“ Understanding what it actually means when you tell ChatGPT, “Hey, be my doctor.” What’s happening there is the language model is taking the meaning of like, you’re an expert doctor or, whatever that you’re saying there. And then imagine just like all the stars in the sky and it’s going out and it’s finding like where’s that group of like the galaxies of all the stars that are talking about like what an expert doctor has said, and then how can I tie all of those things to the meaning of what Joe or somebody else is asking me for. And then it takes all those things and transforms that into the response that it gives you.

So, you can use the right words to kind of put the AI into a space of its training data where it would have the best answers or information about that. But still, you don’t wanna just rely on that training data.” – Brian McAnulty

Joe Casabona: A few weeks ago, I ruffled some feathers on social media. I said, telling people you wrote your book with AI is like telling people you ran a marathon with your car. Wow! People really took umbrage with that statement. And while I stand by what I said, I thought it would be a good idea to talk to an AI expert. So I reached out to Brian McAnulty of Heights Platform to set me or set everyone else straight. And you know what he told me? We’re all doing it wrong. We shouldn’t use AI as a proxy for Google. We should use it as an actual assistant. And for that, we need to give it a ton of context.

Look for these top takeaways: Instead of thinking of AI like a tool, think of it like an instrument. You need to learn how to play it, personalize it to play your kind of music.

In the long run, you won’t be able to use AI to cheat because it’ll be obvious. What makes your content unique is your personal experiences. And since AI works on how you train it, you can have as many AI assistants as you want. You don’t need to just have one. And you shouldn’t just have one. You should have multiple specialized AI assistance. Bryan’s gonna tell us how we can do that in today’s episode. And be sure to stick around for the pro show.

Bryan and I spoke for like four hours on the day we recorded this because I also went on his podcast. And so, I’m gonna give you a little behind the scenes look on how that was, and my first experience with using SquadCast and how that went.

So if you want to become a member, you can sign up at [streamlined.fm/join] or if you’re listening to Apple Podcasts, you can just hit the subscribe button and become a member right in the app. But that’s it for this. Let’s get into the intro and then the interview.

Hey, everybody. And welcome to How I Built It, the podcast that helps busy solopreneurs and creators grow their business without spending too much time on it. I’m your host, Joe Casabona. And each week, I bring you interviews and case studies on how to build a better business through smarter processes, time management, and effective content creation. It’s like getting free coaching calls from successful solopreneurs.

By the end of each episode, you’ll have 1-3 takeaways you can implement today to stop spending time in your business, and more time on your business or with your friends, your family, reading, or however you choose to spend your free time.

All right. I am here with Brian McAnulty. He is the founder of Heights Platform, and we’re gonna talk about AI today. So I’m pretty psyched about it. Bryan, let’s come out. First of all, how are you today? Let’s exchange nice things first. How are you doing?

Bryan McAnulty: I’m doing great, thanks.

Joe Casabona: Fantastic. So let’s dive right into it.

I don’t think we need to define AI. We don’t need to tell people what ChatGPT is. Let’s just start with this question: What is an effective way solopreneurs can use AI?

Bryan McAnulty: Yeah, it’s a great question. So, from what I’ve seen working with AI, the average person is using it the wrong way, and they’re treating AI like a source of knowledge rather than using it as a tool to kind of augment your abilities and help you make decisions.

And what I mean by that is if you’ve used AI, you’ve probably experienced something where you ask her the question, it gives you this answer. It looks plausible. It looks believable, but it turns out it’s actually completely wrong. And so if you’re treating AI like Google, you have the scenario where either, one, you end up with an answer that may not be correct, or number two, you end up with something if you’re asking it to do something creative. It could be doing that same exact creative output, what seems creative for everyone. So, it’s not actually really creative or unique.

And the better way to use AI is to give it the proper context about you and your business and have it answer from that. And there are research papers out there that show when you give AI all the context and say, “Hey, look at this, and then here’s the steps for you to evaluate it, for you to transform it. Give me information about it or feedback.” Then it’s able to use that and combine that in a way where it’s actually not going to make things up or at least very, very much reduce the chance of it making something.

Joe Casabona: Gotcha. I think that makes a lot of sense, right? It’s almost like asking some random person on the street if that mole on your skin is skin cancer versus asking like a dermatologist, right?

Bryan McAnulty: Exactly. Well, I think that, like the perfect analogy, most people try to use AI for something in their business, right? We’re talking to solopreneurs here, and it’s like going to somebody on the street and saying, “Hey, come in here. You’re gonna work in my office next to me, and now go and do this.”

