Mega Prompt Framework
Tired of aimless AI conversations that go nowhere? Zach Hammer reveals the secret to focused prompts that get results – the Mega Prompt Framework. With components like persona, task, and goal, you’ll structure prompts like a pro. Emulate the right expert, clearly define what you need, and watch AI leap into action.
Can’t get ChatGPT on the same page? Zach drives consistency with context and constraints. Lay out format rules, tone guidelines, even templates to train AI like an obedient worker. Now prompts pump out polished products with minimal massaging.
Struggling to repeat that magic? Let mega prompts be your master template. Imbue prompts with your business knowledge DNA so clones execute flawlessly. Plus, learn helper prompts to build more prompts faster than ever. Zach unpacks it all so you can let AI handle the heavy lifting for good. Stop spinning AI wheels – with Mega Prompt Framework, get moving in the power lane today.
Other subjects we covered on the show:
- The benefits of structuring prompts for consistency, repeatability, and handing them off to others.
- Using personas based on celebrities, executives, or personalities to influence the tone and style of responses.
- Leveraging past experience and strategies to guide prompts towards known effective formats.
- Providing templates as examples of the desired structure and sections.
- Optimizing for reduced back-and-forth by capturing expertise upfront in the prompt design.
- Crafting prompts that can be easily refined by non-experts through fill-in structures.
- Developing helper prompts to assist in building new prompts or refining existing ones.
- The caveat that too broad a prompt may lead to inconsistent results vs capturing a specific outcome.
- Separating elements that will change each use (like target/product details) from the consistent framework.
- Techniques like Markdown formatting to convey templates and structures without explicit examples.
AND MORE TOPICS COVERED IN THE FULL INTERVIEW!!! You can check that out and subscribe YouTube.
If you want to know more about Zach Hammer and Charlie Madison, you may reach out to them at:
[00:00:00] Zach Hammer: What’s up. Welcome to another episode of Real Estate Growth Hackers. Today with me, I have my good buddy, Charlie Madison with Realtor waiting lists. Charlie here is a developer, a Realtor himself as well as just all around smart entrepreneur. Welcome to the show with us, Charlie.
[00:00:23] Charlie Madison: I am excited to be here.
[00:00:27] Zach Hammer: I think Charlie is likely going to be joining us pretty consistently, at least for a number of episodes here.
[00:00:33] Zach Hammer: So, we’ll be talking through some good concepts together. Today we’re going to be talking about how you can structure your ChatGPT prompts for maximum effectiveness repeatability and just really making them something that you can deploy in a way that’s going to get you a good results, all the time.
[00:00:53] Zach Hammer: So that’s what we’re gonna be diving into. Let’s talk a little bit a context here. So, for anyone who isn’t [00:01:00] familiar, ChatGPT, it’s a an AI tool put out by open AI that allows you to have a conversational flow with something called a large language model. That lets you ask questions and get meaningful responses back that are surprisingly Intel intelligent.
[00:01:15] Zach Hammer: And in terms of how you interact with this working with these large language models, you use something called prompting and all that is it’s just a fancy word for the questions or the inputs that you put in in order to get the results back. So when you ask a question, a really simple prompt would be something like I don’t know, give me a really simple prompt.
[00:01:35] Zach Hammer: Charlie, what’s a simple one that somebody might do?
[00:01:38] Charlie Madison: Write me a Facebook ad for a home buyer.
[00:01:43] Zach Hammer: Perfect. Okay. So that’s a really good example of a simple prompt where we’re saying fairly clearly what we want. Charlie actually even went a little bit more advanced of giving a specific person that we’re targeting. Some people don’t even go that far. So, yeah, so we’re talking about how we take something like, write me a Facebook ad [00:02:00] for a home buyer.
[00:02:01] Zach Hammer: And turning it into something that is, it is really structured for success to get the result that we’re looking for. And so, yeah, let’s dive into a little bit about why this matters. Okay. So, first off, chances are, if you are looking to leverage AI in your business. You are probably trying to get some level of larger output compared to your input, right?
[00:02:24] Zach Hammer: That’s one of the biggest reasons why we’re using tools like this. We’re trying to reduce our workload while increasing the power of the results that we get from it. And one of the things that is typically the case, is that you want to be able to do these things repeatedly. And repeatedly have success.
[00:02:42] Zach Hammer: So does that hold true for you, Charlie? Is that something that you’ve seen with AI that you’re often using it for even some of the same things over and over again?
