(upbeat music) Welcome to The Prociptical Show. Boxes where we believe that people like you can get the important things done without sacrificing your health, family, and things that matter to you. If it's your first time listening, welcome to The Show. I'm Tan Fam founder of Asian Efficiency, based here in Austin, Texas. And I do this podcast to get it with my co-host, Brooks Duncan, all the way from Vancouver, Canada, Brooks, how's life? Life is good. World Cup is getting close. Getting excited, everything's good. How about you? Do you have tickets for any games yet, or are we hunting? Yes. Canada versus Switzerland. I manage to win the Hunger Games and get tickets to that one. So that'll be exciting. That'll be fun. Yeah, I scored three tickets for all the games from the Netherlands, because I grew up there. But in fact, I haven't told my brothers those yet. So if they're listening, this is maybe the first time they might be finding it out, but I'm gonna be missing one of them, 'cause I have to go to a wedding for one of the matches. So I already prepaid, so I have an extra tickets, but I'm sure my brothers will put it to good use for somebody that they know. But today we're gonna be talking about AI. So if you're new to the show, first and foremost, welcome here on the productivity show, Brooks and I talk about all things productivity. And we're known for three philosophies around productivity. One is happy people are productive people. Two, one, two week a week is all takes a massive productivity gains over time. And then three, everything we want to do on the podcast, on the newsletter, and of course, should be simple and actionable, because that's how we can get started in build momentum. So if you ever wanna get in touch with us, feel free to reach out to us via email at
[email protected]. And we also like to start off our podcast with our top three favorite resources. So since we are talking about AI today, Brooks, I figured I'm gonna share some of my favorite AI resources. The first one is a platform called Lindy. So it's spelled L-I-N-D-Y Lindy. And this is my favorite platform for building AI workflows and agents. I've been using it for a little bit over a year now, and it's become kind of a master of the tool. And now I also teach like life classes, and I'm heavily involved with the company, not like in a formal way or contractual way, but I just do a lot of free classes for them, for their audience. It's just because I've been such a user of the tool and big fan of it. And it's a great way to build AI agents and assistants. So I have my primary AI assistant on this platform. And when I typically build AI assistants for others, I recommend people use Lindy. So big fan of it, we'll link to it here in the show notes. The second resource I wanna recommend is perplexity computer. So you might have heard of perplexity as a AI search engine, and it's great for that. I use it as well for that purpose. They have this new feature called perplexity computer that is their version of having an AI agent in the cloud that can build apps and tools for you and do extensive research for you as well. So the best way I can describe it is, hence the name, a computer in the cloud that can just do stuff for you. So I use this all the time for building apps and tools, for prototypes, I use a lot for finding data and doing research online. The technical term is technically scraping. But scraping has a bit of a bad connotation when people think about scraping and they think about copyright infringement and downloading stuff, right? But scraping could also be, for example, hey, I'm trying to find all doctors in Austin, Texas and trying to figure out who has the best and highest reviews because that's probably the doctor I wanna go to. That's a form of scraping as well. So perplexity computer in my experience has been the best in terms of research and then also for online tasks that you want an agent to do. And the third and final resource I recommend is Whisper Flow. So I'm a big proponent of voice first workflows. So instead of typing in email, I am now drafting emails or sending emails just using my voice. And the app that I like to use for this is Whisper Flow. So it's a really smart dictation app. And where things are gonna be going is that as you use your voice, you can start executing commands over time as well. We're kind of there and we're starting to get there. But Whisper Flow is my app of choice for any sort of dictation that you do because it's a fast way to prompt and also a fast way to send out emails as well. So interesting when you watch YouTube videos about AI and all that type of stuff or like online classes or whatever, like pretty much everybody uses Whisper Flow. It's like, I don't know if it was the first one, but it's definitely the one that has taken the most mind share for people. And all of the kind of like high level AI users, I put that in quotes, seem to use your philosophy of voice first. So I think there's gotta be definitely something to that. - Yeah, I think they were one of the first and kind of grab market share from people who talk about it on X and YouTube. And I first discovered it that way and just kind of used it since. And what also helps is if you refer somebody and they'd not be using it, you get a free month. And so that helps as well when you have a great affiliate over for all program. And I think honestly a lot of like YouTube recommendations are primarily driven by that. And so I like to hopefully offer recommendations to people where that's kind of agnostic, where for example, I'm a power user of Lindy, Proplex2Computer and Whisper Flow regardless of the affiliate programs that they might have. This is something I would recommend for everybody to go check out. And we'll link to it here in the show notes as well. So you can see it there. So Proplex2, let's talk about AI here. Something that I've noticed that has been a shift in AI maybe in the last six to eight months is that we are going from the AI is telling you what's possible and what you could do to now AI actually doing it for you. So for example, in chat, UBT or cloud, you could say, "Hey, draft me an email to make this sound more professional or summarize this email for me, tell me what to do." And then chat you be, "It's pretty good. I didn't tell you, hey, here's the email that you should send out." And this is the AI telling you what to do. Now where things are currently today and where things are going even further is you can now have the AI actually do it for you as well. So instead of you telling AI, "Hey, go draft me this email and make it sound more professional," which