How to turn Claude Code into your personal life operating system | Hilary Gridley

The opportunity cost of my time has never been higher. Also, as a new mom, it's never been harder. It has never felt harder to get out of my time what I want to because my attention is super fractured.

How do you decide what to automate?

For any possible task, if I were 10 times better at it, would it have 10 times the impact? If the answer to that is no, then I just automate it. And if the answer to that is yes, Those are the things that I want to put more time and effort into.

And you don't have to start with a big complex Python script or anything like that. You just have to start with a problem statement.

You learn by doing. And so every day my Claude gets a little bit better at helping me manage my time, helping me do work because it is observing what is really happening. So we adjust as we go and it takes the cost of maintaining the system and the cost of setting up the system to zero because Claude Claude is just doing everything for me.

Welcome back to How I AI. I'm Clara Vo, product leader and AI obsessive, here on a mission to help you build better with these new tools. Today we have our first repeat guest, Hillary Gridley, who was on one of our most popular early episodes teaching us how to be a better manager with AI. Now she's an entrepreneur and a new mom, and she's back to show us her personal anti-system system For using AI to manage her day, her to-do list, and get everything done through that little alien in our computer, Claude Code. Let's get to it. This episode is brought to you by WorkOS. AI has already changed how we work. Tools are helping teams write better code, analyze customer data, and even handle support tickets automatically. But there's a catch. These tools only work well when they have deep access to company systems. Your Copilot needs to see your entire codebase. Your chatbot needs to search across internal docs. And for enterprise buyers, that raises serious security concerns. That's why these apps face intense IT scrutiny from day one. To pass, they need secure authentication, access controls, audit logs, the whole suite of enterprise features. Building all that from scratch, it's a massive lift. That's where WorkOS comes in. WorkOS gives you drop-in APIs for enterprise features so your app can become enterprise ready and scale upmarket faster. Think of it like Stripe for enterprise features. OpenAI, Perplexity, and Cursor are already using WorkOS to move faster and meet enterprise demands. Join them and hundreds of other industry leaders at workos.com. Start building today. Hilary, welcome back to How I AI. It's been almost a year, and I'm going to flatter you a little bit. You were one of our most popular early episodes about how to be a better manager with AI. So I am super psyched to have you back on the show and show your new set of workflows and AI tools that help you in your kind of changed life now. So bring us up to date. What's been happening in the last year?

A lot has changed. A lot has changed in my life and in AI. Talking about custom GPTs almost feels so quaint now, like what we talked about a year ago. But I also have a little baby now. I was pregnant when we were first recording that podcast. And so it's been really fun exploring these tools with a whole variety of new demands on my time and my life and just with the new capabilities that are available and how far these tools have come since then. So lots have changed and I'm really excited to show you what I've been up to.

And in addition to having a new member of the family, you're also focusing full-time on your own business. Isn't that right? So you've gone from—

That's right.

being a product leader in an organization, thinking about how to manage employees and those constraints on your time to being an entrepreneur. And again, the constraints on your time don't go away and you have lots of pressure and demands from both clients and the work that you need to do and, and your family. So you've built an entirely new system and we are upgrading from custom GPTs to something a little bit more advanced. Isn't that right?

That's right. We are upgrading.

And what I love about what we were talking about before we started recording is, yes, what we are going to show is going to be a cool, pretty complex way to automate your personal and professional life. And also you have a philosophy on how complicated this can actually be to match your life. So tell us a little bit about how you think about the system of setting up your personal AI.

The biggest thing that has changed for me since becoming a mom Is I feel like the opportunity cost of my time has never been higher. Like, I just have no tolerance for wasting time in things that are not like either enriching my life or helping me towards some goal or spending time with my son or whatever it is. But the problem is like also as a new mom, I feel like it's never been harder. It has never felt harder to get out of my time what I want to. Because my attention is super fractured. I'm tracking about 1,000 more life admin things and logistics than I ever have been before. And I'm historically not great with logistics and life admin. You know, I'm not sleeping as much, like my brain's a little bit foggier. So it's been really fun for me to explore how AI can help me with that and help me sort of keep all the pieces together moving forward. So that I can get more out of my time, not in like a, you know, hyper-productivity optimizer kind of way, but like, you know, like reclaim time so that I can go to the market and get a baguette and have some nice cheese and like enjoy life and not spend all my time juggling all of these logistical plates.

So what you're telling me is you're just prepping for the post-AGI world where we're all just, all of our stuff is taken care of. We're living in an abundance world, abundance utopia, and we just have to figure out how to fill our time with robot-baked baguettes and delightful leisure activities.

