful things.

But it's not a done button. You don't just say, do the thing for me. And then it spits back the output in a good enough quality that you can actually use it. You have to give it parameters. You have to give it context, all those sorts of things. and that sort of transforms the way that you interact with all of these tools, but with co-work specifically. And like small time creators like me and Mike,

we have so much donkey work in our lives. I think this kind of job really comes with it. So that makes us very eager to solve this problem. This episode of the Focus Podcast is brought to you by Gusto. Go to gusto.com slash focused for payroll, HR, and benefits simplified. Get three months for free. As the new year begins, a lot of us are trying to get our business operations together. And honestly, having payroll benefits and HR handled by Gusto

feels like starting with a clean desk and an organized inbox. So you can focus on actually growing your business. When you're running a small business, handling HR and benefits is a huge time sink and it's dangerous if you don't know what you're doing. Gusto will solve that problem for you, saving you time, money, and risk. Gusto is an online payroll and benefit software built for small businesses. It's all in one, remote-friendly, and incredibly easy to use. So you can pay, hire, onboard, and support your team from anywhere.

They've got it all. Automatic payroll tax filing, simple direct deposits, health benefits, commuter benefits, Workers' Comp, 401K, you name it, Gusto makes it simple and has options for nearly every budget. And if you've got a question, you get direct access to certified HR experts, to help support you through any tough HR situation. It's the number one payroll software, according to G2 for fall 2025, and trusted by over 400,000 small businesses.

So try Gusto today at gusto.com slash focused and get three months for free when you run your first payroll. That's three months of free payroll at G-U-S-T-O-com-slash-Focused. One more time, gusto.com slash focused. And our thanks to Gusto for the their support of the focused podcast and all of relay. Okay, so you want me to dig into the donkey work threshold? Let's do it.

Yeah. Where is the line where, okay, this is officially donkey work? Yeah. Well, first thing, I think we need to understand some underlying technology with Claude Co-work. Historically, we've had chatbots, which are great. Like, you can feed it text and it can fix it, and it can make it sound like a pirate. It can do all kinds of great things in chat.

but it has two real limitations. The first is memory. It doesn't remember anything very long. I'd like to think of these AIs as like having an amnesiac intern or an employee who can't remember anything. They're very good at following the manual, but they forget the manual every 15 minutes. So that's a problem. And then the second is access.

If it's locked in the chat window, then you've got to copy it and paste it. and it can't really do a lot of the kinds of work we want. This was a tough problem like six months ago. And then Anthropic released MCP, Model Context Protocol. And so MCP allows you to connect the chatbot to external applications. Like as we're working on this, we just got one from Fantastic Al. They just released their public MCP. So now your Claude, your co-work, or your chat,

chat or code can understand your calendar, can see what's going on in your calendar. If you've got Google calendar, you've got the same thing. There's a lot of ways to do this, but that's just one. And third-party app developers are jumping on the bandwake and they're building these MCPs to connect them. So suddenly we're getting access. The biggest one of these is called clotting Chrome, which allows it to drive the browser. So if you do browser-based work, that's your key. So we've got access now. And then the question is memory. Well, Claude Co-work, the way it's engineered is when you started up, it says, okay, where would you like me to look? You know, what is my memory for this

session? Now, there are a lot of applications out in the world to help you take notes and keep track of information. For 15 years, I've been talking about text-based notes apps because I think they just seem inherently better. Mr. Schmitz is a big fan of a Obsidian and does a lot of teaching on that. Obsidian is a great one. Note plan is another one on the Mac that saves markdown files to your local drive. But suddenly these apps have their moment because when you turn on co-work and you just point it at your Obsidian folder or your Note Plan folder, then it's just got a bunch of

text it can read. There's nothing faster for these things to work with than text. You know, it doesn't have to plug into an MCP and go digging through resource files. it just reads the text. And suddenly, now the thing has a massive memory. As much memories you want to share with it, which is your Obsidian database, your No Plan database, or you could even just use a folder full of text files,

but I'd recommend using No Plan or Obsidian. And so you have the two big problems solved. It's got memory and it's got access. And now you've got wings to build your robot. So that's the thing. You put those two together, the existence of memory, the existence of MCPs, which are growing every day, and then you can really do something.

Like, just a simple example. I have that blog post problem. I was talking about the block wall of me posting to WordPress. I have taught Claude Co-work exactly how I want blog posts posted at WordPress. And that includes things like, what are the categories that you check? What are the tags? Who does it go to? Is it go to labs members or does it just go to the public?

There's a bunch of little fiddly bits you do when you pull, a blog post. Well, it has created a memory, a text file in Obsidian for me that has all of that documented. So as soon as I finish posting a blog post, I say, okay, ship it. It says, okay, boss, and then it goes in, reads the manual. It reads the instructions for how to do that again. And then it opens up clot in Chrome. And then it goes to my website. It logs in. It grabs the text out of the Obsidian document. It grabs the tag information. All. the published date, all that stuff, and it just takes care of it, and it's done. And when I look back at it a few minutes later, I've got a blog post posted on the website. And it's that easy. So now, instead of going

through all that myself, I just type ship it. And in that case, it's using the memory of what it's got of the skill to do it and the information I've given it in a Bissidian, and it's using the MCP, the access to do that. So you couldn't do that before with just a chat bot. So it's really changed the game now. So suddenly we have a lot more power available to us via AI than we did before. Yeah. And the thing that makes that powerful is the way that you've been working with those notes in Obsidian, where it's not a repository for every piece of information that you ever stumbled upon that might be useful someday that you stuck in an archive.