we will do now, it can then also put that email in your email inbox as a draft ready to go. And if you like it, you could then say, "Hey, I like this, go send it, and then it will send the email for you." This is now possible as of maybe less than a year ago or so if you were really on the cutting edge. And what I consider coming edge is relative to everybody else, because I live and breathe this. But for most people, this is really cutting edge, meaning the AI cannot do tasks for you whereas for most people, they still are dark edges, this may be a strong word. But they just use AI as a chatting tool or as a search engine when in fact it can now execute tasks. And I think that's a big revelation for all of us here. - Yeah, absolutely. If I think of my AI use over the last year, let's say, it was definitely 99% the ask questions. Find out information, get recommendations, research something. And it's been great for that and it continues to be great for that. However, I think in the last probably month or two, it's definitely shifted. It's a lot more of, hey, what can I build out to make my life easier or my job easier? I also think of AI now as almost an unlocker as well. And by what I mean by that is, let's say there's something you wanna do in your life or your job, like an example of this actually happened to me today where I use this learning management system and it's kind of an old junky one. And I was hitting this limitation like, there's this certain thing that didn't quite work the way that I wished it worked. And in the past might have been like, oh well, that's just the way it is. Or you might have gone to the next level and gone to a clot or a chat you beat you or whatever. And gone, hey, ask it like, hey, what are my options here to do something about this? And it would be like, oh, check this setting, do this setting, which is kind of to your point. However, now I think a lot of these tools, you could then go next level. And this is what I was starting to do this morning. I haven't got too far in it, but it's like a possibility machine. It's like, okay, well, this thing I'm doing isn't working the way I want it to. Can you build me something that does work the way I want it to? Or can you make up like if this thing uses plugins? Can you make a plugin to make it work the way I want? And it's just like a possibility unlocking mechanism that we've really never had before, where it's allowing you to do things that you couldn't do before. So it's just so fascinating the number of levels that have been kind of opened up in the past six months, even or so. Yeah, so what's your tool of choice when you're doing stuff like this? I'm mainly, and I'm not saying this because they're a sponsor, but I'm pretty all in on cloud right now. Like I do have a chat, you be de-account, I do use it. But between cloud, cloud, code, work, I use a lot, and then the cloud code a little bit as well. Those are my three primary tools. However, I am running into something that I think Lindy really unlocks. And,
and things like OpenClaw, which we've talked about on the podcast before, unlock as well is this. I'm realizing that a lot of what I want to do kind of relies on my MacBook Pro being on and online, which it often isn't, right? Often I have it in the living room and close it overnight or whatever. So I'm starting to get at that point where I'm like, it would be nice to have something all the time that could do whatever I want all the time and I'm kind of like passing into that next level, which I know you've passed a long time ago and you've seen people passing. I'm starting to get there as well and so is my son. Yeah, I think that's a really good insight there because first of all, Brooks, what I recommend you do ASAP after we finish recording is you go find a Mac Mini because they're back ordered right now and it might take a few weeks or a few months depending where you are. But the reason people are buying these Mac Minis is for exactly this purpose. Is that you have an always on machine that can just do stuff for you even though you might be closing the laptop or your computer and the beautiful form of the Mac Mini is that one, it's tiny, two, once it is set up and up and running, it doesn't need a keyboard, it doesn't need a monitor, it doesn't need any of that. You can just kind of remotely log in and do stuff or even from your phone, especially now with Claude with having, as long as you have the desktop app open on your Mac Mini, then you can use something like this patch, which is their newest feature where you can just, it's kind of like their command center in a way where you can just tell it to do something and you could say, "Hey, what was that PDF file that I downloaded yesterday from this company? Could you pull it up for me and summarize it?" While you're on the phone, on the train somewhere, it can then open up your computer, go to your downloads folder and pull up that PDF file and then summarize it to you essentially what's in there. So this is where things are kind of going, which is kind of cool. And OpenClaw is kind of like the first AI agents platform that allowed you to have this form and function of always on and it's just doing stuff. And something that makes OpenClaw very unique is what they call their heartbeat system, which basically means that it can wake up on a certain interval and just do stuff. So the heartbeat system is actually just a markdown file. It's like maybe 10 lines or 20 lines depending on how much you put in there. And you could say, "Hey, every 30 minutes, "I want you to wake up and then check my email inbox "and is there something urgent from a client or a VIP contact?" Then send me a message on Slack. And now you can just go about your day and it will wake up every 30 minutes or so, not like strictly on the strict interval but roughly around that time frame. And if something urgent is in there, it will just notify you then. So then you kind of feel like the thing is proactive. This is why people are so in love with OpenClaw. It feels proactive when in fact, it's just a, you know, in technical terms, it's a cron job in a way where it just wakes up every 30 minutes, which is a schedule task essentially. And so now it's doing stuff for you, not just doing the research and telling you what to do, but it's actually doing it for you, checking your email, drafting emails for you. Before we start recording, I had to reschedule an appointments with my doctor. And so I told my OpenClaw, like, "Hey, I have an appointment next week on Thursday, you could see it on my calendar. Could you email them and ask them we could move it to a few days