I mean, I like to work. That's the thing. Like, I want to free up time to do work that makes me excited and that I love doing. But then I also want to go have a picnic. So it's really, you know, within me there are 2 wolves, and neither of those wolves are life admin. And so Any help that I can get on that front is amazing. But the problem that I've seen with a lot of systems when I watch other people use AI and use Claude Code in this way is it involves a lot of setup and it involves a lot of organization. And they're like, look at this amazing system I built where it pulls my top priorities and it pulls all this kind of stuff. And I'm like, you've already lost me because the whole point of this is that I don't want to put work into maintaining a system. I don't want to do a bunch of setup. I don't want to like, like I just want to get started and have my problem solved. And that's what I'm going to try to show you today is how I approach that in a way that, you know, we talked about as sort of like the anti-system system. That's my philosophy in all of this.

Well, and I think from a personal productivity perspective, there is this spectrum that I talk about, which is I think some of us are like Notion boys where we want like tables and pages and linked assets and Kanban boards and all this stuff. And then some of us, and that's me, are just like a plain text file on my desktop that says things I need to do. And depending on where you fit on that spectrum, you're going to want to set up your AI in a different way. And I, like you, I'm a little bit more, um, stream of consciousness approach to how I want to think about structuring my day, which is not at all. So I'm excited for you to show us a couple of the tips and tricks you use and things you've built, particularly with Claude Code, to get you there. So where are we going to dive in?

All right, first I want to show you something to set the stage, which sort of gets into what we were talking about. This is a picture I took, like, I don't know, you can tell from the quality, it was probably 15 years ago. Do you know whose desktop this is?

No.

You're never going to guess in a million years. So I'll just tell you. This is Al Gore's desktop.

That's not what I was expecting.

I just thought, not in a million years. I went to see Al Gore talk, and he flashed this up while he was loading up his PowerPoint. And I was like, this is the most amazing— what an amazing peek inside this person's brain. This person who's obviously accomplished so much, and like, this is the state of their computer. And I don't, like, that's not a criticism. Like, I felt seen and I felt like I finally have somebody who it like doesn't seem like they have it all together behind the scenes and is yet able to make it seem like they have so much together. So this is like, this one goes out to Al Gore and his extremely chaotic desktop because I think that is sort of like the state of, How I felt about all the stuff I was trying to keep track of when I started setting up this system.

I revise my statement. We're all on a spectrum of Notion boy to Al Gore desktop. That is my new definition for how organized we are relative to personal productivity.

Exactly. So the first piece of this that I wanna show you is just how I plan my day. And when I think about planning my day, like again, it's not the like big, uh, 3 priorities that I absolutely need to make progress on that I need help remembering and scheduling. I know what my big priorities are and I do them. The problem is all the stuff around them that I'm always forgetting. And so the first step of that is just how do I make sure that nothing escapes the leaky sieve of my brain for things that I have to do? Or ideas I have. So for things that I have to do, I just have a shortcut right on my phone on the lock screen where I just double tap the back of the phone and then it pulls up a dictation box and I just say, reschedule pediatrician appointment. That's one of those things that I'm falling asleep and I'm like, oh my God, I have to reschedule the pediatrician appointment. So I tap the back of my phone, I say it out loud, and then it gets added into my inbox. This is no AI. We have not even reached AI yet. This is just a shortcut on your phone. You can set it up very quickly if you just go to Shortcuts, add a dictate text, uh, and, and then you also set up the accessibility back tap trigger. That is accessibility within your settings. You go down to touch, go to back tap, uh, click on double tap, and then find your shortcut here. So super simple to do.

If Apple's watching, I need a chatbot to navigate iPhone settings for me.

I totally agree. So I can actually, I can share this whole Figma file if that would be helpful for your listeners. I'll put it at writerbuilder.com/howai, and you can also share it in your, in your show notes. This is like so easy to set up, but it's just clicking around that like is very hard to do. So you can just follow the steps there if that's helpful. But now this is the fun part, which is #HowIAI. Do people still, I guess people don't still use hashtags.

I mean, it depends on, yeah, I'll throw a hashtag out there every now and then.

#HowIAI.

But I spell it out. I literally write hashtag.

Perfect. So, okay, I'm going to open Claude Code, which is where I run my entire life. Now, if you are new to Claude Code, You are like, I don't understand. You are planning your life out of the terminal of your computer. That's confusing. Again, I'm just gonna very quickly show you just to prove how easy it is. We're not gonna dwell here, but you have your background. You just click this little magnifying glass, start typing terminal. Terminal pops up, you type in Claude, you hit yes. Don't ask questions.

Dangerously skip permissions.

I, I don't dangerously skip permissions because I'm actually a bit of a scaredy cat, believe it or not. If you have never installed Claude Code, it is also super easy to do. Just go to the Claude document. It'll give you a little line that you copy here. You copy that, you literally just paste it in, hit enter, the robot beep boop beep boops, and you're good to go.

Yep.

