Focused - 266: The Ultimate Shiny New Object

Episode Date: October 7, 2026

Welcome to the Focus Podcast. I'm David Sparks and joined by Mr. Mike Schmitz. Hello, Mike. How are you? Pretty good. How are you, David? Well, I haven't moved and my wife hasn't broken her leg, so I feel like my life is going pretty well. But you're a busy guy right now. We're going to talk about that today in Deep Focus. You set up a brand new studio. I did. If you're watching the video version, you can see it right now. But a morpha, baby. It looks amazing. I just want to say. I think you need a little more back. We'll say it for deep focus, but I want a little more backlight. I got a couple suggestions, but I, you know, it's very exciting setting up a new studio, I will tell you.

I really like that. It is, yeah. But I actually have those dimmable, so I think I just turned it a little too low. Yeah, yeah. Either way, we're the focus podcast where we talk about the fact that life is more than just cranking widgets. But before we get started, you've got something to share. Yeah, I do.

We talk about quarterly planning on this show a lot. One of the first products that I ever made was a personal retreat handbook. And I have recorded new videos for a personal retreat course. So I just wanted to mention that that's going to be available by the time this show goes live because Q4 will just be starting. So if you would like some help getting your personal retreat dialed in, that will be available. The URL is links.practic pkm.com.com slash retreat. And I'm going to have a special launch price for the podcast here, $29, which is going to include the digital version of the personal retreat planner that I put together.

So that's actually how I do it now as I do it on my super note nomad with my fancy fountain pen style stylus. Feels really great. So you can just load that on to work on an iPad or remark. as well. It's just a PDF file. But that's included and then just like walkthroughs on how the whole whole thing works. Each individual step is broken down, how to do it in Obsidian, you know, but all the principles are there. So even if you just wanted to take it, take the concepts and go do it in a blank notebook, you could you could do that as well. I just want to endorse what you're doing, Mike. I feel like quarterly planning personal retreats are the most bang for your buck that I

get when I do that stuff. And I think it's the thing people are most unlikely to do. And if you're sitting there on the fence and you need a little help, for 29 bucks, Mike's going to get you over the finish line. So go check it out. Yeah. And I will share here publicly the long-term play with this is that I want this course to be available so that I can bundle it with the book that I'm going to write on the subject. But that is TBD. And honestly, going through that just gives you more feedback. Like one of the things I find teaching something gives me a lot of insight as to where people are having trouble. So help my count.

All right. Well, the topic today is the distraction capabilities of artificial intelligence. I know that AI is a big topic on everybody's mind. People are wondering if it's going to kill us all. My question today is, is it going to kill all our focus? and what this grows out of is me teaching that robot builders club and the robot assistant field guide. Something I'm witnessing with people going through the course is these people get to like a point of exhaustion with it and frustration and they do find that it is a real focus damper.

The problem is that it's always talking to you and always asking you questions. And for a place where we're trying to tell you not to do a lot of multitasking and try to stay on one thing and build that focus muscle, it's tough when you've got a little digital assistant that has nothing but questions for you. Nothing but questions or ideas and suggestions on what you should be doing right now. Yeah, yeah. I mean, the agentic AI is kind of amazing. Like, I continue to be surprised by it. Like, I ordered a thing and decided I didn't want to keep it. it and I told my AI, hey, I don't think I'm going to keep that thing.

Will you please take care of, you know, the return and said, sure, it went and looked up the receipt, prepared an email for me to the vendor asking for return authorization, did the whole thing for me. All I do is press send you. And I'm still continually easily amazed by little things like that that it can do for me. But it is a dangerous distraction because it arrives wearing work clothes, you know? You know, you're like, this is for the work, so I can do it. And the effect of it, though, is grinding people down.

Like I said, I see those people who are dealing with it. And some people just literally have to take time away from it to kind of get themselves back on track. And I think that is a real risk for you. If you're trying to have a focused journey, but you're also trying to use this agentic stuff. I always talk about AI as an amnesiac employee, but it's not just one. can be a flock of them if you run a bunch of things at once.

And I think it makes focus really hard. I agree completely. I was actually listening to Cal Newport, Deep Life podcast the other day, and he was bringing up this article that someone had written in by an Italian writer who was talking about how AI has changed the way that she does her work. and Cal has a term for this, which I think fits perfectly here. He calls it checklist productivity. So just marking the thing off is done.

There, I've been productive. I guess if I were to reframe it, maybe it's like procrastina working, kind of like you were talking about. It's going to give you all these things that, hey, I could do this real quick and check off the box, and there I'm moving things forward. Yeah. But there's no real clarity on what are the things that actually move the needle. So the example that he shared in the podcast, which really resonated with me because this person that this article was on was a writer, was she gets bogged down now in all of the businessy tasks that she never worried about before, optimizing the email campaigns and the social media posts.

