You Couldn't Do THIS With ChatGPT A Year Ago...

All right, in today's video we'll be looking at features of ChatGPT that weren't available a few months ago that I actually started using regularly. Namely, I'm talking about the Mentions feature and the History feature which I've been playing with more and more. And I've got to say I found some fantastic ways to implement it in your everyday ChatGPT workflows. Now, these features and techniques discussed today will help you get more out of pretty much every ChatGPT conversation. But at the end of the day, you will have to apply it to your very own context. I will give you some examples what I have done, I will give you some examples of what community members have done, plus I will provide you with a written step by step guide that you can follow to get the exact results I will get today yourself. But at the end of the day, this will be an instructional video that will show you new techniques and ways to use these brand new features that you might not have considered. And we'll be talking about strengths and weaknesses of these features. So with that being said, get ready to manipulate some context inside of ChatGPT, because that's what this is all about. But both Memory and Mentions are all about manipulating context, at least in this video. And as you might know, context is what shapes the quality of what you get from ChatGPT. So with that being said, let's start by talking about these features one by one, what exactly they do, and then we'll move on to talk about how I started to integrate these features into my workflow and why this is actually an improvement over a plain interface that doesn't have these. Alright, so first up, let's talk about the Mentions feature. This is very simply explained if you already understand GPTs, if you're completely new to ChatGPT or how GPTs work, you can check out some other videos on my channel. But basically the Mentions feature allows you to have one conversation and then bring a GPT with a different set of instructions and unique capabilities into that conversation seamlessly. This really easily allows you to mix and match different capabilities within one conversation. And as you know, some of the best results from ChatGPT come from giving it detailed context or having a longer conversation and then performing your final step as then the conversation contains all the info that is going to be helpful in getting you what you actually need. So this can easily be demonstrated with a conversation between two GPTs. In this case, I'm going to take the Innovator by Yvid Mollick, a popular AI Blogger and teacher. His very insightful post on a regular basis. And he built this innovator which basically takes a multi step approach to crafting ideas. It's not just give me 10 ideas, but it asks follow up questions and intelligent follow up questions. So let me just start the quick conversation here with the Innovator and as you can see it already points out here by, by using. I could link other GPTs, but we'll do that in a second. First we'll generate something useful with the Innovator here I'll just very simply state that I'm looking to help people learn practical AI skills and tools to boost their everyday productivity. Then the innovator comes back with some possibilities. These are clarification questions so it can give you a better answer. Essentially it's helping me to explore the context that might be relevant before it gives me some ideas here. I'll just say I like these questions, but make them more business specific as that is usually a use case that makes a lot of sense for most people, whether it's marketing, entrepreneurship or just getting ahead in your own career. I'll just say yes, these sound good. And here we're basically just using this GPT and using it how it's meant to be used. So the GPT does its thing here. It generates 10 ideas, then 10 more ideas by searching the web and then it puts all of that into one document. And here we're taking advantage of the instructions of this GPT and the web search and the code interpreter all at the same time. There you go. So next to the doc, I also have the ideas right here. And now we get to the interesting part part which is the fact that we can now use all of this information to go further and to turn this into output that actually matters to us. And this is a pattern, this is a formula that I've been using over and over again ever since the release of GPT4. And this mentions feature makes that really simple because otherwise you have to start with a certain prompt sequence and then you need to move into a different one and you need to bring it all together manually. And all I simply have to do here is talk with one GPT and once that conversation comes to its conclusion and I have everything I need, I can use the same conversation and the context window of 32,000 token tokens that is available in GPT4 to continue the conversation with, let's say in this example, the Copywriter GPT. This is just a popular GPT from the GPT Store. You could have your own writing GPTs, or as I'll talk about later, at this stage you can literally use any GPT that transforms it into a different format or that achieves some results. Because the formula that we're following here is essentially one, generate context and then use the mention feature to transform that context into into something else. Now, the context might be anything you desire, anything you're interested in, any interaction with ChatGPT, any GPT doesn't matter. And then the transformation step comes in the form of a second GPT. And this is where these GPTs become actually really useful. Because to transform text, you could keep it very simple and just say, hey, turn all this information into an essay, or you can go ahead and you could prompt it in a very specific way. Maybe the way you write your essay, maybe the way you write your newsletter is by specific to you. And if you're a power user of ChatGPT, you will know at this point what use cases you like and what prompts you like to use. You will have customized these to yourself over time. And this feature allows you to access all that in a heartbeat. And that's why I started using it more and more. So to continue with our example here, now I have these 20 ideas. And now I'm going to say write me website headlines for each one of those ideas. Now, this is a simple prompt, but I'm using all the