How She Used Claude to Build a Million-Dollar Business: Prompts, Agents, Automations & Real Workflows | CatGPT

This episode is brought to you by Google Chrome. You think you know a browser, but Gemini and Chrome, that's new. It can help you with practically anything on the web, like restoring a vintage motorcycle from a 50-page restoration block, or finally break down that long article you've had open for weeks. Gemini and Chrome is here for it. Ready to make anything online make sense? There's no place like Chrome. Check responses, setup required, compatibility and availability varies 18+. When you need to build up your team to handle the growing chaos at work, use Indeed Sponsored Jobs. It gives your job post the boost it needs to be seen and helps reach people with the right skills, certifications and more. Spend less time searching and more time actually interviewing candidates who check all your boxes. Listeners of this show will get a $75 sponsored job credit at indeed.com slash podcast. That's indeed.com slash podcast. Terms and conditions apply. Need a hiring hero? This is a job for Indeed Sponsored Jobs. A.I. A.I., as it stands today, is like having 100 mediocre interns at your fingertips. I think right now trust is at an all-time low with A.I. tools. You made $100,000 with A.I.? Show me. Literally share your screen and show me proof is in the pudding because there's so many grifters out there on the internet. One thing we hadn't really publicized is that we were roommates together for a week in the south of France for Cannes. I think it was just like a tornado was in our Airbnb. I still think we need to pay $2 bajillion and stay in a hotel next year because that shit was dust. I feel guilty because I don't really understand what my A.I. usage's environmental impact is. And yet I continue to do it anyway. You could just use A.I. only when the sun goes down like me. Save energy. Should everyone be paying for an A.I. platform at this point? The crazy part about this moment in time that we find ourselves in right now is like we are all writing the guidebook together in real time. I'm Sheree. I'm Jean. And I'm Kat. And we're the Tiger Sisters. We are your Wall Street friends. We are your Wall Street and Silicon Valley big sisters. And we're a top 10 business podcast bringing late night sister talk meets boardroom strategy. Today we're joined by Kat Goetze. You probably know her as Kat GPT. She is a founder and creator and the Internet's sharpest voice on how to use A.I. to optimize your personal life and how to use A.I. to build a million dollar business. Kat used A.I. to turn her business, physical phones, into a million dollar business in less than a year. Today we are tapping into Kat's playbook on the best prompts to use as a coach to improve your life, how to improve your business, and also what the most common pitfalls are in using A.I. so that you can avoid them. We'll also talk about her best blind spot prompts using A.I. to actually challenge your assumptions instead of only validating them and a very, very special one time only special segment called Roots. We're going to talk about Roommate Revelations where Kat, Jean, and Sheree are going to talk about what we learned about each other when we were roommates together in the south of France this summer for a week. Yay! Thank you so much Kat for joining us. Thanks for having me. I'm so excited. It's going to be a great conversation. Yeah. Okay. So let's start this episode by revealing a little bit of Internet history. Oh God. No, it's only good. It's positive. It's only good things. I trust you guys. Which is that Sheree and I. Yeah. Sheree and I were one of the very first people to get our hands on physical phones back when they were just prototypes when we used them in our photo shoot in 2024. That's right. Oh my gosh. I almost forgot. We mailed you guys early, early prototypes. Early, early. Editor, throw those photos up. Yeah. For real. That's amazing. That's right. The lore runs deep. I can't even remember. I don't even know if they were functional at that point or they were just design prototypes. I think they were design prototypes and shout out Maggie who brought them over and we use them for our photo shoot. That's so awesome. They were functional for the photos. That's right. They serve the purpose. And the photos turned out great. You guys look hot. The phones look hot too. Thank you. I agree. So Kat, diving into part one, how you used AI to build a million dollar business. You built a million dollar business using AI and you were selling landline phones, which is such a funny sentence. And now you've evolved physical phones to the ultra successful product and into e-commerce. So how did you get to Kat Labs, which is a full-fledged experience and company? How did you use AI to build your companies? And specifically, what did you use it for to supercharge your multiple businesses? Yeah, totally. So I think first and foremost, I want to set the scene by saying, if anybody thinks that simply by using AI tools, all of a sudden a business is going to fall out of thin air or become fabricated in front of them, that is absolutely not how it works. Just because we have AI today, you still need the ability to build a million dollar business. And that's what we're trying to do here. We're trying to build a million dollar business. We're trying to build a billion dollar business. on and was like, okay, how can I bring all these pieces together to actually gather value for my team and really put something load-bearing for my business on this tool? And essentially what I did is for CatchyBT, the media business, I'm in a Slack channel with my management team who's constantly pitching me brand deals and saying like, hey, would you want to do this deal with this company or this deal with that company? And it's so much to keep track of that by the time, a couple months later, I'm getting followed up by them and they're saying, hey, we need XYZ video for that brand deal that we talked about four months ago. I'm like, wait, what brand deal? And I'm like searching through the Slack and it's just super inefficient. So what I did was I went into Cloud Cowork and I said, okay, look at this Slack channel. I want you to crawl the last nine months of history of me getting pitched brand deals from this management company and organize every single one into a Notion database that includes the name of the company that the brand deal was with, how much money it was for, what the deliverables are, what the dates are, what the price is, all this stuff. And it took that and it churned on it for a long time because the amount of threads and back and forth, you start a conversation about one thing here, two other people say something, and then three days later, you're following up on that and it's super disorganized. But the intelligence behind these systems and all these models have gotten good enough that it can actually parse out all of that complexity and noise in conversation. And it actually turned it into a Notion database that then I could go through and much more easily scan and search to find the deals that I was looking for. And most importantly, the thing that I did after that was I set an automation so that every day at 9am, it goes into the same Slack channel and checks for any updates on previous threads that it knows haven't closed out yet, or checks for new threads of new deals that have gotten pitched, and it constantly updates that Notion database. So essentially, I went from being like, I wonder what are all the deals we've been pitched, or like having no way to kind of audit or inventory what we've agreed to or what I've been pitched, to a very reliable, constant source of truth about what the inbound for my business looks like. Check responses set up required compatibility and availability varies 18 plus. Need a hiring hero? And then talk us through the part where you said you set an automation. How does that happen? Or how'd you do that? Yeah, totally. Once again, like any main frontier lab has these custom has these desktop apps that are very easy to use. But yeah, in if you go into the cloud desktop app on the left hand side, you'll see something that says scheduled automations or automations or scheduling. I forget exactly what it says. But