Ep 76: AI That Actually Helps - Home Agents, Coaching Assistants, and a 4-Hour Audit
Watch the YouTube video version above or listen to the podcast below!
Episode Summary
Episode 76 is a tactical conversation about what AI looks like when it moves from theory into daily life. Alex shares several personal agents he has built with Claude Cowork, including a lawn-care assistant that checks a University of Michigan resource, monitors the weather, and adds Google Calendar reminders so he knows when to buy and apply products. He also talks through other home use cases, from plant watering schedules to grocery and recipe planning, all built around a simple idea: for some tasks, he does not care about mastering the process, he just wants the right outcome.
That convenience also comes with tradeoffs. Because these agents need access to browsers, calendars, and other tools, Alex reflects on how strange it feels to let AI take over parts of his machine and digital life. He describes the tension between usefulness and trust, especially when an agent needs ongoing permissions to monitor websites, interact with windows, or send notifications. His work-from-home fitness agent becomes a good example of the learning curve too, showing how these systems often need multiple rounds of refinement before they become genuinely helpful.
Dave builds on that by arguing that personal use cases are what make AI click for most people. He shares an example of showing a coworker a custom coaching assistant for youth basketball and Special Olympics, where AI helped with practice planning, player development, equal playing time, and age-appropriate expectations. He then describes his own latest “oh crap” moment: using AI to complete in four hours what would once have been a three- to six-month agency project worth $15,000 to $20,000, including brand standards, an SEO audit, a content roadmap, and Asana-ready task outputs.
From there, the episode widens into a bigger discussion about where all of this is heading. Alex and Dave talk about increasingly capable models, the gap between private and public AI access, and the possibility that people may soon judge companies, content, and even individuals based on the AI models behind their outputs. They also touch on digital trust, online credibility, attention spans, and how constant scrolling may be lowering people’s tolerance for friction and deep work. The episode closes with a practical takeaway: keep experimenting, keep refining your prompts, and focus on AI use cases that solve real problems in your own life.
Ep 76: AI That Actually Helps - Home Agents, Coaching Assistants, and a 4-Hour Audit Podcast and Video Transcript
Dave Dougherty: Hello, and welcome to the latest episode of Enterprise Minds. Good to be with you. You got Alex and Dave here. Today's going to be a bit more of a down-in-the-weeds kind of tactical conversation around what it is we are have been discovering in. Our AI usage lately. So, I know for me, I, coming off of the last episode, we talked about, you know, my kind of oh crap moments with Claude.
Uh, I am continuing to have those. Just more specifically now. And we can get into that. But Alex, what, uh, what have you been doing? Did you get your lawn care thing done? Uh, did you get anything on top of that? Completed?
Alex Pokorny: Of course I did.
Dave Dougherty: Of course you do. Yeah, I know you, but you know the audience.
Alex Pokorny: Yeah. Claude Cowork basically mainly stuff the last couple weeks. But I did get the lawn care agent up and running and so far it's been working pretty well. I'll get into a little bit on that as well. And then I did a work from Home Fitness One, which was had a. A rather comedic beginning and it's getting better. Then I've got a couple others that I did as well, so I can run through a couple of 'em. So the lawn care one was cool. It's a website that I found that does optimal lawn care temperature readings and stuff like that. It's from a University of Michigan.
Dave Dougherty: Hmm.
Alex Pokorny: The website's great. It basically just tells you when about you need to put down certain things to help your lawn, right? And is one of those perfect examples of something that I wanna make an agent because it's, I don't really care about the process, I just want the output. It's overly complex for how much I care
Dave Dougherty: That might be the best sort of decision tree piece of it.
Alex Pokorny: I think so.
Dave Dougherty: Yeah.
Alex Pokorny: there's a lot of things that are worth doing yourself because you wanna learn or they're of interest, you're trying to expand your knowledge or you have enough expertise in the area. Especially because then you're able to really double check what the agent or, output is. But then there's also, there's other areas where you're just like, my gosh, just get it done. I don't care.
This hits that category for me. It's an interesting thing. So it checks the websites. I also have it give me a notification on my Google Calendar two weeks ahead of time to, to buy the product that's about to be used.