later?" And it was just like, "Okay, cool." Email sent, right? And it was like, "Great." I didn't have to review the draft. I've done that a few hundred times now, so I kind of know how it writes and I've given it enough instructions. And so we're in this phase right now, where AI is not just telling us what to do, but it can actually do its as well. And I think that's the big unlock people are starting to see. And to your point, if you close your laptop down, you have no possibility of taking action, unless, you know, it's somewhere in the cloud or it's always on somewhere, right? And one of the things I always like to say in my workshop is if OpenAI and a Thoropik all went down today, how much would you be affected by it? Like, how would you feel that day? And most people, when they first come to the workshop, obviously they're there to learn, you know, and so most of them would say, "Hey, you know, I probably wouldn't be affected that much." Like it'd be kind of annoying, but not really. Me, on the other hand, I would be highly annoyed because that means 40, 50 agents start working. So I'm not getting production done. And then two, most of the work that they do now would just not be possible because I heavily rely on this. And the reason I bring this up during the workshop is that there's a correlation between leverage and your dependency on the AI tools anyway. And what I'm really trying to say is, if and Thoropik and OpenAI are down and you don't feel bad about it, that means that you have no leverage. However, if you feeling like, "Oh my gosh, you're like, this is not good, this is not great at all," that means you have very much leverage using AI. And that's actually, again, this is my personal opinion, that is a great place to be because AI is built and developed so that you can create this leverage, which something is, you know, I think you're still a competitive edge and most people don't have it yet. And so, you know, one hand, I'm also a believer of you don't want to be too reliant on something. And so this is one of the reasons I start buying like a Mac studio. So I have like AI locally on my computer. So if the cloud servers were to go down, which they have been in the last few weeks, like I can still move on and move things for a little bit slower, but not as bad. So, but the big takeaway from this is everyone should realize, like, "Hey, there is a relationship between how much you use and how much leverage you have as well." So one of the things I think we could do in this episode, we've done AI episodes before episode 600, which I'll link to in the show notes. We did a whole deep dive on a lot of these AI topics, kind of demystified things like, you know, we talked about OpenClaw, we demystified things like skills and stuff like that. One thing that I do want to say before we get into, I think it would be good to give some like really solid examples of your workflow and how you are, you know, you're talking about that leverage, right? Like, I guess the next logical question is, "Okay, what are some examples of this leverage? Like, what is it doing for you that I think would cause pain?" So I would love to unpack that in this episode. Before we get into that, one thing I do want to mention 'cause you were talking about how, you know, you set all these things up and there's skills that make it happen. And again, we talked about that a lot in episode 600, so I'll link to that. But one of the things I really, I don't think I really, kind of understood is that, and you see these examples of skills, right? People will show it. It's just a markdown file, you know, with all these like human readable instructions. But when you see this stuff, you might think to yourself, "Oh man, that means I needed to like learn how to write these things and what to say and all that type of stuff." And what I didn't really understand is these skills things that you're seeing and that is doing these things for you, you don't actually need to know how to create them. (keyboard clicking) It's that time of year again, when we start thinking about gifts for our parents and the problem is, a lot of times they have everything. Well, a great option is to give them the gift of lifelong learning about topics they're interested about. A great example could be Joanna Gaines' class called "Designing a Home That Tells a Story." That's one of the great things about our parents. They have the story to tell, and their home is a great way to express it. Thanks to Masterclass for supporting the productivity show. Masterclass fits into real life with audio mode, short lessons on your phone or TV. You don't have to carve out hours, you just have to start. Plus, Masterclass makes a great gift for the upcoming Mother's Day. Right now, our lessons get at least 15% off any annual membership at masterclass.com/tps. For example, one tip from Joanna's class is to pick three words or phrases that describe how you want the space to feel, and then work from there. It turns out there's so much parallel between designing your home and what we talk about here on the show. Unlike other learning platforms, Masterclass puts you in the room with the people who defined their fields. We're not talking just experts, but the best in the world. An audio mode turns your commute or workout into a classroom. So you can download lessons for offline access, and you or your parents can learn on your own terms. Plus, there's no risk. Every new membership comes with a 30-day money back guarantee. Masterclass keeps adding new classes, so there's never been a better time to get in. Right now, as a TPS listener, you'll get at least 15% off any annual membership at masterclass.com/tps. That's 15% off at masterclass.com/tps. Head on over to masterclass.com/tps to see the latest offer. I found out the hard way that owning a home is amazing until it's not. One minute you're sipping coffee, and the next, you are ankle deep in water from a burst pipe. Repairs don't care about timing, and they definitely don't care about your budget. Regular homeowners insurance usually covers the really big stuff, but not so much the day-to-day wear and tear like plumbing failures, HVAC breakdowns, electrical issues, that sort of thing. You're often on your own for those, and that's where home serve comes in. It's like a subscription for your home. For as little as $4.99 a month, they've got your back. They've helped homeowners like you for over 20 years with a trusted national network of 2,600 local contractors. Now, my wife and I have owned homes for 23 years now, and there have been so many times that looking back, I think I wish I knew about home service.