when you click on that button, it'll basically just say like, what automation do you want to schedule? And you can just describe in plain English, exactly what you want it to do. And it might ask you a couple questions be like, okay, just to confirm this is what you want. Okay, here are the tools that I'll need to run. And it's truly that simple. And then it creates this little automation, there's a toggle on and off. But there's no like with Zapier and some of these other tools we've used in the past, you'd have to like, be very specific about the things that connect and where you're pulling data in from and how you're going to enrich that. All of that goes away. You literally just use plain English to describe what you want to happen. And it will do it. That's really cool. So what you just described, that is an AI agent, right? Does that count as an AI agent? Yeah, absolutely. That's what I mean, when I'm like, people make it sound scary. And you know what I mean. But I think a lot of times when people will use AI agent, they're referring to something that will go off and then do additional like what I described is fairly, it's fairly easy to understand. It's like very directly replacing an amount of like tactical kind of clicking around labor that I would otherwise have to do. Sometimes I think when people refer to agents, they're talking about coding agents, which are actually writing code, right? But they're all agents. Nonetheless, the way I think about agents is just an AI that can do stuff. Yeah. And I think if you look at a bunch of these different AI platforms, I like Clawed because it does prompt you for like, it asks questions back, so that it can get more specific. But looking across all these AI platforms, and you bring up Clawed specifically with this co-work example, can you walk us through which ones that you use for what or what it should be specialized in one if we really like it? Or how are you thinking about which ones we should use? Because there's so many out there. There's so many out there. There's so much noise. And one of the most common assumptions that I hear from people is that they have to know all of them. And that literacy with AI is equivalent to literacy with all the tools. You have to know all the tools if you're good at AI. I think that is probably the most wrong assumption. That's very common in this field. The reality is, I think you should pick one and go really deep on it and learn it really well. So if you're going to take Clawed, for example, just click on every button that you see in the interface and get to know what it does. Because the fact of the matter is, I think we're seeing all of these different frontier labs start to converge and commoditize the intelligence layer. And so now what's becoming different is the product layer and what people are able to get out of each of the different desktop apps and web apps and tools. And we're seeing Anthropic come out with Clawed design and all these different things. But for the most part, they're kind of all copying each other. They're all sort of the same. So yeah, if you take Clawed, for example, I think just getting really good and getting to know Clawed and Clawed's voice is a great way to just kind of go all in on one thing. Mm-hmm. And for you personally, at this point, are you still using multiple tools for different things where maybe you've like trained one up in one specific area, or are you kind of trying to just spend more time on improving the one that you spend the most time in? Definitely the latter. Okay. I just spend, for me personally, I've put so much context into my conversations with Clawed that that's where it knows a lot about me. I will say that's also not a really good reason to stay with one though. Meaning- Mm-hmm. If there was a product advantage of going to somebody else, exporting the knowledge that I have from one tool to another, isn't that hard? Yeah. You can actually just say like, "Tell me everything that you know about me," and it'll just spit it all out. And then you can just grab that and copy and paste it and drop it into the next tool. Granted, you're going to lose some depth, but no matter what, every tool that you're using in a conversational interface is summarizing and compressing all your conversations from the past. So it's never going to have like perfect memory and know absolutely everything about you. So I think the switching costs are a little bit overblown. Should everyone be paying for an AI platform at this point? That's a good question. I think so. I think so. I think the difference in the pro version, which is like $20 a month or whatever, versus the free version is significant in terms of intelligence and also just level of access. You won't run out of credit so quickly. From max to pro, I understand if that's not in everyone's budget, because that's literally an order of magnitude higher. But I do think- I think that for professionals, especially if you can expense it through work, it is absolutely worth every penny. What level are you using now? I use max. Yeah, because I use it all the time. Because she's Catchy PT. I'm on the max plan. Yeah. So I pay for the highest tier of Catchy PT and Claude. I really love the example that you gave us of this agent that you made, and you made it sound really simple to create. Do you have any other examples that you could walk us through of recurring projects or these kind of automated workflows that you've made for your business? Yeah, totally. Another one that I would describe is somewhat related, but it's basically an accounts receivable automation. So in my media business, we create sponsored videos. And then after that, the brand sends money, essentially. And sometimes it's hard to know, okay, wait, when are we supposed to get paid for that thing? And did we get paid? And I am not an accountant. And I also don't have a full-time finance person on my team to follow up and do accounts receivable. And so basically, we built an automation that looks at the database of payouts and matches that up with an export that I grab once a week from my bank statement and basically matches the two and says, okay, did we receive, did the literal cash land in the bank to match up with what we allegedly were sent? Because ultimately, the bank is the source of truth, at least for me. And so I'm just very much automated. And the way that it marries it all up is just within Claude Cowork. And then it sends an automated message to me on Slack every week, just letting me know, hey, these payments are going to come. These ones are looking late. And it gives me and my team something to reference when we go back to the management team and we're like, hey, just want to follow up on this payment. And it saves us so much time in terms of manual effort. I will say we have run into snafus with this automation. And I mention that because I think it's really important to call out that these systems are not perfect. And that for any responsible business owner, you're not going to put something as important as like final say financial accounting on an AI agent unless you're like 100% sure that it will not hallucinate or make mistakes. So we still look at all these things that the bots are doing and occasionally we find mistakes. And so I mention it because it's also like it is an ever evolving thing where we're working with it to make sure that we're mitigating as many mistakes as possible. But those things do crop up. Ultimately, what we find is that we still save a lot more time by using these tools and then going back and double checking their work than we would if we had to do everything from scratch. But don't believe everything you hear. There's a lot of people on the Internet who are like, it works flawlessly every time. I think responsible business owners should know that you should always double check and make 100% sure. And only when you've seen it be successful time and time and time again without any mistakes should you kind of take your eye off of that. When you do find these mistakes, do you try to then train the model to be like, hey, you made this mistake. Make sure you don't make it again. Or are they still making the same mistakes over and over you find? Yeah, it depends on the kind of mistake. So, yes, oftentimes it's as simple