YouTube/podcast-style video by Justin R. Lester that teaches practical ways to get better results from ChatGPT and similar LLMs. Lester presents five core recommendations: (1) set the role/goal/format in the first prompt so the model 'plays' a specific role; (2) provide rich context (industry, audience, tools, deadlines) because "context is currency"; (3) don’t treat LLMs like Google—ask for summaries, applications, and actionable syntheses rather than raw links; (4) iterate—use follow-up prompts to refine tone, format, and specificity; (5) use in-conversation adaptation and (if available) memory to persist preferences. He closes with practical reminders (garbage-in/garbage-out, be creative, an action challenge) and a call-to-action to subscribe.
Chapters
Foreign. Welcome to tech with that tech pastor. It's Justin Lester and I'm so glad you guys are tuned in with us again today. I thought we'd try something a little unique, a little bit different. I wanted to do something like this. I went into my chat GBT and I asked a question. I said, if there are five things you want people to know how to use you, well, what would you tell them? How would you teach them? And the conversation got a little weird. And I want to have the conversation with you because I think one of the things with the advent of ChatGPT 5 or whatever else is coming next as we look forward to AGI, which is coming really, really, really soon. How can we use ChatGPT perfectly so there's no longer five different models to choose from? Because you only get ChatGPT 5 to choose from. You're making images everywhere. You don't have to worry about ultramany high. Your code is in there. Like everything is built into one model now. And that all changed when ChatGPT went overhauled. So I actually beat what are the five things you want people to do and ask you. You know what the first thing it said to me? Well, you'll learn today in our podcast. Today as we get into our time together, don't forget that like button. Smash that subscribe button, make it do backflips. You guys already know the drill. Thank you guys for tuning in. If you're on Spotify, itunes, whatever it is, be sure you are tuned into our podcast. Let's have some fun. Welcome to Tech with Ethic Pastor. Today, all about what ChatGPT wants us to know about ChatGPT to use it properly. That rhymed. Let's get it. So first of all, point number one, it's gonna give us a cheat codes. Point number one. The first thing ChatGPT wanted me to know how to use it properly as it continues to evolve is the first thing you type sets the whole game. So if we. If you give it tell me about marketing, you'll get a boring marketing, one on one answer. For example, that's why I can't stand preachers who use it to preach. If you ask it to write a sermon, it's gonna give you the most basic thing that it can find. It's the most popular thing. Think about it. If you always want the most popular thing, is it always the right thing? That's not necessarily so. If you ask it for a basic marketing response, a basic preaching response, that's what you're going to get. But if you say you're my marketing strategist, offer me sustainable coffee brand targeting millennials. Give me a three month social media plan with daily post ideas and simple captions. Now it's pausing, it's thinking differently because it's thinking like a marketing strategist. So now it's not only going to research marketing, it's going to research strategy, it's going to research culture, it's going to research coffee, it's going to research research sustainability, it's going to research scalability, it's going to think about social media, it's going to think about all the tools that coffee we get on that work with social media, it's going to think about the people you're trying to reach, it's going to do all of that. And now you're cooking. Because the first instruction tells it the role, who you're talking to, what the goal is and the format. And ChatGPT can work magic with good bones. Number two, the second thing ChatGPT wanted me to make sure we all knew is context is currency. So imagine meeting someone for the first time, you never talked to them before, and you tell them, write my autobiography. And they don't know anything about you, right? If you go into your chat and it doesn't know anything about, you're wasting your time. So if you tell it, if you tell it your industry and your audience and your style and your preferred tools and even your deadlines, the answer jumps from a generic tone to a good one. So think about it like this. If you go into your chat and you make sure you tell it, hey, listen, this is what I, this is the group I'm always talking to but always asking me before you write something. These are the tools I prefer to use if I am using something, if I'm going to stack something, my deadlines, my busiest day of the week is normally this week. My lightest day of the week is normally this week. This is what my week format normally looks like. This is what my industry is. That changes everything. That's why I always challenge people. When you're using chat, always tell it to ask you three questions before it responds. Because context is your currency. The more you give it, the better your returns are going to be. All right, number three, stop treating ChatGPT or Claude like Google. It's not Google, right? It's not here. So you can dump a list of links. It gives you answers. Now in the answers, it gives you sources. But it's not just a list of links. That means that your AI tools can summarize, it can draft, it can role play, can edit, it can plan, it can teach it and write speeches, it can write, we can make PowerPoints and we can do so much stuff. But if you all you want is links, you need to go back to Google. So if you want to research, you're going to get a research, you're not going to get the links, you're going to get summaries of where it's at. So now the challenge for you is how you prompt that research. So if you want the top five articles on remote work trends, then tell it that. And tell it not only the top five trends, but then three ways