Yeah. Which is one of the issues I've had with some of the AI tools that have been released as plugging. that work inside of Obsidian is I don't want it to just, you know, search everything. I want a very specific subset of information when I am in a specific context, when I'm trying to do a specific thing. There are certain things that are relevant and basically nothing else is. And so I don't want you to just, you know, pull from everything that I've ever collected and say, you know, here are 10 things that you should be thinking about as you're studying whatever

whatever topic, that has limited usefulness for me. But if you are the way I've been using obsidian and I think you have too is you've created these artifacts of important documents. It's almost like a living will, SparkyOS, I know is the term that you've used.

But it's like this is the way that I think about things. These are the things that are important to me. Which for me, tying into the life theme stuff is a lot of like big picture core values type stuff. It's not necessarily just the standard operating procedures. When you send an email, this is the way that you do it. When you write a blog post, this is the formatting. These are the buttons that you push. That sort of stuff is good. But the real thing here is not, you know, if you've

just been sending stuff into an archive and then you think you're going to appoint Claude, co-work at it and get valuable output, you probably have a little bit more work to do. But it can also help you refine this as you go. The thing that's going to make it work, though, is that perspective of, you know, these things are, are not just little bits of information that are in a filing cabinet somewhere that I can dig up if I need them. But, you know, what are all the, the little things that can be useful for me in terms of creating something new? On that point, like I have a folder in Obsidian called Meditations. And I stole it from Marcus Aurelius, like everything. But there are hundreds of essays in there that I've written over the last 15 years.

It's like when I bump into a concept that I'm trying to figure out what I think about, I always write an essay about it. Some of them I've handwritten and scanned in. Some of them I've written on a keyboard. A lot of them I've dictated because that's the way I like to input text. But there's just hundreds of essays. And that is kind of my belief system. I've never shared it with anybody. Like nobody reads it except me.

But the process of creating it helps me in my own mind get clear how I think about things. you know and some of them are really short and some of them are thousands of words but that's there my robot has never I've never directed it to read that maybe it has I don't think it I don't think it would unless I directed it to it has nothing to do with what I'm doing with the robot the robot's doing donkey work you know it's just sometimes stuff out of that will lead to some donkey work usually it's more related to production and kind of life management things but this is not what I'm using it for I'm not using it for the Sparky OS management or development. It's just, this is a thing that's to the side, that it's a tool that can go and do donkey work for me, so I don't have to. I would argue, though,

I'm curious your perspective on this. If you don't have that piece at the top, that's like the top layer of the PQM stack framework that drives everything that I do. And what that does is it provides the clarity for what are the things that are actually worth doing. Yeah. You know, if you've done that sort of. of work at the beginning, then you can create the documentation so that the donkey work the robot can do easily. But if you are starting from a place where you haven't really thought that stuff through, I think it's possible that you procrastle work on building your robot assistant to do things that really don't matter. Yeah. No, I mean, the fundamental precept here is you have to figure out

what's important to you. Like, exactly. None of the stuff matters if you don't know, you know, use whatever tortured analogy you want about the latter against the wrong wall or whatever, but you've got to figure out what's important to you. But one way I have kind of included it is because I built out a task system with the robot. And one of the things I did, because it knows of my, in my life, it's the role system. It's based on these roles I fill in my life. It knows my roles that are most important to me. And I said, index my whole task system against my roles. So go through each one and assign what you think is the appropriate role to it. And if you see something that does not fit on my existing roles, flag it.

And so that was a little bit of interaction between kind of my foundational system and this robot. But that's about as far. And it was interesting because they did find some tasks that really didn't fit in any roles. And so I deleted them because I don't want to do that stuff. The roles are the basis. So there's some use for that. But the real foundation for what I'm talking about here with this robot assistant is donkey work and just get that stuff get the laundry done so I can do the art it's that simple

yeah and it doesn't just need to tie to obsidian you know I'm looking at the list of connectors that are available in here and it's insanely long I mean you can connect to notion and Gmail and canva slack and all these other other tools but the the tendency I think and this is kind of when I first looked at co-work I was a little bit overwhelmed by all this stuff that it that it could do. And I realized that I was being pulled in a direction where it's like, well, let's just connect all the things and let's automate all the tasks. And that doesn't work. And I would make the argument, the best way to do this is with a local text bank. However, you do that. Because the MCPs are great, but sometimes they break. And sometimes they're slow. And like, it's really nice. This text is so easy for

this computer to digest and read and process. But the big point is you can set this up today. This technology now is there for us. It's been democratized for us, normal folk. And you know, you've got to figure out the skills to get it going, but it's not that hard. And once you do, then you can do some great stuff. Yeah. And you don't need to have it do everything at once. I will admit looking at your list of skills at the beginning of the robot assistant field guide, it's like, wow, he's been doing this for a long time. But as you show in the field guide, you actually are using the way, the back and forth that you have with co-work to create that stuff. So the place to start is really to identify even just one specific use case of donkey work where it's like, I am spending time on this thing that is stealing away from the most valuable thing. that I could be doing in terms of my probably work.