as just like, hey, I forgot to mention when we get paid out for this particular brand partner, it goes into this bucket. So you don't need to worry about that. Or like I forgot to mention, don't factor in external transfers because those are just moving between my accounts. Like things like that that are just missing. But when it's an actual mistake, there's kind of two things I like to do. The first thing is to just tell the model, hey, this was the success criteria. This is what you did instead. Do you understand the delta? And don't do it again, essentially. Is that the language you use? This is a success criteria. This is what you did. Yes. Yes. I think the more clear and direct you are with these models, just like with an employee, like it's easier to understand. You're using computer language with the computer. Who you are. Yeah, exactly. Programming language. Exactly. Like computers are like inherently deterministic. Obviously, we're talking about a large language model. So like that's weird. But in general, like think about it like a programming language. Like the more concrete and specific you can be, the better. Compura. Compura. I brought that up last night and no one knew what I was talking about. Okay, good thing now she knows. Everyone just started. It was like, okay. Compura. Sharia dinner. Compura. And then the other thing that I recommend. I'm doing is looking up the.md file. Do you know what I mean by this? No. Tell us. Okay. So when you're interacting, like let's say in Cloud Code or Cloud Cowork or ChatGPT Cowork or any of these other tools. So wait, would we call this intermediate level? Yeah. Or advanced. I think that's intermediate. Yeah. Yeah. So basically what I mean by this is like when you go and I wish I had my laptop, I can show you what this means. But like in Cloud, if you turn on Cowork, it's going to be like, do you want to select a folder to work in? And you're like, what folder and why would I need to do that? This is what before when I was referencing projects where it can be helpful. Because what's going to happen is if you say, I'm about to start working with Cloud and we're going to start working through accounts receivable and finance stuff. I'm probably going to come back to Cloud and want to keep talking about this later or like next month or in a week. So what I'm going to do is before I just start running off the bat and having this conversation, I'm going to create a new folder on my desktop. Like literally how you would save like photos. Okay. Okay. A new folder anywhere on your computer and you're going to call it finance or whatever. And then in Cloud on that little button that says select a folder, you're going to click on that. You're going to navigate to that finance folder and you're going to select it and you're going to click continue. What that's going to do is it's going to centralize all of the context and conversations from the one you're about to have about finance. Plus any future conversations where you go back and select that folder again into one place on your computer. Which means that Cloud. Or whatever tool you're using is going to start writing and saving files directly into that folder that will help it remember in the future what that folder is all about and what this project is and what you're working on. So without you even prompting it or knowing that it's doing it, these tools are writing markdown files, which are essentially just very plain text word documents. Like there's type on a page explaining, okay, Cherie really wants to talk about finance. Sounds like she's got a couple tools going on. These are the three that she's mentioned. She really wants me to double check to make sure that nothing, you know what I mean? Like it's writing its own memory. Yeah. And then it saves it into that folder. And so when I say a.md file,.md is like the shorthand for a markdown file. It's like.xlsx is an Excel doc. It's just the whatever that's called, the suffix. And you can actually go in and you can go to that file on your computer and open it up and read it. And so sometimes if you read through it, you'll realize, oh, it hallucinated because. It saved the wrong logic here. And so you can either modify it by asking Claude to change it or you can literally just delete it and type it in. Or modify it yourself. Yeah. Interesting. I like these examples about the accounts receivable and also the first example you gave about matching up the brand deals and creating a folder to keep track of what's coming in and when. Those are pretty straightforward because it deals with numbers and updating those numbers. When you think about using AI platforms. For more creative work or also things that are more strategic, or we said in the beginning, challenging assumptions. How do you change your prompts and how you approach one of these platforms? Yeah, totally. So a couple of things come to mind right away. Number one, I absolutely love to ask for it to ask me questions. So I'm constantly telling Claude or ChachiBT to please interview me or ask me any follow up questions to ensure that it fully understands. What I want it to do before I have it go off and do whatever thing that is, especially if it's like any kind of creative or kind of more open ended thing that's less about like, hey, crunch these numbers or check this thing. And it's more like, I need to work with you on this and we're going to kind of be a thought partner together. The reason why I do that is because these models thrive on context and on clarity, but sometimes it's hard to be. You're sitting there thinking like, what's all of the possible context it would need to know? Like you're sitting there and you're racking your brain, but it's hard to know everything. Yeah. And so it's a hack to just be like, just tell me what you need to know. What else are you wondering? And it will, it does a phenomenal job of identifying what the holes are in what you've just asked it to do. And it'll just tell you what the holes are and you could just answer it that way. I love this best practice. So then what do you say? You're like, oh, please, I want you to do this. And then what do you write at the end? Like, what's the phrase you add? Ask me any other questions you need to know. Exactly. So it'll, it'll be like, I'm trying to think of an example for this. So perhaps like writing a job description, right? It'll be like, I start with the wireframe. I'm not just going to have it write a job description for something for me because that's. Really important for my business. And I want to make sure I have my thumbprint on what I really care about. So I'll start with it. Like I'll give it 15 bullet points of like, here's the things that I want, but I'm not going to sit there and write like the full bio and description and everything. So I'll drop that in and go, I want you to help me write a job description for my company. Here's the starting point paste in the 15 bullet points. Ask me any questions that you have in order to make sure you have the context that you need to complete this job. Okay. Ask me any other questions you have to ensure that you have the context to complete this job. Exactly. And like, I know that can be helpful to know the exact sentence, but also if you word it slightly differently, or you just say like, ask me any questions you have. That's totally fine too. Yeah, it'll still work. And then when you write the prompt, do you include, this is the success criteria or like how often do you use that sort of language? I feel like the success criteria language and being like super, super specific is more for when I'm trying to automate something that I need to run over and over and over again, without me looking at it, if it's just a one-off task and I'm, I'm going to be right here with it and it's, I'm not going anywhere. I'm, I don't worry about that as much because I'm going to be involved the whole time. Yeah. I guess it's more when you have like a determinate outcome versus an indeterminate outcome. Yes. Yeah, exactly. So Kat, I have a more advanced question. We kind of did beginner, intermediate and now for advanced. Yeah. Sound the horns. So how do you, we talk about like, we're like, Jean, you like actually like did typing. Hands as you're like typing out on this coming from Sheree, the