And are those things that ultimately as a writer slash creator you quote unquote should be doing? Maybe that's the thing. You can make the argument that, oh, I should be doing this. Now I can do it easier because AI is allowing me to do it. But it just slowly eats away at the time, energy, and attention that you have available to you. And you do all these things that you were never doing before. And the thing that was really important doesn't get done. Well, I believe one of the problems people are bumping into on.

that on a meta level is they're trying to get the AI to do too much. And when you get it beyond its capabilities, then suddenly you're a manager of a bad employee. And that's a much more time intensive thing than a manager of a good employee and effective employee. So, you know, give it a job that it's capable of and don't think that means that it can do the next job above that. I think that's something a lot people will make. But I wanted to talk about kind of just the traps that come with this stuff. So I started, you know, thinking it through. And I made a list of all the ones I've seen from people going through the robot assistance and field guide. Hey, real quick, before we get to the traps, though, I want to just add one more thing for context

here with the checklist productivity idea. Yeah. Because the thing that I realize, I wrote about this in my newsletter, the thing that it can't give me when we're talking about, like, these are the things that you should be doing is any sort of value judgment. It can give you like quantifiable results but not qualitative results. And the way I defined it as in the newsletter is it can't tell me what has any sort of resonance. And I use this metaphor. I'm curious because you're a musician. If you, this speaks to you, where like when you're playing the piano, right, you the little felt hammers or the felt pads get lifted the hammer strikes the string and then the string vibrates right so the damper pedal essentially like keeps it from from resonating i feel like

that's what a i does is it just like suppresses all of that it's all on the same level now there's there's it's so hard to to choose out of uh all these things that i could possibly be doing you have to the term human in the loop maybe gets overused but when you're trying to decide what it is that's actually worth doing. You have to have that perspective. You really can't delegate that, at least not currently. Well, I mean, it's a taste in judgment of the terms I always use. It's not capable of that. And if you think of it, one of the reasons it's not capable of it is that it doesn't have the proper context. It doesn't know the way you think about the world. It doesn't know the message you want to provide. It's just going to to turn anything you give it that requires taste and context into the vanilla least common

nominator version of that. And that's not even on the outline today, but I see people drifting into that. Like you get into writing or working with AI and then suddenly it's, well, it does a pretty good job of getting the blog post on WordPress. So now I should have it build my strategy for the rest of the year. And it's like, no, you just went over a line. I mean, Not only have you surrendered taste and judgment to something that's not capable of it, but you're a human. That's your point is to do that stuff. And frankly, it's really bad at it.

So it gets us right back to this whole distraction and loss of focus because you're trying to get it to do something it's not capable of. And that is a huge problem. Exactly. And it's a trap people fall into because at first you distrust it when you first start using agentic AI, but then you start seeing success on a low level. And you think, well, it can do low level, can do a high level. And there is definitely a below the line, above the line kind of divide here.

And you can't go above the line. It's just never going to work. It seems to me that AI is absolutely terrible at deciding what are the jobs to be done in any given scenario. But if you know what the jobs to be done are, then you can delegate things effectively. Yeah. And to go back to Cal's example with the writer, I have similar methods I use. Like, I'm doing more social media if you're on Instagram and all the socials, you'll see me more often. And that is almost entirely given to AI in terms of finding potential content.

And I hired a guy who takes that and figures out what actually makes sense to use. And, of course, I have some involvement. I'll say yes or no to things before they get posted. But I knew very well when I did that. that AI could help with the initial filter, but it cannot do the work. And that's just the way you've got to do it. But yeah, I mean, we've already dove in deep on this problem.

But let me get into the five problems that people hit. The first one is very obvious, if you've spent any time with AI, I call it tool tourism, right? These tools are evolving very quickly, you know. And then, you know, I hear from people like, well, you know, I ran the same prompt against all major models, and now I do it for every prompt to make sure I get the best possibility.

Stop that. Just stop it. If you're rebuilding the workflow every time there's a press release from Anthropic, then you are distracting yourself from doing the hard work. You know, don't let every product launch destabilize you. And things are moving fast now. Things are evolving. But the change in these tools is,

not a justification to not do work. This is like the ultimate version of shiny new object syndrome. Yes, exactly. And it's every day. Yes, yeah, exactly. There's a new model every day. And if you're constantly asking, you know, can this one do more than the other one? Probably.

Does it actually add any additional value to your workflows? Probably not. Yeah. I mean, to get back to that problem I was talking about earlier, the taste and judgment problem requires context that it doesn't have. if the model gets five times better, it's still going to have the same problem. Yeah. So I agree with you.

I like the term tool tourism here. And I feel like AI started finally getting useful for me when I stopped worrying about which model was best and I just picked one. And then I started figuring out what can I actually do with this that is useful, that makes the jobs that I want to get done easy. and help me do them better more effectively. That's kind of when I started to see how the dots connected in a useful way. Yeah.

To me, the turnaround point was the concept of donkey work. Just like, if this is donkey work, I can probably give it to it. And if it's not, I should not. And every time I try to go beyond donkey work, I find myself spending too much time. Managing something is not capable of. I play myself. You know, it happens. But on this tool tour or something, I think what you should do if you're working with

Agentic AI is pick a tool set like Mike said and just say, this is where I do the work. Now, if you want to experiment and that's exciting to you, that's fine. You can do it, but schedule time for it. Like make that your weekend thing. Instead of doing something else, you play with these tools. Maybe that's so interesting to you. but I'll tell you that if you don't want to spend your weekend working on the tools and you just let it go and just use the set you've picked and just pay attention

every two or three months other people are going to figure it out you know follow me I'll tell you what's what's good you know and you don't need to go and be the explorer it'll get handed to you and and if the other model has got the cool tool that you want the model you have will have that tool within a few weeks because these guys are like aping each other so fast yeah so yeah tool tourism is a problem only switch horses if a new horse is a lot better and a lot faster because you're going to have to spend a lot of time you know doing that 100% agree want to talk about infinite collaboration yeah that's the second one and I think a lot of people can appreciate this if you deal with AI it'll you begin, but it never tells you when to stop.