I mean, I'm not sure you probably could use this in your personal life as well, but I think it's primarily professional. Professionally is the way we're talking about this, right? Sort of. I mean, I also, like one of the things that I like to do is I like to log things. Like when I have communications with a sponsor or when I buy something or when I have an expense, I always want to log it. And I've done this manually my whole life.

It takes a lot of time. I wrote fancy Apple scripts to create links to emails and do all this crazy stuff so I can get back to this stuff. And now, like, it logs all, everything for me automatically. I don't even think about it. But also, like, personal expenses. I bought a new stropping stone yesterday for my shop. And the receipt came in. So I just went ahead and forwarded it to the robot and said, log it.

And so now I've got a collection of woodworked. working tools I bought and said, oh, would you like me to create a spreadsheet for how much money you spend on woodworking? I'm like, yeah, okay, go ahead and do that. But it's like, you know, so I've got some personal stuff going in there on a logging capacity, but most of my personal stuff doesn't really need this kind of treatment. Right, right. Well, should we talk about some of the ways that we're using it? Yeah, let's get into it, man. And let me just qualify that I've been working on this for months and mine has come a long way since I first started. When you start doing this, I recommend you start, like Mike said, with one or two things,

build it slowly. I kind of broke it up in categories. And for me, the first category of things I've got it doing for me is communications. Email is always a challenge for me. I like corresponding with listeners and lab members. But email in general to me feels like a burden. You know, the old E.B.Y thing about the morning mail is my enemy. me and it can really distract me. And so I've got it doing email triage for me now. It goes through

and checks for me for anything that's like critical. Like when I do a startup and shut down routine with the robot and the first thing it does is just, you know, did you have an email coming last night that you absolutely need to deal with right now? Or at the end of the day, it does the same thing. And that's really good because then I don't have to go waste a bunch of time in the email. I just have it surface what's there. communication-wise, it also does Slack, because it has a Slack MCP, and I always am concerned, gang, I'm not good at that communication piece. I don't check text messages during the day.

I don't, and it relay we used Slack, and we had a thing happen where the editor needed me to do something, and I didn't see it for a day and a half because I didn't open Slack, and I felt terrible afterwards because it created more work for other people. And so now it checks for me in the morning and the evening. It checks Slack and see if there's anything in there that I need to check. And so it kind of fixes that for me. And then, of course, it also logs all the email stuff that I go. I have it at the end of the day read all my outgoing email for the last 36 hours

for anything that needs to get logged. It's like, oh, you wrote the sponsor about, oh, you wrote the sponsor about, you know, something they're doing. You want me to log that? Yeah, because for each sponsor, I have a record. with a list of all the dispatch log of all the communications. So if there's ever a question, I can go back and look at it. I think it's the old lawyer in me coming out, you know.

Another communication element for me is I have people that help me on the back end. And as much as I like robots, I want humans in the loop. And when I have an automated blog post go up, it doesn't just post. Another person reads it other than me because at that point, I feel like I know what it says and I can't, I'm incapable of proofreading. So a notion record gets modified so that, person could get notice, oh, this is up in WordPress. I need to go check it before it gets published. So I have a lot of communication stuff being handled with it that is in my mind donkey work.

The proof reading that you just described, I'm glad it's not just me that has issues with that. Yeah, it's true for all of us. Because first I write it in my mind, then I usually speak it with my lips and then my brain just connects my memory with what's on the screen, not the actual words, happens all the time. Yeah. Yeah, do you want to go back and forth or do you want to crank through your list and then I'll share mine. Okay, I can continue if you'd like. I've got, the other one is I call daily management stuff. And that's like the calendar tracking and some management. I really have come to appreciate the robot assistance on the startup and shutdown routines, you know, especially because for me, the big one is shut down. And that's when I do

go through email in the evening. And so I can, like, crank through email while the robots processing the shutdown for me and just throwing up, okay, these are the things. What do you want me to do with these? And the way I trained it is, like, when it's processing email or any of these tasks I say, give me a suggestion and what you think you should do with it. And then I tell it if it's right or wrong, and over time it updates itself. It's one of the real tricks of this. If you go through the course, is you've got to keep updating it so it learns with every session, but it's got really good now at knowing what I wanted to do with different kinds of information. So mainly, I'm just like answering email, looking over to the right at the robot and occasionally saying yes,

you know, then it goes and does a bunch of stuff. One of the things I do like is I track website links for the week because at the end of the week in the Max Market Labs, I do a like a 10-minute podcast for the members called The Lab Report. And I update them on Apple news and rumors from the week and a couple other things. And people like that. And it just gives them a one week summary if they don't, you know, watch all this stuff carefully. But the workflow I developed was with Reader, you know, from Readwise. And Reader has an MCP, you know. So now as I'm going through the week and I check the feeds, I can just tag something to Lab Report or I could tag it lab report and blog. And then at the end of the day, it goes through and checks the MCP and says,

okay, you tagged four things for the lab report this week. Okay. So then it has a text document for Friday for me to record the lab report and it collects those links so I can go read them and record it when the time comes. It says you tag this one for a blog. Yeah, okay, I want to write a blog about that. Okay, I'll create a blog content item for you. So it just does that for me. It's like that kind of donkey work that just keeps kind of getting out of my life. Do you, one of the examples you have in the robot assistant field guide, which knocked my socks off was the daily briefing. Do you still use that or how has that workflow evolved since you put that together?