computer engineer, computer science engineer, major. No, it's more casual than that. How do you use one of these AI platforms when you're doing speak to text or dictation? Or I don't know. I just see some of these videos on like Instagram where I'm like, Ooh, that's so sexy. Someone's like talking to their AI and like walking across the street and like it's so integrated into their workflow. They're like strutting and like working at the same time. Like, how do we do that? And does it actually work? And what's that all about? And they're like, ah, yes. I literally made a t-shirt that said hot girls don't walk and look at their phone, but we do walk and talk to our phones. You just can't look at it. We walk and talk to our AI agents. Okay. So, uh, the, yes, one like hot take that I have is that in the future, we're going to be doing a lot less typing and a lot more talking to our devices. Will it be through glasses? Doubtful. But I do think that dictation has gotten so freaking accurate, like biblically accurate that you could just talk way faster than you can type previously when we didn't have large language models and you weren't talking with AI chat bots. It's like, yeah, dictation is great, but like, what's the point of like typing up a big rambly thing? Cause now you have to go back and edit it before you can send it to anybody else. But now because we have these AI chat bots, you can load that baby up with rambly context. it's still gonna be able to process everything you're telling it. like it's kind of crazy if you go into any startup in the san francisco bay area right now and you open the doors of like a we work or something like that every single desk has a gooseneck microphone have you seen this no and it's because they all have this application called whisper flow and it's literally just like a super simple app you download it off the internet and you turn it on and you have this little like gooseneck mic and people are just like um upload the context in the dot md files and it's like everyone's whispering in the office because nobody's typing anymore say this episode is brought to you by google chrome you think you know a browser but gemini and chrome that's new it can help you with practically anything on the web like restoring a vintage motorcycle from a 50 page restoration block or finally break down that long article you've had open for weeks gemini and chrome is here for it ready to make anything online make sense there's no place like chrome check responses set up required compatibility and availability varies 18 plus when you need to build up your team to handle the growing chaos at work use indeed sponsor jobs it gives your job post the boost it needs to be seen and helps reach people with the right skills certifications and more spend less time searching and more time actually interviewing candidates who check all your boxes listeners of this show will get a 75 sponsor job credit at indeed.com slash podcast that's indeed.com slash podcast terms and conditions apply need a green hero this is a job for indeed sponsored jobs but it's how it works like that's that's how everyone's doing it so yeah i'm big big big big on dictation like once you do whisper whisper flow you'll never go back i'm constantly talking to my ai the thing that i do is like in the car i'll turn on one of the agents and i'm talking to it and it responds to me and it's just like i'm talking to someone on speakerphone and it'll be from like anything from like hey i'm talking to someone on speakerphone and it'll be from like hey i'm driving a trader joe's right now and i'm like about to make dinner help me brainstorm this stuff or i'll be like hey i'm coming out of work it's just like a normal conversation yeah and i can have it in the car where i am like hands-free 100 like well i have to like be like eyes-free looking at the road totally i do the exact same thing yeah the times i use voice the most are context dumping for something like what i just described or it's like a big task and you're like i'm not going to sit here and type out all the context that i need to share with you i'm just going to give you a bunch of rambling thoughts and you turn these rambling thoughts into a job description or something like that or like you just said driving the other hack that i'll do for that is i find that i end up using driving time a lot to like go down weird like nerdy rabbit holes like i had this thought the other day while i was driving where i was just like wait how are billionaires leveraging their assets to like get loans and how does that not just put them in more debt and like how does the flywheel continue like whatever who am i going to call to ask about this the things that go on in cat's head while she's driving in la so far cry up there you know what exactly what color should i paint my nails next yeah um so anyways i like but what i do for that is yes i'll i'll like dictate with chachi bt or what have you i find that chachi bt voice is really good i like it so it's it's really good for that but instead of having the british guy no but i don't have it talk back to me because i don't like how it shortens all the answers are really short the things that i'm going in on i'm like no i want you to give me like a lesson right now like we're going to do math like let's go in yeah and so what i do is i just dictate i do speech to text sends the text but what i append to the end before i hit release is i say give me back the response in the form of a transcript that could be read aloud to me because if you don't do that it'll give its answer you know with like a bunch of headers and then like number one number two number three and it breaks out so it's more easy to scan if you're reading but if you're driving it's like header one idea header two like it enunciates each of those things and it makes no sense so you have it read back as a transcript and then if you scroll down there's a little play button which nobody ever clicks on but if you any chat gpt response you get you can click play and it'll just voice over everything it said yeah and so then it'll talk for like six minutes huh you like it more than the conversation because the conversation isn't doing it because it's too truncated yes interesting i haven't tried that before i'll have to try that and you don't like that it interrupts you no not that it doesn't interrupt me but that it's too short it's like way too short like it's more conversational like it'll say something for 30 seconds then you have to respond and then it's just like a very back and i hack is for language practicing i've used it to practice um like mandarin and also cantonese and you can tell it very specific with like dialects like where you want them from like hong kong cantonese versus like canto cantonese that's really really cool and it gets like the nuances has it has it helped no but the thing is i feel like it can't talk slowly uh no you can tell to slow down really oh they must have improved it then because i've asked it in the past to slow down and it's like like yeah no problem but it still doesn't really cost me fuck okay ah yeah i could do that for you that's the other thing is like it's really chai gpt voice really annoys me because it's like ah totally understand that's a great point no problem absolutely and i'm like shut up you're so annoying what in a british accent but in a british or that's again if you're like just give me the transcript and then read it it'll just talk like a normal person i see i like that yeah that's cool hey guys so you might notice we are in an incredible studio right now thanks spotify and if you are not noticing because you're listening watch on spotify turn on your video see this amazing studio check out the lighting guys it is so good when you're watching on spotify you get so much more of the in-person experience with jean and me you see the banza you see our facial expressions and it's way more engaging turn on your video and check us out now back to the show okay cat so let's move on to how you use ai in your personal life because one thing i learned about you when we were roomies together in france in france in france um is that you are super organized and just like i feel like you're like really optimized in your day-to-day life so i'm