back then. For example, Vancouver doesn't get much snow and it doesn't get that cold, but there was this one winter where our furnace died and we were without heat at the worst time. It was so annoying and expensive. I could have just contacted HomeServe back then. So help protect your home systems and your wallet with HomeServe against covered repairs. That's homeServe.com. It's not available everywhere and most plans range between 499 to 1199 a month for your first year. Terms apply on covered repairs. In a lot of cases, the AI, whether it's cloth, whatever, creates them for you. All you need to do is say what you want to have happen and it writes it in the best way possible. So when, and I wanted to mention that because when Tang gives these examples of a lot of the things that he does, you don't need to know this arcane language to prompt engineer these markdown files to make it happen and these skills. It's more about defining what you want to have happen and the tools will get you 90% of the way there. Would you say that's accurate? Yeah, 100%. So nowadays, we can talk to AI and just normal, hillbilly language of me, if you may well. There's this funny thing right now where people have their AI talk back to them in caveman language because caveman says, I understand. I do. Thank you. And the reason for that is to save tokens because they just surround the language of what they say back to you. But that's not the point. The point is what used to be a very valuable skill, which is still is in some degree, is like programming or software engineering. And coding has been mostly replaced with AI nowadays. And this is being dispersed every day as these models are getting better. And so in order for us to have all these little systems and workflows, we had to have disability to code. But because AI is so good now at coding, you can just describe in your own words what you would like it to do. And because we are now in the place where the AI is not just telling you what to do, but it can actually do it for you now. And it has this capability called tool calling, which basically it can call all their apps and tools and naively integrate with them and then use them in a really simple way. Now you can actually combine different things to have the AI kind of like create a series of steps or create a workflow now. So what's a practical example of this? So every morning at 8 am, I get an email and in that email is my plan and my meeting log for today. So it tells you, Hey, Tim, good morning. Here's the weather today. I notice that you're playing paddle at three o'clock. Today is going to be roughly 27 degrees Celsius. So make sure you drink a lot of water and electrolytes before you play and after. Right. And then here are your top three meetings. You're meeting with Brooks today. You're recording a podcast with with him. Here's the topic about AI. So I want to make sure you know that you're some things, you know, you guys are going to be talking about second meeting you're having is with Mary. You guys are going to be talking about the content schedule. She needs like two things from you. One is the affiliate links for different programs and then two banking information. So you get paid for those signups. Right. And then number three, it could be like, Hey, you're having lunch with this guy, Evan. It's going to be a dislocation based on your last five to seven emails with him. This is what you guys talked about. And by the way, I checked your text messages as well. You guys have been talking about X, Y and Z. So you want to make sure you understand that context before you go into that lunch. Right. So this whole thing lands at my email inbox every morning. And I also get a slack message as well. So if I don't check my email, it's in slack there too. And what does it do? It checks my email. It checks my text messages. It checks the weather. It does research on every single person that I'm meeting so that they know and understand the context. And then they put this like briefing together that I get every morning. So that when I just kind of like drink my matcha, I just sit there and read and go, okay, this is what my day looks like. Here's where I'm meeting. Here's why we're meeting. And so when I'm on back to back meeting sometimes, I'm so glad I read the briefing because when I go from one meeting to another, I know roughly what I'm getting myself into. And this eliminates a lot of work that, for example, executive systems used to do a really high level executive assistants would oftentimes spend 10 to 20 hours creating a meeting brief for their boss or the people that they're working for to just go through their email inbox, look at their LinkedIn, do some Google research and try and figure out why are they meeting and how can I summarize this in a way so that this person can read the meeting brief and then be fully prepared, right? And now that is being generated in like two or three minutes. And the compute power to generate this is maybe like a dollar or something like that. It's insane leverage in that sense. And so you can take this as a step further where this is what I call the meeting prep agent. So every time I have a meeting, 30 minutes before the meeting starts, I get a notification on my phone that says, Hey, here's your next meeting. Here's who you meeting. Here's a quick summary. Here's what you're talking about. Here's some things you need to discuss. And so if I'm on back to back meetings, if I cancel or end and wrap up one meeting, I can look quickly look at my phone scan in like 10, 15 seconds, what my next meeting is about and then jump into the next meeting and go, Okay, Hey, how's it going, Brooks? I know we're talking about X, Y, Z today. Let's go. And so this is a real pain points that I had and I know a lot of people have, which is when they're on so many meetings, they often times have no clue why they're running into a meeting and they go, Oh my gosh. What are we doing? Oh, yeah, let me look prepared. When in fact, I'm not. And as you guys are talking, I'm just going to do some chit chat and you know, when back some time, while I'm trying to figure out why we're all here and what we're trying to do here. You know what I find really interesting about stuff like this to about using AI for for this type of thing is like, I don't have a morning briefing quite the same way that you do. But one thing I've started to do or one thing I've done forever is when I'm doing my journal, I will look back at what my tasks that I completed yesterday. And I had this whole whole setup where it would like position my my windows on my monitor and open up, open up my journaling app and open up on the focus, which is my task manager and not like a completed perspective. And then I would like look and I like a lot of times kind of copy and paste over between the two. And so to me, I thought this is actually a pretty high leverage way to save me a bunch of time is to have I use cloud core for this to to basically connect to my army focus, pull my completed tasks and then you know, redo them in past tense. So if I if my task yesterday was right XYZ, it would say wrote XYZ and stuff like that. So that's basically the only instructions I gave it is, hey, pull my completed task and make it past tense, which is great. It did that. But one thing it also did is it cleaned it up and made it clear what it is because a lot of times I'm just like my tacit on me focus are just like using a web clipper from G or something like that like or something where like I know what it is, but it just like has a bunch of G-RTK codes and all that sort of stuff. And cloud would summarize it and put it in like human reading language, get rid of all the craft. And so it's really clear what it is. And then it would also be like and this is something that I've really noticed too is it it always kind of like goes the next level. So yes, it lists list my tasks and stuff like that. But then it would say like, hey, it looks like you really focused on so and so project yesterday, you know, that type of thing. It kind of go kind of like what you were saying about oh, make sure you drink lots of water. Like it kind of like kind of knows what would be helpful to you without even you having to ask. And I think the more you I'm going to guess the more you do this stuff, the more those kind of like next level opportunities pop up. And that's one thing that I think I wouldn't have even thought to have it analyze my tasks for for trends, but it's it's doing it. And I think these are the things that you kind of don't really get until you start using it. You know, 100% and one thing that would take that to another level would be persistent memory. So for example, let's imagine every time with co-work, you're analyzing your tasks. And then at the end, whether you're prompted or not or maybe it invokes a particular skill where it always does it, it will do an analysis of your task. And then and this is the simplest form of a memory would be like updating a text file or a markdown file and just write down here the top three things that Brooks did today. And here are some things that are really important to him that he finished or here's a project that he just wrapped up. And so now you might have like one text document that is kind of like a running log of your memories. And so every time the AI is working on something, it can reference this text file and go, Hey, let me figure out what Brooks is currently working on or maybe this project that he has on the stuit list now is actually finished already. So I can actually mark this task as done because in the memory, it shows that this project was wrapped up. So why should I then tell Brooks to remind him of this task? Right? And so the way I kind of like figured this out was I just started to realize from using this that my AI was kind of stupid in the sense that it didn't know what I knew.