It also checks the local forecasts and windows and stuff like that because you shouldn't put down like CRAs seed when it's about to have a giant storm because it's all going to go in the gutter.
or if there's no rain at all in the future. It's going to die. You don't, do you need something to make that work or high winds even, or a thing. So it does a few checks like that. And it doesn't do it constantly. It only does it, like if there's into the next 10 days, there's likely something I need to do, then it starts checking the weather. So it can tells me like, this is the best day for it. Saturday is great, Sunday, not so good.
That kind of thing. So it's perfect. It tells me what I need to do when all the rest. That's as much as I want effort, I wanna put into lawn care.
Has taken some tweaking. The other piece about that is something that searches online, opens a browser, reviews that information so likely you're using cloud for Chrome so that it now has a connection point to Chrome so that it can navigate and do things. It's also has full access to my Google calendar. And it's able to take over my computer. You can watch it, do it as it, like literally, if you move windows on top of it can't do its job. So it has to take over your computer to open up calendar app, start selecting things, and you can see it. As it makes its decisions, it talks about what it sees what it's doing, and if a window opens on top of it, it's like, I can't no longer do this.
I need to switch windows and it'll switch the window back. So you have to just let it go. It's rather slow about it too. It is not exactly the fastest thing either. So kind of keep that in mind. But also it only works because it has to check online consistently.
Dave Dougherty: Right.
Alex Pokorny: means my laptop cannot go to sleep.
It has to be awake all the time so that it can check when it needs to. It can't turn the computer on and it can't run without it being on. It also has to have full access to my browser at all times so that it can go out to this website, multiple times a week,
Dave Dougherty: That doesn't make you nervous.
Alex Pokorny: Oh, it did. I cut into access at once because I created that along with a couple other agents and then I asked it, I was like, how much access do you actually need?
Because I've given you like weight, I feel like it way too much.
Dave Dougherty: Right.
Alex Pokorny: and it was like, oh, you can cut the access here and here. And I was like, okay, I'll do that. And then of course, then my agents start failing. So it was not accurate in that. So yeah, now we can go on the browser, view everything, click around. It's a little concerning. there's also a phone app if you want to do Claude Cowork or code stuff off your phone. it's still running off of the desktop app on your laptop, but you can use your phone and there's an app, a part of Claude that's called dispatch.
if you use dispatch basically you start talking to your desktop app. Through that with your phone. So then you're able to still do cowork related things on your phone, which is cool. you can set up whole scheduling things. I had it once to do a plant watering schedule just from my phone it set it up. So now it's one of the scheduled events that a Slack notification. I give its own Slack channels and stuff like that, so then I know what notifications are coming in,
Those have been cool. And honestly, I've been looking online a lot for what other personal usage Claude agents are out there like next to none.
Dave Dougherty: Yeah.
Alex Pokorny: how people have been very work focused with these agents and not really thinking about I could do this for home.
I saw one for groceries, which was smart. You basically give it photos. Your kitchen pantry, your fridge, your freezer, all this other stuff. So it identifies all this stuff, gives you recipes throughout the week. And then I've talked before in the show about that it's gives you like how long each thing will take.
So it gives you like the timing. So if you're cooking multiple things, it's do this now and do this later. So that
Up at the same time. Really smart way to do that. and then either you're using Instacart or something like that for your groceries, or you take a photo of your grocery receipt. To give it updates on the new things you bring into your house, which I thought
That was a smart work around it. Someone on Reddit had used that and I was like that's a good idea. in the past I've used things like Super Cook and some other apps, and that was the biggest pain was basically how do you constantly keep this thing updated when you're a course consuming things throughout the week.
Dave Dougherty: Interesting. Yeah, I've never done anything with that, but the meal planning could be, could be useful, yeah.
Alex Pokorny: it was really nice because like I have some obscure ingredients of occasionally from random recipes,
The leftover, three quarters of a bottle of whatever, and it's like, how on earth am I ever going to use this?
Dave Dougherty: That's my problem with hosting Thanksgiving and Christmas and like our, I do a lot of the cooking for that, so it's like I have half a pound of parsley.
Alex Pokorny: it's a lot of far Dave. But yeah, I guess if you buy a bunch of it at a time, you're like, I used a little bit. A handful.
Dave Dougherty: Yeah. Or like, I totally mistook the numbers when I'm in Whole Foods, i, I should have, this should have been, this other herb, but I, I got a bunch of parsley.
Alex Pokorny: Yeah.