curious how do you use ai as part of that system that you've built it's funny that you say that because i feel like i sometimes feel up here's what i think i do i think i spend a good amount of time deciding what i'm going to be spending my time on up front yeah so that i have a lot of free time to like breathe and be creative later so it's funny that you say that i eat that i'm optimized sometimes i feel like i'm not but thank you um okay how do i use ai in my personal life like to stay organized essentially yeah and you can even talk about maybe just the system that you've built like really sort of briefly and then maybe if there are any ways that ai plug into that yeah um well essentially i don't i honestly don't know if i'm the best person to answer this question because i really don't use ai much in my personal life at all like my ai usage is pretty much entirely isolated to my business i'm trying to think like okay of course these there's the occasional like my stomach hurts what should i take like but that's more kind of just like loose you could also just google that you know what i mean like yeah i wouldn't consider that very like intermediate or advanced ai usage um and i think we're all kind of used to it but i think it's we're all doing that anyway nowadays um there's probably more stuff that i would recommend against with using ai in personal life that i would recommend for people to do as an example like me and my partner are in this argument what should we do i think that these models have a real tendency to reflect us back at us and an unintended side effect of that is that it can really hype you up and sometimes to a detriment where you don't see other points of view and i think it's a really slippery slope especially for people who really just want to hear their own ideas reinforced and i actually think what's far more valuable is going out into the real world and finding friends and a friend group and people in your life who can like push back on those ideas and like provide another point of view i don't know if that's what you meant by personal life yeah um but that's kind of how i interpret it whenever i hear ai in personal life like little alarm bells start going off in my head i mean there have been points when i've used ai in that way and made sure to call out i'm like what am i not seeing here like challenge my assumptions like this is my point of view this is the argument that we had and this is how i'm thinking about it like play devil's advocate and challenge me i'm like am i being too harsh or like how are things actually totally going on the other side i think one one way to do this is you just always ask for the bull case in the bear case sorry that's again a very businessy thing but it's like no that's good give me the supportive and the like devil's advocate point of view on everything that i give you like on every and just but the thing is you have to remember to ask for that every time yeah it's risky because if you forget to ask for it again listen i think sometimes you do just want to be like soothed a little bit and it's going to be like isn't she a bitch or whatever it's like helpful to have yeah she is she's so annoying yes you're so right queen i think some people don't even realize it's doing that they think it's objective and then that's where they're like they feel very self-righteous about everything because now they've got a chatbot in their pocket it's funny because sometimes you people say like when you talk to a friend and you have a an issue sometimes the friend is like okay do you want me to give you like solutions or do you want me to just listen and empathize or do you want me to like give you opposing viewpoints it's kind of like the default for chat gbd is always to empathize unless you ask it to do the other thing yeah that's actually a great point yeah i think this is a good way of putting it i will say claude does a better job of providing that opposing viewpoint or to ask for it than chachi bt chachi bt no matter what is going to hype you up claude will be like you sure about that okay Like it's kind of it will it will push back. I think its default is much more likely to lean that way. Interesting. Kat, we did an episode on Tiger Sisters maybe a few months back now on how to use A.I. to be hotter, healthier and happier and happier. And in the comments of that episode, it was like our first foray into like talking more about A.I. officially. And in the comments of that episode, people brought up the environmental impacts of using A.I. And I just kind of want to get your opinion and thoughts on like how can we reconcile the use of A.I. every day that helps us with our productivity and our work and the negative impact on the environment? Totally. I'm really glad you brought this up. I'm actually doing a campaign right now with Anthropic because part of the thing that I feel really challenged about right now with my own A.I. usage isn't merely I feel guilty because of the environment. The environmental impact of A.I. it's I feel guilty because I don't really understand what my A.I. usage is. Environmental impact is. Yeah. Like for me, the gap isn't I know it's bad and therefore I do it. It's like I don't even know what the impact is because I've seen such wildly different sources saying things from it's the scourge of the earth and it's burning the atmosphere on a scale we have never before witnessed to like everyone is overblowing this. And you literally. It uses less energy than like a hairdryer. I'm like, how are we getting such varying views on this? And so Anthropic right now is currently doing a campaign called Hard Questions where they're actually welcoming their audience and asking the public to ask them their hardest questions about A.I. so that they can go back to the research and come back and give them those answers. And so I'm currently partnering with them. I don't know when this episode is going to come out. It might be done already. But if there's still time to like collect these questions from people and that's my question that I provided them. So that we can actually like come back with those answers. Yeah. Do we need a third party on that? On that research study? Very fair. Paging McKinsey. Yeah. I would say the EPA, but maybe not the EPA these days. But yeah. Who is objective on this matter? Harvard, get in there. Yeah. Somebody, please. But it's a good but it's a good call out. Yeah. Unfortunately, I wish I had more information about exactly like what the environmental impact is. Yeah. I do believe it's getting overblown, but I don't I cannot sit here and tell you why that is or like show the facts. So I really don't know. I think it's fair to say that you're in information gathering mode. I think that's a totally fair answer. Like you're not like, oh, I know what the answer is. Right. I mean, I think all of us are like, we don't really know what the we don't know what the carbon footprint is. We don't have that information. Yeah. And I don't know if it's like we don't have it because nobody has it. Again, it's just like there's so much for like. Different, disparate sources of truth on this and nobody it all seems to be so different. Yeah. Or you could only you could just use AI only when the sun goes down like me. Save energy for you. It's the little things. I love that. That was a joke. Oh, I couldn't tell. I thought you were serious. I was joking. Okay. I thought you're like, I'm a night owl. And I was like, oh, that's cool. You're you're so you're going along with it. Oh, nice. All right, Kat. So we're moving into a section where we're. We're talking about practical prompts from your playbook. So as we're talking about prompts, I want to hear like something that anyone listening or watching this episode right now, if they were to do something immediately after, what is one thing that they should do in their AI platform? Yeah, totally. So I think when we think about this in terms of like, what's one prompt that everyone should do? It's kind of tricky because, like, the reality is your prompt should be entirely different depending on what your needs are. If you are a teacher. You're a prompt that is extraordinarily valuable to you is going to be completely different from somebody who's an entrepreneur. So it's not like it's like