And so if a project was done, it didn't sometimes, didn't know that. And so I was trying to figure out how would AI know that this project is done? I would have to either tell it, or it would have to inform it somehow in a smart way or reference it somewhere, right? So the simplest example I can think of to illustrate this idea is oftentimes, if I'm done with a meeting, I have to send a follow up email. Now the AI will actually now draft the follow up email for me. So as soon as the meeting is done, assuming a transcript of the meeting, it will use that transcript and draft a follow up email. Say, hey, it was great meeting you today. I promise you I would follow up with this proposal or this document. Here you find it attached. And then it's still up to me to have final say because I'm the one reviewing the email still. And I still have to send it. But the issue I was having is that it would create a task in my task management system to say, send follow up email to Brooks. And it wouldn't know and it would keep reminding me to send that follow up email. And even though I might have sent it already. And so I was trying to figure out, OK, how do I fix this? And what I ended up doing was I created a nightly job at 4am where OpenClaw is going through my task list. And one of the data points or data sources, in this case, or inputs, it will look for us. It will go through my send folder of that day or the last 24 hours and see if something matches my task. So if one of my tasks was send to follow up email to Brooks, one of the things it will do is it will actually scan my send folder and have two email inboxes. So it will scan both and see if something matches that particular task. If it does, it will mark that task is done. And so now when I wake up to my to do list every morning, stuff is being marked as done that was in fact already done. I just maybe forgot to mark it as done, which happens all the time. And so now that I have this little quote-unquote backup workflow, that does it for me where it checks, that's great. Then I took it a step further and this goes back to AI. It's not just telling us what to do. It's not doing it for us. Let's say I have a task that says something like, hey, I need to do some research on pricing for whatever things cost nowadays before I go on the trip, right? Let's say trip insurance or something like that. It would then create a-- if I had a source somewhere, it would then-- every day at 4am, it would find this task and go, hey, can I do this task on behalf of TAN? And if it can, it would do it and then update the task with the notes. And so now, for example, if the task was, hey, do some research on pricing on trip insurance, it would still have to figure out, OK, what does this trip that TAN is talking about? So it would then look at my calendar. It would look like a memory log and kind of figure out what that is inferring from. And then it would basically call some tools like perplexity or brave and do some research on trip insurance and then update the task notes with, hey, I did some research on this. Here it is. So by the time I wake up and I'm trying to prioritize my to-do list in the notes, it's already saying I did this research for you. Here's it. Here it is. Go review it. And I'm like, great. All I have to do is QA. That's for review it. And I'm good to go. So now I'm waking up to do this kind of like half done. And this is one of the most-- I don't know how to describe it, but one of those powerful feelings in the world when you're just going to bed. And you just know, things are being worked on while you're asleep. It's like the David Allen, the mind like water thing. It's this feeling that, oh, yeah, my AIA agents are half things handled like I can sleep peacefully. Yeah. And it's as simple as if you had a MacMini running, this is why you want to have a MacMini or just a computer that's on. Then you can tell it to do stuff like this. It's instead of you sitting behind a computer, prompting it, you just scheduled the task to write that prompt and just execute that prompt. And the prompt could be as simple as go through my task list, check my calendar. And if something was marked on in my email inbox or in my send folder and I did it, market has done, could be as simple as that. And we'll just run every day at 4am. And that's why we want to have machines on 21st 7. And that's why MacMini's are sold out everywhere. But by the way, on the topic of persistent memory and text files and the markdown and stuff like that, I've been talking about Obsidian for probably years now on the podcast. And I finally found like a really strong use case for Obsidian is just like the AIA is writing these text files that we talked about making this gills and stuff like that. But Obsidian turns out is a very great window onto plain text markdown files. So you don't need it. You can just do it through things through the finder or Windows Explorer or whatever. But Obsidian makes managing and looking at all of these various text files. It's really easy and nice. So who knew Obsidian would become an AI gateway? Yeah, I'm actually, in fact, looking into it, it's one of those things I have on my list of things to explore is reconcuring using Obsidian now because these AI and models are just so good with text that if you are just maybe 10, 20% organized with your files, that's all you need because they can do everything else. Finding it, summarizing it, creating relationships between documents, it can do all that for you now. And so the only thing that we have to do now is really just externalizing our brain and just somehow putting it into Obsidian so that they can do everything else for us. And I think that's a great place to be. So by now, you probably know is that AI seeps into pretty much every episode of the productivity show these days. And the reason for that is simple. Tan and I keep finding workflows that make us faster and better with AI. We can focus on making our overall jobs better by taking stuff off our plates. A recent example for me, I had an annoying and mind-homing workflow, taking information at a JIRA, updating PowerPoint slides, updating as Page and HubSpot. I finally said, that's it. And I used Clawed to take the JIRA information, figure out where to update it in PowerPoint, and fix any formatting issues that pop up. It's so great, and it saves me so much time and mental energy. So speaking of Clawed, Clawed is the AI for minds that don't stop at good enough. It's the