And if you think about it, that is completely different from you hiring somebody, them knowing what your company is, them knowing what the role is. And now you say, “Hey, do this task.” There’s all this context that they already have about you behind there.

Whereas ChatGPT, it is this general tool. It’s almost interesting to think of it as an instrument, like a music instrument rather than like an AI or a computer. And you have to understand how to kind of play it and how to use it the right way. And so, if you’re going to get ChatGPT, it doesn’t know if it’s supposed to be a your like content reviewer or your doctor or whatever that is. You have to be very explicit in explaining, “Hey, this is me. This is everything about me. This is everything about my problem, and this is what I want you to do. This is everything about how I want you to do it, and what I expect you to give back to myself.”

So if you’re someone who has hired people before, delegated something to a virtual assistant or another teammate, and gone through the processes of like writing out a standard operating procedure or something for your teammate, then you’d probably be one step ahead of everybody else as far as how to better work with AI because it’s really the same kind of thing that you want to write all that out in order to get AI to perform best for you.

Joe Casabona: I think that’s a really good, no pun intended, a really good context to put it in, right? Because as you’re talking about that, I think back to when I hired my first VA and I told her to just go off and do something and the results weren’t very good, right? But then I recorded myself doing that task and I gave her that and I said, watch this video, write out the steps, now do the task. And tho like that she’s really effective at doing those things. And now she understands a few years on how my business works.

So I’m curious about this context setting because I caused a little bit of a kerfuffle on the internet. Imagine that because I said something to the effect of…But basically said, had two different versions of this depending on the social media platform I was posting on. But I basically said telling ChatGPT to act like a doctor with 20 years experience makes it as much a doctor as telling George Clooney to act like a doctor makes him a doctor. And I think on maybe LinkedIn, I went on to say this just kind of feels like a cheesy thing that “prompt engineers tell you to do to make them look smart” And so, I get the context part, right? You want to provide as much context to AI as possible, but like telling it to act like a doctor, is that the best way to provide that context?

Bryan McAnulty: So, yeah. There’s parts of this that work and there’s parts of this that are helpful and there’s parts of this that you might be missing if that’s all you say to it.

So the way that a lot of people use AI is they’re relying on the training data that is there and the knowledge that it already has. And until yesterday, open AI and ChatGPT, it was, it had this cutoff for most people of like 2021 for its training data. So it was a couple years behind the present where we are. Now, that has been updated now until April, 2023. So now we’re in the same year at least.

Joe Casabona: Wow!

Bryan McAnulty: Which that’s great.

Joe Casabona: And this is for like 3.5 and four…

Bryan McAnulty: but, still…

Joe Casabona: I guess, let me bang in here real quick and say like, we’re timestamping this pretty hard, which I think we need to do with AI, right? So we are recording this in November, 2023.

Bryan McAnulty: Yep. Yeah. Right after open AI Dev Day.

Joe Casabona: Yes. So, oh! So that’s awesome.

Bryan McAnulty: Yeah. So, Yeah. So now it’s more relevant like current events, things like that. And now, like you can use the browser tool where it can go and look up something to get additional context. And so, but that additional context and things like that, that’s important. You want to, you wanna give it as much and not rely on that being in the training data because you don’t know what is in the training data. And that’s one of the ways that people say like an LLM and ChatGPT. It’s like a black box that you don’t know what it’s getting at and how it’s getting to that.

But, essentially to distill this for the non-technical folks in a way of understanding what it actually means when you tell ChatGPT, “Hey, be my doctor.” What’s happening there is the language model is taking the meaning of like, you’re an expert doctor or whatever that you’re saying there. And then imagine just like all the stars in the sky and it’s going out and it’s finding like where that group of like the galaxies of all the stars that are talking about like what an expert doctor has said. And then how can I tie all of those things to the meaning of what Joe or somebody else is asking me for? And then it takes all those things and transforms that into the response that it gives you.

So, you can use the right words to kind of put the AI into a space of its training data where it would have the best answers or information about that. But still, you don’t want to just rely on that training data. So, if you are actually working with your doctor and like you have a report from them, if you have like history and documents, you would want to give all of that to ChatGPT. Say, “Now, you’re an expert doctor.” And then you would wanna do other things like explain exactly what you want from it. And so maybe that is not just saying like, tell me, diagnose me, but maybe it’s saying things like what are possible, like ways that you could diagnose this, what are like pros and cons of this different thing and really describing exactly the format of what you expect to get back from it, and then it can do that, and then taking that. And then comparing that even with maybe a different prompt you try or something else.