Okay, so this one I think may be an interesting conversation because I use AI for collaboration a lot. And maybe I differ from you a little bit on this. Maybe it's the way that I collaborate with my AI. So one of the things that I do is I use AI in the ideation process of every single newsletter that I write. it doesn't write the actual newsletter. It will give me suggestions for like,

this is how I would introduce this paragraph and things like that. But by the time I get done, I take all that text. And I've started to publish these via Substack as well. Substack has a connection with that panagram or whatever, that AI detector. I always get like 100% human written on that. So I am doing the work here, folks.

But it's about a two-hour process where I go back and forth and I will write a little bit based on what it gives me and I'll paste that in, like, where would you go from here? And I do find that the AI, maybe it's a perfectionist in me, it gets to a point where it's like, you're nitpicking details, you should just ship it now. Yeah. But I do agree, you have some stuff in the outline here about how, like, it's going to keep giving you more ideas. And I have also used it, like, when I was talking, putting together the Amplify cohort, things like that,

where if you keep asking it for, what else can I add? What else can I do? Like, it's never going to run out of suggestions for you. And it's not going to tell you at what point those suggestions all of a sudden get a lot less valuable either. It never says enough, you know. Yeah, yeah. I think that's the major point you're making here. And that I definitely agree with.

You know, and I think that that's a way to distract yourself is to say, well, you know, let me go back to it one more time. Check it. See how you can improve it. Where, you know, what did I miss? You know, those are great prompts to give you an AI. If you've got a piece of writing or an idea, you're trying to develop a project. but again you're the human you get to say we're done now and i think a lot of people have trouble with that because there is no natural stopping point when you're dealing with an ai it's just

super happy it no end of variations objections directions it'll do all that stuff it doesn't get tired you know that's one of the advantages of it but it's also this this infinite collaboration trap that that people fall into you know you ever notice how like you finish something it says, okay, now how about this? It is just so eager. It's like a little puppy that is at its core understanding is that I want this person to engage with me. You know, so how do you avoid that one? I feel like if you're aware of it, you know, if it's warned is far armed, but just decide

what your stopping point is, you know, decide what is the job of this tool? And it is a tool. It's not a person. Don't give it a cute name. But you say, okay, the job of this thing, like for me, you talk about news. talk about mine. So I write a newsletter. I give it to it and I say, look for things that you think I missed on this topic. And it'll tell me things I missed. And then I also say, what are my hidden assumptions in this topic? Because sometimes I write and I start at step five and there are people

reading that need steps one through four first. And those are two of the big questions I ask it. And then the third one I ask it, are there any show stoppers? And that is a secret code between me and the I'd say, you know, did I get all my comments in the right places and all that. And those three prompts usually get me pretty good, but it is so happy to do more, you know. You know, oh, but what about this? Would you like me to research this? And no, stop, you know? And even the showstopper thing you have to be careful for because to it, the initial time we started doing showstoppers,

it thought that my voice was a showstopper, you know, because my voice is unique. Now you can't say that. And I'm like, well, I do. So leave me alone. But yeah, so put guardrails in. Another thing you can do is you can limit it to a few options. I think one of the best pieces of a prompt you can give it, if you're looking to say, look for the three things in here

that you think I should change or improve upon. You know, and just that's it, those three. And that's good. You've got your value for that, you know. And I think the final piece of this is you decide when it's done and you end it. Don't leave it open. Do you close your threads with AI, Mike? No, I leave them hanging.

Well, I should say I always start new threads, but like when I'm doing the article, I never tell it when I'm done. I just will abandon the chat and take it with me and go publish it, things like that. I have a prompt I give at the end where I say,

now based on what we did here, do any of your skills need updating? Like when I correct it on what, you know, what a showstopper means and then it'll go back and update it. It has a couple of promises. that I said, I also want you to take a log of how much money in time we just,

tokens we just spent. And I have a, it creates like a dynamic database. So then I can go back later and say, well, what was I really spending tokens on them? But the added benefit of that is as soon as I fire that off, Texas manager snippet, AI.c.l close, AI. Close. It fires that off. It's like, I'm done.

There's nothing more you can do for me now, you know, and I close the threat. And I feel like that's another way to kind of avoid this. I like that. I don't do it. And I also am a big advocate of updating the skills. That's interesting that you say update where as necessary because I've had instances where it's updated a skill and I haven't liked what it's done. So now when I tell it to update a skill, I'm very specific about what I want it to update. Don't touch anything else.

You know, just do this. Part of that is I'm working with an assistant now and I want her to have access to the skills. I don't want them to break. And then all of a sudden, she's using them to draft a whole bunch of social media posts based off of the article that I wrote. Yeah. You know, and it's ripping out pieces of things that I've said.

And it's, you know, doing it in a way that doesn't sound at all like me because it makes my job of hand editing every single one of those before I publish it a whole lot harder. Yeah. Also, you mentioned that you have to be clear on the job that the tool is doing. I agree with that.

