To create an AI assistant that thinks and communicates like you, you need to train it with your personality data (such as MyersBriggs, DISC, or Big Five assessments), provide writing samples, create a 'nuke word list' of AI-associated phrases to avoid, and calibrate its voice through iterative testing and refinement. This process involves uploading your personality profile, having Claude research your communication preferences, creating a skill that captures your unique voice, and continuously refining the AI's output until it authentically reflects your personality and communication style.
Deep Dive
Prerequisite Knowledge
- No data available.
Where to go next
- No data available.
Deep Dive
How To Build A "Digital Clone" AI Assistant In Claude | Andy O'NeilAdded:
What if your AI assistant could actually think like you? Not just spit out generic responses that make decisions the way you would, communicate in your voice, and create content that sounds authentically you. Today, I've got Andy O'Neal with me, and he builds what he calls digital clones inside Claude AI.
He takes your personality traits, your beliefs, even your decision-m style, and weaves all of that into a custom AI project that actually works as an extension of you. Hello, Andy. How are you today?
>> Hello. I'm good. Thank you. Thank you for having me.
>> Thank you for being here. We are so excited to hear about this. What are we going to take away by the end of this video?
So what I have created uh which uh you'll be able to watch in throughout this uh session uh and take away is a skill for Claude that will take your personality data and put it into your communication style.
I am so excited about this because there's so many times you see people's content, you're like, you know, chat chat GPT wrote it or you can just you just get this feel that there's not that personality like you can feel that the actual thing that you're reading or viewing it doesn't feel real and alive.
So, I'm super excited about this. Can we just dive straight on in?
>> Sure thing. So really what I what I identified several months ago, I've been using Claude exclusively and I canceled my chat GPT account in April of last year, so April of 2025. And so I've been all Claude since then and recently Claude Code on top of that. And so one of the things as I I write content with with Claude and what what you get with ChatGpt or Claude out of the box is this generic AI voice uh that people are getting pretty good at identifying by now. Uh can we can we say hello M dashes and horizontal lines? uh you know that's a that's an immediate you know giveaway that it's AI. And so what what we're trying to do here is we're trying to skew away from that default voice. And there's a lots lots of ways that I've tried and I I still use to skew away from that default voice. Uh one of those things is providing writing samples. Um, I'm not a big writer, so I have a few things that I've written in the past, but not not in depth of a lot of writing samples. Uh, I also use what I call a nuke word list.
Basically, the instructions are don't ever ever ever ever use any words off of this list ever. Um, because they are the words that are typically associated with AI. and then uh adding in special instructions to not do this and not do that. But really the the personality training is what made a huge difference for me in uh in the writing style. And when people actually read people that know me read something that I wrote with AI, they actually think I wrote it. And so it's it's worked really well for me in my uh content development.
I love this. I'm super excited. And I even love that you have a hit list of words that you don't use. And I'm like, I want this list already.
>> Yeah, they're they're handy words for sure.
>> Well, let's throw them in. Can we do this with you today? Can we like see your process of how you actually make it make it all sound like you?
>> Absolutely. So, I am going to move some things around here.
And am I showing up? It looks like I'm showing up. There we go. Got it.
So, the first thing I did uh because I have some of this and I have lots of memories in my cloud is I opened an incognito chat. Um this is the first time I've actually used incognito chat.
And then I gave system instructions that just said, "Pretend like we're brand new and we've never met." So that's what's on the screen right now. Um just a fresh set of instructions or fresh kind of a fresh slate to work from. So I have uh the skill that I'm giving away and and I'm giving away no email transfer or anything uh because I like to give stuff away that I create to people. Um, I'm I'm giving away a skill that builds your skill. So, the process we're going to go through is we're going to start with that skill. It's going to inter interview me through a process and then it's going to give me a chance to rename the skill and reinstall the skill that's been customized to me. Uh, what that means is you can use that starter skill as many times as you want.
So, I'm simply just going to say we're going to access the personality skill. So, it's going to go read that and what it's looking for is a profile that's been set up in that skill. And because it's not been ran, I actually deleted my skill uh for my personality from Claude so it would not confuse it.