So that one goes out to everyone who hates setup and has thought that maybe Claude Code has annoying setup and that has kept you from using it. Not the case. All right, so now we are going to get into plan my day. So all that I do is I type plan my day and I hit enter, and the robot starts beep booping again. Now what it's doing here is following a set of instructions that I have given it, and I'll show you in a minute how I actually give it those instructions and how I set that up. But for now, I just wanna show you like what it looks like, um, and how it actually helps me plan my day. So it's pulling from a few things here. You can see it's pulling from my reminders. And basically what's going on here is, you remember I had that list of reminders that I was capturing from my home screen or my lock screen. Claude is in the background taking those reminders and putting them into a doc where it's organizing them based on category, basically. And this is just a Markdown file. So it's just a text file that Claude edits for me. I don't touch it. And it lives in a folder. This is a folder where I just keep everything that Claude ever needs. So this will come up a few times in my demos, but for the purposes of this part, it is just a folder with a text file in it.

And I have to ask for folks that wanna go a little bit deeper, I see Obsidian. Are you using Obsidian? How are you opening this? Are you opening this in Cursor? Like what, what is the system behind the system or literally just files?

I, yes, I use Obsidian to edit Markdown files. I do that more if I'm writing.

Mm-hmm.

Um, I'll have Obsidian open in like the left 2/3 of my computer and the terminal open on the right side of my computer. And I'm actively writing in the, uh, in the text in Obsidian. And then I'm also going back and forth with Claude to talk about things that I want to edit. But I don't actually ever open this, so I'm just opening it here to show it to you.

Yep.

Like Claude is going and looking at it and it, like, it's none of my business how Claude gets its work done.

I wanna pause there because I have this opinion as well, which I think is really interesting working with agents. Um, and, and I'm thinking about OpenClaw, which is really funny.

Yeah.

Which is, I think of my OpenClaw agents as employees, and I would never, you know, unless the law required, I would never go into their email and be like, how are they doing their work? Or into their to-do, to-do list and being like, how are they tracking their tasks? That's just not something I would do as an employer. And so it's really interesting how that parallel kind of moves over to agents, which is like, yes, it is actually none of my business. how Polly gets it done as long as she gets it done. Now, every now and then we gotta spot check that she's following the rules or something goes sideways. It's nice to be able to inspect. But yeah, I agree with you. This is, this is their system, not mine.

And it's, I mean, it's so task dependent. Like for most tasks, a 2% error rate is totally fine. And occasionally there are tasks where a 2% error rate is not totally fine. And then it's like, yeah, maybe you trust but verify a little bit more. But for the most part, for the vast majority of what I'm doing, the cognitive effort that it takes off of my plate to not have to even worry or think about this is 100 times more valuable than like, you know, a couple times a month one thing falls through the cracks. Like, I can live with that. It is certainly better than whatever was going on before any of this. When not only was I constantly thinking about all of this, but also like 50% of things fell through the cracks.

Yeah.

So like, deal that. It's just such an improvement. All right. The other thing that it is pulling from here is a list of preferences. So this is really great for me as a mom because there's a lot of constraints in my day that make it hard to schedule things well, right? Like I need to be pumping regularly. I need to be feeding my son. On the weekend when I don't have childcare, either my husband or I need to be with my child at all times. And so if the AI is going to make a helpful schedule for me, it is very important that it knows these preferences. Now, again, I don't like setting things up, so I never went in and was like, let's think about what my preferences are. Like, hmm, what are the hours of the day when I'm most effective? Like, I don't, I don't wanna do that. So this has just been the AI observing me over time and changing based on things that it is observing about what I'm getting done, what I'm not getting done, what's actually happening, which is great because it's actually my real behaviour as opposed to me being like, I want to make sure that I get a walk-in for at least 60 minutes every single day. And if you tell that to the AI, it starts scheduling a walk-in and it's like, If that never happens, you don't want that in your preferences. Even on your head, that's your preference. You know what I mean? So it has all this stuff here that it has just been observing. This one makes me laugh because I was complaining about how if I do certain types of work before bed, I stay up too late because I get in flow.

Yeah.

And so, noted.

Thanks.

Thanks, Claude. So that's what it's pulling from. And then it goes ahead and it makes, or it pulls what's on my calendar. So it's like, all right, here's what your Wednesday already looks like. Here's what's already on your calendar. Let's pick 1 or 2 things. What do you wanna get done? And I'll talk, like I use dictation for this. I use dictation for basically everything. So me talking to my computer, I say, I definitely need to make progress on the baby passport. And otherwise, I wanna use any block I have to prep for going on a podcast. Spoiler, it's this podcast. I needed, I needed to take a lot of screenshots.

You know what's really, I just have to, I have to pause and Hillary make you laugh, which is you and I are truly the exact same person. Same doctor's appointment, same schedule, same husband situation. We, one of us gotta be watching the baby. Literally after this podcast, I am going to get the baby passport done. It takes so much effort. So this is, I am your audience of one for all of this. Amazing.