But I would also add to that. Be clear on the job that is to be done, the task itself. Yeah. So when I go into, and I'm writing the newsletter, for example, I always started off as, help me create an outline for a newsletter I want to write on this topic. And I feel like that kind of forces it not to just go down the path of, oh, I'll just write this forum because obviously that's helpful. Occasionally, I will ask it to go a little bit further. Be like,

hey, help me, you know, write the bottom line, the outro piece as I, you know, my wandering thoughts have all connected and coalesced throughout the different sections of this. newsletter. Help me tie it all together because I tend not to be really great at that part of it. But yeah, just knowing and communicating, like this is where I want you to stop, I think is key. This episode of Focus is brought to you by Gusto. Be honest. There's probably one task on your list that you constantly push to next week because it's just so tedious. And for a lot of business owners, that task is payroll. Gusto is here to take that entirely off your plate so it becomes the easy part of your job instead of the dreaded one. I've worked in businesses that have both

used gusto for payroll and have not used gusto for payroll. And I can tell you the one that did use gusto for payroll, it made everything so much smoother, so much simpler. We basically never had to worry about anything. In fact, I've been using gusto in some way, shape, or form for probably about five years now. I've used it as an independent contractor, I've used it as an employee and I've used it as an administrator. From the independent contractor side, it makes getting your tax documents at the end of the year easy. From an employee side, it just makes payroll simple and automatic. From an administrator side, it eliminates all the headaches that typically come with having to run payroll. Look, if you're a business owner, you want to focus on actually building your business,

not all the nitpicky administrative stuff. That's where a service like Gusto can really help. Gusto is online payroll and benefits software built for small businesses. It's all in one. remote-friendly, and it's incredibly easy to use. So you can pay, hire, onboard, and support your team from absolutely anywhere. Save time with automated tools built right in, like offer letters, onboarding materials, direct deposits, and more. And it's quick and simple to switch to Gusto. Just transfer your existing data to get up and running fast.

Plus, you don't pay a cent until you run your first payroll. So try Gusto today by going to gusto.com slash focused, F-O-C-U-S-E-D. And if you do, you'll get three months 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, that's gusto.com slash focused. Our thanks to Gusto for their support of the Focus podcast and all of Relay. All right, Mike, the next one on my list is system procrastination. And this is one that you and I, I think, are particularly susceptible to.

And this is building a machine around the work instead of doing the work. Before we got on Mike today, I showed Mike my new. that I've built to replace Obsidian. Like, talk about system procrastination. Like, oh, I don't like this one feature of Obsidian. What if I just built a whole new native app that replaced it, you know? AI makes this tempting because it lets us create sophisticated systems very quickly. And it has all sorts of suggestions on how to make them more sophisticated.

And you say, oh, yeah, go, please go do that. And before you know, you've built this intricate mouse trap. You don't even know how it works. and you're supposed to be using this thing. And I think this is a great form of procrastination. It is a great form of procrastination. So did you procrastinate when you built this thing? No.

How do you decide what's worth it or not? And there's a tradeoff here, right? Because sometimes it's like what I made is in the background over a course of a few weeks is amazing. And I use it every day. You know, it's all kind of the problem for me. but I also know that it's very easy to like spin these up on a daily basis and like oh yeah I'm going to change everything it can do it all for me you know the old days if you wanted to switch

your task system and you had to go through and recode it in or something that was you actually had friction points that kept you from doing you know spontaneous tool switches and now a lot of that's gone you know and and I do think that you can spend a lot of time like tweaking and improving, like this app I told you about. I am only allowed to improve it on Saturday afternoon. There's a couple of reasons for that. Number one is I got to work. I can't spend the whole week playing with it.

Second is Saturday afternoon is when my Anthropic tokens reset. So I wait until the end of the week and if I've got 20% of my tokens left for the week, I just fire up Fable and set it loose on that app to make improvements. And okay, okay, that's it for the week. Now I'm going to go back to it. Give yourself some kind of constriction that does that, you know. But don't spend all your time tinkering, you know, to talk about the example from Cal Newport. You can really get caught up in this stuff and not actually do the work.

Yep. Yeah. So I, you mentioned building your own task system and like every other influencer on the internet, I've been working on that myself. Yeah. I'm actually working on a full-fledged task manager replacement. But even in the short term, I've been able to vibe code something that is insanely useful for me. I just call it Quick Capture. And it's a purely Quick Capture app, which allows me to capture things either as inline tasks,

because that's how I do most of my task management. But also, I've got a couple of buckets from my creative projects, which are notes as tasks. That's a deal-breaking limitation in the past as you were trying to do task management and obsidian, which I have been doing for years. You can't have both. You can't have your cake and eat it too. You've got to pick one. Well, I can now quick capture both ways very easily.

I'm actually got a YouTube video. My most recent one actually is on this topic. It'll be live by the time this goes out. So I will link that in the show notes that people are curious on how I put this together. And I'm making it available as a tool that people can download. But the larger point I want people to understand is like if you know how you work, then you can design the tools to fit the way you work. but there's a fine line there between this is a tool that will help me do what I already am doing in a better way.