It's going to go read that and see that it needs to ask me the questions. So, it's checking if there's a profile right now.
>> Oh, so it's just going to prompt you.
>> It is. Yeah, it's made to to ask questions and and interview me basically.
So I have um I I initially built my skill on uh the MyersBriggs Types Index or that's not the right name but MyersBriggs personality index. And um in preparation for this and building this skill, I actually went and did some other assessments. So you can see some popular assessments here. You don't have to have all of these. Uh if you have uh in my opinion, if you have the MyersBriggs um if you go to 16personalities.com, take the free test, pay $9 for a few more uh for one more other assessment.
I'll show you what that looks like, what the results look like. Um that's plenty to get started. Um I also took the DISC and the Inagram in the Big Five. So we're going to put that in there, too.
But basically this is whatever personality data you have. You can just do a dump in here and it will use all of that.
>> I am going to get over here. So I've got I load up here. So I've I've screenshot all this. So this is the big five um their data. This is the disc I took. Uh this is my Inagram score. Uh what what it all means. This is the 16 personalities MyersBriggs profile. And again, I I think this is probably the best one to start with. Um and uh so here is here's the data you get for free. So you get four letters plus a fifth. Um typically MyersBriggs was I think first developed with four letters, but there's been a fifth added. And then here is the this I'm going to show you 12 metrics you get with the $9 thing if you add on and take the extra $9 assessment. Um you get these four influential traits, these four and these four. A couple to note uh about me is my emotional intelligence is uh very low.
That means um I'm also very introverted.
And if we go back here, perfectionism is very high. And what what that means in a MyersBriggs test is I've got to make sure everything's right before I do anything. So before I go on this podcast, I got to make sure everything's perfect and everything's set up. Like I freak out about being having everything perfect. So just a little If your perfectionism score is high, that's that's what that means. So I'm simply going to drag those images I showed you right in here.
And once those upload, we're going to let we're gonna let Claude read those.
It'll do a perfectly uh it'll do a great job reading those directly off of there.
It knows what to expect. It can see there's my numbers coming down there.
There's my disc profile in sorted by match.
Big five. Have you ever heard of the big five?
>> I I took it this weekend. I never heard of it.
>> Was it good?
>> Well, I mean, it's it's it tells you um like like it breaks down your people into like five categories. So, mine is openness and neuroticism, which means I tend to stress out more than other people, which probably goes with my 75% or 90 97% uh perfectionism score. So, all right. So, that's all uploaded and read. Uh, just make a It's It's asking me which way it wants to go on the Inagram.
There's kind of a secondary thing. Uh, make an educated guess because I don't know which one's the best.
>> This is so cool. So, I guess if you're someone who's into human design, you could add your human design stuff in. Or if you're really big into astrology, you could add your astrology stuff in. So that's super cool.
>> Let's see. Let me run the research phase.
I'm going to stop that and I'm going to turn research on.
Continue research. All right. So, this is where it's going to take my my design, my all those results uh that that it's looking at and it's going to go research it. So what does a what does an ISTP MyersBriggs that is also a DC what does that mean and what is you know when we look at this and this what does that mean and so um goes through and does quite a bit of research and this is the research that is going to um really help with your tone as you write content.
All right. So, right here, um, have all the research I need. Let me build a comprehensive profile and inferred communication preferences. When I was building this, it was like, answer these questions, like open-ended questions.
And I said, no, I want you to write them, and I want to correct them if they're wrong. So, these six questions are written based on the uh, personality data we've given it. And, uh, I I did this with a client this morning, and he wrote he read this, and he's like, totally me. Yep. Totally me. Yep. Oh, this is holy cow. I can't believe this.
Like he was just blown away by this. But basically, it's going to be like what's your reading tolerance? You know, what's your information format? Like, do you like long walls of text or do you like bullet points or somewhere in between analogies and metaphor? Humor. Uh I I love humor and dad jokes and puns and things in my writing.
um decision support uh and context depth, you know. So, how how deep does this need to go? Uh I'm going to say this tracks.