collaborator that actually understands your entire workflow and thinks with you, whether you're a debugging code at midnight or strategizing your next business move, Clawed extends your thinking to tackle the problems that matter. That PowerPoint example I gave earlier is such a good example because the old workflow wasn't hard. I could do it. But now, Clawed does it for me. And I can start looking at the inputs and outputs and focus on making them better and more effective. Clawed co-works, professional outputs, are impressive. Creates polished documents, spreadsheets with working formulas, and presentations directly in your file system. And now, Clawed lets you set up workflows that run on a schedule. So daily reports, weekly summaries, recurring data polls, no need to do coding and set up weird crime jobs or anything like that. Core handles it. For problems we're solving, get started with Clawed at clawed.ai/tps. That's clawed.ai/tps. And check out clawedpro, which includes access to all the features mentioned in today's episode. So another thing I want to ask about, and you've even kind of mentioned that I think is like Twitter/X is such that's like the hotbed right now of just information for better or for worse about AI on the good side and the bad side. But one thing I have been seeing a lot lately is this concept of having a digital chief of staff. And people are saying, hey, depending on how you value your time blah blah blah. They're saying even with all the expense of using clawed or the high clawed plans or the high open AI plans or whatever, it's still totally worth it compared to hiring somebody to be your quote unquote, chief of staff. Do you know anyone who has done that? Do you do that? What do you see as the role of a digital chief of staff? And if you have made it one, what are you having your chief of staff do? Yeah. So I've built my own digital chief of staff, as of let's just call it, Spring of last year. So Spring 2025. And the reason I got so gong ho about this was, so I did a dinner. I organized a dinner where I invited three or four people that I thought were the smartest people around AI. And the question I asked at that dinner was, hey, do you guys still have an executive assistance? And everyone said no. Everyone got rid of their executive assistant and went fully digital. And these were the people that are on the most cutting edge of AI. And I was like, oh my gosh. This is like a glimpse into the future. And this is already around the time when I was playing heavily with Lindy. And I started to notice that I was like, offloading so much work from my real human assistance that I started to realize, hey, I think I'm on this path of completely replacing a lot of work that this person does. And so I wanted to ask a dinner if this was the case for them, too, or if I'm just being crazy right now. And so it turns out, yes, they were. And it turned into this little evolution of, let's give it a D atman work that this person would do to then have a more time for other things and more high level strategy. And then it kind of evolved into what I now called digital chief of staff, which is actually a-- this is called a suite of agents that I now deploy for founders and executives that, funny enough, in Austin, where I live, where I'm mostly known for is that I deploy these.
digital chief of staffs for people in like two weeks. And so if people were to hire me, they would say, "Hey, Tan, I heard you have this digital chief of staff. Can you show it to me?" So I showed them a demo. I sent them this like Google Doc in like a short video of what it can and cannot do. And then it basically does three things. One is it manages your email inbox. So for example, when you log into your email inbox, every email is pre-drafted in your voice. So if someone says, "Hey, Tan, are you available next week on Wednesday for lunch?" It would say, "It would actually check my calendar and see if I am available." If so, it would draft an email, say, "Hey, I'm available 12. We'd love to see you." And let's do this. So that's one. It manages your email inbox. The second thing is it does a meeting brief and a meeting prep. So I want to kind of told about earlier of like 8am, you get a meeting brief for the day, and then 30 minutes per for every meeting, you get a prep note. And then the third thing is usually some form of note taking. So it's like you have a AI noteaker, something like Otter or some people might use granola, or they might use their own version. And then processing that transcript into admin task that is now done for you. So oftentimes it's, for example, updating your CRM. So he had a call with a prospect. You would update the CRM automatically using this agent to say what is their current status, how much are they committed for for whatever reason, or maybe they changed locations, and now that's being updated as well. Like it does that automatically for you. The other thing you would do is, for example, is update your to-do lists based on your meeting, update your project management tool. If it's an internal meeting, it might send a follow-up email to everybody on the team or within the company automatically. So it takes from note taking all the way down to whatever actions people would do post meeting, and just having it done for you. And so I also don't know how people live without one. No, this. So for me, it's like, I've said it at once. I haven't running for over a year now, and it's just a part of how I operate now. So the people you set these digital chief of staff tools up, you know, it hits the three main heavy hitters, right? The email, the meeting notes, the recordings, all the deck counters up. Do you find based on, if you hear back from them, aside from that, like, you know, being, they wouldn't want to live without it, do you find that they then go on to find other things for it to do, or are people mostly like, hey, these are my biggest pain points. I'm good with this. Like, what sort of things are you seeing with the chief of staff users? I would say for them, this is usually the gateway drug for other things. So once they see how this digital chief of staff frees up so much of their time, they then get more curious about AI, and they get more curious about, hey, what else can we do now? And the thing I've learned from teaching thousands of people around AI now is that one, you don't know what you don't know. And two, the only way you can see the power and discover the power of AI is when