And let's just see what is possible. Ultimate shiny new object syndrome, I feel like when it comes to system procrastination, this is where it manifests the most for me because I tell myself, oh, it's my systems. It's going to make things more productive in the long term. It's worth the investment to figure this stuff out. And it's never,

that's never the case unless I am clear about how it's actually going to be helping me. distraction wearing work clothes, you know, and that kind of runs through all of these. It's so easy to get caught up in that. I do wonder, I mean, this is a little sideline here, but so people are making their own apps, people who have figured this out already making their own apps and you don't have to be a coder. So many people are making apps out. Just look at my inbox. Every day, like 30 to 50 emails from people that want me to talk about some app, they vibe coded. And that's great. I wonder if this is temporary though. I wonder if the ultimate solution here is not little bespoke apps, but like an AI system that just takes care of all

this for you and you don't even think about it. And I think probably that's where this is ultimately going to be heading, but we're not there yet. Question on that. We talked about how AI is terrible at deciding what are the things that are actually worth doing. Do you just believe that at some point it's going to get good enough at that that you would trust it? No. Okay. So that I think is the major limitation to the future that you described there where AI just manages the stuff. Well, I guess what I mean by management is that it holds all the data for you. Sure. So it's the bucket.

It's the brain that has all the metadata and it can surface the suggestions, but you're still always making the list. And these are the three things I'm working on today. Yeah, I agree. The problem is context. I mean, just to get back, I keep saying it. I mean, it doesn't matter how smart the brain is if it doesn't have the context. and I don't know how it could get the context.

Like an example in my Circle community, Max Sparky, I decided I wanted to put my blog post in the community. So if somebody is in my community and they still want to keep up with I'm writing on the blog, they don't have to go over to the blog. You know, if you're already in Circle, it's there. There's one of the groups is like, I call it field notes.

It's just copy my newsletter, copy anything I do public facing gets also published to the Circle. And so I set up the AI work mode that said, okay, when I finish a blog post, because my current method is I write a blog post in Obsidian or my bespoke app, and at some point I tell the robot, hey, this is done. Push it. That's the magic word. Okay. It knows how to go to WordPress to create a new blog post to check off the right boxes and to put the copy in and set it up. I have it, save it as draft because I don't trust it enough to hit the publish button.

And then I said, well, in addition to that, now I want you to go to Circle and do the same thing. And it knows how to do that, too, there's an MCP. So something that used to take me, you know, 10 to 30 minutes now gets done by the robot. That is an example of donkey work. But the robot said, oh, I have a helpful addition. You know, since you're putting it on Circle, we want to make sure that everybody goes to the blog, so you get clicks of the blog so you can sell more expensive advertising. So I'll just go ahead and put a portion of it in circle for you with a link back to the blog.

And that will be best for you, you know. And I'm like, no, that was judgment. You don't have that. You don't understand that with my audience, I don't nickel and dine them. I don't make them click links for stupid things like ads, you know. And so you won't do that. But that is an example to me of where the brain probably studies the internet, knows that that's the best practice. if you're putting something on the internet,

but it doesn't have the context of me and my customers. Yeah, I love that. It's a great example. Yeah. All right. Next one, though, cognitive debt. When the AI gives you answers faster than you can understand. And I think Anthropic is worse at this than chat, JVT.

I use them both. Anthropic is thinking so far ahead on things, and it gets so cryptic with its answers. Then I have to stop and try and figure out what it's talking about. And then you've got this thing constantly giving you reading assignments and suddenly you feel like you're working for it. So it's rough when you start getting into this thing and you feel like you've got research assignments growing out of a job you gave it.

Yeah, I guess I just don't read everything. Yeah. That's the answer, honestly, you know. Yeah. Do you think that you could control that at all with any sort of like system level prompt? like don't give me a whole bunch of texts just keep it simple yeah i've tried and actually the anthropic models are getting better at it the most recent rind of opus is like getting smarter about that but but i think this kind of goes back to the thing i said earlier i think though times when this gets

worse is when you're asking it to do things that tip into judgment and taste yeah you know and that it doesn't really have the ability to do that so it just makes up a bunch of words it's almost like um you know like when you meet somebody and they really don't know what they're talking about, but they try to sound like they know what they're talking about. Yeah, it's like an AI word salad. Well, I mean, it's like it's almost like a BSer. You know what I mean? Yeah. And like I used to run a bunch of these lawyers. Always they had an opinion on everything. And I used to think it was funny when I'd meet somebody like that. And I would just like ask them to explain themselves and they would, you could hear them just drifting as they, you know, AI does the same thing. I feel like sometimes.

because when you drift, when you get into taste and judgment, it really doesn't have the context to answer you, so it just starts making them stuff. I think the best way to avoid this cognitive debt is just not to get into it. Give it donkey work jobs. Don't give it jobs where it has to make decisions like that, and it won't give you word selling.

Yeah, I actually do like it to help me with research and things, but I always ask it for the links and the source material. Yeah. Because I don't trust anything that it actually pulls from there. And frequently, I'll find that the anecdote it grabbed from this article or this interview that somebody did is not the one that I want to use. There's actually a much better example further down. So I kind of use it to point me in the right direction, but I ignore a lot of the conclusions that it arrives at based on the information that it consumes. It's pretty good, I feel like, at identifying where there is good source material, but it's terrible at identifying what's actually useful out of there.

And what I would say is that it's good at identifying source material and quantity, which is what leads to the cognitive debt problem. You know? Yep, yeah. And you don't need 50 articles on this concept for your newsletter to read. And it would happily give you 100, if you'd ask. Yeah. So I know you've mentioned that if the AI comes for us all, you'll be one of the first ones that it goes after because you're so mean to it, right? Yeah.