And do you mind kind of just talking a little bit through what that is? Yeah. So you know how like if you're the president, you walk in the oval office and someone hands you a piece of paper that gives you everything you need to know for the day. I've always kind of wanted that, right? I wanted somebody to do that for me. So now the robot does it for me. Once it goes through, because I do the heavy lift at the shutdown. So it knows what are the tasks for today or my available task? What are the appointments? What are the appointments for the next couple days? What are the upcoming deadlines in the next week? And it generates an attractive report for that. Like I customize

I told it the fonts I wanted to use. And it's a real process, you know. But at the first thing in the morning, it does one more sweep and it prints that out for me. And then I've got it on my desk, a piece of paper. Just kind of a fun thing. A lot of people are doing that with OpenGlaw, too, where it's automated. With this, I need to trigger it. But it's very useful to me.

And the important thing I want to say is, like, this is an assistant to me. It's not my boss. So, like, when it gives me a list of available tasks, it doesn't mean I'm going to do them all. And in fact, I've told it that. It's like at the end of the day, the beans is like, okay, I gave you, these were the things.

Did you do them all? I'm like, no, that's not your job to check up on me. Your job is to tell me what's available to me. I was joking with somebody in the Max Sparkett Labs. I feel like if there is a robot like uprising, they're going to kill me first because I am so mean to this thing. I don't treat it nicely.

I just tell to do the work. It's a computer algorithm. it's not a person. That's one of the reason I call it the robot assistant. I'm not going to give it a cute name. This is a tool I use to get work done. But ultimately, for me, it has to go on the field notes. You know, I write the stuff down that I want to keep. But I do like having the little one page around my desk. But yeah, so the startup and shut down stuff is really kind of nice, because it does feel like you have an assistant kind of helping you do that stuff. Well, I had to laugh one point in one of your videos. You made fun of the fact that you actually said, please,

you're prompt the robots. I don't think you're quite as ruthless as maybe. I try not to. I try not to, but sometimes it comes out. Hedging your bets. Yeah, there you go. Another category for me is content publication. And I kind of talked about that already, but some of the content publication I do is very complicated. In the Max Mercury Labs, we published like two videos a week to members to like some kind of learning. And so that has to go to the labs, a memberful site, has to go to WordPress, has to go to Notion. It's got to do a lot of things. And that skill now is just rock solid. It just does it for me every time. I don't think about it. And that's really nice. Another category that's great for me is Max

Sparky, is file organization. Because I've been teaching how to name and organize files for 20 years. and it's always been a battle because, you know, when you get a 100 documents that all say scan number X, and you just want to name them, like, there's been ways to do it in the past, but they're very complicated, and you have to make kind of arcane rules, and they're a little brittle,

and they don't work that well. And now it's just a solved problem with this AI, because it can just look, like the way I taught it is, look, read the first page of each PDF, and give it a name that starts with the name of the date of the document, not the date scanned, but the date of the document and the category of document and the name of the document. It's like, okay, and it'll do that to 10 documents or it'll do it to 1,000 documents,

you know, and then you say, okay, now come up with a folder structure and organize them by folder and do that too. That is a problem that we've been trying to solve for a long time. That's all there. Another thing I do with it is sponsor stuff because I have sponsors for the blog. and the newsletter and some other things. And I was always a little flaky at that. Like I felt like too often I'd be writing the sponsor Friday night for a Monday morning

post. So now the robot prepares the email for me two weeks. It tells me when they haven't responded to the email. It does the invoicing. I've got a fresh books account. So it goes on through Chrome and it creates the invoice and it checks each box that I want to do. And like that whole thing is just kind of like off my plate now and much better.

So just so many things. Customer supports another one like that I talked about, like, you know, resetting passwords and, you know, dealing with download codes or whatever the problem is. It knows how to do all that stuff. Like the code we talked about today, the robot focused, the 10% code, my robot went and set that up. I didn't set it up, you know. Nice. So I'm doing a lot with it, Mike.

And I'm finding I'm kind of getting to the edges of what I needed to do. Like most of the stuff I've kept in my play. I want on my plate now. But it's pretty great. And it's just shocking to me. I didn't realize six months ago that by now I would have this much automation with AI that is this useful. Yeah.

So you are definitely the Yoda when it comes to the donkey work approach to AI. I honestly have not used to use co-work a whole lot. I've used Claude quite a bit, but it's primarily been with the chat and then also. the code piece. I'm inspired by your robot assistant field guide, and I feel like some of the stuff that I've done as one-off things in chat could be automated using co-work, and I could create essentially standard

operating procedures or rules for automation based off of that. So I'm going to start doing some of that stuff with some of the back-and-forth creative projects that I've been doing, and maybe you can tell me if I'm using this wrong. But maybe this isn't fit with with co-work. But I want to develop, you know, some of the personalization in the context in terms of the skills. So one of the things that I do all the time with, with Claude, is I don't use it to write my scripts or my newsletters or anything like that. But I always get feedback on it.