But in general, yeah, you’re gonna get a way better result by giving it your actual data and information. So like if you had all, if your doctor could really go and give it all of the reports of all of the information, be certain that it knew about all those things, then it would perform a lot better.

Joe Casabona: Yes. And if you listen carefully, you just heard the cries of thousands of data security experts. But that was a little bit of sarcasm. But I think you’re right. You need to give it as much data as possible, right? To give you a good answer. So like when I made my flippant tweet or whatever, right? About acting like a doctor, like someone argued back with me and said, “Yeah. Well the difference is that ChatGPT has access to thousands and tens of thousands of medical journals, and George Clooney only has a script.” And I’m like, But ChatGPT doesn’t have everything. It needs to make it a doctor, like we shouldn’t be treating it like a doctor. And so, I like what you’re saying. Yeah.

Bryan McAnulty: I mean also like what you could say, so, like a reason, like for your case is that it’s pretty good to understand what somebody means, but you might not be an expert at prompting it, knowing how to get exactly to those points of information that it has. And so if you say, Hey, you’re a doctor, but, yeah. Whatever. And then this other text in there that could have it find some information about pretending to be a doctor, like somebody might be on some forum post somewhere on the internet, which would, in that case be completely not helpful and potentially even harmful because it’s just talking about something that somebody else made up who was not really a doctor, right?

Joe Casabona: Yeah. That’s a really good point, right? Or it could be like WebMD where literally everything is cancer. (I’m sorry, I brought up cancer twice now. I don’t know why.) But I like what you’re saying though, because while my comment was glib, I think I agree with the, tell it as much as possible for what you’re trying to do, right? You can’t just be like, act like a doctor. My foot hurts. What’s wrong? Right? Because it’ll be like, oh well, maybe you twisted your ankle, or maybe you have gangrene, right? Like, maybe it’s one of those. So, I like this a lot.

So let’s say our solopreneur now understands that they need to give it the proper context, and they need to use it almost like an instrument. Before I get into what I think is gonna be a very inflammatory question for you, what is, maybe give me a couple of use cases, like what’s one way that you, outside of Heights Platform, right? What’s one way that you, Bryan, use AI on a regular basis?

Bryan McAnulty: Sure. So, ways that, I mean inside the Heights Platform is definitely very context aware of everything. So, it’s never like out of just asking ChatGPT about what it knows. but for myself, there are a couple different ways I use it.

One of the ways that I provided context really easily, I think that’s gonna change with some of the recent open AI announcements. But until now, and I think what will still be powerful for people is there is a free note taking apps called Obsidian, obsidian.md. And this is like one of those second brain type tools, kind of similar to something like Notion. it works for Mac and Windows, mobile devices and it’s free to download. And then it has this plugin store of like community plugins. And there’s two plugins in particular that are really, really powerful. One is called Smart Connections and the other is called Co-Pilot. And these are both free.

But you can plug in an open AI API key, or even some of them work with like, if you wanted to go through the trouble of setting up like your own local LLM AI model on your computer directly if you’ve got a computer that could do that. And then what happens is that can take all of your notes. And then be able to use those as context in a conversation.

So for me personally, I have things like my podcast. I put all the transcripts in there. I separated it actually into all the things that only I said. So I’ve got everything I’ve ever said there, and I can ask things like what’s something surprising that I’ve said, like on my podcast, and it’ll actually be able to find that for me as opposed to me like telling an assistant or something, can you go through, like, read all of our transcripts and like look for something surprising.

And so things like that. Picking out pieces of content. Understanding and connecting to things that you’ve written or talked about in the past is really powerful because a lot of these, like note taking apps, you would have to go and like have this whole time consuming process of tagging everything and saying, oh, this note’s related to this note…

Joe Casabona: and making your knowledge graph or whatever. That’s like their big thing. Yeah.

Bryan McAnulty: Exactly. So Smart Connections, that plugin can do that for you. So now when you open the note, you’ll see all these notes on the sidebar and it’ll tell you that they’re related. They’ll even show like a percentage of like how related they are.

And this is cool because before AI, you had like keyword based search where you could type a keyword to say like, where did I say like this word? I’m looking for this note of this thing, but how many times have you come across this point where you’re thinking of something that you wrote or you said, you know, you wrote something about this, but you have the wrong word in your mind. And so you can’t find it in search.

And the benefit of AI is it’s not using the word. It’s using the meaning. And so it’s able to then find like, okay, that’s the word you’re saying. Well this is the note that has the similar meaning even though you actually used a different keyword. So it’s a much better way of searching your own content in that sense.