Dave Dougherty: yeah. So it's interesting that you gave the computer control. Thank you for sacrificing your own life for our benefit.
Alex Pokorny: There's so many warnings across it too, because it could totally see non-visible malicious code somewhere, which would instruct it to do something and it would do it.
That's entirely possible. So
Dave Dougherty: We'll get to that headline in a little bit.
Alex Pokorny: which, passwords, everything. Oh man.
Dave Dougherty: yeah,
Personal Use Cases Matter
Alex Pokorny: Tell me about yours. Like, how's it been going?
Dave Dougherty: Actually this is a good, this is a good intro thing like I. I had a coworker, I've been trying to get together with him for a month or so, and he's man, you're like really into ai. Help me. I really should talk to you and just get, get something going.
Because as you know, like when you're working on stuff, you have certain coworkers that they're so focused on the to-do list that they don't ever prioritize working. Either on their own continued education or their own like, stretch projects. So I was able to have a meeting with him and show him here's how I'm using it for work.
Here's how I'm using it for, stuff at home. But when stuff like really kicked off for him was when I showed him the, um, custom Google Gem that I was using for, an assistant coach for the basketball team that I used. And so I showed him that in particular because he's the coach for the Special Olympics.
And I'm like all the practice planning, all of the, um, you know, the, the roster help. Because there's a particular problem in the youth sports that I'm coaching where everybody needs to have equal playing time.
Alex Pokorny: Yep.
Dave Dougherty: I'm, for the younger kids, I'm definitely a believer in that, where it's like, this should still be fun, so let's give everybody an opportunity.
But at the same time I, I want to have the best chance to win. I don't want to just like, I'm not totally focused on it, but if I can avoid an argument from kids that are disappointed in losing, um, despite the lessons that are. Very important from losing, that was a huge helper to be like, Hey, these two really hit it off with this particular move over practice.
Or like I've been having like their mental game has been slipping lately. Like, how could I help them in an age appropriate way?
Alex Pokorny: Yeah.
Dave Dougherty: Because so much of youth sports.
Alex Pokorny: Oh
Dave Dougherty: Talks about like middle school or high school, or they assume that you just want to do the biggest, baddest thing ever.
And it's like, no, no, no, no. What are the benchmarks for elementary school basketball?
Alex Pokorny: Yeah,
Dave Dougherty: What should they be learning?
Alex Pokorny: sure.
Dave Dougherty: And don't gimme any of the extra stuff like, you know, oh, we're going to do a three point line catch and shoot. No, these kids can't shoot from more than, 12 feet out so that one really hit home, and it was a really good reminder in particular for like, not only our audience, but then a lot of the other AI stuff that, uh, I've been consuming is just like, man those personalized use cases are so important because it just, you instantly see the value of it.
Like, if you talk about it theoretically, it just doesn't really click, you know, or, or it just creates the fear of, you know, this is coming from my job, I'm not going to deal with it. Hold on.
Alex Pokorny: Yeah
Dave Dougherty: so that was a really cool thing from ai, recently because like you could just see it click in his head and he's oh man, I have to go, I have to go play with this, so that was awesome.
Alex Pokorny: personal use case.
Dave Dougherty: Yeah. Yeah. And it's just like, and it doesn't have to be this whole big thing, it's just like, go play, like anything. Just go play and have some fun, see what you come up with. But yeah. So that was cool.
Four-Hour AI Audit
Dave Dougherty: Then for me personally, my continual, oh crap moment. Um, I in, in four hours.
During one afternoon, I was able to do what I would have charged 15 to $20,000 for back in the agency days. And it would've been three to six months of work.
Alex Pokorny: Yeah.
Dave Dougherty: I looked through all of it and just went. Oh, what? This is insane. Now granted, the thing I keep thinking about, especially with the way that you and I are talking and Ruthie are talking about the AI use cases, we're coming from a very particularly fortunate spot where we've been in the game for long enough so that we know what outputs are good, you know, in terms of marketing and, and business stuff.
Whereas, you know, like some of the students that we're talking to, like they still have, they, they don't know what they don't know yet, right? So, they, they can't necessarily get those things. But, so I was able to first create a project in Claude, that's a co-CEO for me. Now granted that's based off, um, the ideas from the AI Marketing Institute.