saying, like, what's the best Google search someone could do? It's like, what do you want to know? You know what I mean? So I think instead, I'd like to reframe that a little bit. It's like, what are some great meta prompting techniques that anybody can use? And a meta prompting technique is literally just like how what are ways that you can prompt that improve the results of what you get back, regardless of the use case. Love that. And so an example of like a meta prompting technique that I am obsessed with and I use all of the time is when I am trying to use AI to make something sound better or more broadly to like take a rough version of something and then polish it and make it more mature. So let's go back to that job description example. Right. I know I'm hiring a community manager and I've written down 15 bullet points, but I need to like actually turn this into a legit job job description. Now, what I'll do is. I'll go to Claude, for example, and I'll say, this is a very important job description. I want you to read it, assess its goal, analyze it for shortcomings against that goal, and then rewrite it to address those shortcomings. And then what you do is you provide a rough draft of whatever it is that you have. So if I just have 15 bullet points, I'll include that. But if I have an email that I'm writing to somebody for a negotiation, I'll include that. Anything that's like sensitive. And I need to get the framing and the tone right. This is a perfect prompt for that. And what it's going to do is by framing the initial prompt is I want you to first assess its goal and then analyze it for shortcomings against that goal. You're actually prompting the AI to analyze it at the overall structural level as opposed to just at the sentence level, which is what it's going to default to if you just say, make this email sound better. So that's an example of like a meta prompting technique that will instantly improve the results of all the results. Of everything that you get back. It reminds me of it sounds like you're having the AI basically do like an SAT question where it's like analyze the intention of this and then assess it for success against the goal that you've assigned it. Yeah. Like it sounds like it's a writing section of the SAT. Exactly. Yeah, exactly. You can take the SAT. And so do you use that? So you basically use it on anything that's external facing. It sounds like. Like where I feel like I need help. Yeah. A lot of external facing things I'll just shoot off myself. But if it's something like really sensitive or dense or I just want to get another pair of eyes on it and I don't have anybody else near me to help, then yeah, that's something that I'll do. And you write this is an important document. Yeah. I'll just write this is an important X. This is an important document, an important job description, an important email. Interesting. Whatever it is. Yeah. It's really fun to hear all the like tips and tricks and like the little things that different people do. Yeah, I know. It's fascinating, right? Another one I learned recently, it's more relevant to people who are using AI for like coding and stuff. But my buddy showed me how he will go back and forth with AI and it'll be like, sorry, I can't do that. And it'll be like, yes, you can try harder. And then it does it. And then it just does it. It's like it's so like that's another technique that people can use. Again, it has more to do with like when you're trying to get it to do something agentic and take action. Sometimes it's just like. Like I've done this before where I've I saw other people online using cloud to analyze videos and then like categorize and organize their videos. And mine came back to me and was like, oh, sorry, I can't I can't read video. And I was like, yes, you can. Like, I know you can do this. And then it came up with a creative solution where it basically took the video, broke it up into screenshots and took like frames from each five seconds. And then it was able to like infer what the video was about by taking those screenshots and the screenshots. It said captions on them and everything like that. So I was able to fully get it like to the end goal of what I needed it to do. It just had to come up with a creative solution. So sometimes it just needs nudging. You're like, why are you lying to me? Yeah. AI platform. Just gaslight it. Right. Yeah. Well, so you got an idea from talking to one of your friends about how they use AI. So like, are you learning about meta prompting through your own trial and error? Or where do you go to get all the goodness, the tips yourself? Are there like video platforms or channels that you recommend? Yeah, I recommend Catchy BT. But seriously, like from friends, like the crazy part about this moment in time that we find ourselves in right now is like there is no playbook. There's no guidebook. Like we are all writing the guidebook together in real time. And so the most valuable insights that I personally am getting from how to use these tools is from the people that I surround myself with who are also experimenting in real time. And when we come together and we share. With those learnings, we're able to actually like compound those and learn from each other, which is exactly why I built Cat Labs. Like Cat Labs literally exists so that I had an excuse to bring all my smartest friends together and be like, so we're going to hang out and talk about meta prompting and just like share screens and talk about what we're doing. We're literally hosting an event tonight where the event is called How I Built This. And everyone has to go up with their laptop, plug in their laptop to the TV, screen share. And they spend one minute showing what they built, but then they spend four minutes screen sharing into cloud code or whatever tool they use to build it and literally scrolling through and showing us here is exactly what I said and the exact skills and methods and tools that I used in order to make the result that you just saw. Because I think right now trust is at an all time low with AI tools and people on the Internet and elsewhere are saying, OK, you made one hundred thousand dollars. with AI, show me, prove it to me, literally. share your screen and show me proof is in the pudding because there's so many grifters out there on the internet who are saying that because then they're trying to sell you something yeah but like who's actually doing it for real i think that is the value of actually turning the camera and being like we're gonna talk the talk and walk the walk and part of that comes with showing what it can't do yet you know that's why i point out things of like look we've had some snafus with the finance thing and we can't fully take our hands off the wheel yet like that's a good call i do think that these models are the dumbest they will ever be which is absolutely bananas because if we made absolutely zero intelligence improvements on these models we would still be figuring out how to deploy them across the economy and it would be transforming the world that we live in for the next 20 years without a doubt like and yet these models continue to increase in their intelligence exponentially over time like it's i think the world we're about to inherit is going to look so wildly different from where we are today so there are very few things that are important right now than for people to like get really literate with these tools and understand what they are and are not capable of this episode is brought to you by google chrome you think you know a browser but gemini and chrome that's new it can help you with practically anything on the web like restoring a vintage motorcycle from a 50 page restoration block or finally break down that long article you've had open for weeks gemini and chrome is here for it ready to make anything online make sense there's no place like chrome check responses setup required compatibility and availability various 18 plus when you need to build up your team to handle the growing chaos at work use indeed sponsored jobs it gives your job post the boost it needs to be seen and helps reach people with the right skills certifications and more spend less time searching and more time actually interviewing candidates who check all your boxes listeners of this