And then like for more, I guess like approachable uses of like, well, what’s a prompt of something that I can do or that I personally would use often? it’s things like about, I talked before about like having it help you to make decisions. So for me it’s, I really believe that many of the things that you do as a creator, as an entrepreneur, it’s not the actual time of like creating that work or outputting it that Is taking the time of the whole span of what you’re trying to create. It’s the time in between of the actual creation and the implementation where you’re trying to make a decision about how to do something. So how many times have you been sitting there trying to write something, trying to build something for your business?

And the first step is you’re trying to think about like, well, what do I even do? What, where do I start? Then you start doing something and then you make this big burst of progress, and then you get hung up on this other decision. And you’re sitting there thinking for a week, well, what’s the best way forward with this next thing?

And what I use ChatGPT for is to help me make those decisions faster. So, telling it like, here’s where I’m at. This is the problem that I have. These are the choices that I could make. And show me like a table of pros and cons. Show me like all the possibilities. And then tell me which decision would you make? Like which way do you think would be better? And almost using it as like this mentor that’s able to look over and review what you’re doing to be able to say like, Hey, here’s this possible option. And imagine basically you have somebody you could always talk to like pitch these different ideas to.

Joe Casabona: Yeah. I like that one a lot. I’ve used that similarly for stuff around my podcast, right? Like, I’m thinking about doing an episode on this. What angles should I cover or, you know, even like, a listener avatar, like, you know, my podcast is about this. Here’s my expertise. Who should I make a really good podcast for? Or something like that, right? Or like, who can I make a podcast for that really benefits? Yeah. Stuff like that is really cool. Yeah. I have two follow up questions.

Bryan McAnulty: Yeah. And you could do all of it yourself too, but the idea is now with AI, you can just get that instantly and you can get every possibility instead of having to think of all of it.

Joe Casabona: Right. Like ChatGPT invented a real person, it seems like for me, right? Like Jenny’s a single mom to A 9-year-old daughter. She works full-time and wants to spend as much time with her family in the evening, but also wants to have a podcast. I’m like, dang, that’s like exactly who I’m trying to talk to. Like, so it was shockingly good. And then this is like ChatGPT four. so just really good stuff. okay. Two follow up questions. Do you listen to the podcast Cortex?

Bryan McAnulty: No, I don’t.

Joe Casabona: So, really good podcast, I think is by CGP Grey and Mike Curley, but they, I think they mentioned Smart Connections, like Grey is like all in new obsidian.

Anyway, quick side quest there, and then I want to, (cause you mentioned Heights Platform), this is, your company slash platform. You mentioned it is context aware. I wanna ask a question around like all of these kind of bespoke services that feel, I’m not saying yours is ’cause I don’t know. I didn’t ask you ahead of time. They feel like ChatGPT wrappers almost where the magic is in the prompt, right? So like Capsho, they have their own AI services for podcasting and the magic is in the prompt. Is that kind of how Heights platform works and, is that how you think a lot of these other tools that are context aware work so that their end users don’t necessarily need to provide a ton of context?

Bryan McAnulty: Okay. No and Yes. So, the yeah, there are, for the second part, there are a lot of tools out there where basically it’s just a prompt that they are kind of running for you in order to provide some kind of information a little bit better. But there’s some cases where there’s really no reason that you couldn’t just take whatever prompt they have internally and put that into ChatGPT, and then get right the same result. So, there’s value there, but It’s minimal compared to what you can do is something more customized.

And so the way I like to think about using AI as an entrepreneur is instead of the knowledge as we talked about, instead of using it for its own knowledge, thinking of it as building your own AI team. And so with Heights AI, what we did is thinking like, okay, well a solopreneur who’s building an online course, a community probably, the first two people that you would wanna hire is an assistant, like a VA and a coach. And so we thought well, why can’t, how can we build these things with AI to give every sole entrepreneur their own assistant and their own coach?

And what Heights AI? We have Heights AI Chat and Heights AI Coach. What they’re doing in the background for both of ’em is pretty different. But in the case of Heights AI Chat, that is like a chat bot that understands the knowledge of Heights. And so that part is pretty, I guess like people have heard of this kind of thing before, so like if you can ask it a question about where do I find like the pricing settings for my course, and it’ll be able to tell you all that because it knows all the information from our knowledge base. And so, and this is a case where like we don’t know what ChetGT knows about Heights, but we’re providing it directly, our up to date knowledge base, so it can use that to formulate an answer.