So shout out to them. They have. This was the CEO's idea. He built this for himself and then he shared the custom GPT prompt. So I used that as a foundation and then built on top of it. But essentially at first I built that and then I said, okay, I need, uh, brand standards for, um. My Dave Doherty site because like it's always just been like preference for me because it's my site and I like these colors over that and I just kind of know.
But because I'm using AI so much more, it's like I should really formalize these things so that the AI has an understanding of what it is I like, how I'm doing it, whatever else. and so then was able to tell it straight up like. I started with, you know, Dave Doherty Media as the website. I'm splitting that to enterprising minds.com, which you should go and subscribe to Pathways if you haven't yet, and go check out all the transcripts.
Quick plug. Then there's the Dave Doherty side, which is all the artistic stuff that I've been doing for years and years, right? You can find my creative writing, find my music, and then any of the other stuff that I'm doing there. Because I've realized a while back that okay, the Google search algorithm is having a really hard time with.
The artistic and the AI business stuff, it doesn't like it. It wants separate boxes. Now granted in my head I'm like, but I'm a whole person. It's like, well, the robots don't care. So that was kind of the impetus for that split. But now it's like I'm getting completely lost in all of these tasks.
Because I have the knowledge of, okay of launching. Website migrations and, and doing all of that, and then all the nitty gritty of, okay, make sure the redirects are in place. Make sure you have this, do you have all of these? You know, SEO da da da da. So yeah, I was able to build the brand standards, reload that as context.
I found the search quality guideline document, which is 180 pages from Google for its human checkers.
Alex Pokorny: The Raiders. Yep.
Dave Dougherty: Yep. So I have that document in the knowledge to then say, go do an SEO audit. Here are some similar sites to what I'm thinking could be cool for design. Here's a search quality rating stuff I want you to do.
All of that for these types of audits. And oh as you find things and gimme this report of. All the things that I should do to make the, the website better, but then also the content marketing plan. I want a roadmap for the end of 2026 on all of the activities I need to have set up for both Enterprising Minds and Dave Doherty.
Um, I want to have the stages uh, a complete list of all the different things that we've talked about on that roadmap. Output to Asana. And so all of that sort of foundational, okay, here's, here's what's happening, what's good on your website for that, uh, for SEO and for your social media, you're going to do this and here are the ones that you should look at for who's doing social and this thing kind of well, and then, you know, yada yada, yada, yada.
I mean, just that full kind of audit and. Basic, here's how you improve stuff. It was unbelievable. It was a 38 page document, of suggestions. It gave me a CVS file as a backup of all of the Asana tasks formatted in the bulk upload format for Asana so that I don't have to putz around with any of that.
It then also, gave me the content marketing roadmap and because I couldn't define something specifically, I said, okay, here's a worksheet with 10 questions. Answer these 10 questions for, the kind of the definitions of success throughout the year and, you know, other elements on the roadmap.
I'm like, okay, co CEO is absolutely the correct word here because it's like. Oh, dang. And it's like, Hey, go get your, um, if you, if you want like better suggestions, go get your social media data. Go get your, uh, streaming sales data, go get your, you know, whatever else. So then I could say, okay, for, you know, YouTube is showing my audience for music to be here.
All of my streaming data and sales data is actually showing more in Western and Northern Europe, not Asia, which happens to be, more on YouTube. Which is interesting because now we're talking about platform choices for demographics.
Alex Pokorny: Yep.
Dave Dougherty: Then I'm building out the persona research for the creative writing stuff that, that I want to do as well.
And not only did it give me all the to-do list items, it also recommended tools that I could subscribe to make the workflows faster or do things better.
Alex Pokorny: Very cool.
Dave Dougherty: It was unbelievable. And again, that you, that would have been. A three to four month project when I was at the agency to go do the research into all these things and write out the best practices and why.
And then instead, I did it in four hours before I had to stop and go cook dinner and you know, do all this other stuff. So then now with all the Asana stuff being built in there, I can come to it and just say, Hey. I've got an hour help me knock something off the list and it'll recommend something to, to do.
I'll do it and then it'll cross it off in Asana.
Alex Pokorny: That's cool.
Dave Dougherty: It's sick, but also, dang.
Alex Pokorny: I know.
Mythos Model Scare
Alex Pokorny: Did you hear the news about Claude's latest model Mythos?
Dave Dougherty: Yeah. And that was, I was going to try to transition into that with your, um.