All right. So, it's going to build the full profile.
While it's doing that, may I ask?
>> I tend to be someone who when I'm talking to people, I try and have a very conversational style so that it's easy to relate to like so that we can relate, you know, build rapport, that sort of thing, >> right?
>> If I want my writing to sound like me, but like my professional version of me, which there is a professional version of me, but at the moment I'm obviously in the conversational version of me. But if I go into a business meeting, you get professional version of me.
>> How do I get content to still sound like me, but the nice like the professional version of me, not the chat GBT version of me? Is there a way that it can find that middle ground of okay, it's still the same person if she's in a conversation, the same person if she's in a business meeting. It's not chat GBT.
>> Right? So, uh, one of the things I would say is, um, there there's a step here and then and I actually skipped it, but there's a step here where you can put in your personal beliefs and other information. So, I think that's where you would put in um here's here's writing samples of me professionally and here's writing samples of me personally. And uh you might even throw those into claude or chat GPT and say analyze these writings and ex and describe describe them and then do that with your personal and what you'll get is two descriptions of that content and then you can throw those descriptions into here.
And then similar >> sorry similar to that is I have some some people that I I send emails to long detailed emails and I have figured out by their content that they are actually the exact opposite personality type than me. So what I've decided is I'm not going to write as me. I'm gonna write my email. And then I what I do is I tell uh Claude, rewrite this as a this personality type. And so it rewrites it.
There's more story and more fluff and more like, you know, the opposite of how I would write. And that's actually the email that I send to them because it's in their what I call personality language. And so it's going to be easier for them to read than if I send my version of it.
Now, I've seen that like I'm the person who might like write two or three very warm, very friendly sentences and then the person will write back and be like yes or no. And I'm like, this is a wonderful answer person. I'm like, all right, I don't need to be warm and kind.
They are missing that whole thing. So, >> I love that because that even in your writing and in your emails, that's helping you build rapport because you're speaking in their language.
>> Right. Right. and they're going to understand it and they're going to read more of it and, you know, they're going to take away more of it than than if I sent it in in mine.
>> No, I love that. I love that. What I was going to say before is when you've got those two reading styles, if you would writing styles, you know, the more serious professional business version as opposed to the more casual this is my life diary version, um the conversational version, would you, if you were making your content, just make sure that your um whatever you're using, whether it's Claude or Chat, would you make sure that that's mixing through a combination of both in your content moving forward so both sides of you are coming out or would you get it to try and find a middle ground?
I I would try to get to find a middle ground. Um if if you don't want a professional voice and a personal voice for two different uh I was kind of thinking two different two different bodies of content you're writing. Um, if if you're just writing for one, then I think I think you probably what I would do is I would give I would have Claude analyze the content, both types of content. Describe it in detail and then throw it together and tell Claude, I want something in between. Interview me on how to do that.
Ask me questions about how you I want my personal slash my personal hybrid voice to sound. That's that's how I would do that.
>> Yep. And then that way your brand voice is always consistent, >> right?
>> Yeah. Cool.
So this is the this is the markdown file that is created from the research and uh just goes through and this is how this is what's going to direct the skill on how to um to write. So here's an interesting one. Patterns to avoid.
Great question. I hope this helps.
That's really interesting point. That makes me want to gag. Yes, please avoid those phrases. Um, so this is this is me in a nutshell as far as communication.
And so we're going to we're going to go another step to create the skill. But if you're using claude code, what you can do is you can take this version and go put it into your project or global claude.md file. And that way this will this will be used. Now, keep in mind when you put it into your cloud.md file, excuse me, that's going to use tokens every time it retrieves it. So, you probably want to shrink this down and just get the most critical information to put in there. And that way when when you when you go back to your project or when if you use if you put in your global claude file, then every chat will include this and you won't have to call that skill.
>> Cool. Now, >> all right. So I want to uh or rename it calibration.
Yeah. So this is a calibration test. You can ask questions. I'm going to rename it as.