[00:02:49] Charlie Madison: Yes. I mean, for example, you got a Facebook ad for a buyer and then you’ve got a Facebook ad for a seller and then you know, respond to this email, respond to that [00:03:00] email. So like I found the things that. Maybe I like writing emails, terrible at it. So I haven’t responded to a lot of my emails. I’m not a writer.
[00:03:10] Charlie Madison: So yeah, I’ve got, I’ve tried ChatGPT on a whole bunch of things and I’ve got a handful that I’ve found, it’s really efficient.
[00:03:20] Zach Hammer: And so, yeah, I mean, that’s the key idea here is like, typically in your business, you’re going to have if you have something that you decide, you know what, maybe I should ask ChatGPT for this. Chances are that’s going to come up again. And so it’s worthwhile to know how to structure a prompt so that you can get consistent, solid results.
[00:03:40] Zach Hammer: So, let’s, you know, double click a little bit into your example about a Facebook ad for a home buyer or a seller, right? Whatever. And do you, Charlie, do you already have some ideas of what makes for a successful ad, right? Like you’ve been at this game for a while you have theories and strategies that [00:04:00] you’ve deployed and sort of learned.
[00:04:01] Zach Hammer: This is a way to make this successful. Does that sound true?
[00:04:04] Charlie Madison: Yeah. A hundred percent. A hundred percent.
[00:04:06] Zach Hammer: Perfect. So one of the things that we want to solve for is we want to make sure that we structure our prompts in a way that we could set ChatGPT up to give us back the things that we already know work, because if we don’t, part of what’s going to happen is it’s going to give you kind of pulling a random collection of things out of its hat to say, okay, here’s what a great Facebook ad looks like.
[00:04:32] Zach Hammer: And then the next time it’s going to look completely different than the next time. It’s probably going to look a bit different from that. And sure there’s going to be some level of similarities, but if we know, this is a structure for an ad that I really want it to follow. I know I’ve gotten good results with this kind of hook at the front end, or it needs to mention these specific points or I have a formula that I want it to follow.
[00:04:51] Zach Hammer: It can be a really useful to, to make sure that happens. And if you don’t give it enough context to do so, it will take liberties with that and give you something different every time, [00:05:00] which if we’re looking to, you know, build out a system that gets us consistent results you know, for our team. For our business then we want to be able to reliably do that. And so, that’s really some of the reasons why we’re diving into this. The other reason for this is that when you structure your prompts properly. The other thing that you could do is you can often hand them off to somebody else so that they can run your prompts in a way where they get the same result that you would even when you’re not there when you’re not around.
[00:05:27] Zach Hammer: So it makes it something that could be outsourced, that could be leveraged even further because you do a bit of work on the front end to sort of download DNA into a prompt so that prompt likely gets back a consistent result as if you were the one doing it. So does that all make sense, Charlie?
[00:05:42] Zach Hammer: Anything there that you think I need to cover further?
[00:05:45] Charlie Madison: Yeah. So one, I’m excited about this because it takes me a lot of work to like fine tune a good prompt. And my secret trick now is to copy your prompts because you have [00:06:00] like, your prompts are like amazing. And the way that I like to think of them is like a really great repeatable prompt is like a really great worker.
[00:06:12] Charlie Madison: That’s actually going to do it 95 percent right, most of the time, you know, so it’s just like, you know, receptionists, they ain’t going to do it right all the time, but this worker is going to do it. Mostly right, like highly mostly right, most of the time if it’s done well, and that’s, I mean, think of how cool it is to have workers that do that for you.
[00:06:39] Zach Hammer: Exactly. And I mean, something that you mentioned, I want to double down on that too. It is worthwhile to understand you know, just like everything there’s degrading returns on this, right? So you will never get a prompt so dialed in unless the concept of what you’re looking to achieve is really simple.
[00:06:56] Zach Hammer: But if you’re doing anything somewhat advanced and nuanced [00:07:00] you’re never going to get a prompt to be, like 100 percent perfect where what it outputs doesn’t require any change, modification or input you’re going to, you should expect that there’s going to be a little bit of massaging, but what it’s going to do is it’s going to get you past that blank page.
[00:07:16] Zach Hammer: It’s going to get you to the line where all that needs to happen is a little bit of tweaking, a little bit of modification generally in order to get that right result. And what we’re looking to avoid is we’re looking to avoid the typical way that people leverage ChatGPT, which is they ask a really vague question at the top.