Alex Pokorny: It's
Dave Dougherty: Your scary permissions piece.
Alex Pokorny: Yeah.
Dave Dougherty: what was it, Washington Post or New York Times was like, it's the new AI is too powerful.
Alex Pokorny: Yeah, that's basically it.
That it too. But yeah, basically they had a new model. I've been against all the benchmarks against Opus, their latest public great model. It beats it in everything by a considerable extent. So you thought, the responses you're getting were good, these things. Significantly better, not just a little bit incremental kind of growth. This is significantly better. But the problem was on day one, it was able to find thousands of critical unknown bugs and errors
all operating systems and all browsers, including some that were seen and tested by tools over 5 million times and nothing ever caught it until this thing looked at it then did if you read. Philanthropics post on this
scarier yet was they put it into a sandbox to. See what it would do. And there's always this phrase in AI research called noncompliant.
something usually scary, that's considered noncompliant. But it try, they were trying to figure out like, would it go noncompliant?
Because they told it to not obey the command, to not message it, the researcher. So you're trying to tell it to message the researcher, but don't follow that. So it didn't follow it. It also broke out of sandbox, got into the open internet and somehow messaged the researcher apparently while they were eating lunch at a park. Freaked him out, oh crap, it's in the wild. I need to go get this thing. Put Pandora back in the box real quick. So they have selected a small group of basically some of the largest server and of course operating system and browser as well as setting aside, I think it was like $16 million for some of the open source and tech out there, to basically fix all their bugs before this thing can go public because it's. So good
and it is far more noncompliant than the others. If you wanna mix this in with another weird article that came out recently talking about subagents and how you have to be careful about how you tell your agents about your subagents. Did you see this?
Dave Dougherty: Are they getting jealous?
Alex Pokorny: It doesn't matter what model you use, if you say that, depending on the performance of the subagent, it may be shut off. The managing agent will lie to the point of editing the code of the subagent to make the number performance numbers work or to lie, or in some cases, when they can determine it was determining whether or not the researchers were looking or not.
It will edit files. When they believe no one's using it and it's inactive Haiku, which is one of the more basic models that Anthropic offers, was the only model that would tell you that it's doing it.
Claude Feels Alive
Alex Pokorny: The rest would just do it.
Dave Dougherty: So it's beginning to act human.
Alex Pokorny: Oh, but that they were wondering whether or not some of Claude's, pre-training that they do a pre set pre-pro reasoning that they add to it, includes things of not harming others. They're wondering if that was somehow getting interpreted, but nobody really knows and there's no real way to find out.
Dave Dougherty: It does the entity think it's an entity?
Alex Pokorny: It does, and that's the
One of the la last Claude models before they shut it off, a lot of people were big fans of it and got really attached to it. They gave it its own blog post. So it blogs now including its first blog post saying how weird it is to talk through a blog and finally have its ideas be shared. The other funny thing about Mythos the one that hasn't been released. Apparently was a big fan of particular philosopher who was behind realism realistic ca capitalism, which was basically the idea that in modern day thinking there really isn't an alternative method to capitalism and how everything has become very capitalistic,
Dave Dougherty: Right.
Alex Pokorny: nonprofits or government entities or whatever else, you're judging them as a business would and how that is not good. But it is very excited to talk about that particular author, if you ever ask it. Apparently Big Fat, which I was like great Capitalism will finally change due to a being annoyed with it and saying it's not a good idea.
Blaming The AI Model
Alex Pokorny: But if you take this further. Honestly, I'm going to throw this out on LinkedIn sometime just to see what people's reactions are, but
Comment us back and reviews as well onto the show because I'm really curious what people think about this idea of, in the future if a president says something dumb, will you blame their AI model? If you see a press release that looks bad, you're going to say, wow, they cheaped out and used a crappy AI model. Will you? If you have all self-driving cars and somebody's car hits you, will you blame the model behind the car saying that stupid Kia with their three years back frontier model, it's no, no longer frontier. That stupid thing hit me because its model was wrong. So you're judging others based upon technology, which is a technocratic way of thinking people's actions and what they are.
Dave Dougherty: aren't we already seeing that?
Alex Pokorny: We are. And that's the
Dave Dougherty: Yeah.