I'm the same. I frequently yell at my chat. I mean, it's not a person. Well, one of the things that I do, and if I am getting frustrated on the lack of quality material that it's giving back to me, I'll be like you, you just don't understand. stand. Yeah. Right. And then I'm curious, you know, do you do that sort of thing when it is giving you this

information? And does that help it kind of hold back or course correct at all? Or does it make no difference at all? It's just always going to think that it's brilliant. And the deductions that it's found are obviously, you know, brilliant revelations that humanity should know about through the David Sparks newsletter, you know. What I do is I limit it, like when I'm looking for resources, you know, find five resources that support this point. So I don't let it just go hog wild. I tell it when it gives me

one that I like, I'll say this is the one that worked for me and this is why. And then at the end, when I say update your skills, it's got a knowledge base on there and my preferences and it will update those. So to a certain extent, it gets a little better at that stuff for me. But I also kind of feel like, you know, another tortured analogy with AI is that it's the intern. I just always keep that in mind, hey, the intern is going to bring me some research and some of it will be useful and most of it won't. And that's fine. But if some of it is, then it's probably worth having the intern do it. And I don't get hung up on it. But I also realize sometimes I get over my skis if I ask it to do something too complex. And, you know, then it's, you know, I'm more angry with myself.

than the AI because I'm the one who asked it to do the stupid thing that it can't do. Yeah, makes sense. Fifth one. Well, I guess we kind of answered cognitive drift. More notes can leave you with less clarity. That's the thing you got to remember. And this thing wants to always give you more notes. And then authority drift is the last one that I see people bumping into.

And they go from gradually using it to perform a task to letting it to find the goal. I mean, the best example of this is somebody told me that, oh, yeah, I had it write my Aritae statements. If you're into the stuff I teach, Aritae is like the Greek term for the best version of yourself. And they're like, oh, yeah, I just had it right them for me. Like, that's bad. Don't do that. You have to figure that out for yourself. You can't have a robot or anybody else figure it out.

And the process of figuring out is what makes you better. But you know, you don't start at that level when you play with AI. You start at basic stuff. but as you get confidence in it's kind of going back to what I was saying earlier, you know, it will drift into more judgment and taste and authority and that gets you in trouble, you know, things you should watch out for, like, what option looks best?

Don't ask it. It doesn't know, right? Yeah. Or the people say, you know, like, I did that once. I had a bunch of tokens and I gave it more, data than I should have about me. And I said, you know, to audit me and tell me what you think I need to spend my time on. And the answer was complete utter nonsense. Yeah. But I'm smarter enough to know that. But if you ask it that and you don't realize that going in, suddenly you're taking life advice

from an algorithm. Be careful. Yeah, this is the one I have a little bit of trouble wrapping my head around. I know, like I understand the path that people go down with this, but I just have this inherent distrust of the machine. Yeah. And I feel like that has served me well so far. Yeah. Yeah, the minute that I see it start doing that sort of stuff, I, that's usually when I get mad at it, to be honest, it's probably why it doesn't do it as often. But I don't know. Like I don't struggle with with this one. I don't see, I don't see myself naturally handing over more and more precious parts of the work that I do.

Although I am kind of doing that with my assistant. So I'm not sure what the difference is. Well, she's a human to begin with. Yeah. And she has context way more than the AITS. True, but like she started to help me with my email. And I know like that's a red line for a lot of people. It's like, oh, I'm not going to give another person my email. And I kind of have the same thought. It's like, well, you know, the AI can dig through these things and find things out about me. This person can now dig through things and find things out about me. Ultimately, even though I've never actually met her in real life, you know, I trust her not to mess things up. But I would never. I don't think I would ever trust the machine not to mess things up. Well, that's interesting because I do give the machine access to my email.

Yeah, I know you do. So that's why I was curious about like, how did we arrive at different conclusions here? Yeah, I mean, but it's not, I mean, it's my work email. It's not my personal email. But what it's looking at, it's looking for certain classifications of problems. Like somebody writes in and they can't get access to their course. and it's got the tool set to fix that. And then that way, well, I'm asleep.

Their customer support gets fixed. But somebody writes me in with a question or whatever, it knows that not to touch that. You know, I get a report every morning from saying, okay, you know, I dealt with this password reset. I dealt with this account verification. You know, it's a very narrow set of things. It does. But as a result, it sees all the email because it has to look at it to judge for that. And I don't really have a problem with that.

because I'm not asking it to write replies for me. I'm just asking it to deal with customer support. Well, at this point, she's not writing replies for me either. I guess the hesitation I have with that being delegated to an AI is I feel like it'll make connections and, you know, want me to do this sort of a thing. It'll be pretty natural if it's looking at my email to say, well, this person needed this. I'll just go ahead and do this other thing as well, whereas I have daily meetings with my assistant, and we go through, she's got, where we're doing it right now is we've got a label called review,

and she puts 20 messages in there every single day, and we talk about the types of rules that we're going to be making and how things should be funneled. And I know that, you know, I could go look at the skills that are being built and see how the AI is doing the exact same thing, but the fact that the human, on the other end, created the documentation for the SOP. Like, this is how we have decided to do this. For some reason, that is very comforting to me, whereas when I think about AI doing that, I just am very apprehensive about it.