[00:07:32] Zach Hammer: It gives you back a really vague answer next, and then you keep going through a series of conversation in order to try and get it to actually refine to what you’re looking for. We’re looking to minimize that conversation flow and make it so that we just have one thing that we put in.
[00:07:47] Zach Hammer: That nine times out of ten gives us back a thing that’s mostly the right way, literally on that next answer is typically the way that this works. So, that’s kind of our goal here. And all of this comes down to something that I’ve [00:08:00] developed and researched over time that I call a megaprompt.
[00:08:03] Zach Hammer: Okay? And so what a megaprompt is it’s essentially All of the different components that these large language models understand effectively in order to get a consistent, repeatable result. There’s some key aspects of these prompts that you can leverage repeatedly in order to get those results.
[00:08:19] Zach Hammer: And there’s really, there’s one, two, three, four, five, there’s five key components, six, maybe and some of them could be broken down into into sub steps. So, what do you say, Charlie, should we dive in and start
[00:08:31] Charlie Madison: talking about the points?
[00:08:32] Charlie Madison: Let’s dive into your magical mega prompt. I’ve used it and it does feel like magic.
[00:08:38] Zach Hammer: Let’s do it. Let’s do it. All right. So first off the first step of your prompts and literally you can change the order of these around a little bit, but I do tend to find this order is about what I want to follow when I’m structuring out a prompt. So, it may not be 100 percent important. You want to cover each of these ideas but even the order might [00:09:00] make a difference for you as well.
[00:09:00] Zach Hammer: So keep that in mind. But the first one is we’re talking about the persona that we want ChatGPT to be thinking like, so this is actually something, all of this stuff, it’s not just based off of random ideas that I’ve come up with. This is, there’s a lot of people who have studied like, what are the things that actually make a difference on these large language models, where does it have a measurable impact to get it to think in a certain way versus another?
[00:09:26] Zach Hammer: And this is one of those things that does make an impact. So getting ChatGPT to not think about everything like, like everything, but rather to think like a specific person or a specific type of person can make a difference for the kind of output that it’s going to generate. So, examples for this would be things like if you’re having ChatGPT write a Facebook ad, this isn’t too hard.
[00:09:51] Zach Hammer: You’re going to ask it to act like an expert copywriter and Facebook advertiser, right? You’re going to, you’re going to ask it to act like somebody who has the [00:10:00] skillset of somebody who would do a good job at the thing that you’re looking to do. So, it’s not super complex but that’s the idea is you want to think.
[00:10:08] Zach Hammer: Who is somebody that’s good at this task? Let me tell ChatGPT to act like them. And what’s cool about these large language models is that it can actually be a little bit nuanced. So like as a, for instance, you could give it a general description like that, but you could also say like who is literally a person that I think would do well with this.
[00:10:27] Zach Hammer: So, you know, in the real estate space maybe you really like the way that Tom Ferry, you know, talks about things and verbalizes things. And so maybe you want your ads to sound like Tom Ferry would write them. So you could literally tell ChatGPT. I want you to write this as if you’re Tom Ferry writing a Facebook ad, or, you know, maybe not Tom Ferry.
[00:10:46] Zach Hammer: Maybe it’s somebody else. Maybe you want to sound like Tony Robbins and that’s who you like to evoke. And so you can tell it. I want you to write this as if Tony Robbins was writing this Facebook ad. Right. And so, I prove, what’s that?
[00:10:59] Charlie Madison: The [00:11:00] same wild guy.
[00:11:02] Zach Hammer: Yeah. Yeah. Or the Slapchop guy.
[00:11:07] Charlie Madison: Exactly.
[00:11:07] Zach Hammer: Billy Mays here. Yes exactly. But yeah, so you can be a little bit nuanced. The nice thing about this is that there’s some good solid principles, but you don’t have to be perfect with them in order to get a great result.
[00:11:20] Zach Hammer: So if you tell it to act like somebody that seems like the right kind of person. It’s probably going to help even if it’s, you know, not the specific way that I would say it or the specific way that Charlie would say it it’s still going to be useful. So, again, our first one is simulating a persona. Okay. So whether that’s a job title, an expertise, or even a person that’s our first concept.
[00:11:41] Zach Hammer: Does that one make sense? Anything unclear about that one, Charlie?
[00:11:44] Charlie Madison: Very clear. You choose a personality. It can be someone famous and it’s going to be directionally correct.