And I will occasionally use it for giving me ideas. Specifically with YouTube, I'm pretty new to YouTube still. and I just don't think in terms of the YouTube algorithm and what makes a good title, what's a good hook. I'm still learning that stuff. So I'll, you know, I want to make a video about this. So give me, give me 10, you know, title ideas. But I recognize that I could create skills based off of the ones that I've already published

on my channel along with some best practices that I try to keep in mind, you know, keep it under 55 characters, stuff like that. So I wanted to help me write the titles and the hooks for YouTube specifically, social media format or social media post formatting for like LinkedIn and things like that. But yeah, I guess the other things that I've got here kind of don't fit into the donkey work category. But you describe with the naming and categories for a YouTube video, like if you script your videos, right? Yes, yes, I do. So you've got a script there and you could make a skill that says read the script,

give me 10 title ideas, give me tagging ideas, you know, because the script is the ultimate source there and it's your words. So let it kind of help you do that and then run them. Like a reverse version of that that I do is when I finish recording a podcast, I take the podcast transcript and I give it to the robot and I say look for any newsletter ideas or blog ideas on things I said during the show. Because I've never been good at capturing because we talk about a lot of stuff on the show that could actually be useful for me to expand on the blog.

And so it goes through and it finds, you know, three or four ideas and it says, these are the ones. I'll just say, yeah, numbers one and three do those. And it'll make, once again, it doesn't write it. It creates, I call them content records. And the content record is like an empty shell. but it's got all the front matter listed and it's it's tagged for the blog and you know tags the status of it so I know if it shows up on my radar or not and so I it's kind of knows

to set up the shell for it so that I can just go in and write it when I'm ready and it's kind of nice because then I'm like okay I want to write blog post today look through my list and give me you know recent ideas that I've had and those will come up and then I'll turn it into a blog post and that's kind of nice Yeah, and I do do that with the scripts. I'll upload the entire script and use that as the context. But I've been doing that via chat. And it occurs to me with co-work, I could say all of my video scripts are in the video scripts folder.

Look at the most recent one and give me some title ideas because it can look at the modified dates and all that kind of stuff. And I would actually do more than title ideas. I would say, you know, make the appropriate metadata, you know, all the little. little donkey work pieces of this. You know, give me ideas for a thumbnail image. You could really just kind of let it. And then that would be a process or a skill you give it and say, okay, I finished the script.

You type, paste in the name from Obsidian. Do your thing. And then all of a sudden you've got title ideas, thumbnail ideas, all the stuff you don't want to do. Yeah, yeah. So there's lots of, I'm, my eyes have been open to the low hanging fruit that is right in front of me. So I got some work to do.

But primarily the way I've been using it is as a collaborative, creative partner, as I mentioned. And then also using code to solve specific paper cut type problems in my workflows. So one thing in particular, I may have mentioned this previously, but I developed an app and I'm calling it Feed Deck, which is basically my specific way that I want to deal with RSS feeds. Yeah. So it's sort of like a social feed aggregator. It brings in all the posts from Blue Sky from Mastodon. It'll bring in like subredits that I follow. It'll bring in specific like YouTube channels I'm subscribed to and then RSS feeds and puts it all in a chronological timeline.

And I basically want to just scroll through that quickly. It's kind of like just browsing the headlines. And then if I want to see a preview of something from my keyboard, I can hit tab, get the preview. And then if I want to save it to read later, I hit Sust. and it sends it to InstaPaper because InstaPaper works really great. Sidelode it on my SuperNote Nomad. That's where I want to read stuff. Yeah.

You know, so I like built this app to fit my workflow that, you know, previously I would have just tried every other RSS reader under the sun until I found one that didn't hurt so much. Yeah. So now is that feed deck? Is that running an obsidian?

Is it an obsidian plugin or? No, that's a standalone app. Actually, I've shared it with a couple of people and they're like, hey, yeah, I'd use that. So I actually set up a test flight for it.

I maybe we'll release it as an app at some point. But it was just something like, hey, I wonder if I can build this dream workflow that I had. And there've been lots of other like one-off apps, quote unquote, that I've built just for the way that I work. Like I built one called book quote

that creates those readwise style quotes, but there's a space there for the bookworm episode. I can put in the URL for the episode that we talked about it. I can put my star ratings in there, things like that. Yeah. I built one which takes the keynote template files that I had for my LinkedIn carousels, which I, you know, I'm trying to eliminate all the friction before I start, like, really going hard with that particular social media platform because YouTube is primarily my,

my place that I show up in terms of social media. But the carousel generator, essentially, like, I don't have to worry about formatting text anymore. I just type the text in, select the template that I want. BAM, it puts it in there. It's like an easier version of Canva based off of my specific brand guidelines. Yeah. I struggle with screenshots sometimes. Like there are certain products that you just can't find an Apple frame easily for. So like I have my specific ones that I use all the time and I have the ability to throw a gradient behind it real easily.

And then I've been using this app called Granola to create meeting summaries of like Zoom meetings and things like that. But they want $15 a month now. So I've vibe coded a menu bar app that does the same thing and gives it to me and better markdown that is more consistent with the type of formatting I would use inside of my meeting notes and obsidian stuff like that. So you've got really hard at clock code is what you're saying. Yeah, but I think I've also done like one-off stuff with both cloud code and chat. So like you're talking about MCP. And the real value of that I'm realizing is once you get it set up, you can have this system that you build that formats things a certain way every single time.