I tweet for one person, Claire, and it's—

That is Claire Vo. Amazing. Okay, so what I like about this and what you've shown is, okay, so it's pulling from your preferences on schedule. It's got a kind of a skill or a script to pull in your reminders so it knows kind of what's a top priority. And then it's scaffolding out a day that should work for you to accomplish the things that you want to accomplish. And then it looks like it's making a recommendation on where to start. Yes.

And so I appreciate this because I don't want it to just take what I say literally, because I say things all the time that I don't actually get done, where every single day I'm like, today is the day I'm going to get the passport for my baby. And then 3 days go by, or 3 weeks go by, and it hasn't happened yet. And a big reason in general that I have found I tend to procrastinate things is because I try to take off, like, too big of a bite at one time. And so I have this big task which involves making an appointment, doing paperwork, going to the post office. Like, it's, it's too much for me to fit in the margins of my day. And so what the AI has figured out about me nicely is that if it can just break that bigger thing down into the very first step, it can slot that into a 15-minute piece that I have free in my schedule. And then I actually start to make progress on these things that are otherwise just weighing on me. So it says it here, you are not doing the passport, you are just making the post office appointment. That's it. 10 minutes off your plate. Tomorrow you can gather the documents and fill out the forms, which is just like, it's such a small thing, but it's been so transformative for me in terms of how I actually make life admin feel like less of a just overwhelming burden.

You know, I've had an executive assistant a couple of times, and so I've seen and know a couple of these tips and tricks to getting administrative work done, one of which is just drop a task on your calendar to get done and block off time. And I think, you know, management by calendar is a really interesting use case for AI because like it or not, setting up calendar invites is pretty tedious. Like you have to write the title and the name and you pick a time. And when AI can just do that for you, it's, it's just a totally different form factor of to-do list that I personally really like and have seen a lot from a couple people who Love to lean on AI for life admin.

Yeah, I agree. I think 2 things are true. One, you cannot say you take something seriously if you are not putting your time into it. There are a lot of things that I say like, oh, I care about this, I want to do this. And if you look at how I actually use my time, it does not reflect that, which is frustrating to me. And I agree with you that the solution to that is you have to block that time out, which is calendar management. But calendar management fails my test of like, is this tedious life admin work? Like, oh, you gotta go. Like, people who are like, I used to go in and add all these things to my calendar and take things off my to-do list and put them on my calendar. Like, absolutely not. I was never doing that. But now that the AI can help me with that, I'm like, this is so obviously the right way to, to run your life and try to like actually get better use out of your time. Both because it, you know, it helps you commit to what you're actually gonna use that time for. But then as you'll see, it becomes much easier to observe the delta between the things that I say I wanna get done and what I actually get done to figure out like what, what is going on there and how can I fix anything that is not reflecting what I actually want it to be.

This episode is brought to you by Lovable. If you've ever had an idea for an app but didn't know where to start, Lovable is for you. Lovable lets you build working apps and websites sites by simply chatting with AI. Then you can customize it, add automations, and deploy it to a live domain. It's perfect for marketers spinning up tools, product managers prototyping new ideas, or founders launching their next business. Unlike no-code tools, Lovable isn't about static pages. It builds full apps with real functionality, and it's fast. What used to take weeks, months, or even years, you can now do over the weekend. So if you've been sitting on an idea, now's the time to bring it to life. Get started for free at lovable.dev. That's lovable.dev. So speaking of that delta, how do you loop back behind a day and say, did I actually book that post office appointment? How did, how did this all go? How do you do this like quality control loop on your Plan My Day agent?

Yes. So another thing you will find about me when you're seeing these demos is I always resist the urge to have Claude connect all a bunch of stuff in the background and get this information itself, mostly because I don't know what my tools are going to do when they're hanging out without me and I'm not in the room. Freaks me out. But I do believe in my heart of hearts in the Yappers API. Which is when I'm looking at things on one screen and I am talking about what I'm seeing to another screen, and that helps me capture virtually everything I'm doing. So because, as I said, I, throughout my day, I just have Claude open in the terminal on the right third of my computer, and then whatever I'm doing is in the other side of it. Because the terminal is such a multipurpose tool, I can use it to do literally anything that involves my computer. It means that I'm then like talking to it all day about the things that I'm doing. And then Claude observes that and takes notes. And so I'll show you how it does that. Just quick thing on this, as you can see, it just drops everything on my calendar for me. The hippo emojis are things that Claude adds, which helps me kind of keep track. And then this is also helpful because back on my lock screen, I can always see what's coming up. So I always have that persistent reminder of how much time do I allocate for this? Do I need to move on to something else, which I am not good at? But the other thing that Claude does is it creates this daily note for me. And again, this is just a Markdown file. It lives in that same folder with all my other stuff. You can see here the folder structure. And so it has my schedule that it has put into my calendar. And the other thing this has is a log.