your business can apply it and three trends your business can see so that it can make money off of what it finds. Now you're making AI work for you. Number four, iteration is the secret sauce. Here's what you don't see. The first answer is not the best answer, right? The first answer on your chat is just the starting draft for your chat. So when you tell it to make your first answer funnier, or you tell it to make your first answer out of sports analogy or turn this into bullet points, you're sharpening it. You're teaching it what funny means to you. You're teaching it what short form answers mean to you. Because the people who win with AI don't stop at round one. The people who win with AI have prompted and prompted and prompted and reminded and followed up and uploaded and sent things and uploaded and sent things and then found the answer that fit them perfectly instead of just the answer they received. This is not a rush job tool. And one of the poorest things I've seen so far is that people have used AI as a rush job tool. It's not a rush job tool. It's a way that can help you as an assistant fit them perfectly. So think about your secretary, think about your co workers. You don't just take the first iteration like you ideate in the staff meeting, but you don't take that first thought in the staff meeting and go put a million dollars behind it. You take the first thought in the staff meeting. Now you're going to hr, now you're going to the team, now you're going to marketing or you come back to the leader. Now you're following the meeting. The same thing you do with ChatGPT, same to a Claude. You, you're thinking and then you follow up and then you go to the leader and then you follow up and then you go to this person there's constant follow up, constant engagement to make it the best it possibly can be. Lastly, point number five, you can learn within a conversation. Now your chat will not remember your life story between chats. It won't unless you have it. Memory. If you have memory set up or you paid versions of things, you get memory. But inside one conversation it can adapt. So if you want to keep the conversation going and save your conversations, adapt and see your preferences on the fly, I would challenge you to tell it to commit certain things to memory. Because if you don't like something it says, you can literally say go back earlier in our chat and reference it if you want it not to be formal or if you want to be formal, if you want to change the tone midway all of that. Now I would highly suggest getting memories onto your chat because the more you have memories, the more you can say hey, reference our memory, reference our past conversation, reference our conversations on this day, how was I feeling on this day last year? Now it'll help you because now it can remember and searches memories and go. But if you don't have some paid version, doesn't have memory, so then I would keep things in longer chat. So you can always re reference things in your chat, right? Those are my five major points. Here's another thing. I want to just give you some random stuff I'm thinking about. Garbage in, garbage out. You hear me say this all the time. Vague, sloppy, non put together prompts with no direction, you're going to get crap back. Intentional prompts, you're going to get some really good things back. So treat your prompts like you're briefing an actual human expert on what you do. Because you wouldn't email your lawyer saying help me. You email your lawyer saying hey listen, here's the problem, here's the constraints and here's what you desire as the outcome. Where are we on this? And hopefully this is not a situation you're in. And the last thing I'm going to tell you, you're allowed to be weird. This is your stuff. Some of the best work I've ever gotten has come from the most random stuff. Like think about it like this. Write a press release on Shakespeare using Red Bull random what is creative? Or teach me using to use SEO or GSO with NBA metaphors or plan a birthday party for my cat, but make it sound like a Fortune 500 product launch. Like it sounds silly, right? But think about the creativity because the more creative you get, the more creative your output. I think many of us are only using 1 to 2 or maybe 3% of the power of a lot of these large language models. They've been trained on tons of of data and the Internet is this oyster. Do more than make funny cat videos and do more than hey, plan this birthday party for my 5 year old. Plan your birthday party for your 5 year old. Like a fortune for hard to start up using NBA metaphors the entire way. How interesting would that be? All right, so that's my thing. Here's so here's my action plan for you based upon this today Chat GPT when I asked it to make tell me how to use ChatGPT if I were telling it to a 5 year old, it said these things. Here's my challenge to you. Get your role, context and format together. Iterate and don't abandon and then be as creative as possible. If you try these three things today, I guarantee you'll see better results immediately. And the challenge. Take one thing you've been putting off on AI and try it out. Like try something creative, try something funny, try something silly, try something stupid, try something and I promise you you'll be surprised at what you get back because these things are waiting to work for you. Make AI your assistant, play with the models, play with the memories, play with the channels, play with the chats, play with all of it. And the way I love to put it, you have your foundation model and you have your layer and you have your application layer. Pay for your foundation, don't pay for your application layer, pay for the foundation layer, your chat TBTs, your Gemini's, your quads, your deep seeks, your groks, your metas, your hugging faces, your all the stuff on your hugging face. Open source, pay for that, don't pay for the job and I promise you have something good. All right, come back next week. Tech with that tech pastor. So glad you guys are tuned in. Like subscribe, hit the button, do backflips, peace.