The next part is we’ve also trained it on all of the marketing information that we have shared or written about, or talked about. So you can ask a marketing question, say like, okay, well now, what is the best pricing for my kind of course? Like, what’s a common business model for a fitness course or what I’m doing here? And it will be able to look through our content for that and give us, give you an answer that would’ve been an answer we would’ve given you versus what ChatGPT would say.

So that’s the one part. And that’s pretty simple I think, to understand ’cause there’s a lot of tools that do something like that where they’re for customer support or therefore chatting with your own data with a PDF or a YouTube video.

The other part is it’s acting actually as an assistant by having the context of all of the courses that you have, and all the community and products inside your Heights Platform account. So you can say something like, take my guitar course and set it to published. Have it launched 30 days from now. And then make the price $199, and it’s gonna be able to know by guitar course. Okay. You mean this guitar course here that’s talking about mastering rock guitar in 30 days? And then it’ll show you that and say, okay, here’s what I’m gonna do. And then you just have to click okay or reject to have it actually make those changes for you.

And this is where it gets really powerful because now, instead of it just telling you like, here’s where you can find the buttons to do that, yu don’t even have to learn the software anymore because I’d say I can actually go and make the changes for you.

Joe Casabona: That’s pretty awesome. I like that a lot. And so, well, I wanna ask my next question, but I do wanna take a break for our sponsors. So let’s do that and then we’ll get into the main event.

All right. We’re back. And I wanna ask what I believe in my core to be true in some cases, but not all. Isn’t using AI cheating? And, I don’t want this to sound like using it for everything, but I’ve seen people say, oh yeah. You know, I use AI to write 20,000 words for me every week. Or I wrote my book, like, AI helped me write my book. And I’m like, well, did you really write your book? if AI wrote your book, right? Isn’t that like kind of like saying I used my car to run a marathon? So, I’d love your take on that.

Bryan McAnulty: Sure. So I think personally that in the long term you actually can’t cheat with AI right now. And the reason for that is if you say, “okay. I’m going to make ChatGPT go and write this entire book for me, or write this entire script for my course for me, then the problem with that is that anybody else could do exactly that same thing. And what’s gonna be the difference in their output versus yours? Yes, ChatGPT says something slightly different every time, but for the most part, it’s really gonna be the same. And as a creator, as a solopreneur, especially as a course creator, most of the value of what you have to offer is not just from like assembling, “Hey, here’s the steps of what you have to do”, but it’s your unique message and your unique experiences that go into that. And if AI is making it, then that’s just simply not there.

So, in the long term, like people are not going to really care or want to read anything that’s completely AI generated when there’s no context of why it was even generating that. And like I’m sure everybody listening to this can point to some place where like you’ve seen on social media, there’s some kind of reply and you look at it and you’re like, this is an AI generated reply. And for me it’s like, I don’t understand why somebody would go through the trouble even to have an AI reply like that because it has no connection to what your thoughts were about, the actual post or topic, anyway. And so if you think people are gonna follow you because they see that and they see your profile, you are really mistaken I think in having the wrong way to think about that. Because if it’s just this generic reply that says,”Oh, that’s so cool!”, it has million emojis after it, like, I can’t even think of the way ChatGPT says it ’cause it’s really bad, like how these, some of these replies…

Joe Casabona: it’s, I should have blocked a guy who responded to all of my posts on LinkedIn with an AI comment. And it was so obvious because it was like if I said, you know, let’s take the doctor post, right? He’d be like, “Great point, Joe”, telling ChatGPT to act like a doctor really is like telling George Clooney to act like a doctor. And I’m like, yes, that is what I said.

Bryan McAnulty: Yeah. Just summarizes what you said in like one sentence with emojis and hashtags added to it, right?

Joe Casabona: Right, like, it’s so obvious. Yeah.

Bryan McAnulty: Yeah. So on that side, I would say like people will find out quickly and realize that this is not a long-term strategy of how you should be using AI. On the other side of things, what is so incredible about how these language models currently work is everybody has access to literally the top state of the art AI that is available right now with GPT-4. So it’s not like, oh, well I wish I was Apple or some giant company to be able to access this and or you have to invest so much money to be able to use it a little bit. It costs almost nothing like it’s $20 a month for a ChatGPT plus subscription. And then you can use that to do so many things for you, like I gave the example before of the like helping to make you make decisions faster for like the content creation side.

I’ve also experimented with things like having it like review a video that I’ve done for YouTube where you take the whole script. And I took a script from Alex Hormozi of a video that I saw he did, that I knew was really, really good. Got a ton of views. I asked ChatGPT, like go through this, lLike tell me about the points that were like helping to build suspense or keep the person entertained. And then I said, “Now, rate each of those from like a 1-10 scale. And it doesn’t really matter how good ChatGPT is, at necessarily rating those things.