Mm-hmm.

And so I have told Claude, just observe what I'm doing. And then it has what I said I was going to do on the day, and then it has what I'm actually doing. And so I can check in periodically and say, what are you observing? What's working? What's not working? I have this end-of-day reflection that I sometimes do, but sometimes don't. When I remember, then that can be helpful. I'll often ask it questions like, what are you seeing in terms of the gap between what I'm trying to get done and what I'm actually doing? Or even just like, what are your observations? Because I'm curious and I'm nosy and I want to know what the AI thinks about me. So it pulls out these patterns that building is crowding out writing, or that you're doing a bad job timeboxing something, or that you say most days list 3 priorities, but only number 1 gets real time. That's fine if you accept it, but then the real question is, are you picking the right number one? So this is when I said earlier that I don't really believe in spending time orchestrating this whole system that you use for your Claude. Part of that is because you learn by doing. And so every day my Claude gets a little bit better at helping me manage my time, helping me do work. because it is observing what is really happening. If I had tried to orchestrate all of this upfront, I would have been wrong about most of it. I thought that saying pick 3 priorities was a great way to do this. I thought that having this daily reflection at the end of the day would be a great way to do this. Claude is pointing out that neither of those things are really working, so we adjust as we go. And it takes the cost of maintaining the system and the cost of setting up the system to zero. Because Claude is just doing everything for me.

I wanna just pause here 'cause I think you just showed us so many rich things and I wanna make sure that folks didn't miss a couple of the key points. And so one, it's just like reduce the friction, reduce the friction, reduce the friction, which is, you know, double tap the back of your phone to just capture to-do lists, make it really easy for Claude Code, which as you said, through the terminal can basically do anything your computer can do. To access your to-do list in whatever form, your calendar in whatever form. Don't go through the effort of defining your own preferences or building out these like complex prompts and instructions. Just say like, as we go, start tracking my preferences and it will start building out those files. Create a file space for it to operate in, which, you know, this just can be a folder markdown. Markdown files. And then one thing that I want to make sure people didn't miss, because I love this idea of the Appers API, which is, you know, you could have gone through all this work and be like, okay, I'm going to build this web app and this web app is going to like OAuth into my Google, my Google Drive and my calendar and all this stuff. And it's going to know when I'm working on documents and it's just going to be able to track everything I do per day. And you're like, no, bro, let's not do that. I will literally just as I write a blog post, say to Claude, I wrote a blog post. Or like, as I got the passport done, just say to Claude, I got the first part of the passport done. And this idea of we do not need, you know, we have so many debates on like MCPs and APIs and CLIs. It's like, literally just look at the screen and be a proxy for the software with your meatware. Like, use your mouth. And, um, this is where I— we just had an episode with Figma and they're like, you know what, nothing beats An opposable thumb. Like nothing beats dragging a design around. And I still think nothing beats just eyeballs and language to describe what's going on in your life. And so keeping things really loose can be a really powerful way to get leverage out of technology without having to use technology at all in that interim step.

Yeah, it's 2 things I believe. One is complexity has to earn its keep. And so if I like, I'll, I can show you some other workflows where I do have my Google connected. Uh, and as you can see, I did, like, I, I eventually connected my calendar here, but I only do that after I've tried the jankiest version of the workflow for like a week and been like, oh yes, I'm going to continue using this. This is very valuable. And what I have found is I have a hit rate of like maybe 20%, like a lot of things where I'm like, oh, I think it would be helpful to set up this type of workflow. I don't actually end up using it. And so if I spent all this time upfront getting all of these things connected, like A, it's a waste of time. B, if anything breaks, then it becomes a double waste of time cuz I have to troubleshoot a thing that wasn't even helpful in the first place. Um, and so I'm like, I'm always trying to find the absolute simplest version of the thing and then like, and then I enrich it and then I enrich it and then I enrich it.

Well, and in addition to the Yappers API, um, I, I wanna call it something else that We haven't seen in your flow, but we've seen in a couple others like Jesse Janae's OpenClaw flow, which is, and I'm gonna call it my mom sharing news API, which is screenshots. And so, yes. Love you, mom. But my mom likes to read articles and then like screenshot like 3 pieces of the article and send it to me. And at first I was like, can you just like copy and paste me this text? And now I'm like, oh, mom's a genius. Like that is actually the way to share context. Is just screenshot your calendar and say, hey, what would you do with, with this if you're not ready to hook up the API? Or, you know, scribble a note while you're at the doctor's office and put it in and say, hey, remind me about this next time we go to the pediatrician's office. I think just finding the lowest friction way for these AI assistants to get the information it needs to do a good job, whatever works for you, just use.