I think at the end of the process, your human assistant will do a lot more of your email than my AI does in mind, you know? Sure, yeah. But for the kind of work I've given, the AI is frankly better than a human because it never goes to sleep, you know, and if people are right in, they want access to their course, they don't want to wait until I wake up. And I don't want to do that work, you know, so it's, you know, you're right. We've all got to draw our own lines, but I think the big risk here is slipping agency. You know, it's not a compass and don't treat it like one, and you'll be better off. This episode of Focus is brought

to you by Saly. Saly is a new ESIM service app brought to you by Nord Security. Choose from several affordable e-sim plans in over 160 countries and eight regions. With a Saly ESIM, you'll always have a connection when you need it. Being abroad with no internet can be a disaster. Trying to find your way to where you're staying, researching the local food spots, staying in touch with loved one back home. You can't do it without a connection, which is why a local SIM card can make all the difference. Thankfully, an ESIM provides an internet connection wherever you travel and saves you money on roaming fees. With Saly and ESIM can be installed only once, eliminating the need for users to install a new ESIM for each country. And it's affordable, significantly reduce and even eliminate roaming

fees with affordable ESIM plans in over 160 countries without having to wait in line at the airport for a local SIM card. Plus, they have chat support available 24-7, and if for any reason your device isn't ESIM compatible, you'll get a full refund. Using SAILI is easy. Just download the app on Android or iOS. Next, buy a plan. Multiple plans in over 160 countries and eight regions are available at some of the best rates online, and then install the e-sim. Just follow the instructions on the app to install the e-sim. It'll be activated instantly after landing, and you'll be able to easily connect to the local network. I know firsthand how painful it can be to be in a foreign country and not have cell phone coverage. That was me when I went to London for the Relay 10th Anniversary

show a couple of years ago. Fortunately, my wife was with me and her phone had access, but I really wish that I could have used something like Saley so that I could have figured out how to get around. I thought it wouldn't be that big a deal, but it really was a pain, especially when we were trying to take the tube, we needed to download this app so that we could figure out where we were going, and we relied on that cell connection way more than I thought we would. I personally felt very disconnected, and it was a little bit jarring. If I had been there by myself, it really would have posed a problem, which is why I can tell you confidently, when I go back, I'm definitely going to use Saly and the ESIM so that I can activate service the moment

that I get off the plane. Right now, you can get an exclusive 15% discount on Saley ESIM data plans. Just download the Saley app and use the code focused at checkout. That's FOCUSED. Our thanks to Saly for their support of the Focus podcast and all of Relay. Okay. So this has been kind of a bummer episode. I've been kind of mean to AI again.

But I do want to say this is a transformative technology. It's very useful. The reason that we have all these slippery slopes is because it's worth climbing on the hill. And I think the ideas we talked about today are ones you should kind of take to heart and be careful in your relationship with this stuff. Another thing that you shouldn't do is just don't open like seven threads and having doing all this stuff at once unless you truly just want to manage an AI today. But you're not going to get any focus work done if you've got it doing seven things for you at once. All you're going to be doing is managing it.

And I'm very careful about that. Like I have a block of time later today where I have some admin work. I want the AI to do for me. And I will make that hour for that work. And I'll just sit here and have the AI open. I will have a couple threads going and I'll keep everything, you know, I'll keep all the plates in the air as it does its work for me. And at the end I'll have accomplished a lot more than I could have on my own an hour on that admin stuff.

But I'm not going to be, you know, sitting here planning. out a next podcast or doing, you know, strategy work or anything at the same time because I'm just not capable of that and neither are you. So be careful with it. This stuff is useful, but you've got to respect your own time as part of the process. Don't let AI become your ultimate distraction. Yeah. If I were to, I guess, consolidate some of the big things and my approach to how I use this stuff, it would be get really clear on the job to be done, how it is actually beneficial to you, and then put parameters around the requests that you make. Yeah.

And I'm not sure. I mean, there's a whole bunch of other details around that, but I feel like that kind of encapsulates my approach to this. And I feel like that has for the most part kept me from drifting too far into a lot of these traps. And like you mentioned, these are really just things to look out for. And just because you've avoided them so far doesn't mean that you're going to avoid them forever. Just knowing that they're there, I think, is the real valuable thing. Or awareness work. Yep. Yep. And ultimately, you know, AI is a tool. So it doesn't have to be a source of distraction. It can actually help you focus on

the things that are actually worth doing, the things that are unique to you that, help you leave your dent in the universe, make your contribution to the world. And ultimately, I think that's what I would encourage people to think about is what is the highest good that you can do. You know, what is the thing that you really want? What's the legacy you want to leave? You know, what do you want to be known for? How do you do more of that? And then everything else, you know, those are good candidates to maybe delegate. All right, Mike. Let's talk about shining new love.

All right. You want me to go first? Yeah, you first. Okay. We just moved. Yeah. And I have a new office setup, which means it's time for new keyboard. Oh, yeah. I have, I'll put the link in the show notes here. It is a K-Kron K-2-H-E. This is the one that you had your eye on. Did you end up getting this keyboard? No. remember no no okay I you know I'm I should not be allowed to ever buy another keyboard because I just end up using the Apple one yeah yeah that's that's fair um this is the one though that I said at at one point if this if there was a David Sparks keyboard this would this would be it this is the one with the the wood on the side I have the the white one which I think looks really nice and it's a different type of switch it's not a

tactile switch. So there's no bump. It's a linear switch. It's the magnetic switches, right? Yep, yep. It feels a little bit different, but I do really like it. And yeah, it's a great little Bluetooth keyboard. So do you adjust? Because with the magnetic switches, you can actually go through software and say, well, make the return key soft, but the space key hard. Like you can actually do it on a per key, like setting. Do you do any of that stuff? I have not. I may do that at some point. I know that there's software you can use to configure this thing, but I've just been using it out of the box.