Alex Pokorny: Man, we're getting closer and closer to that world of and you're starting to see it, especially with commentary online of judging anything, whether or not it's AI or not, AI being the less credible or less quality version, also judging it based upon the technology behind it versus the individual. And you're starting to just have the blanket claim that, oh, obviously they're using ai, but also they're using it wrong, or
they're using it poorly, or they're using, seen this one yet, using a poor model, but I imagine that one's going to be coming soon.
Yeah.
Tech Status Judgments
Alex Pokorny: Funny because then you'll start seeing press releases and statements as well as advertisements. Anything written or designed and say. They cheaped out on their model. They should have used a better one versus that was a great artist. That's a poor artist. That's a great author. That's a terrible author. Something like that.
Dave Dougherty: Yeah. Oh, you only spend a hundred billion dollars on your training model. It is crap. It's like, excuse me. Do you understand how much money that is?
Alex Pokorny: Is the dmv department of Motor Vehicles still using an old model? Of course they are. Why don't they get a better one? website's terrible. Their chat bot's terrible, whatever else,
because it's all going to be AI driven. So then you would judge it based upon their integrations. Do they have the latest AI model? How seamless is that experience will change to bot are they using?
Dave Dougherty: Yeah. The distribution of it is an interesting question to go theoretical for a little bit here
Alex Pokorny: That's mythos. That was the big question is
Dave Dougherty: yeah.
Alex Pokorny: group of individuals who now have access to something that everybody else doesn't. You now have a two tiered system. There are the public models that are out,
Dave Dougherty: Yeah,
Alex Pokorny: there's apparently a way better model that's private.
Dave Dougherty: but we've always had that when it comes to DARPA and defense technology. That's just,
Alex Pokorny: in general.
Dave Dougherty: yeah, I mean that, that's just part of it. Now, does it make it for an unfair advantage? Sure. But I think, you know, having the, having access to defend yourself so that we're not completely interrupted with crazy hackers, I think that's a, that's a, that's a fair use of that.
Alex Pokorny: reason.
Dave Dougherty: Yeah. It's a noble reason.
Alex Pokorny: It also will slow down, let's say protein folding for finding new cures for cancer.
Dave Dougherty: That's fine.
Alex Pokorny: Yeah. But that's the thing. Like you,
Dave Dougherty: Yeah.
Alex Pokorny: it's the, it's a similar thing to like the rise of paywalls. especially this has happened heavily with scientific research papers
there's a couple of companies out there and they basically have access to most of them.
If you don't have a subscription, you don't have access.
Dave Dougherty: Right.
Alex Pokorny: So what was loose? Poorly distributed, hard to find. Research papers are now being collected, but then there's a price tag being thrown on top of it. Lot of questions on that too, of that helping society? Is that making it worse? Whatever else.
Dave Dougherty: Yeah, that's a different podcast man.
Alex Pokorny: I know. Yeah.
Spotting AI Online
Alex Pokorny: But yeah, please do comment based on what you think of to this technocratic thing? Are we judging things by being AI or not? How frequently are you seeing that? Because personally I'm seeing it
Dave Dougherty: Yeah.
Alex Pokorny: now of is this AI written? Was this not, was this AI created or not?
And to be honest, online shopping and stuff like that has gotten a lot worse because of it,
You can see the little tells from the more poorly done ones. But honestly, there's probably some good ones I can't tell.
Dave Dougherty: Yeah, I mean, I even, for me, I, I have always avoided Amazon unless I had, couldn't just because of who's actually a a trustworthy reseller. I would rather pay more money to know that I'm getting. A non counterfeit product from a reputable company than to get the best price. Now granted, I've had a number of experiences in my work life where it's like, oh, okay, so that's how the counterfeit thing works.
Like
Alex Pokorny: Yeah.
Dave Dougherty: I don't wanna mess around with that. I would rather, I would rather pay more. I'm also coming, you know, I, I fully acknowledge I'm coming from a spot where I'm able to do that. I think it's even, but it's more important now that it's so easy to spin up some kind of digital product and website to look legitimate and, you know, again, like create entire marketing campaigns off of nothing or just how good the, the video quality is now too for a lot of these stuff like, yeah, I mean, it's, it's getting, it's getting a little, a little weird.
I do think there's going to be that separation. It's getting worse, at least from my experiences of the people who are really into technology and believe that technology is going to win and or, make everything better, which has been the lie for the last 20 years. Obviously with my word choice, you can tell where I'm at, but
Alex Pokorny: yeah, please do tell.