And so this going to create a copy of it and then we'll have the opportunity to upload it to Claude.
Let's see. We'll write a calibration test.
Ah, sure.
So, for the people still using chat GPT, can they still do this process inside of chat?
>> So, this markdown file, you should be able to take that and put it into a um I don't know, our custom GPT still a thing? I haven't used it in a long time.
um you know the equivalent of a custom GPT put this file in there and then you know you may have to ask chat GPT we need to format this differently and that kind of thing but yes you could you could definitely use this data >> amazing >> all right let's see explain how what prompt should we give this um explain why the sky Eye is blue.
>> So starts out with wavelength and visible light. So I've got I'm very sciency in my personality when the Earth hits atmosphere.
Yeah. So this is a very sciency uh explanation. Uh even the uh the nanometers that scatter roughly 5.5 times more red light. So what what my personality type I like data like show me data. Like I don't even know what 700 nanometers is, but I still want to see it. I want to see the data there, you know. And so uh this this this lands pretty good. It's not very long. Gets to the point. So why blue and not violet?
Two reasons. Boom. Boom. It answers the questions. Yeah, that's good. That's good.
>> Yeah.
>> So, say that started talking about how the sky is blue because the unicorns are up there running around and it's all really great with the fairies and something that's completely not you.
>> It's time to start over.
>> Go back to step one.
>> Okay. Maybe not completely not you. But even if there was little tweaks you want to make, do you just work with it to make those those tweaks in terms of a new prompt or how would you get it to just adjust that little bit?
>> Um, I would tell it uh, you know, this this calibration step is is for that to say, "Wow, something's something's wrong here. Where did you where did you get that?" Um, you know, maybe I pasted in a client's MyersBriggs by accident or something that that's unlikely to happen, but something weird happened.
Um, actually over the weekend, I I I was screencasting or screenshotting the example data. Uh, and when I went to have to get it ready this morning, the the data was different. I was like, what's going on? I looked and it was the example data I had put in over the weekend. So, I mean, things like that can happen. And so it it wasn't enough of a difference that I was like, "Holy cow, what's wrong with this?" Um, but it it would have skewed it in the wrong direction.
>> Amazing.
>> And I love that also you said, "Where did you get this from?" And I like, "Oh, that's super interesting. I would never have thought to ask that."
>> Proper skill.
>> Yep. So, it's going to package it into a skill.
So, the skills should have a couple documents inside. And if you're new to skills, um I'm just barely not new to skills. So, uh something I learned that a skill is actually um when we look at this, we're going to see a skill. We're going to see two files inside the skill.
And I was concerned that I had to download three files. But a skill is essentially a zip file that contains those other two files, which I had no idea. I asked Claude a few things to finally figure out that that was the case. So, uh, you only have to download manage one skill. Those other two skills are tucked inside.
>> Yeah. No, that's great information.
Thank you. When you said skills, I thought you were more meaning relating to you as a human and your personality.
So, that really helped clear that up.
Thank you.
>> Yeah, this is um Let me get this saved.
There we go. Okay.
So, right here I can copy to my skills and we'll go look at where the skills are in Claude in a second. Um but this is my skill. If I click here, let me look at it.
I get all right. It's not going to let me look at it. It has let me look at it in the past. So, I'm going to copy to my skills.
It doesn't like that. Let's go exit.
How do I get There it is. There's the X to get out of there. All right, let's go down here to the uh your your profile image and go to settings >> and then skills are in capabilities.
Now, interestingly enough, the client that I was doing this with um this morning, his capabilities page and skills page looked entirely different.
We were kind of fumbling around to find it. It was that different. So, I don't know if he's on a newer version or if I'm on a newer version, but uh this is this is where those live. Actually, this is what his look like. There they are.
Okay. This is what it look like.
>> All right. So, so here is here's that that the the uh starter skill. And I'm not seeing it didn't it for sure didn't download it. So, I'm going to stay right here and I want to go in my browser to retrieve that.
>> Yeah.
>> The the fun that happens on live um live calls.