you see use cases that other people are doing. So nowadays, when I teach AI, a lot of it's not really technical or mechanical or tactical at all. It's really showcasing, here's 15 ways I've deployed AI for other people. And oftentimes when people see that, they go, oh, if you did this for this person or for this company, I can see how that could apply to my situation. And when they see that idea and that vision, then they go, hey, I want that. How do we make this happen? And then you start talking about mechanical things. Like, oh, this is how our problems might work. Or we got to connect this AI to this tool. And some people want to get really tactical and get into the weeds. And some people are just like, no, I understand the high level that AI is really good and powerful. And I just want you to do it for me. So everyone has a different range. But typically, once it's set up, they just want more of it because they start to realize, wow, I'm getting back 10, 20 hours a week. And this thing is non-stop. Like, what else can I do? So when these things are set up, this is both for you and maybe for people who you know that use it. Like anything else, you set it up. It's going well. But then eventually, once it's like real life, then you start noticing little things. Like maybe, maybe, like, hey, my agent is booking all these meetings. But I'd really rather not have things between 12 and 3 or, you know, like, little changes you want to make. How do you find, or what's the best approach to making these type of changes if there is things that you want to change with your agents? Do you have to go in and edit the markdown files? You just have to tell the agent, like, hey, don't book me any more meetings at this time. Like, what are some of the tweaks that you see people having to make once things are like in real life? OK, if you asked me six months ago, it would have been, go tweak a file or a setting. Nowadays, it's tell your agent. So if you're like, hey, I don't want any more meetings on Mondays between 12 and 3, you just tell your agent and go, OK, got it. Remember, I won't do that again. And you're done. So that's the cool thing about AI is it's growing, it's evolving, it's moving fast as well, which is also very overwhelming for many people, including myself. I'm overwhelmed by it every day. I kid you not. I'm overwhelmed by whatever is going on with AI. There's this funny meme going on of, you know, anthropic right now, we're pretty much shipping and update every single day. And you're just like, today I'm overwhelmed. Tomorrow, I will be overwhelmed. Yesterday, I was overwhelmed because anthropic is just shipping like crazy. And so if you ever had a bad experience with AI, I'm going to guess it's probably six months to over a year ago. And if you did the exact same thing that you did then now, I can almost guarantee you that you'll have a much better experience today. And so I always want to remind people and encourage them to go, hey, if you had a bad experience, try it out again, maybe on the same platform or a different platform. And I think you'll be pleasantly surprised that it can do now what you actually wanted it to do back then. And I always would like to remind myself, too, this is the worst it will ever be. It will only get better. If I think about where AI was six months ago, a year ago, year and a half ago, two years ago, it's totally different. So much better. There's so much smarter now. And so I encourage people to really lean in and start using it. So speaking of getting better and final question for me, maybe a lot of things you want to talk about, which is fine. But the final question I've been wondering about is-- and maybe email is not a great example for this, because me, personally, I would still want to check every email going out under my name. That's just the way I am, but for the most part anyway. But how do you get comfortable once you start setting up these agents and you go beyond what I do, which is telling it to do things once. And then next level is agents are doing things for you. But eventually, you're getting to the point where agents are making decisions for you, where it's not even asking you. It's just like doing this or that. How do you get comfortable or how do you advise other people to start getting comfortable with these agents making decisions on their own? Because at least, if you-- we've talked about delegation on the podcast for around-- probably five years now, if not more. And that's a skill of being comfortable with letting your employees or contractors make decisions on your behalf. But at least it's a human. How do you advise people to get comfortable with with the AI agent making decisions on your behalf, which may or may not be the same decision you would make? Yeah, I've spent a lot of time thinking about this because even though somebody who is living and breathing this, there's still a lot of things that I would never let an AI touch. And then there's still a lot of things that even though I could let go of control, I'm still not willing to. So for example, I've had it draft so many follow-up emails, like hundreds at this point, that at this point, I might maybe change a word once or twice every 10 emails or so. So let's just say 80% of the time, it's good to go to send emails as this. And I'm still tweaking every week to make it closer to 100%. And even if it were 90% or 95% or 99%, I personally would still want to see every single email before they go out. And so there's a lot of times situations where I want to know when my name and reputation is on the line and it's going out to other people that I have this review process in place. Even though I know 99% of the time it's pretty good and good to go, I still want that as a form of accountability, but also as a quality control. So my personal rule is if anything is going out towards other people, I have to have final say and final review. And I will oftentimes push that button as well. Then there's some things that are just co-entered, semi-automated or fully automated. And also maybe our customer facing, but are maybe a little bit more, for example, transactional. So for example, when someone buys something on Asianeafishans.com, I don't personally send that email out saying, hey, thanks for making this purchase. Hope you enjoy it. Here's your log in details and so on, right? I could have done that, but I've decided, you know what? I'm willing to give up control there that people are going to get it.