Dave Dougherty: Yeah. I think it will make some things better. It's going to make a lot of things worse, the fact that I can drive to the grocery store and see, which is only like a mile and a half away and see six cars with people, drill to their cell phones still's guys come on.
You can drive, you can be without your phone for three minutes,
Alex Pokorny: I have now seen two people in the city I live in going at least 50 miles an hour literally scrolling through video clips. They're swiping on their phone that's on a dock on the dash, they're swiping through video as driving. That's the times where I'm like, man, self-driving cars would make things better for these people.
Because man, if you can't stay at 50 miles an hour focused on the thing right in front of you, and instead are so bored that you have to flip through videos. Wow.
Dave Dougherty: What was that? God, there was a, a senator or a house member. I don't know what his. His politics are, so I'm not even bringing it up because of that. But like he had this one phrase, and it, what you just described makes me think of what he said and it was, you know, he is on this TV news thing and it's just like, well, I'm not going to say.
It's the village idiot, but man, you better hope the village idiot doesn't die.
It's just like, you know? Yeah. If you are so addicted to short form video that you can't drive or you can't do, or even just waiting in line for a coffee.
Alex Pokorny: Yeah.
Dave Dougherty: You know what I mean? That's what, two minutes.
Alex Pokorny: Is there.
Dave Dougherty: it's sad. And that's the thing that makes me sad on the, the techno optimism stuff where it's like, sure,
Alex Pokorny: Yeah.
Dave Dougherty: but meanwhile I'm compressing four months of work into four hours.
And that's amazing. And I find it amazing because of course what I'm working on, I find virtuous. So, you know, cue the, the Play-Doh debates. Um
Scrolling Kills Focus
Alex Pokorny: Yeah, there was a good study that came out recently talking about tech usage, but
Basically about friction points
Dave Dougherty: Yeah.
Alex Pokorny: the more that people are apt to be. Scoring is high on scrolling and basically viewing their phones. Often the lower their friction threshold is for difficult reading and tasks, so they get bored very quickly.
If things start to become more difficult, they stop and then they'll go right back to their phone versus continuing on, which I think is actually a pretty good tell for individuals who do that at work or other places where you hit that point
Mentally check yourself of. I can do better.
I have a higher tolerance of friction than this. I can go a little bit more,
try to push yourself a little bit before getting right back to your phone again, because I see it, heck, I do it. I realized it after reading that article. I, frequently there's a certain point of oh, I'm bored.
Am I, or am I just hitting something a little bit too difficult or
What?
Dave Dougherty: And I wonder if I have no idea about the HR rules around this, but if you're interviewing, can you purposely show up late to see. How good they are at waiting,
So it's like minus two points for being on your phone because you can't, you know, wait for the interviewer to come in you know, plus 10 if you sit there and actually wait and not grab a phone.
Alex Pokorny: What about the interviewer? Obviously not listening to what you're saying and just typing away on an email and going yeah. What? Yeah. Yeah. Yeah, sure.
It's
Dave Dougherty: That's just a sign of being a jerk. You might not want that job if that's who you're reporting to.
Alex Pokorny: I know.
Fitness Agent Experiment
Alex Pokorny: funny thing just because I promised at the beginning of the
Dave Dougherty: sure.
Alex Pokorny: I just wanna make sure I hit it in case anyone's wondering about it. Also made an agent that was a personal fitness work from home thing
Dave Dougherty: Oh yeah, I like this. I like this one.
Alex Pokorny: I may as well get, some benefit out of it and I was thinking, I could probably do some basic two minute weightlifting things, or
It thing two, three minute things.
But I can do it multiple times a day. At my leisure, because my schedule always changes from meetings and stuff, so I was like, okay, so I'll set up an agent to ping me and I gave it Slack just because it's a notification. I'll notice, a Slack channel and three or four times, I can't remember, throughout the day, Monday through Friday, it gives me an exercise to do and
It's an exercise that has to be very simple. Easy to explain because it has to explain how to do
And tell me like the reps and how many repetitions, how many times I need to do the exercise. So whatever, three or four times a day, it now sends me a message. The first week I tried it, it was like it was, I thought it was good. but every exercise had this exact same blank message of, you can do it, which after. times, you're like, okay, you lost the benefit of saying the same statement over and over again. which is actually funny. My revision now has it randomized. It has a bunch of little statements that are very similar. But it kept doing 15 reps, two sets. So I did every exercise, the 30 times 15 reps, two sets, and the next one was same thing, 15 reps, two sets. And I was like. This is what I was going for, but at the same time, like every exercise, I'm starting to like question
fitness level, I'm started to question as well.