>> No, this is good though because this could happen to us. So, you know, it's great because what if it didn't turn up for us when we were trying it, >> right?
Um, see chats.
Okay, so here's a funny thing. Since I was in the incognito chat, none of that chat is available to me, which makes entire entirely perfect sense. So, it appears that I have lost my skill. So, all that to say, um, if once you have your skills, so I'm going to re-upload mine. Uh, so again, this is this is the starting one. Personality adaptive communication. This is what that's going to be called when you download it um from from the blog blog site that I'll provide. Um, so go right here and we're going to upload a skill.
Click here.
and skills. And they're all spot again.
Nope, that's not the right one.
All right. So, I'm just going to go ahead and up upload and replace this. Uh, it's going to overwrite this one, which is fine. Um, but my my skill would be saved to my machine somewhere. I'll have to go find it. Um, and you would just simply upload it here. Hit the plus and upload it. Uh, if you're not on an incognito window, you'll get that button that says install skill. Just click that and it'll actually put it right in here. Uh, and like I was talking about, here's here's the skill. Here's that markdown.
And then here are your references.
Personality profile.
And and this one you can see this one this is the original. So it says to be completed yours will have data here uh the percentages >> and uh in the framework this is kind of a standard framework it it hangs on to so that's how you create that skill and um pretty easy I try to make it easy just to walk through step by step >> so that that you can build it.
>> So then yeah from there what do we do?
We do we just open new project and get it to write things.
>> That's right. Just open new project and uh uh tell it in in regular claude you will need to call the skill. Um and again we talked about uh claude code putting it into either the global or the project claude file. Um and then you just call it you start start typing asking it to build things.
Can we do a test? Like, do you need to say use the skill or does it just do it automatically?
>> Uh, you do need to call the skill.
>> Okay, >> give me a second. I can find mine.
Sorry. Hope you're going to post this.
>> No, this is great cuz otherwise we won't know how to do it ourselves. So, I love that you're walking us through all this.
>> All right.
Actually, I think I know where it is.
So, what applications would you use this for? So, you use it for content. I feel like you're using it for email.
>> Um, use it for if you had to write a report for someone. Um, what sort of applications can people use it in the real world?
uh you know if if I would say a good application would be um proposals get you know getting business um there's there's probably enough information out there on uh different people to to get an idea I mean if if you get it close like for example if you're as introverted as I am and if I can shift that voice close to a you high high energy CEO that's loud and like that's borous and like you know obviously that's not my personality type. I could at least get it closer to that person's type so when I write it it's bold and you know it it sounds like something that he would um resonate with.
>> Yeah. Um I've and I've I've not I've not done that well in the past and you know I've gotten feedback saying well like hell is numbers and facts and figures but like you went right to the solution and you didn't talk about the problem and how to solve it and I was like well yeah >> now I know I'm like well that's how I'm wired and that's that's my that's my default. So, um, you know, and and having Claude helps me kind of get around that now.
>> Um, but but definitely, you know, even if you can get closer to that person's personality type, there's even a a Chrome extension that works with LinkedIn and u it's got free credits, so you can go to somebody's LinkedIn profile and click it and it will estimate their MyersBriggs personality type. So that's another way to get it, too.
>> Yeah, that's super cool. Thank you for sharing that. Love it.
>> All right, so I found my my So my ISTP personality.
>> Yeah.
>> So to get to this, I'm going to say um open ISTP skill.
I'm just going to look through and see what that's all about.
Cool.
And then would you use this for any content you're writing? If you're say doing a Facebook um post or even LinkedIn, would you use this for all content?
>> So I I purposefully skewed to more of a an extrovert uh profile on LinkedIn.
um >> just because you know there's a lot of I I'm trying to reach the business owners and the CEOs and the executives and the founders and typically um you know in my understanding is they are they are not built like me. Um I am I'm not the innovator in the go create a new thing uh something big guy. Uh so I will I will skew that direction. Uh, so it'll take the ideas I have and it'll write to to a uh at least at least to the point where the first letter is an E, which is extrovert.