[00:11:54] Zach Hammer: Perfect. Perfect. So our next one is that we’re going to actually tell it the task that [00:12:00] it’s going to do, right? So we’re going to say, this is what I want you to do, right? This is the task you are going to, write a Facebook ad. And you might give it some qualifiers, right? You’re gonna write a compelling Facebook ad that is designed to get the person to click and register, right?
[00:12:19] Zach Hammer: That’s designed to get the person to click and become a lead. I want you to write an email that effectively responds to this inquiry and makes it clear that I am not interested, right? Like, whatever it is, right? But you’re just going to lay out very clearly what is the actual task that you want them to do.
[00:12:40] Zach Hammer: There’s not a ton in that. It’s just very clear descriptive what you want it to do. So, what’s what’s an example of another task that you might have A.I do that you’ve found success with?
[00:12:50] Charlie Madison: So, I’ve had success having it cancel certain subscriptions that needed. You know, one of my favorites though, was I actually [00:13:00] had a client that asked to cancel but really I thought what they needed to do was pause. So I had ChatGPT write a great email that says we’re happy to cancel you.
[00:13:16] Charlie Madison: We also have this program where we can pause and keep your line in where you’re at. What would you like to do? And it was done really well. And the client did pause.
[00:13:29] Zach Hammer: Awesome. And that actually leads us into the next one pretty well as well, these are very related. So, we have our task, which is what they’re going to do what you want this prompt to achieve, what you want ChatGPT to actually create for you or respond with and then the next thing is the goal.
[00:13:45] Zach Hammer: So what is the desired end results? What is the desired outcome goal or objective? And this could include any number of things. This could include the goal is to create a compelling ad that gets somebody to click. The goal is to create an email [00:14:00] that makes the person feel heard and understand that we’re happy to help them.
[00:14:05] Zach Hammer: But results in them also understanding that we can pause instead of cancel right that makes them feel good about it, right? So you could basically just say if this went perfectly, what does that look like, right? If at the end of this process you want to give ChatGPT the idea of when it’s done well, what is that end result?
[00:14:26] Zach Hammer: And so, those relate to together pretty well. Sometimes you can integrate the goal a little bit, in terms of how you’re describing the task. But it’s important to very clearly know that you want to give it some sort of goal. And so that’s our third one. So far we’ve got persona.
[00:14:40] Zach Hammer: That we’re emulating. We’ve got the tasks that we want them to, that we want it to do. And we’ve got the goal or successful outcome of that task. The next thing that we’re going to do is actually really important for getting a consistent, reliable result. This is one of those tricks that even if you don’t use the rest of this this [00:15:00] one thing can actually make a difference for the quality of the results that you get from any prompt that you do moving forward.
[00:15:06] Zach Hammer: So we should probably stop there, right?
[00:15:09] Charlie Madison: I think so.
[00:15:12] Zach Hammer: All right, cool. Let’s just stop. Let’s stop. Okay. Well, we’ll keep going. We’ll keep going. We’ll go into the next one. So the next one one of the things that they have found in studying these large language models is that you can get a better output from ChatGPT. By asking it to think step by step and literally, if you use those words, if you stay, I want you to think about the step by step, the fact that this works is honestly, it’s one of those things that like makes me a little bit scared about like, what it is that we’re actually working with here, because I know the reality is like, Honestly, I know I think better when somebody asks me to think through something step by step and it sort of blows my mind that like some of the same things that [00:16:00] apply to humans actually apply to working with ChatGPT as well.
[00:16:03] Zach Hammer: So the next thing that we’re going to do is we’re actually going to try and lay out the steps that we want ChatGPT to follow. So, you know, we’ve given it the task, we’ve given it the goal, now we’re going to give it the steps.
[00:16:15] Zach Hammer: So the steps to complete the task, like step one, I want you to understand my product, right? And so maybe you’re like, you’re doing an ad, so I want you to understand my product. Step two, I want you to craft a compelling a compelling ad that speaks to this person about my product. Step three, I want you to revise that ad and make it even better, right?
[00:16:39] Zach Hammer: So, the steps tend to be a little bit more logical. Actually, let me go ahead and pull up one real quick to give some better examples of that. But while I’m doing that, why don’t you go ahead what are some of the prompts that you’ve given that you think steps could help and what would those steps be?