But then you could give it your own video script and say, “Okay. Now, let’s do the same thing. Now, rate mine on a 1-10 scale.” And so again, if it knows where to give the perfect 10 or not, that’s not really important, but what’s important is the difference. And I could see my script was clearly not rated as well as Alex Oz’s. And so now I can compare that and understand and get this third party review of knowing what I could potentially improve instead of saying like, “Oh man! Where do I find somebody to like review my YouTube content to go and hire to do that?” You can get some kind of feedback basically instantly.

Joe Casabona: Yeah. It’s almost like a flashlight, right? Like you’re looking for something in the dark. The flashlight’s not gonna find it for you, but it’s going to illuminate certain areas where you can potentially find the thing you’re looking for.

Bryan McAnulty: Nice. Yep.

Joe Casabona: I also want to throw in here that, you know, you mentioned like unique experience. What makes your work valuable?

Bryan and I have done a podcast swap, and so I would strongly recommend you check out, Bryan’s podcast, the Creator’s Adventure because I talk all about storytelling, and why creators need to tell stories. And so I think this slots in very nicely with that.

But, you know, I made this point about like telling ChatGPT, like saying like, “Oh! ChayGPT wrote my book.” I got into a spirited debate with a friend of mine, and then another dude chimed in and he said, “If my readers can’t tell the difference between what I wrote and what an AI wrote, why am I gonna waste my time?” And I said, “Brother, that’s a bigger reflection on the writer than it is on the reader. Like, sorry to tell you, it means that you’re a bad writer.”

Bryan McAnulty: Yeah. Right.

Joe Casabona: Like, if your readers can’t tell the difference, you don’t do anything to differentiate your writing. So, but I like this. And again, this kind of, you know, I kind of like set you up with an inflammatory question here because we’ve been talking about how the average person is using ChatGPT or AI in general, wrong and they’re using it as a source of knowledge. They’re not necessarily giving it the right context. And in this case, they’re basically saying like, oh, well ChatGPT can just write the same words I wrote or I would write but it doesn’t, right? It has the information, and it’s up to you, a human being to structure that information in a way that works well for other human beings. And that’s just not something that AI can do. And I mean, I don’t wanna say can, won’t ever be able to do, but it certainly won’t have your lived experiences.

Bryan McAnulty: Yeah. And at the very least, even if we say like, okay, well maybe it can do some creative things. It can’t do it the way you want without your direction. And so at the very least, you have to act as a director in giving it the right way to do things in the, and as we talked about before, like defining the processes of what are the guidelines of things that are important to you like even with, I’d say iChat that we talked about, like it can be kind of like a support for our replacing part of our support team.

But first of all, it hasn’t, we haven’t fired anyone on our support team. The roles have just kinda shifted. Now, we’re looking at like feedback that comes in of negative feedback if the AI got something wrong or couldn’t answer something. And then our support team is helping to make, I’d say AI better in improving all that kind of context in the background. And it let us focus on like more important questions that the AI wouldn’t be able to answer instead of like, where do I find this button that we’ve answered a million times before? But what I’m getting at is in that case, it’s not coming up with something new. It’s basically synthesizing the information to understand here’s where that information is, here’s how I can kind of transform that into giving you the answer from what is already there.

Joe Casabona: Yeah. I love that. I love the director analogy, right? And I think that like Hayden Christensen, I’m a huge Star Wars fan. Hayden Christensen, perfect example of this, right? Like people hated him in Star Wars episodes 2 and 3, but like George Lucas, for everything he’s great at is a terrible director. And he didn’t give Hayden very good direction. Now in the Obi-Wan series, right? And in the ASCA series, he got great direction and he delivered really impactful performances, I think. So, I really, I love that analogy. And then it also gave me the opportunity to mention Star Wars. So, all good there?

Bryan McAnulty: Yeah. I haven’t seen those yet.

Joe Casabona: Oh my gosh! Well, I don’t know what the statute of limitations on spoilers is, but if you haven’t seen this…

Bryan McAnulty: I saw the Mandalorian, but that was it.

Joe Casabona: Okay. Well, I think it was obvious that Hayden was gonna be in Obi-Wan.

Bryan McAnulty: Yeah, yeah. I’ve heard that.

Joe Casabona: Yeah. It probably was a surprise that he was in a Soca, but it was like an unavoidable surprise once it came out. So, I won’t say anything more about that. Sorry if I spoiled it.