But I have MCP set up and I used it to update 1,400 daily notes because they were formatted, a specific way and I needed it and formatted a different way. I moved actually my daily questions scores from tags and the note to properties at the top. So I've been, I have to create the recurring systems basically for that sort of stuff. But the one-off things I've been doing. I've used chat, for example, to create summaries of like explainer videos. I just redid the whole member guide inside of the library. And I wanted to create these explainer videos like this is where you go to do this sort of thing,

hot seat coaching, all that kind of stuff. And so I recorded all these videos. And then I uploaded the transcripts one at a time in the chat and said, you know, give me bullet point summaries of this. And then I took those summaries, pasted those in the lesson pages and then added the tags to the appropriate spaces, you know, links, that sort of thing. And that's the sort of thing like in the list of things I needed to get done, you know, this week, that was a one-off task. But if I think about it for a little bit, I realize that there are a lot of instances of that type of thing.

Like that's that exact workflow or 95% of it happens over and over again. So very low-hanging fruit in terms of building the systems to do that. Just have it created the skills database or the index, I guess, is what you call it in the robot assistant field guide and set all that individual stuff up so that I don't have to think about it again in the future. Yeah, so let's talk about that. If you want to set this up, what's involved. So we talked about the two pieces, access and memory. So the first thing you need is a memory bank. You need somewhere that the thing can look when you open co-work and what do you point to that. You could do this with just a folder full of text files. You could also like set up the

MCP to Notion if you're a big Notion person or something. I recommend that you, is either Obsidian or Note Plan. Note Plan is a native app on Mac. If you want native apps, that's excellent source for it because it's just, again, folder full of markdown files. A lot of people like Obsidian.

I know Mike you like Obsidian. In the course, I teach it through Obsidian because I wanted it to be an app you didn't have to pay for to figure this out. And Obsidian's great for it. But a folder full of markdown files

is the memory. So you set up Obsidian or Note Plan and treat that kind of is the thing that you pointed to when you're getting started. And then so you start up co-work, you pointed at that memory bank and then you get started. Now, there's a couple things going on here technically. At your source folder that you pointed at, Claude wants to create a thing called clod.md. And that's kind of like it's basic operating instructions.

And that's the first thing the robot reads when you started up. So you start putting basic instructions in there, you know, how to, does it refer to you? What is its purpose? You know, kind of just basic stuff that you have it doing. And this all evolves over time. But what you find out quickly is that you can't just load everything into that Cloud MD file because as you teach it more skills, it needs more manuals to read. So then you start adding skills. The term they use is skills, and I think that's a fine term. So like talking about Mike's thing earlier, he could make a skill to read a note for a YouTube script and add five title ideas, tag ideas, thumbnail. It goes through and has a list of things

it does. When I talked earlier about, you know, the complicated nature of putting a video up for the Max Sparky Labs, there's a skill file on my computer where it knows all the details at each website that it needs to go to. What boxes do I check and how do I do this exactly the way Sparky wants it done? And so each one of those is a separate skill. At this point, I probably got north of 70 skills that I've taught it over time. And so the second layer problem there is you get so many skills, it doesn't know which one to go. It's amnesic. It's going to forget. So then you also have it build an index. So if you look at the hierarchy, it's clod.md.md is the basic instruction. One of the things that says is always read your index. And then the index is a list of all the skills it has and how they

work and what kind of words you say when you want it to use them. And it should always have knowledge of everything on the index because it has enough memory to remember that. And then so when I say, okay, now go do the post the video thing for the labs. It goes to its memory, active memory, it says index. Oh yeah. There's a skill about that. Okay. And then you'll watch it. And it'll say, okay, I'm reading the skill about that now. Okay, now I remember how to do it. I'm going to go do it. and then does it. You know, so you've got to build out that three-tier system, claw.mdd plus index plus the skills file.

Now, in 10 years, are we going to have to have all that structure? No, it's going to be way better. It's going to have better memory. But for now, this is the way you do that. And that's what this course really goes into is how do you set those things up? You know, the course has got three hours of initial training, which really kind of walks you through all that set up. And I built a skill that you get with the course,

and it's called the assembler. So when you watch video two, then it triggers video. You say, okay, I just watch the second video and then it does the assembler for that. And it helps you set up a new skill or set up your index. It goes through and does it. And as you go through the course, there's additional things like one of them. It builds like an auditor for you to check them because you want to kind of make sure the skills stay current. But the course idea is you go through these three hours of training and you've got everything you need to get going.