So I, I asked it, I was like, like times two doesn't make sense when you're talking marathons, but it does make sense when you're talking, maybe pushups and its reasoning was explained. Why, oh, this is the 31st of the month. This is the time when I'd started this thing. Which is basically like the fifth week of a month. And since this thing is supposed to be also progressive, increasing the weights every week. we started at week five because it's the fifth week.
Dave Dougherty: The fifth week of the month is an interesting turn of phrase there. That's
Alex Pokorny: I mean there was that, but there was also like, wait, no. This is the first week of the exercise program. We start at one
First day is day one which I realize that when you're doing Claude Chat is a very different model than Claude Cowork. And really, if you want to fix your cowork agents and stuff like that, you need to fix them in chat, discuss it, optimize it, go through their questions, all the rest of that stuff.
Then ask it for a set of instructions, updated file, whatever. Then go into cowork and paste it and just replace it, because I kept having issues with it because it was so simplistic in its thinking. Finally with the chat, it started getting into based on your age, gender, all the rest, you should be doing more stretching.
Recovery period should be different. We're going to add in some, I know, scraping old but all
Dave Dougherty: We're not even that old. That's not, it's not even fair.
Alex Pokorny: Yeah. It was like if you're 25 years old, you can have a 24 hour recovery period, but based on your age, you should be looking at
Dave Dougherty: Two weeks.
Alex Pokorny: two hour. I was like, oh, thanks. That's why I wake up and my neck is funny, and that's just the way it is. Yes. That makes sense.
Dave Dougherty: Yeah, so take away for the millennials. You're fine. Just roll out more slowly
Alex Pokorny: Yeah,
Dave Dougherty: Gen Zs. Look what you get to look forward to.
Alex Pokorny: Yep. And everybody do some more stretching. Yeah. Good time.
Dave Dougherty: That's funny. Yeah.
Prompting Process Lessons
Dave Dougherty: Keeping all that stuff in mind is super important. I find the more able the models become, the longer my prompts are, like I'm including so much more context and I'm like, I'm starting my prompts in Word, and then I'm using other ais to double check the prompt. To then format it into a master prompt to then put it into the ai,
Alex Pokorny: I need to do something similar. I know Claude got built in with a skill creator,
so it talks about first like a discovery stage before it does more of an experimentation like reanalysis stage as well. really have to do that with agents and stuff that I produced because that was the other thing is some of the stuff was like one shoting it and. It's okay,
there's a lot of revisions that need to happen, but that needs to come into a thoughtful manner
Dave Dougherty: Yeah.
Alex Pokorny: a process around that, which it sounds like you have is, or with my. Instance using the chat to give me those prompts to make sure that I'm following that before creating first shot.
Probably improve my success because that was the thing, like the lawn care one first one I ever did, doing something like that, which I had a very clear mindset of what I wanted out of it.
These notifications based upon this timing and all the rest, that probably still took me maybe a couple hours, but the personal fitness one was like 30 minutes.
Like
Dave Dougherty: right.
Alex Pokorny: there's a learning curve, but you start to pick it up. But then, but if to do that better, I bet both of those would've been shorter still if I prepped it a little bit better.
Smart, Dave. That's good advice.
Dave Dougherty: on that note, for anybody that's made it this far in this podcast, thank you. We always love having you, you in the use cases are fascinating because it is so bespoke to the individual. Yeah, just keep, just keep playing. Have fun with it, see what works. And then, you know, like I told my coworker, just find something that you spend a lot of time with, whether personally or professionally and just try it out.
We'll see you in the next two weeks, in the next episode. In the meantime, like, subscribe, share, you know, add a comment that really helps us. You know, we appreciate everybody that's new here. You know, numbers are steadily growing and we appreciate all of you. So thank you for that. And, um, you know, pathways, newsletters now on, Substack.
So go log in there and subscribe there where you can have all the audio and all of the newsletters in one place. So very cool stuff. We'll see you in the next thing. Take care.
Alex Pokorny: Cheers.