[00:16:53] Charlie Madison: So one of my favorite prompts is to get the demographics for [00:17:00] my client. And so I’ve got this long post that says, you know, tell me everything about this client. So tell me everything about someone that makes 75 to $150,000 a year in Nashville, Tennessee that wants to move to a bigger home. And then I’ll say, step two, tell me everything, someone 150, 000 to 250, 000 a year.
[00:17:26] Charlie Madison: Same thing. Step three, tell me everything 250, 000 to 500, 000. Step four, give me the similarities between those three people. Then the last step is craft an ad that focuses on those similarities. Is that a good example?
[00:17:47] Zach Hammer: Absolutely. That’s a perfect example. So yeah, you’re thinking through just very logically what would you do step by step to go through this process, right? You know, first you’re gonna list out this thing, then you’re gonna list out that, then you’re gonna list out that, then you’re gonna go through and sort [00:18:00] of, blend that information together, discern some information from it, and then you’re gonna take action on it, right?
[00:18:04] Zach Hammer: So yeah, the same way that you might think through that. That’s the same level of what you could do to get better results from these prompts as well. So here’s an example from a prompt that I have, this is a prompt that I put together. That’s designed to create a good introduction method or a introduction message for me.
[00:18:21] Zach Hammer: So I very often part of what I do I introduce people a lot in my day to day business life. I find, you know, like, hey, Charlie’s working on this and he mentioned that he needs help from you know, with this specific idea. So I’m going to connect him to say Raul.
[00:18:35] Zach Hammer: Raul’s a buddy of mine and I think Charlie and Raul would get along well together. So I want to introduce these guys so that they can connect, right? So I do that a lot. So anything that I do a lot. I try and see if I can make that process easier on myself. And so I have a prompt designed to create the kind of introduction messages that I typically put together.
[00:18:52] Zach Hammer: And so the step by steps for me on this one the first step that I give ChatGPT one, write a message, introducing two individuals [00:19:00] who you believe could benefit from connecting based on their areas of expertise. Two highlight each person’s unique skills and what they’re great at.
[00:19:07] Zach Hammer: Three, emphasize the potential mutual benefits of the connection. Four, make the message conversational and encourage them to take the conversation private. Five, provide additional value by offering to facilitate the conversation or make introductions to other relevant individuals. So I’m laying out, in this case, these step by steps.
[00:19:26] Zach Hammer: Really, they’re more so following through kind of the structure of the template message that I tend to send, where if you look at the messages that I send I’ve had AI analyze it and say, what am I doing here? And it says, ah, this is what you’re doing.
[00:19:37] Zach Hammer: You know, you’re doing this, you’re doing this. And I was like, okay, cool. So that’s apparently what I want you to do. And so that’s what those step by step instructions look like. So that’s another example for you. But the key idea. And again, this doesn’t have to be perfect, literally by saying, here are the steps, or these are the steps that I want you to do to complete this you’re going to get a better result and a more consistent result by sort of laying out a flow for it to go through, that makes [00:20:00] sense.
[00:20:00] Zach Hammer: It doesn’t have to be perfect. There isn’t exact right steps or an exact right way to think about this. It can be a little bit loose and that’s going to be okay. But that’s the idea is steps. So, far we’ve got persona, task, goal, steps. The next thing that we’re going to talk about that you could give it, this is actually somewhat optional, but I do find if you really want to get something repeatable, it’s very useful to have this section which is constraints.
[00:20:27] Zach Hammer: So the constraints are going to be anything that you want it to know. So if you’re having it craft a Facebook ad, maybe, you don’t want a long form Facebook ad. So you want to tell it I want this to be two to three sentences max. I want those sentences broken out over multiple lines, or I need it to be this number of characters or less because I’m doing a Twitter post.
[00:20:49] Zach Hammer: You know, I need to make sure that it rhymes because I want it to. Right. Whatever constraints you might have where you could sort of help it to say. [00:21:00] You know, out of all of the options that would technically fit everything that you’ve done so far, I want you to just cool down to the ones that fit these constraints.
[00:21:08] Zach Hammer: That’s going to be a big part of actually helping helping it to be consistent for you.
[00:21:13] Charlie Madison: Can I share what I like about that?
[00:21:15] Zach Hammer: Absolutely. Go ahead.
[00:21:16] Charlie Madison: What I like about that is, I don’t know if you’ve ever seen the video of the dad that, he tells his son to walk him through making a peanut butter and jelly sandwich. And so the son says. You know, stick the knife in and put it on the bread. And like, there’s so many assumptions that the son has that like the dad just makes this terrible sandwich and the son’s about to like lose his mind.