The 10th video in the course is like a 45 minute. video where I walk through a day with my robot so you get a good idea. Some people are telling me that they wish they had watched that first because it gives them kind of context for the course, but I think a lot of it wouldn't make sense if you hadn't watched the beginning of the course. And then after that, we're going to do a 10-week workshop. I mean, this is an extended workshop series, and we're all going to get on the phone together, on the Zoom together, and we're going to just, I'm going to explain deeper, like, how do you run a task system or what is it? You know, each week we're going to cover a different element of it. With the courses I make, I always give a pre-release to the

MaxRocket Labs members they get in early. And we just had a call on it yesterday. And it is shocking to me how many of the members just with the three hours have built like really impressive robots. Like one of them is a Red Cross volunteer and does like really high end work for them. And like, he's running it off his robot. I mean, it's just, it's awesome for me to hear how people do this. It's a strange field guide for me because it's not teaching you the ins and outs of an app. It's basically giving you a construction kit to build a robot. And then your robot at the end of your work and your experience is going to be different than what I'm showing you in the course because you needed to do different things that I need

mine to do. It's just a weird kind of trippy thing. But it's a lot of fun and go and go check it out. I mean, we've talked enough about it today. But the 10% off focus robot. and if you want a little peek at where things are going, I think this is the way. Agreed. This is a good one.

And there have been a couple big things in there. One of the aha moments for me, just to double-click on something that you just mentioned, was the different layers of the context. I've always viewed context of working with one of these LLMs as you upload a file. And it can go so much deeper than that. and that really allows you to create these systems,

which is sort of blowing my mind and also really exciting, because that's generally how I think about things. I've got a lot of work to do, but I'm really excited about the premise. And then the other thing that really clicked for me, which you just mentioned, was that manual essentially in the guide you talk about giving it the manual every time it needs to do something.

You don't have to manually upload those files every time. You can create the systems where it knows where to look for all the stuff, that it needs to do, and that kind of is a game changer with this stuff. Yeah. I mean, it's interesting because to get into the weeds of it, in Claude, they have a system called skills, where you upload the skills to your account. And that is kind of like the genesis of this idea. But the way it works with Claude is their read only.

The robot can't change its instructions. You upload them. And when I first started on this journey, I would say, okay, I gave you a skill to post a blog post, but now I want to change it. So let's rewrite the skill. And then I will download it as Markdown. Then I will re-upload it to you in the back end. So now you've got an updated version. I did that for like a couple weeks. And then it occurred me, hey, wouldn't it be great if I didn't have to do that every time I change the skill? Because you're constantly changing these things as you develop it. And so all I did was I put it all in Obsidian. And I gave in my claw.md dot MD file, it says

obsidian is a source of truth. Don't read your internal skills. Read obsidian. And then it's found. And now it's constantly evolving. These instruction sets we give. And something somebody asked me, I think it's not obvious until you start doing it.

But I don't write them. I just tell Claude what I wanted to do and it does it. And then it writes them. But they're in plain English. So I can go read them. And if something's wrong, I can literally just take a text editor and fix it. Or I can just tell it, hey, that step you did,

that's wrong. you should do it this way. And then it'll go and refigure it out. And the other thing is weird is when it started connecting them. Like I've got one, I talked about the one that where it goes through and kind of looks at a transcript for ideas for blog posts. I've got another one where I have it read the transcript of the podcast to give me edit points and links I can include.

Like I've always been bad at remembering to add links. So now the robot just extracts them all, looks them up, gives me a list. I can add them to the show easily. And same thing with edit points. It's when we're, when we podcast, I am the worst person at writing down timestamps when something happens. But now with me just yelling out edit point while we're talking and the fact I just did that, it's going to really screw it up now. But it's going, it'll go through and read the transcript and look for areas that I think

that should be edit points and give me a list. And it has timestamps on them so I can do that. But anyway, so I made these two skills to do these two different things on two different days, to solve two different problems. And then it said, well, these are related. do you want me to run them in series? Like, oh, yeah, that makes sense. Connect them, please.

So it's trying to be helpful. But at the same time, you're the boss and make sure you're the boss. But that's the whole system is you give it memory and you give it access and you're careful about building it out slowly. And before you know it, a bunch of donkey works off your plate.

So this is really exciting. I mean, I found myself waking up at like five in the morning just thinking about something else I wanted to train it and just getting it out of bed because I couldn't stop. This episode of Focused is brought to you by Incogni. Data Brokers are in the business of collecting

personal information, which can be an uncomfortable thought for a lot of us. Nobody likes the idea of other people having access to our personal details. That's where Incogni comes in. They reach out to data brokers on your behalf, request your personal data removal, and deals with any objections from the other side. The thing I love most about Incogni is how much time it saves you requesting those removals for you. Logging into the incogny dashboard, I can see that it has actually saved me 297 hours. That's how long it would have taken me if I would have had to reach out to each of the people and requested my personal data removal that incogny has handled for me. I can also see an overview of the status of those removal requests. I can see that 396 are

completed, 238 are in progress, and I can even see a log of the activity showing which sites have honored those removal requests and which ones are resisting. With Incogni, you can protect your privacy in three easy steps. Just create an account, grant them the right to contact data brokers on your behalf, then kick back while they get to work. And I love how Incogni has both individual plans, which will send repeated automated removal requests to data brokers and people's search sites, but also family and friends plans,

which you can add up to four loved ones to your annual subscription and keep their personal information off Data Broker and people's search sites as well. So visit incogni.com slash focused, FOC, USED, and secure your data today. Sign up and enjoy a 30-day money-back guarantee. Protect your privacy with incogny and experience peace of mind knowing your personal information is safe. Take your personal data back with incogny. Use code focused, via the link in the show notes, and you'll get 60% off an annual plan.