I think that's 100% right. And I tell this to people too when they say, I've come up with all these great ways to use AI in my personal life, but I can't use it at work because I'm restricted in terms of the permissions I have or the tools I can use. And I'm like, but no one's gonna give you access to all of the data at your company if you have like a half-baked idea for how it might be helpful. And so if you can make, again, like a janky version of this brilliant workflow that exists in your mind, that is held together with screenshots and prayers and you talking, prove that that is actually valuable and then bring that to whoever holds the keys to the castle at your company and you will get the permissions that you need. The problem is if you haven't done your own proof of concepting and you're just letting yourself be blocked even just to prove it, you're kind of you're kind of proving the problem, which is that you haven't really thought through the problem that you're trying to solve with this data.

Well, and I, you know, I think we could spend another hour and a half going through specific workflows. But what I want to really talk to you about next is how do you decide what to automate? And, you know, you opened our, our discussion with saying there's just very high opportunity cost on your time. And I feel that when you can like experience the infinite love of looking into your newborn's eyes, or schedule doctor's appointments. Like, you're clearly— you're the, the value of your time is very important. Then layer on top of that being an entrepreneur where there's like literally a dollar amount attached to your time that can only be created if you go to work. Um, it, it does feel existential to try to figure out the right things to automate and the right things to keep your human time on. So what are some of the frameworks you use to even just decide what to do?

Yeah, let me show you. I have like a little diagram. So basically the way I think about this is like for any possible task, if I were 10 times better at it, would it have 10 times the impact? And if the answer to that is no, then I just automate it. And if the answer to that is yes, Those are the things that I want to put more time and effort into. And 10 times the impact here, in a work context, if I'm managing a team and I'm trying to decide for people on my team, do I want you spending time on this or is this busy work? That framework works from, if you get 10 times better at moving pixels around a PowerPoint deck, you're not going to be 10 times better at your job. But if you get 10 times better at pulling important insights out of a body of user research that we can make better decisions about our product on, that I want you to focus on. And so that is something that I would not want to automate. But even within a life context, to me, the impact there is it goes beyond work, right? It's like, I mean, I spent a year building a digital therapeutic to treat depression, and something that I deeply internalized from that was like, you have to carve joy out, or you have to keep joy into your life. You have to actively protect joy and all of the things that make being human fun and all the things that make your life worth living. So I also think about that of like, is this something that is going to bring me, like if I were to really invest in baking this loaf of bread or something, is that something that's going to really enrich me and my life, or is that something that's going to feel like a chore? It can be very task-dependent. It can depend on what mood that I'm in, but that's the framework that I use. Even within a given task, you can further break things down. I said even within this example of I'm giving a talk, and this I can walk you through how I would use AI to develop a talk. For something like that, there are parts of it that are uniquely me and that are uniquely like, if I get 10 times better at this, my life will become 10 times better. And that is like how I come up with the ideas. That is how I craft a compelling narrative that people are gonna like respond to and find interesting. That is not building the deck. And so like even within a given thing, you can always break it down into discrete parts and make decisions within those about where you are most valuable and where the AI is most valuable.

I think this is just such a useful framework because I agree. There are still things that I feel like it benefits me and benefits the work to not put AI in the process. And writing is absolutely one of those. And so for you, as somebody who has chosen to double down on writing as a profession, you know, I tell everybody, look, my, my, my tweets are lovingly crafted with my human fingers because I don't think if I, if I get 10, 10 times better, at, uh, at, you know, posting, I think I will have 10 times the impact. If I get 10 times better at writing long-form content, it will have 10 times the impact. Same with this podcast. Like, if this gets, you know, the interactions, the guest prep, all that stuff, it gets 10 times better, the content gets better. But like, cutting, you know, cutting, um, transcripts into like 10-step workflows, I can only get so much better at that. That's not where We're gonna add the benefit. And so it's like you do have to do some of this automation. And so if I were gonna give folks a takeaway, it is take a couple tasks in your life where you're really either spending a lot of time or feel overwhelmed and break them down into this framework at 2, um, at, at the double level. I love like, is the task itself at all in this category? Yes or no? And then if it's yes, what tasks within that category can be partitioned off? And then And then automate them.

Yes. And one other thing I will add to that, especially if you are a manager of people and you are thinking about how do I develop the people on my team, keep in mind that this framework is very— it can change depending where you are on learning curves. And so I don't actually regret all of the time that I spent moving pixels around PowerPoints because it turned me into the kind of person who you could tell me, Hilary, you have to give a talk in 30 minutes, and I'd be like, no problem, I got it. But I didn't start there. And so for me, there's a lot of steps of the giving a talk that I've reached saturation on in terms of more reps is not helping me get better. But for somebody on my team, they're often at the bottom of that. And so that thing of like, oh, I feel like I wasted all this time because I was moving pixels around a PowerPoint is like, You, you can't evaluate that, you know, without the context of where you are on that learning curve.