[00:21:44] Charlie Madison: And it’s a beautiful example of step by step versus assumptions. And what I love about this is it removes a lot of the crazy assumptions that [00:22:00] ChatGPT would have, like, for example, my client that paused ended up pausing instead of canceling. I know this person really well.
[00:22:09] Charlie Madison: I know why they needed to pause and so the first message was like dear madam and like i’m not talking to her. So like one of my constraints that ended up like be short punchy Personal and friendly, and it got it 90% of the way there. And then I changed a few things and it was perfect.
[00:22:28] Charlie Madison: So I love that, you know, there literally are constraints that are part of this. And so having this block so that we can kinda take off the assumptions that are gonna throw it wildly out to left field is really helpful.
[00:22:45] Zach Hammer: Yeah, absolutely. And it makes a massive difference just giving it those constraints on the front end because it really can like, you know, maybe the way that these large language models typically see a Facebook post where it’s like the average Facebook post[00:23:00] is long form, or something like, I don’t know if that’s actually the case, but maybe that’s how it typically sees it.
[00:23:05] Zach Hammer: And so it, it always tries to give you back that long form thing, but maybe you’re like, man, I never talk this way. Like, I always try and keep it shorter, you know, simple points. So you tell it to do so, and it still may not be perfect, but it’s going to be a lot better than if you just kind of let it guess and it’s going to have a better chance of getting back what you’re looking for.
[00:23:23] Zach Hammer: So within this realm of context and constraints, there’s a couple of like sub modifiers to this that I tend to put in their own, like, on their own line and in their own bucket and how I’m doing them on prompts, but really they’re just like another element of context and constraints.
[00:23:38] Zach Hammer: So one of them is output format. So if you know that you want like a blog article with headings and you want it to leverage markdown in the process of doing that then you could say, use markdown headings to format, to make this easier to read et cetera. Right.
[00:23:55] Zach Hammer: So you could say that sort of thing. You could say I want you to format this like it’s going to [00:24:00] be a text message to somebody that I know. And then it’s going to adapt the formatting to make sense for a text message and not assume that it’s going to be a blog post, or maybe I want this message to be formatted using proper formatting for Slack or for Discord or for Facebook or whatever, right? I wanted to use emojis or I don’t want it to use emojis. So this again, you can see this still sort of fits into the constraints and context. But, when I think of output format that let’s be put my brain in the right spot to think the end result needs to kind of do these things.
[00:24:34] Zach Hammer: So that’s a useful, helpful one there too. Another one that’s really useful this can go into the simulate person persona, but it’s often useful to break it out and separate it, which is the tone or style of the thing that you are outputting. So, I could say, I want you to act like an expert Facebook, you know, copywriter very familiar with writing high converting ads, but I want you to make the ad sound, [00:25:00] tone, style, like, Tony Robbins after he just got out of a cold plunge or something, right?
[00:25:08] Zach Hammer: Like, so you could tell it how you want it to sound. And that can be separate from the persona. So, again, the same way that persona could be either based on like a celebrity or based on like an archetype or an expertise tone style could be a similar thing where you could say, I want it to sound professional.
[00:25:24] Zach Hammer: I want it to sound friendly. I want it to sound short, punchy, succinct. I want it to sound mildly sarcastic. I want it to sound a little bit, a little bit snarky those kinds of things. You could leverage the tone style to do that. And that can actually create some really great results where you get something that without that tone style sounds very like, very average.
[00:25:48] Zach Hammer: It’s solid. It makes sense. It’s good on paper, but you tell it to make it snarky. And then it’s like, Oh man, this thing sounds so passive-aggressive.
[00:25:56] Charlie Madison: It puts personality into it.
[00:25:58] Zach Hammer: And it’s amazing. [00:26:00] Exactly. Exactly. And so, tone or style is a great way to do it. I mean, like, so, humorous you can ask it to do it, like it’s a rap. You can ask it to do it, like it’s a poem. You could try and get it to do a haiku those sorts of things, right? There’s all sorts of things that you can leverage that section for. So that’s all within this realm of context slash constraints. Another thing, I actually don’t have this in my document, but I should add this.