But, so let’s move on to this. You’ve convinced me that using AI isn’t cheating. If you can believe it. I wanna move on as we close out here to, you mentioned Open AI’s Dev Day. They made some new announcements for us. And I think this kind of fits into the context of what you were talking about with the Heights Platform Coach specifically. So, first of all, maybe you could tell us a little bit about the Heights Platform Coach and how that works? And then what does open AI’s new announcements, new as this episode is being recorded, what do those mean for us?

Bryan McAnulty: Yeah. So what I was getting at before is the wrong way of using AI is using it for the knowledge it has currently. The right way is to think of it, especially as an entrepreneur, as like creating this AI team for yourself. And one of the things that kind of a limiting mindset people have is thinking like, “Oh, I have ChatGPT, I have this AI assistant. Well, you can have as many AI assistants as you want. It doesn’t have to be one. And so this, when you start to think that way, it opens up a real lot of possibility and realizing that there are these different roles and micro tasks in what you’re doing in your business that you could somehow involve AI in.

And right now, the current way that the models work is they perform great with these specific tasks. So, if you don’t already have like some operating procedures and things written out for your business, if you are a solopreneur or haven’t hired any contractors, it’s beneficial to start writing those things out and documenting them. And thinking about like, well, which things here like would the output just be, could be text that would provide me what I need. And so something like reviewing content that I mentioned before is a really powerful example.

And with Heights AI Coach, we took that kind of a step further. And Heights AI Chat, that’s your assistant. Heights AI coach is your coach. But the problem with going to ChatGPT and saying, “Hey, be my coach is similar to the doctor problem because you talk to and say, “Hey, be my coach.” But now you have to get it to ask you some questions or figure out how to help you. And oftentimes as the entrepreneur, you don’t know the question to ask to, even get that answer that you need.

And so what we did with Heights AI Coach is like Heights AI Chat, it knows about all the context of what you’re building inside our platform, the digital products, courses, community, everything that you have there. And then we have this onboarding process where you give it some information to say like well, this is my business model. This is like the size of my email list, my social media following, things that I wouldn’t be able to see like in your Heights platform account already. And then you have this introductory chat with it where instead of you asking it a question, it’s gonna actually ask you questions, and it’s gonna collect this information about you to then understand how it can help. And unlike ChatGPT where you have to keep saying something in order for it to do something for you, Heights AI coach is actually gonna run on its own every single week. It’s going to continue learning about what new things that you’re doing, and then it’s going to actually send you an email and say, “Hey, here’s some tasks for this week and here’s some recommendations” and Basically give you that every week to guide you towards whatever your specific goal is.

So, ChatGPT, you’ve gotta talk to it. Heights AI Coach, once you kind of tell it what’s going on, it can kind of learn on its own and then just help you on its own. So imagine like you get to wake up in the morning and then you get an email from your coach except, wow, it’s AI.

Joe Casabona: Wow. That’s amazing. And I don’t think I made this analogy during either of our recordings, but it makes me think of like, when you just say to ChachGPT like, “Hey, be my coach”. it’s kind of like hiring Lebron James to be your hitting coach in baseball, right? Like I’m sure Lebron James is probably plays a good baseball game compared to the average person ’cause he is athletic. but, yYou know, he is not an everyday major league baseball hitter, right? I’d rather have Aaron Judge be my hitting coach because he’s in that every day. He knows the context.

And so like having Heights AI Coach taking all of your con, like this is the important thing, right? This is like the overall lesson for anybody who is thinking about how they can leverage AI better is you can’t just like bang into ChatGPT once a day, ask it a bunch of things and bounce, right? Like you can, but the most effective way to do it is to let it know as much as you are comfortable. Let it know as much as you’re comfortable telling it about you, right? About how you work, what your days are like, your schedule and things like that. And then it has all of this context to work on.

And so with that, Byian, why don’t you tell us a little bit about open AI’s new announcements from Dev Day because I think these are really interesting.

Bryan McAnulty: Yeah. So this is really excited. I’ve been a bit overwhelmed in the past, 24 hours thinking about how we’re going to implement this and what it’s going to mean for entrepreneurs. And I was actually fortunate I got to hear about some of these things before they announced it the end of October because somehow somebody accidentally leaked the system prompt for some of this new feature into my ChatGPT account and I happened to see it.

Joe Casabona: That’s awesome.

Bryan McAnulty: So I already kind of knew what was coming here, but the idea is like it’s not quite going to be able to do like things like what Heights say, AI coach does ’cause it’s not gonna have that built in context of learning on its own and like running automatically. But we had to do quite a bit of work of managing what it’s going to know about and when in order to have it actually be valuable and functioning the way that it does. And like give also like the context of this is how we would coach the person.These are the things that we would do, and having it act like us instead of general ChatGPT.