>> Yeah. Okay, cool. So, how are you doing that? If you were to do that, would you be using this test as we're doing here now? Like this skill, I mean, as we're doing now, then you would give it the prompt for the content you were right, and then would you say, can you put this into more of a um this type of personality or how are you doing that?
So it it depends uh how much I know. Um so the the the guy I was talking about earlier, he I I guessed his personality type as an ENFJ and he went and took the test and he he when he saw me the next time, he said, "I can't believe it. Your AI was exactly right." So in that case, I know that's his his MyersBriggs letters. So I would come here. I'd probably build a profile for him. Uh, if if if the frequency of messaging him warranted it. Um, if I just knew he was high level leader and very introverted, I would probably just give that to Claude. Um, >> yeah.
>> And maybe have Claude interview me, ask me some questions about this person to kind of nail it down a little bit. But it it wouldn't have to be as perfect as a personality profile. It just needs to skew away from from my personality uh to do that. So, and quite honestly, I've um I've had Claude in talking about personality is like, "Oh, do you want to take the MyersBriggs personality tape?"
And I was like, "What? Sure." And it was it was asking me questions and and got um pretty close to my I think it I think it got my four-letter uh MyersBriggs by asking Claude. So Claude knows a lot about personality and these tests as well.
>> I love that. Cool. Yeah, that's good to know. Um, and as well also I'm just thinking because we've got a skill in there for you. So if you're writing something, but you could potentially get it to build the skill of the highle CEO extrovert. Like you could get it to build that opposite personality so that you can just be like, "Okay, today I want to use my personality for this writing." And then like an hour later you're like, "Okay, I want to use my extroverted targeted personality." So you can just swap between those skill sets that are already in your cord. Yeah.
>> Yes. Yes.
>> Super cool.
>> A LinkedIn post in my INTP personality type about this virtual session.
So that's the that's the text I sent you before we head on here.
>> Yeah.
>> Okay. So it's also pulling in my my my humor profile is called caffeinated squirrel.
That was what cloud came up for me.
>> How did you pee with your humor profile?
I I had some things that I had written and I just described it and had it interview me.
>> Um yeah, so it's it's you know uh there's certain topics I don't want to be funny about, you know, uh death, pain, other people. Um there's uh like like I love puns. Uh I I'm happy with self-deprecating humor. So like make fun of myself in my own writing as Claus's writing about me, you know. Um so things like that. Um because I don't want to hurt anybody with my humor. It's just it's like let's make fun of me. I've got four kids and I realize that to those four people I'm probably the biggest dork in the world. So I'm just kind of leaning into that and and okay with that.
>> I love that. I'm a huge fan. I love dad jokes. Like love them. And so, yeah, I love that you're actually getting your Claude or your AI to actually sound like you in its humor as well. I think that's like something where I really miss a little bit because that fun, playful humor side doesn't come out. And I think that for a lot of people potentially who are watching this, they're also kind of thinking, how do I get it to like have humor? Like, how do I get it to be funny like I am in real life?
>> Yeah. And there's a um a LinkedIn guy I follow that he's he kind of got me on this. I mean, he's he's got two boys.
He's a total dork. And he every once in a while he will do a I'm a complete dork post on LinkedIn. Like he wants to keep he wants to be human on LinkedIn and not just be the professional LinkedIn expert, which is what he does. And John Nemo, that's his name. Um, and so, you know, he's he's upfront with that. When you get his ebook, like the the first picture of him is not a professional picture like this. It is he's got his kids have dressed him in something crazy. And, uh, that's the first picture you see of him in his ebook. So, uh, yeah, he he leans into that as well.
So, here uh, here's the post. I want the I want to try this again. Um, use my multi-STP personality skill to write this 100%.
No humor or other inputs.
>> I was really into the There we go. Matter of fact, voice matter of fact, experience. Experience anchored. Short declarative sentences, concrete nouns, active verbs, no filler, no humor, content creation guide. There we go.