instructions inside of the Copywriter GPT to run this. And it's important to realize that now it ignores all the instructions inside of the Innovator GPT. It takes all of this conversation in as context, but it doesn't use the instructions behind the Innovator GPT anymore. Because now I'm writing copy. And now the system prompts, the instructions of the Copywriter GPT are what matters now. And if you're anything like me, you're all about effectivity and productivity. And this just makes things so much simpler because you only have to solve a problem once you save the solution inside of a GPT and then you can call it anytime. But it's important to keep in mind that this is really a two step process as stated. First, you have to explore the context, and second, you transform all that information into your final form. And there you go. Now I have 20 headlines that are pretty well written because I've been using this Copywriter GPT. Now look, at the end of the day, you'll have to have your own GPTs or build your own GPTs to do these steps. But the point of this video is to become Aware of this very simple technique of simply calling one transformation GPT after gathering a bunch of context with another workflow. Okay, so simple idea, but very powerful. Now the question presents itself. Igor, what do you and your team use this for? What else could I be doing with this? Because this is pretty generic, right? You're just generating a bunch of ideas on a specific topic and then you're writing headlines on them. And the quality here might be higher than if you just prompted straight up. But what other use cases do I have here? How can I apply this to my life? Well, this is where me and the team went in and we actually created a step by step guide on how we use this ourselves and what notes we keep in mind, what while doing it. And while by watching this video, you already learned some of the underlying principles of this guide. This guide uses our specific context, how we use it to actually generate copy for our newsletter that we send out every single week. And we do this with one sales GPT that has all the context surrounding the AI Advantage community, and then a second GPT which is optimized to write the news segment for our newsletter. So the first GPT allows us to craft our context, craft the information that is valuable and comprehensive, and the second one packages it into a form that is a good starting point for a new segment. And we document every single step here with screenshots and all the exact prompts that we use along the way. And at the end of the guide, we have a table with a bunch of use cases that you could be using this for both the context component and the transformation component. We made this freely available in the public section of our community. I spent the last few months building this community and a lot of systems that produce guides and tutorials like this. And I plan on sharing way more with that, with the YouTube for free. So this is the beginning to that. And here are some use cases that you might want to consider for the context part. Well, first of all, you could generate ideas like I showed you here. But then also you could bring transcripts in from videos, podcasts, or any conversation that you record with your phone. Beyond that, if you have a product that you're selling, you could just copy paste the entire sales page and add it into ChatGPT. And that is your context that you can start prompting on top of and then mentioning another GPT to transform it. One more fantastic example is actually going ahead and inserting a scroll screenshot here. So I'll just take the Apple website and take multiple screenshots of this here. And I'll just upload these screenshots to the conversation and then I'll just add a little comment, send this to ChatGPT, and this is the first step and this is the context that I'm providing it with. Again, if you want a more extensive list, it's in the free guide. But trust me when I tell you, this formula of context plus transformation is just so powerful and so universal. This is one of my go to tools when using ChatGPT. But here's the thing, you can take it further than that. This is kind of just the beginning because context is such a wide term, right? You could give it pretty much any information, right? And what we did in the community is another cool thing that I would like to share with you, which is that we actually hosted a challenge in using this mentions feature because when it came out, we wanted to see how our community members use it themselves, how they can put it to work. So we issued this challenge where we said, okay, you're going to build two GPTs with the GPT builder. The simplest way to create a GPT, right? Literally just go to create and using their preset tool. Otherwise I would never recommend using it. But you can can really easily build basic profiles of personalities, right? So in the case of this challenge, we didn't even use this formula. We just crafted an interesting context by creating two GPTs which had two distinct personalities. You know, you could go ahead and craft a GPT for Taylor Swift and another one for Elon Musk and then have them talk to each other. Or you could maybe get a little more interesting with it and take Socrates and Dostoevsky and then have them talk to each other. Now, I think showing you all that goes beyond the scope of this video. But also that guide I'm going to include in the description of this video, you can check it out. Basically, our challenges are always these custom guides on how to do a new skill. And then we have our community do it and people can compare results. We host the event where we pick a winner and talk about what we learned. So if you want to learn this technique, you can check out the guide we had for the challenge. Basically, what you care about is the step by step guide part of the challenge. But what I do want to share with you is the winner of this challenge. And this is the type of learning that I could not have presented to you a few months ago or a year ago where it was just me figuring these things out and teaching you. Now, with the community, I can show you what Nyklass did with it. He was the winner of this challenge and he did something very interesting. He. He basically crafted two GPTs, one that represented Joe Rogan and one that represented Charlie Harper from Two and a Half Men.