show will get a 75 sponsored job credit at indeed.com slash podcast that's indeed.com slash podcast terms and conditions apply need a hiring hero this is a job for indeed sponsored jobs yeah i'm pretty fired up i mean i'm like i feel very inspired to go back home and kind of do like a cleanup session to kind of like dig into what we're using now and set it up in a better way because i think one thing you said earlier that i really loved was the folder setting up the folder ahead of time it's just so funny because it reminds me of like back when you work in corporate there's this kind of concept of like folder hygiene and just general like work hygiene and setting things up so that the system is set up for success in the future when you bring in a lot of different people or once you have a lot of tasks and workflows and all that stuff and this is like the new age version of it where you need to have a set of rules and hygiene and best practices for setting up your ai desktop basically yeah i think also said another way like if you were joining a team you would have an onboarding file or an onboarding doc for yourself or for other people and in that way you need an onboarding doc maybe it's a markdown file for ai but you need to onboard all of your ai folders and tools to make it efficient for your own projects and usage 100 i think using ai tools is very similar to hiring it is just like bringing on someone new to the team the more context you can give it the more tools and access you can give it the better it will be but also you don't want to overload it with so much that now it's like confused about its role and that's where i go back to like success criteria the best performing employees will come when you're like this is exactly what you need to do to succeed if you do this thing you will be successful and if you don't do this thing you will not be successful like it seems overly stern but it's actually incredibly kind because you're being very clear and you're setting you're giving someone the playbook of like here's what i need you to do to be successful most people don't take the time though to articulate what that is and so the employee is just left to like fill in the blanks and interpret and for a role that's like pretty confined and specific that's probably okay but for roles that are like really broad i think that's where things can kind of go off the rails or you'll see ai's that just like start doing random things because they're like i don't actually know what she wants me to do you know hey guys it came to my attention that some of you don't know we have a matcha brand what a matcha brand yes because like some of my friends were talking to me like wait you guys have a matcha brand and it's called sisters matcha jean and i started it in october 2024 and it's something that means a lot to us because i lived and worked on the farm where we source the matcha from this matcha brand is our baby and we are so passionate about it and we really love everything that has gone into it there's just so much love and passion that goes into our company and we want to share it with you guys we source from a small family farm just an hour and a half outside of kyoto called wazuka is the name of the town and it's absolutely incredible we've visited the farm we've worked on the farm and we're so excited to bring it to you so you can find out more at sisters matcha.com that's sisters matcha.com all righty now back to the show in that same vein of being direct to your ai agents or the platform that you're using one thing i really admire about you cat is how direct you are and i know you've had a series of videos recently of this concept go viral and so the question i have is that you don't appear overly preoccupied with making every sentence sound soft and agreeable and i think in one of your past videos you said i think i'm too direct and so can you talk a little bit more about kind of the mindset you have when you're delivering feedback for people who might beat around the bush me a little bit too much um whether you're talking to like an ai agent or also to someone on your team yeah for me and my entire life i've been on the corporate for seven years before i started becoming a content creator and my most common piece of critical feedback that i used to always get was you're a little too direct like could you just kind of like soften that a little bit like you're really just like very straightforward with people and it's funny that now it's seeming like people are really like tell me more how i can be direct online and you're like well what's the success criteria yeah i think exactly so it's just funny like it's all so relative but i do think um especially like as women we're treated we're told that we should like be very likable and like be soft and don't be too hard whatever so um my mentality look this is this is my approach on it i appreciate when people are clear with me because it shows that they think i'm smart and they respect my time when you beat around the bush or you fluff things up it makes me think you either think i'm stupid i'm not tough or that you just like literally don't care about my time because you're willing to fill it with all these words so like i appreciate when people are clear with me yeah and like we can use an example and i can show you how this sounds and i'll show you how i actually do give feedback in real time yeah because i think it it illustrates how much like it feels scary to give direct feedback in the moment but when you're receiving it it's actually not that it's like really not that bad and i think when you internalize that you realize it so as an example like what's an example we could do like let's say like jean i have to give you feedback because what would be something like that like was a mistake like let's imagine we're working agent she messed it up she programmed it wrong she like spent too much money on credits or something like that like you didn't put a credit limit and so you like spent a thousand dollars on credits when you were up at 3 a.m because you only use it after midnight yeah right so the soft way to do this would be like hey jean do you have a second yeah so so like i know you probably didn't mean to do this and like it's totally not your fault or anything like that but the thing is like i did kind of like sort of notice that there was like that thing with the credits that happened and just like i'm sure it was an accident or things like that but it's just really important because like you know we only have so much money and we have to like be really conscious of that so could you just like be really careful next time and that kind of thing is that okay yeah okay i'm exaggerating a little bit but we kind of sounds like how i talk or i need what was your impression of that oh it was so soft and kind of annoying and i was like damn i sound like that okay so here here's how i might deliver something i'm scared i'm scared no don't be scared i'm scared okay um wait what you overspent the credit yeah you didn't do anything this is made up this is a hypothetical scenario you didn't do anything a thousand dollars um jean do you have sec um so listen i saw the bill was a thousand dollars for credits last night did you see that yeah yeah um we have to be really careful about those credits and not spend them um do me a favor um do me a favor do me a favor do me a favor if you notice that it's gonna go over again call me because even if it's late i can help you work on that otherwise we can't have that happen again and if it does we're gonna have to have a conversation do you have any questions no okay thanks keep up the great work you know something like that that sounded really good i'm kind of improvising i could have done that better too but whatever in the moment it's just like i mean it sounds markedly better than the first one anyways right yeah right i think it's like it's kinder it's clearer i'm like jean you're smart you know that was a mistake you didn't mean to do it on purpose but like if it happens again like that's gonna be an issue so like just please don't let it happen again yeah you know the other thing i would probably incorporate that i didn't do that time was like let me in why'd that happen you're not stupid something must have gone