>> Every AI tool has a default voice. The voice that writes all your content. your audience reads the tone the same for every business using the same tool. So that actually got that post got shorter if you notice when it rewrote it. Um >> which is indicative of my personality type as well.
>> Yeah. Okay. So I'm going to have to tell it don't say the 2,000 words I would say.
>> Make it shorter.
No, I love that. That's so cool that it even adjusts to how many words you would use because that is the thing like you know you are going to get the people who are doing the one or two word answers.
You are going to get the people who are writing >> two to 5,000word essays. So even if it can adjust to a middle ground on that sort of stuff >> is really really really helpful as well.
Like even if you can say to it, could you say to it like if you've got your ideal client profile for example, say your um business is structured around a certain type of person.
>> Um literally get it to kind of meet in the middle ground of you know your 100% but meet back at 50% of that where they are as well. So it's a bit more them, >> right? Yep. Cloud will definitely handle that.
>> Yeah, it's amazing. I mean AI in general I'm I'm I'm a cloud guy but you know AI in general can can definitely handle that kind of you know merging of of information. Um and again say ask me questions if you're not sure so I can say hey do you want to you want to use this phrase or this phrase you know and that two or three of those tells AI a lot about your intent.
>> Yeah. even down to like um can you ask it the question of what are the words that I should avoid because they're used in AI all the time so that you can make that list you mentioned at the start like can you literally just ask it about itself >> that we should avoid altogether when writing with AI Yeah.
Reflect. I've been reflecting deeply on.
Yes. Delve. Yep. That has to be on the list.
>> Yes. We'll love this.
>> If you're prepared to spend the time on it, I also love Claude. If you're prepared to spend the time on it, you really can get it to do what you need it to do. Like it doesn't have to just sound generic, but it's not going to do that automatically as to learn. You have to prompt it. Hey, what are the words that we shouldn't be using? Hey, let's not do the dashes and the and the things like that. What else do we kind of need to look out for if we're going to use Claude in this sort of way?
So, one thing I would say, um, so I I built a a project before I kind of before I started building these skills, but I built a project for writing on, um, Medium and I actually, funny story, I I published a couple uh, stories and this guy reached out to me and he was like a writer's coach and he said, "Hey, like you seem to be a a really good writer.
But your story, there's kind of like four points to your story. It's that it like it's not really focused. And uh I I have a writer's coaching thing you could join in. I think I could help you do that. And and I wrote him back and I said uh that was AI written. I didn't write it. And and I don't fess up to that to too many people, but he didn't reply. But I I I was like, "Okay, this writing coach read that and said this human person could use my help and I could help him uh and and reached out to me." So, first of all, it told me I >> I think I was on a pretty good track for writing human sounding stuff. The second thing that told me was when I built that profile to read that Claude never said, "Hey, this is really going to write like write stuff that's got four different purposes instead of one." And so what I did after getting that email is I went back to Claude and I said, "Hey, this profile you made me like here's what here's the email from the guy who said it." And and Claude goes, "You're absolutely right. We did not build anything in here to make you make these stories to have a one clear purpose. And so I tell that story to say there may be things that you don't know you don't know when you build these prompts um that you know I built prompt how to write and I presumed that Claude was going to create me something that had a single purpose. I'm not good enough writer to have sat back and and evaluated and said, "Oh, this this meanders all over the place." I went, "Wow, great article I wrote or Claude wrote with me." And I posted it. So Claude's not always going to tell you the things that you probably need to know. Uh you just may have to learn those along the way. Um, could I have said to Claude, "Is there anything else about this that we could improve to make this the outcome of this a really well-written article? Maybe that would have caught it, but I didn't ask that question."
>> Uh, so I I I shored that up and cleaned it up and now now it has a purpose, a single purpose. Um, which makes me wonder what what else does my prompt not do that it should? But um it's better than it was.
>> I love this. Thank you for sharing because it's really is that question of you just don't know what you don't know.
And like what you said there of just asking it the question of what could we have done to make this better?
>> Cuz we don't even think of it, but maybe it will, you know, or just taking feedback that someone else gives you.