I think that's so important for people to hear because again, we're not saying formatting slides is not a valuable thing for a person to spend time on. It is not a valuable thing for you at this point in your career and your skillset to continue to spend, to spend the time on. And again, I'm just gonna say, Hilary, you're the same person, cuz yesterday I gave a presentation and I mean truly just in time slide delivery. I was like 7 minutes before engaging going live in front of 1,000 people and my slides were very good.

Very good.

Shout out to Gamma for helping me make some amazing slides right out the gate. Great, great use of AI. Well, Hilary, I again, I think we could go through a million use cases. I know folks can, you know, go to your newsletter and follow you on social for use cases. But I want to, I want to zip over to some lightning round questions. And the first one is less of a question and more of a request in the form of a question, which is Can you show us your recording mode skill? Because this is such an important use case for folks like me and folks like you who need to share AI workflows live on podcasts. I know riches and niches, but, uh, you showed me this before the show and I was like, we have to, have to get this in the, in the closer. So can you pop up that skill and what you, what you built?

Yeah, I'd be happy to. So this is, okay, meta lesson here is the amazing thing about AI, especially as a product person, is it completely changes the altitude at which you can solve problems. And so I think about this a lot when I think about how AI is amazing because it can help you build more ambitious things. It's not because you can move 10 times as fast or just like make 10 times as many things within the same paradigm that you were previously operating in. It's because it allows you to operate in completely new paradigms, which allow you to have like a 10x scope with small effort. So I'll show you what I mean by that.

Okay, so I'm gonna start with, with the problem statement for you, which is you are a, uh, lady podcaster with a lady podcaster guest on, and you wanna show all your personal workflows on how AI makes your life with your kid and your husband better, and you don't wanna dox yourself.

Yes.

What do I do? Hilary, save me.

So the first thing I tried to do when I had this problem was, oh, I should just make a copy of my context directory that is all blinded and anonymized. And then I was like, that is such a pain to basically create an entire demo environment for myself and have to like manage these 2 things. Double complexity fails my test. Instead, what I did is I just made a skill called Recording On. And basically when I do that, Claude, anytime it's going to pull up any identifying information, it just changes it before it puts it on my screen. And so it's still pulling from all of my files. It's still pulling from, uh, like all the real stuff. The workflows are all the same ones I follow, but it's just going to like change people's names and, and things like that. And then when I'm done, I just say Reporting Off. And then it goes back. And so it does it. It also tracks like if it starts referring to, you know, this person as person A, it remembers that person A is like the same person. And then nothing, nothing gets changed. If it needs to fix anything that it messed up while recording a demo, it just fixes it for me.

And what I have to call out is this is such a cool way to demo Claude code, which I think is awesome. But now I'm thinking about folks that are maybe selling B2B software who have to demo their app. And my personal workspace, for example, on ChatPRD is the best demo app. It's so rich in information, but a lot of that we pull in like customer insights and financial data and all this stuff. And I'm like, oh my gosh, I need to have like a toggle for recording on mode where anytime I'm demoing live in my app, it does— it pulls from my production data sources but anonymizes them in a way that I'm not sharing customer, customer data. Brilliant. 10 out of 10 idea. Um, we're gonna clip this one for, for TikTok and put it on there. And I think this does just show there are like these micro problems that in hindsight, like what you say, double complexity to solve. I was trying to demo my OpenClaw and I literally had the OpenClaw make a web app of the conversation and redact the data because I couldn't scroll through my Telegram on YouTube and show all my kids' information. And so again, the ability to just solve these tiny, tiny, tiny problems is so fun and so impactful with AI.

Can I show you very quickly how I do it?

Yeah, please.

Okay. So I sometimes see these, like the skill libraries on the internet, and no offense to these skill libraries, but it feels to me like the feeling of somebody taking their toy chest and like dumping it on my head and all this stuff falls on me and I'm like, I don't know what this is and I don't want it in my space. And so a question I get a lot is people are like, oh, how the workflow I set up, for example, for plan my day, how did you do that? And it's so simple, it's so easy, but it does not involve any of these skills from the internet, which I don't even know what those are. I like to say with all of this stuff, I always reason from first principles because I have no idea what I'm doing. And that's how I feel when I talk to Claude. It's just me being like, I don't know how this works. Fix it. I don't know how this works. Fix it. You don't need any specialized knowledge to do any of this. And so just to give you a quick example, one problem I have is returns, right? I hate doing them. I forget about them. It's always at the bottom of my to-do list, and then I miss the deadline. So what do I do? I just tell Claude, hey Claude, I keep forgetting to return things on time. Also, I hate them. Just have to get that off my chest. I want you to come up with some solutions to make this easier for me. What do you got?

I like that word trauma dumping on Claude.

Yes, exactly. The trauma of my return mail. So Claude asked me some questions. I do like the pattern of have Claude ask you questions, but my patience runs out after about 3. So I'll tolerate a couple and then I just say, use your judgment, figure out from here. So it basically asks me a few questions and we kind of riff back and forth on it trying to understand the shape of the problem, which it's funny, people sometimes are like, oh, you're product managing Claude. And I'm like, no, Claude is the product manager. I'm the user.