and semi-automated email, which I think most people nowadays are kind of conditioned and expecting so to do so, right? And maybe that will change in the future where AI is gonna do that for me as well with personal follow-up emails, who knows, right? So anything customer facing, I usually like to have final say, if it's not customer facing, I've kind of slowly, but surely tried to hold back as much as I can, and I'm constantly trying to hold back, and I'm actually now trying to get to a place where I am now to bottleneck for a lot of things where I don't even wanna be to bottleneck anymore. And the perfect example of this is two weeks ago, I was in Miami watching a professional paddle tournaments, I was there for three days. And in my hotel room, I remember, I have this like AI agent, his name is Rames. And there's an open source platform called Rames agents, which is kind of compared to OpenClaw, if you wanna get really nitty gritty, but I was trying to give it a spin and see how good it was. And long story short, what I decided to do was I had this project called Mission Control, where I'm making a personal dashboard for myself and my life, and I had a fully spec written out that I've spent probably like a few days on, and was very detailed. And Rames agents was gonna orchestrate with Cloud Code and with Codex. So if you don't know, Cloud Code is kind of like, the coding agents from Anthropic, and then Codex is the coding agent from OpenAI. And I basically told Rames, hey, I want you to look at the spec and build it out, and I want you to use Cloud Code and Codex to build it, but the way you're gonna do it is, you are going to have them plan every single feature, and then on my home computer, Max Studio, there's a local LLM data running called Quinn, Quinn 3.5, which is, I believe, a Chinese model that's open source. And so Cloud Code and Codex were planned, then they would tell Quinn, my Max Studio essentially, to code it, and it's done, it submits a review, like a PR essentially, and GitHub, and then Cloud Code and Codex had to review it together, and they had to come to a consensus before they built the next feature. And so I'm like, am I yammy, I'm kicking this off, I remember this was like a Saturday morning, I'm kicking this off on Slack, I'm going, hey, here's the spec, here's the pipeline of how we're gonna build it, and go do it. And this thing was coding five days in a row 24/7, and I'm like monitoring on Slack, like I get an hourly updates of what it's doing, and I'm like amazed that this is going on for five days in a row, building out feature by feature, and because it has like a loop where it's reviewing stuff, it will find bugs, it will find features that are missing, it will find things that supposed to be working, but aren't working, 'cause in the spec, I have like certain tests, it has to pass, and so it's running all these tests all the time. And so after about five days of coding 24/7, it was like, okay, it's finally done, here it is. And I run the test, or I run the production live, zero floss, and I was like, oh my God, like this thing worked for five days in a row, it worked as soon as I logged in, it did everything I wanted, it, my mind was blown, Brooks, my mind was absolutely blown, and I was just like, this was insane. - So just think of how many decisions that whole thing saved you by having the models do it, and go on, and I think that's a great example of where things are headed, and there's people way out there, like I always say you're quite a bit ahead, but I think this is where this stuff is going for sure. - Yeah, and the crazy part that I think people should realize is not, yes, it is very impressive, it was running five days on the row, which is very unheard of, 'cause a lot of benchmarks won't get there, it's because we had a third orchestrator, but the fact that I had four agents, kind of like four AIs working together, you had Ramesh, Codex, Cloud Code, and then Kwen all working together, and Ramesh was like the orchestrator. One, that was really impressive, I like didn't believe that that works. And then the other thing is, because these two coding agents were debating each other, like I could see it in the logs of like, "Hey, I think we should do this." The other guy was like, "No, we should do this." And they're going back and forth, and then eventually they come to consensus. I come, "Oh my gosh, like, there's two-year-old, "there's so many decisions and thoughts that went into that, "that I could never come up with on my own." - Yeah, and you were saying a lot of scenarios, you are kind of the bottleneck. So, it took five days of working away, five days that you didn't have to do by the way. It took five days of working away, but imagine if it had to come to you for decisions, would it take a lot longer than five days? So even five days is a massive bonus. - Yeah, and that's when I realized where I was a bottleneck, because in the past, my working pattern was, it would build something, submit a for review, I would review it, and the state of where things are, these things are coding and building stuff so fast now, that I have a hard time keeping up. So I have this huge backlog of things I have to review, and I realized, it's kind of like, you know, our to-do list or our email, sometimes we're like, "Hey, we're always gonna have way more stuff than we can actually do. "We just have to be okay with the fact "that we're not gonna do everything, "and we're just gonna have to start prioritizing our to-do list "and be okay with the fact that there's 100 things on her "that I'll just never get to." This is what I'm feeling with AI right now too. It's like, it will do so much stuff for you, production-wise, and deliver, and deliverables, that you just have to be okay with the fact that you're not gonna look at everything, and you just have to trust that it will actually do the right thing. And the only way you can trust it, honestly, is you use it enough and you use it every day that you start to see that it starts to earn your trust because it's doing something consistently pretty well so that you can kind of start letting the list. Just like what any new employee or team member, they earn your trust so that over time, you can work on something together and you trust that they're gonna make the right decisions, do the right thing, and it's no different with these AI tools. All right, well, I'm sure most of the people listening to this are not gonna be hitting stop on their podcast player and then kicking off for AI agents to argue amongst themselves while watching a pedal tournament. So what's maybe something a little more accessible that people can do if they wanna get started on taking a level just beyond asking, clawed or chat for advice, like maybe what's something that can kind of like take them to the next level, do you think? Yeah, so what I recommend that you do, is always, you know, this is part of our philosophy here, of making things simple and actionable. I want you to start thinking about what is something that you do with AI quite often and turn that into it doing it for you. So for example, if you're always doing research on chat, you be to your clawed, instead of having to do research for you, how can it actually do something for you? And honestly, the current state of affairs is that you kind of have to pay for this, you have to have a paid account, you cannot do this on free accounts. So if you're willing to learn AI, it's worth the 20 bucks a month you're gonna spend. And I recommend you actually play around with clawed co-work. Co-work is probably the most easy and accessible tool right now. So if you download the clawed desktop app and then pay the 20 bucks a month plan, it will release this feature called co-work. And instead of you prompting with AI normally, like you would normally would, in co-work, you would do the same, but then you'd tell it, do it for me. And I bet you'll be so surprised what it will do for you. You go, oh my gosh, I'm never gonna go back now. I'm not just gonna do research anymore. I'm actually gonna tell it to do stuff for me now. And that might mean that you have to connect something like your notion or some database or an email account or Google Docs or whatever it might be. But even then, without it, it will conduct stuff for you as well like save a file, save it in your folder, do stuff. And if you wanna quick one, tell it to organize your downloads folder. You'll be surprised what it will do. And you'll be like clawed-pilled as people would say when you see that happen. So go give it a go. I'm hoping that you'll see the light and that we'll continue to do more episodes here on the productivity show. Thank you so much for listening. And we'll see you again next week.
PodcastTranscript AI