I am living life to the fullest.

And then he let them discuss the challenge topic of which Jobs will AI not replace. And all of a sudden, Joe Rogan was debating Charlie Harper. Now, few things to be pointed out here that he actually did with it. One thing he did is he customized the GPT with this technique that I taught about a year ago on how to use a similar structure to what character AI uses for its character in inside of ChatGPT to give it a consistent personality. Now, this includes opening phrases and a longer description of the personality and the mannerism of the character. This goes deeper than what the GPT builder does for you. So this is one thing he did. I honestly think this is the most underappreciated video on the channel. But basically, if you're creating a character, this does a better job than ChatGPT ever does by itself. Then he used 11 labs to use the voice of Joe Rogan and Charlie Harper to generate those transcripts, and he turned it into a video podcast with this tool that I've never used before called Vidio. And that was his submission. Check it out.

All right. Diving straight into it. When it comes to Jobs AI won't replace. I think we're looking at a future where creativity, empathy, and human connection are the real MVPs. Think about it.

So, in the grand tradition of Charlie Harper, I say bring on the AI. Let it try to replicate the irreplaceable charm, wit, and downright human messiness we bring to the table. At the end of the day, the best things in life, like a perfectly mixed martini, the laughter of friends, and yes, even a little heartbreak, are beautifully, irrepressibly human. Cheers to that, my friends.

And that's just one fun possibility, right? But my point is this. You can adjust this to your context. I hope all of the ideas in the guide and some of these examples spark your imagination and you will be able to adapt this technique to your own life and your own workflow. Because really, mentioning a GPT saves you a lot of time, because all of the work that you put into crafting that one prompt or the one GPT can be accessed in a heartbeat. You just have to keep in mind the context plus transformation. So I have one more point that will help you before we round this out. This is a brand new memory feature, and I have to say I bashed it maybe a bit too harshly in last week's news video where I said it's essentially just a beginner feature that automatically gathers context for you. It's like automated custom instructions. I did point out the fact that it actually doesn't count towards your context window, which is a really good thing because if you go to a custom instructions, there's only so much you can include in here. And this brand new history function which essentially manages your context for you, it saves it in here. It doesn't take away from the limited context window that GPT4 has, by the way, just to reiterate, 32,000 tokens. So how does this help? Well, if we're doing the simple workflow of gathering context one way and then using a GPT to transform it into something else, having a history is just as having another message inside of this conversation. So it respects it as a part of the history that you established here. So this can make your context gathering the first step easier. And particularly I would recommend it if you're going to work on a single project over an extended amount of time. Okay, so if you're not just going to have one conversation that pertains to a certain topic, but maybe three conversations, maybe you know, you're going to be spending the next hour working on a project inside of ChatGPT, in that case, you can absolutely go ahead and save that information to your history so you don't have to keep adding it to the different conversations if you know it's going to be relevant to everything you do. So for example, here, if we're trying to do something with the information of these Macs, let's say all these details about these MacBooks are relevant to me, then what I can go ahead and do here is say, save all the details to memory. And here I'm combining multiple concepts, right? Because I'm using the visual input and the analysis extracted from it and saving that to my memory, which, if I now move on to the Innovator, GPT, can help me in my research and the crafting of my context, because what I save in memory stays consistent. It's like a calcified part of every conversation that is always there no matter what. But you do have to be careful because look, in this instance, for example, it says eager is interested in promoting the latest MacBook models featuring the M2 M3 M3 Pro. And that is not exactly what I meant here. I wanted it to be aware of the technical details. So you have to always double check and Maybe you have to change your prompts up a little bit so it saves the part that is actually relevant. As per usual, ChatGPT is a lot of trial and error, and it's no different in this case. And there you go. As you can see, now it's crafting that memory entry just like I wanted to, with all the little details embedded in the image. And these tokens don't count towards your context limits, so you can have even longer conversations when gener the context in the first step of this two step formula. All right, so my hope is that with this video, you're going to be able to adapt this framework to your very own use cases. I gave you a bunch of inspiration, a bunch of use cases, and the creative way somebody in the community used all of this to create a unique conversation in combination with other AI tools. And as you can see, this is really bringing many things together. Right? It brings the knowledge of GPTs, prompt engineering, other tools all together to create a superior result than people who simply open ChatGPT for the first time, run a simple prompt and are disappointed. This is what intermediate to advanced users looks like and that's what we do with the community. And my hope there always was that everything that we do there is also going to enhance the YouTube channel. This is the first video that I made sharing some resources from there. Let me know if this was helpful. If you learned something, subscribe to the channel. There's a whole lot more YouTube content coming. I'll see you soon.