Maybe that's what's going to help you to sound even more natural or more real or more alive, >> right? and and question clo question the AI you're using whatever whichever one it is question it um if you read a a LinkedIn post that just jumps off the page copy and paste that thing and go over to claw and say why doesn't my why don't my post sound like this um you know and and have your AI tell you why and then work on your voice on how to make them if you're if that's what you're looking for um which I think we all all are striving for that and writing with AI. Uh you know I love how this pops off the page. What do we do to make mine pop off the page like this as well? And so you know I I think at least I was I think people that I that I meet and have calls with are afraid like they ask the question, they get the answer and they they walk away. like press into it and and I don't like this phrase or I don't like this or this is really good but this isn't. You know, press into the answers you get. Uh and that's how you get really good answers by doing that.
>> Yeah, I love that. It sounds like it's all just keep going, keep refining, keep doing it. and even your suggestion of like um copying in something that actually does sound really good and saying, "Hey, how like what what is it about this and how can we make my content more like this?" And I think you see that a lot with you see those people who are content creators and every single piece of their content just lands. And now we have a way where we can actually like emulate that but in our own words and in our own style.
>> Right. Right. Easier than ever before.
And uh AI today, whenever you're watching this, AI today is the worst it's ever going to be. So, you know, it's just going to keep growing and and continuing to improve.
>> It's crazy how exponential it is and how it's like it can be like a week and it's just improves so much. Time is going at such a different rate than it used to, I think, just because the speed of things like AI and technology are moving so fast. Maybe that's just cuz I'm in that world, though. Uh, is there anything else with this that we actually need to know or like any other like tricks or tips you've got for us today?
>> I don't think so. I think we've covered uh quite a bit and um just just keep uh keep using your AI and learning you know uh just for example the skills I when when we originally talked about this session I was going to build a project in Claude and uh the last week I said you know what it seems to be a skill. So that is just a a a an evolution of the AI where skills are becoming very powerful and and changing. So um the process of continuing to learn and the process of adjusting and doing things differently uh is not going to slow down. But I I think I think we've covered the nuts and bolts in this this today.
>> This has honestly been amazing. Like I feel like as a content creator, there's probably been I don't know at least eight or nine things that I wouldn't even think of. As someone who uses Claude like pretty much every single day. So like I can imagine some of the people at home who might be a little bit more new to this would be going I just need to like learn the whole session off the off by heart. Like there's just so much in here. So thank you so much Andy.
It's been amazing. Where do you find you?
Uh, so I I hang out um on YouTube. Uh, great content over there. Um, I spend a lot of my time on one-on-one calls, a lot like this, where we're building automations or AI tools. Uh, that's that's how I, uh, run my freelance practice. I like investing directly into people's lives. And, um, that's that's where I spend a lot of time and and on LinkedIn. So, you can find me any of those places.
I love that. Thank you so much. We will put links to all of that. So, everybody go and check out Andy because this has been amazing. And thank you so much for your time.
>> Thank you.
Related Videos
VALORANT's Latest 'Exclusive' Tier Bundle is Rough...
KangaValorant
17K views•2026-05-28
Flight Attendant Mocks Poor Looking Black Woman — Mid Air Announcement Exposes Her Real Power
SkyboundStories-b4r
184 views•2026-05-28
I FIXED My Friend’s Blown Turbo RX-8… Then Sold It
Cameron-RX8
134 views•2026-05-28
NewsWatch 12 at 5: Top Stories
NewsWatch12
1K views•2026-05-28
Simon Jordan & Danny Murphy deliver PREDICTIONS for Arsenal's Champions League FINAL with PSG
talkSPORTArsenal
6K views•2026-05-28
Botting is OUT OF CONTROL in Classic WoW (Again)...
SolheimGaming
108 views•2026-05-28
The "AI Job Apocalypse" is CANCELLED!
WesRoth
9K views•2026-05-28
STREET FIGHTER 6 - INGRID Story Walkthrough @ 4K 60ᶠᵖˢ ✔
RajmanGamingHD
12K views•2026-05-28