I'm the product.

I'm telling Claude my problems and Claude is solving them for me. So we're just kind of going back and forth and I'm like, all right, I forget what its first idea was. Every time you buy something online, I can add a reminder to your reminders file with the return by date. Yeah, great.

Cool.

Let's try that. So I like your first idea, but I don't want to have to tell you what I ordered. Can't you just figure that out? So often I intang to Claude, can't you just do it? Can't you just figure this out? My favorite question to ask people too, people if I'm managing them, what would have to be true for this to be yes? And then Claude starts thinking and Claude's like, oh, I could do all of this for you. So what would have to be true is I would need email access. And then we're kind of talking about that and I'm like, okay, I'm into that. It offered to sort of integrate it into my daily workflow. But as I said, I like to kind of test run things a few times before I integrate them into something that's working, because then if it breaks, the thing that was working is now broken. So I say, let's just run this a few times. I don't wanna integrate it into my daily workflow yet. And it's like, okay, so you just need a script. It asks me a couple more questions, and eventually it just says, okay, I'm gonna look at how your email address or your email access is set up. Here's what I'm gonna build. Let me build it. Oh yeah, great question. I didn't even think about this. It was like, where do you have to go to do the returns? And I was like, I don't know, man. That's part of the problem is I never know where I have to go for the return. So it's like, okay, I'll take care of that for you too. And then basically what it does, like it builds a skill. So it made /command returns. It writes a skill file, which is just the markdown file. It writes the script, which is the code that it needs to run to do all of this stuff. And here's what that looks like. This is the skill file. So it's just a markdown file and it literally just has English instructions that a person could follow, but it's Claude following it with some code that it will run to do all of this. This is actually funny. It went in and pulled return policies for different retailers and drop-off info for Nike. Cool.

Okay.

And so that just literally me just describing this problem to it. And now all of a sudden it just will do it for me.

I am going to steal this. I love this idea of let's presume I'm going to have to return something and work backwards from that presumption, as opposed to me getting in the panic mode at day 35 being like, pretty please, can I return this Nordstrom Rack? I really don't fit in it. So I think that again, that idea of like you can preemptively solve problems you might have with AI in a very low-cost way. is super powerful, and you don't have to start with a big complex Python script or anything like that. You just have to start with a problem statement. So I'm gonna ask you that as our last lightning round question, which is, what would you say to folks? Because again, you started this saying like, you know, I don't really like a system, I'm coming to this with a beginner's mindset, it's not that scary. And then we've spent the entire episode in like Claude Code in the terminal in monospace font in dark mode. Like, you are a life hacker right now, my friend. And so for folks, for like, you know, again, you want to demystify this a little bit. How do you— what do you say to people about, you know, getting over that fear, that initial hump, getting started and just installing Claude Code for non-code reasons?

I think the best advice that I have heard and that I try to give to people is just like, Try to do one thing with it every day. And you just need to build the muscle memory so that you start to reach for it and you start to have your brain rewired so that you think, oh, the alien that lives in my computer could probably help me with this. And that seems so obvious, but it's like building any habit. You can intellectually understand it, but you still just need to do it every day. And eventually it will start to feel like second nature to you. I was one of those people who was like, I'm never gonna work in the terminal, like just a non-starter for me. And the only reason I did was because I was using Cursor and I kept hitting my limits on Cursor. And then I would begrudgingly go over to the terminal and work in Claude Code until my Cursor reset. But then eventually I got used to the terminal and eventually it was like a week. Um, and then I, once I got used to it, I realized the incredible power that it has. And then I just started playing around with it and then it became fun.

Amazing. Well, Hillary, this has been again just a superstar episode showing how you can track your to-do list, your calendar, build yourself skills, just do everything as you said with that little alien in, in the machine. Mr. Claude, Mr. or Ms. Claude's Claude Code. Where can we find you and how can we be helpful?

Amazing. Thank you for asking. You can find me on the internet. Best place to find me is Substack. I have a newsletter, hills.substack.com, where I write about all this stuff that we talked about here today. I have a course that I teach for managers on how to use AI and a few different events that I run for women especially who want to get involved in AI. But the best way to keep track of any of that is just to follow me on Substack, and I will post about everything there.

Amazing. Well, Hilary, I know your time is worth a lot, so I'm gonna get you out of here. Thanks for joining How I AI.

Thank you for having me.

Thanks so much for watching. If you enjoyed this show, please like and subscribe here on YouTube, or even better, leave us a comment with your thoughts. You can also find this podcast on Apple Podcasts, Spotify, or your favorite podcast app. Please consider leaving us a rating and review, which will help others find the show. You can see all our episodes and learn more about the show at howiaipod.com. See you next time.