wrong like let me in what happened there and just like let that person respond and react um would be the other thing that i would probably incorporate i love it yeah i need to watch this over again and it's not a script but it kind of is so that i like remove the fluff and also the beating around the bush yeah yeah there's also a thing there too of like you're reacting to how that person's reacting so jean is kind of like yeah so i'm not gonna like be super hard on her because she i can tell from her body language that she already gets it um so that's even more where i would lean on the thing of like i just talked to me like what happened what went wrong i think like taking the emotion out of it and showing your team that it's like you're smart this was obviously a mistake like and and like coaching them on how to just like use words to talk through it and not like i'm gonna bite you because you like fucked up or something yeah um is also like a culture thing showing people that like it's okay to admit you made a mistake and in fact especially like for the people i work with i'm like you better own up to making a mistake because the worst thing you can do is brush it under the rug or pretend it didn't happen that's like non-negotiable you can't do that if you're not going to take accountability you can't work at catch a pc okay so i thought we would close out this episode with a fun section which is going to be called roommate revelations we've never done this before and i don't know if we'll ever do it again but one thing we hadn't really publicized is that we were roommates together for a week in the south of france for can we sure were and i thought it'd be fun if we went around and we each talked about something that we learned from each other that you can only really learn by living with each other with someone for a week okay sounds good so i can start it off with something kind of just like light and fun um so what i learned from you a best practice is that when you go to a new city you buy an ice pack and you leave it in the freezer and you use it to help de-puff which is so genius because i would sometimes use like puffy in the morning yeah like you know people are you know and it feels good so i would sometimes take like spoons and put them in the freezer if you use them in a pinch but the ice pack it's it's really thinking ahead i specifically let the record show i specifically do that for tan because my i don't do that in every city i go to but i appreciate that my sleep schedule gets so buff yes if you're flying that it's like i'm just this is a wash so yeah having an ice pack in the freezer so when i wake up i could just hold it on my face and you left that ice pack and now it's at our house you took it home i think we just grabbed oh that's great that's awesome i think roommate revelation i just loved like living together for a week because it just felt like girlhood it felt kind of like sorority because i feel like you sorority sorority because gina and i live together but like when you're living at home like you try to keep things in order and like your mental space is kind of like oh i want things to like look nice and whatever i think it was just like a tornado was in our airbnb just because it was so crazy in can yeah it was right there was like um and i say this with love and like i contributed to it there was like food at like makeup everywhere there was my clothes everywhere my lash my my eyelashes on the floor but like it just felt like girlhood and i don't know i mean this wasn't really a revelation but i just loved um living with you in that week because it was just so fun that we can come together like after a big event and kind of debrief together and be like what did we learn like what happened today especially because we had different schedules um and i think maybe my learning is that like when you go to a big event like find people who um you're going to be able to be comfortable with and you want to see when you go home and you're in your pajamas and you have no makeup on your eyelashes on the floor like just like hang out because so much is going on and you want to find um an ally a friend that you can kind of debrief with at the end of the night i love that yeah i think mine would be i really loved witnessing your guys's sister dynamic because i feel like on the podcast and in everything you guys post online obviously you guys are super warm towards each other but there's also an element of polish but it was fun to see you guys with the eyelash half hanging off and like in the oversized t-shirt and you guys like i think one night you had ordered like really late night like halal or something and it was like three in the morning and we were all scarfing down food at like the plastic kitchen table or whatever and just your dynamic i think jean you were like oh i don't know it was something where i was like are you gonna eat that yeah i'm done okay and you guys were like finishing each other's sentences while like sharing the food that you like you knew she was gonna want so you saved extra for her and like it just made me think of me and my sister we have the same age gap that you guys do and so it made me think about her and just i think witnessing your guys dynamic was great and seconding the idea that having people to come back to and debrief with at the end of the night was so clutch like i wrote down things in my notes app from that conversation that i've not stopped thinking about since can oh my god send it to me i will i totally will they're like direct quotes from you and um it made me really really really want to film it next year like i think we should live together first of all we're staying in a hotel fuck that airbnb but like literally dusty oh and mind you fifteen thousand dollars or whatever the fuck you know what i mean like it was a bajillion dollars you're gonna but i still think we need to pay two bajillion dollars and stay in a hotel next year because that shit was dusty and we're gonna film the whole thing because those conversations i feel like that's the you're it's like the nugget of wisdom that you have when you're like half sleep deprived and you're in a full beat but your hair is in a bun and you're like eating mcdonald's at three in the morning that's the conversation that i want to capture and share with the world next year on like the real post in in so yeah yeah it could be a whole podcast episode in and of itself yeah there's so many things we talked tiger sisters from can come on we'll do it we'll run it back run it back that should just be a podcast in general we only film it at 3 a.m and we're eating fast food while we do it it'd be impossible to book guests or really easy because no one's busy at that time awesome thank you so much cat for joining us so fun yeah and for everyone watching and listening where can they find more of your content more about cat labs if they want to get involved yeah what are your socials and websites that you can share yeah you can just look me up on instagram if you just look up cat gpt you'll find it um or you can go to ask cat gpt.com and where can we find physical phones yes you can buy a physical phone um physical phones.com or look up physical phones on instagram it's all the same thing you'll see this thing thank you okay thanks guys um and don't forget to subscribe to our newsletter the link is in the description of the episode and if you enjoy this episode please rate us and also subscribe thanks we'll see you next time tiger fam bye close your eyes exhale feel your body relax and let go of whatever you're carrying today well i'm letting go of the worry that i wouldn't get my new contacts in time for this class i got them delivered free from 1-800 contacts oh my gosh they're so fast and breathe oh sorry i almost couldn't breathe when i saw the discount they gave me on my first order oh sorry namaste visit 1-800 contacts.com today to save on your first order 1-800 contacts fall never arrives all at once burst the air then the light and slowly the things you love begin to find each other a jacket meets a dress a hoodie slips beneath a blazer an old cap finds a new scarf because you don't simply step into the season you create it layer by layer until it finally feels like fall find fall your way at ross