And the problem with that is ’cause many of you guys, if you’ve used AI, you probably have gotten to the point where you realize ChatGPT is going to forget part of the conversation if you have this very long conversation or this very long document that you give it. And so these different tools have popped up where you’re able to have these different ways of retrieving that information to give to AI assistant or AI tool so it can know about your business and like the things that we’ve been mentioning throughout this episode.

Now, OpenAI has announced this custom feature called GPTs, where you build your own custom GPT assistant, which I love because it aligns exactly with the idea I’ve been talking about of like building out your own AI team. And before, the way that you did that is you had to come up with these different prompts and processes, but the limitation was you would still be using ChatGPT. If you weren’t a developer, you didn’t have a way to say like, oh, well, I wish, besides this prompt, it really knew about all like my transcripts, all of my blog posts, whatever it is that you’re working on.

And now, just using ChatGPT, you can go through a conversation with it to describe how to build this custom assistant. And you have the ability to actually go and upload files to give it the context. So you could say, here’s all my podcast transcripts, or whatever it is that you want to give it before you had to pay different AI services and developers. And there were so many pieces that had to go into creating something like that.

But OpenAI realized that like the idea of retrieving the knowledge, it’s so general that wouldn’t it be cool if somebody could do that on their own without being a developer. And I think if you’re listening to this now, it’s not gonna be right after OpenAI Dev Day anymore, but it’s still gonna be really early, and a lot of people are not thinking about this. So I’d encourage you guys to think about how you can start building your own kind of AI teammates to create these GPTs, not only for your business but also like what would people want in general, because they have like ChatGPT store that is coming out with this and you can actually build these things without a developer and then publish them in the store for other people to use. And there’s gonna be this revenue sharing model apparently that goes along with this. So this could actually be a business or marketing opportunity for you as well if you start to leverage this a little bit more.

Joe Casabona: Wow. That’s amazing. I feel like there’s a lot more we can talk about this, but we are coming up on time. So I’ll just ask you this, do you have any new ideas based on that already?

Bryan McAnulty: So biggest new thing for us is, the second part of that is now ChatGPT, the overall context by itself is gonna be 128,000 tokens. Before, the maximum was 32,000 if you were a developer and had access to that. The regular ChatGPT was 4,000 and so 128,000 I think that’s something like 400 pages at least of text.

Joe Casabona: Wow!

Bryan McAnulty: I did a test yesterday. I think it was more than 13, like 40 minute podcast episodes.

Joe Casabona: Wow!

Bryan McAnulty: My podcast that it, you can just paste in and it will know all of that at once. And so that’s really incredible. And I don’t know exactly how I’d use this for business yet, but now it can also speak and it can also see with the vision API.

So, just in the last 24 hours, I’ve seen some incredible examples of people saying like putting up a sports game. Having ChatGPT basically watch the video, and then speak as like a commentator for the game as if it was like the announcer. And it’s impressive, like within less than 24 hours here, how much people have built and how realistic it actually is. So, I’m really excited to see like what are the use cases that come up now that it can not only write, but, speak as well.

Joe Casabona: Yeah. That’s pretty wild. I am paying for ChatGPT-4 through Raycast AI which is, Raycast my favorite launcher but, it’s like limited, right? ’cause they’re talking to the API. This might make me upgrade to the actual ChatGPT-4. So, we’ll see.

But Bryan, this has been illuminating. So, thank you so much for your time. If people want to know more about you and, how to learn what you’re doing, where can they go?

Bryan McAnulty: Sure. So if you wanna check out what we’re doing at Heights Platform, we’ve got a free trial, no credit card required. You can try out Heights AI Chat and Heights AI Coach for 30 days in building your online courses and digital products. That’s just [heightsplatform.com].

And if you’re not quite ready to do that yet but, you are a solopreneur and creator, I would encourage you to go to [creatorclimb.com], and that is our private community. Over 7,000 creators in there. And that is built on Heights, on our community and course features there. So, you can check us out there. There’s other creators. There’s also our podcast is inside there as well, the Creators Adventure.

Joe Casabona: Awesome. I love that. We’ll include all of that in the show notes, which you can find over at

[howibuiltit/344]. Again, you can find all of the links we talked about over at [streamlined.fm/344].

Bryan, thanks so much for joining us today. I really appreciate it.

Bryan McAnulty: Yeah. Thanks so much, Joe.

Joe Casabona: Thanks to our sponsors. Thank you for listening. And until next time. Get out there and build something.