Check it out now, incogny.com slash focused. Our thanks to Incogni for the support of the Focus podcast and all of Relay. So in addition to your robot assistant, got any shiny new objects? Yeah, it's actually robot related. We had an extra studio display in the house. Long story, but we had an extra one. I was going to sell it. Now I've actually hung it on my Mac.

It's to the right of my main display. And it's kind of just so, like I was talking about doing the, you know, when I have a robot running and doing stuff in Chrome, I want to keep an eye on it, but I don't want it in the way I want to be able to still work. So I hung a second monitor. I've tried this three or four times over the years and it never sticks. This time it's stuck because now I've got a good excuse for it. Awesome.

I think that's a great, an underrated tip maybe is to have like that context window always open because I've learned so much from watching how it thinks in terms of what it's doing. That's the only reason I've been able to have any sort of stuff. success with Claude code because I don't know how to write any of the code is just understand its approach and how it's thinking about things. Oh, okay, so next time I'm going to talk about it this way. So having a window where you can always see that stuff, I think would definitely increase the usefulness for this. Here's a power tip, Mike. When you're training the robot, don't give it details on how to do a specific task. Tell it the result you want and ask it to give you five ideas for

ways to implement this because you know it knows better about how it was how it works and when I try to like prescribe to it I want you to do this this this and this it usually is the worst way to do it it's better to let it come up with the solutions you just tell it the you know the solution you want and let it come up with the procedures yeah that's a great tip um so for me my shiny new object is the result of going to the new media summit. I saw someone walking around with a DJ Osmo Pocket 3, saw that it linked with those DJI mics that I've got, and I already know that I need to get a little bit more serious about short form video. I found a good deal on one, and I got one. Like, you will not regret that.

I have one too, and I use it all the time. In the labs, I do a quarterly question and answer session. Yesterday, I did mine in the backyard and showed off my bonsai trees while I was answering questions. And it's that pocket three. I mean, and you're right, you just get, it's not that expensive. You get a wireless mic that connects to it with no hassle. Yeah. And I'm also starting to do more of the Disneyland blogging stuff, you know, the Disneyland field guide kind of thing.

And I use it for that too. And I know you can do it with your phone, but you know, you do other things with your phone. And having a camera, you can just stick in your pocket and then pull out and just like shoot and run. It's a really great piece of technology. Yeah. And that gimbal is so good. Like it's so silky, smooth. One of the issues I run into when I use my iPhone for the B-roll stuff is sometimes I have to shoot B-roll in places that I don't have a lot of natural light.

and so the action mode doesn't work real great. Yeah. And that thing is, it's basically impossible to screw up a shooting video with that Osmo Pocket. I mean, for someone who does YouTube videos, it's a B-roll machine. Because you can just pull it out and shoot it really fast. But even just like, if you're not doing that kind of stuff Mike and I do, it's really nice having a really good quality recorder that fits in your pocket. Yep, 100%.

What do you read these days, Mike? So we just finished a book for Bookworm called The Science of Storytelling by Will Store. And this is an interesting book because it's essentially a science-based storytelling class in a book form. There's a lot of neuroscience. There's a lot of behavioral studies that go into it. And it's primarily, I think, geared towards fiction. writers. That's not why I picked it up. I just wanted to be able to tell a better story. I think there's a lot of ways that that gets applied in the work that I do that could really just benefit a lot of my

professional workflows. But it's a fascinating book and I've learned a ton from it. So even if you aren't necessarily wanting to get better in terms of writing fiction, because I think that's kind of how it'll hit you at the beginning. There's a lot to be learned from this and it's a really good book. I would recommend this to just about anybody who's interested in the topic. Yeah, I mean, thanks to your recommendation, that's next in my cue. I'm reading a book this week, and it's by Lost Art Press. It's called By Hand and I. And it's about the history of taking measurements. It's a woodworker book, kind of, but I think it's an artist book. And like the way people used to do measurements, they didn't have

refined rulers and tape measures back in the day. So they would use proportions. And they just knew certain proportions. If you look at old furniture often, it's very pleasing because over hundreds of years, craftsmen figured out the best proportions for this stuff. And they just used dividers. And they would figure out, well, this is a three, you know, three units across and two units down cabinet, you know, or something like that.

I really like it. But the way the book is written, it goes through the history, it goes back to Egypt. and talks about the golden meme. If you're watching the video, I have, I'm holding it up. And it's just a beautifully made book. Lost Art Press, books are expensive, but they get them printed in the U.S. And like it's heavy, the cardstock. Like, if you want to see just a well-made book, because, you know, so many books these days are so cheap,

there's just, I can appreciate, to Mike's point, the artifact nature of this book. I just kind of love owning it and I've enjoyed reading it again. And if you're at all interested in that kind of topic, any kind of art, I think you could almost get away with reading this book. It's not too long. The authors do a good job and I'm really enjoying it. Nice. Looks good.

All right. So that'll do it for this episode. Remember if you want to pick up the robot assistant field guide, you can do so via the link in the show notes, which can be found at relay.fm slash focused slash two. 252. While you're there, if you want to support the show, you can sign up for DeepFocus, which is our extended ad-free version of the show. Today in DeepFocus, we're going to continue to talk about my Obsidian Task Management Existential Crisis. Thank you to our sponsors today. That's our friends over at Gusto and Incogni, and we'll see you next time.