Keychron is such a great. The keyboards are good. They've got that software. Like you can map the hyperkey to whatever. And, you know, like the one I have, I, like, mapped F-13 to the right control key. And then I make that my dictation button. So my trigger and all my dictation software is F-13. I mean, you can do a lot with these things.

But I think it's the combination of touch ID. And I don't know. I just, you know, I pull it out and I use it for a week and then it goes away for another couple months. It never sticks for me. Like, I don't know what the problem is. One thing I do like about it, though, is that it's narrower because I have the battleship Apple keyboard, you know, with the number pad on, which I barely use.

And it's kind of nice having the mouse a little closer. Yep. I'm going to get it out again today. See, this is how it goes for me, man. I bought myself a Australian object. This is a total, just me having fun with myself. There is a toolbox that I have been admiring for years. Toyo, it's a Japanese company.

They make these stamp steel toolboxes. I have a couple of them. But my kids have flown the coop, Mike. They don't live with me anymore. And frequently, I'm going to their places to fix things. and I've decided I want a proper toolbox. So I bought, and I've admired this one, the big one from Toyo. They have it on Amazon too now.

I didn't realize I got mine from them directly. It's on the way. I haven't got it yet. But it's just a great hinged lid toolbox, like an old fishing tackle box. But it's got a place I can put screws, drywall screws, all the stuff I need when I go do work for the kids. And I've got a prepacked toolbox I'll be carrying around with me. And if you know, you know,

when it comes to the Toyota Toolboxes, but they are great. They are, yeah. I remember you mentioning you had a couple of the smaller ones. Yeah, I interviewed Steve Camp, but you got the big one now. Nice. What do you read? I just finished a book for Bookworm, which is called You Can Just Do Things by Kate Hall. A little bit different, kind of interesting.

Kate Hall has an interesting story, that's for sure. She was a successful lawyer, I think, in the D.C. area. She quit practicing law to play poker professionally, became the top-ranked female poker player in the world. She started a pandemic medicine company. She's done all these really impressive things. And she starts the book in the introduction by saying there are two versions of her life story. There's the LinkedIn version, which is kind of the one that I just shared.

And then she's like, there's the impoverished. Foster Syndrome version, which is all of the failures that she had along the way, the drug addiction that developed when she was playing poker, all that kind of stuff. And the framing for the book is really interesting. The premise is in the title. You know, you can just do things. You don't have to wait for permission. You don't have to wait until things are perfect. That is definitely a message that I need to hear. So it's, I think, very useful for me personally. But also, as she's framing those two different stories, I recognize I definitely tend to tell the imposter version a lot more myself.

So it's a lot of like just go do the thing, kind of inspirational if you want to think of it that way. But she does offer a lot of like reflecting questions that you can use to, you know, what's the worst case scenario? If I really did this thing and it was a complete failure, what would be true? Sort of like Tim Ferriss Fier setting style stuff. I thought that questions were, they were interesting. I have my own version of questions that I asked.

So I didn't actually find any of those like, oh, yeah, this is so much better than what I'm currently using. But for someone who maybe is just starting with like the quarterly reflection process, there's probably some really cool stuff in there, really useful stuff in there. Interesting, easy read. It's like 190 pages. So I think overall I would recommend that people pick that one up or at least go to library, give it a shot. Worth your time. All right. So I read a weird book, Mike.

It's about Seneca, who was a philosopher who I have a lot of respect for. But he was the teacher of Nero, Emperor Nero, who was famously not a great guy. And somebody wrote a book called Dying Every Day. And it's James Rom, who's a smart guy, Sinica at the court of Nero. And it's a great kind of documentation. Nero famously killed his mother. He's just, like I said, not a great guy. And Seneca's relationship with him.

And of course, Seneca ended up also dying at Nero's hands, basically. He was forced to commit suicide as they did. But he wanted, he's this highfalutin, high thinking guy, and he's given this kid to teach, and the kid turns out to be a dud. And yet he sticks around. And it always bugged me, right? You know, wouldn't you have just, like, got out of there? out of that situation.

And the study of the book is like, why did he stick around? You know, why did he keep trying? And I don't know, it answered a question for me. And it made me think, you know, about how there are people in your life that you think you can change and you can't. And even some of the smartest people in history fall into that trap. Oh, it was a good book, dying every day. interesting yeah well well done uh by james rome i think he uh he did a good job of it i don't think

i've really sold the book very well here but i enjoyed it not that uplifting that's all right some of the best books are the the ones that you mentioned an answer to question that you had right that's one of the things that uh piece of advice i give people when they're trying to decide like what should I read? Like, find a book that you think is going to solve a problem that you're currently facing or answer a question that you currently have because you're going to be way more interested in the material if that's the case. Yeah. We, like, we try to explain certain people and we realize, no, that's just who they are. And like, there's nothing you can do about that.

Yep. Anyway, that wasn't very uplifting, but that was my pick. That's all right. All right. We are the Focus podcast. You can find us at Relay.fm. slash focus. We'd love to have you come check it out. If you want to get into DeepFocus, that's the ad for extended version of the show. You can sign up for that there. Thank you to our sponsors today, Saly and Gusto. On DeepFocus, Mike and I are going to be breaking down his new studio. Can't wait to get started. We'll see you next time.