[00:26:23] Zach Hammer: I’m going to add it right now. Another thing that I like to throw into this context slash constraints I’m going to give this one to you as a bonus. Okay. So this is a bonus. It sort of fits the context and constraints, but it’s also kind of its own thing. Which is give it a template.
[00:26:38] Zach Hammer: You say this, here’s an example of what a good output looks like. And generally when I do that, I tend to have the template be like descriptors, if that makes sense. So, you know, heading that accomplishes this description sentence that accomplishes this, right? So I give to it as an example template typically at a technical level.[00:27:00]
[00:27:00] Zach Hammer: If you’re familiar with Markdown at all, I tend to use a code block for how I do the template. So I give it three ticks and then I use Markdown formatting to kind of break out that template to say, these are the different sections that are in there. And in doing so, I communicate more than just the words are saying, I’m saying everything between these tick marks is the template and in there, I’m using heading symbols and bold and italics and all of that in order to communicate more without actually having to say it word for word.
[00:27:34] Zach Hammer: So, giving it a template can be a great way to get a good consistent output as well. And yeah, there you go. That is the process for how you structure a mega prompt.
[00:27:46] Zach Hammer: Again, those elements are a persona, a task, goal, steps, context and constraints, which can be broken out into the output format, tone, style, and [00:28:00] templates. And when you put all of those pieces together into a prompt, that is typically what you can do in order to get something that repeatedly gives you the same results.
[00:28:10] Zach Hammer: Time after time without having to have a super long conversation context in order to get it to refine it further. Now you might still have a couple of questions that you use to refine with it, a little bit of a conversation that you do to get it perfect. Or you might, you know, pull it out and massage it more manually, but that is the that is the process there.
[00:28:28] Zach Hammer: So what do you think? Now, I don’t know if I’ve ever walked through how I think through that bigger problems before. What do you think now that you’ve heard that Charlie?
[00:28:37] Charlie Madison: So I’ve used your Mega Prompt before, and like I said, it does just feel magical. Instead of, you know, I love just, it creates really clean output. Like, but before then, you know, it may take me 10 or 12 different dialogues back and forth, and so I’ve got like this messy [00:29:00] page and I’ve got to like pick out what’s great which you’ve got another prompt to kind of clean that up which I also love, but I love how like, if I learned to think through problems this way.
[00:29:15] Zach Hammer: Right.
[00:29:16] Charlie Madison: It’s really, it’s a new language. It’s a new way to think. One, it’ll clarify the way I think, but then it’s just a great way to get great results quickly and efficiently.
[00:29:28] Zach Hammer: Right. And you want to know something else? I’ve actually got two, I’ve got two prompts that relate to this structure. Cause here’s something that I learned. I want to create a lot of these prompts. And here’s the thing that I’ve learned. Anytime I want to do something repeatedly, you know what I do? I create a prompt so that I don’t have to do as much of the work in order to do it repeatedly. So I actually have a prompt that I call the prompt engineer prompt. And [00:30:00] or what do I call it? Prompt engineer GPT, I think is what I call it. And it’s a prompt that you could throw in that is very specifically designed to make it easier for you to craft a prompt like this.
[00:30:11] Zach Hammer: You throw this prompt in and then you just chat with ChatGPT for a bit, and then it will give you back a prompt that’s designed to do whatever you told it you wanted to accomplish. But it engages you in a conversation flow in order to get that developed rather than you having to sit down and think, what what tasks do I want it to do?
[00:30:30] Zach Hammer: You could sort of lay it out in a rough draft. And then it helps to start filling in some of those details for you so that you start with something like, I want you to write a Facebook ad for a home buyer and you say, I want you to develop a prompt for me for this. So anyway, it’s a prompt designed to do that.
[00:30:46] Zach Hammer: And then separately, I do have a different prompt. That’s kind of the opposite, which is say you’ve had one of those long conversation chains. I think this is the one that you were talking about. So you’ve had one of those long conversation chains where you’ve gone through that process of sort of refining with [00:31:00] ChatGPT, exactly what you want.
[00:31:02] Zach Hammer: And you get that end result. Like you, you get to the end of it. You’re like, yes, we got there. This is the thing that I want. I have a prompt that’s designed for you to use at that point so that you could take everything that you developed with ChatGPT, any templates, any concepts that you dove into, and then take that and turn it into a well structured mega prompt.
[00:31:21] Zach Hammer: And so, yeah, those are two prompts that that I’ve developed. I actually don’t know. If I’m going to give those away or if those are part of some course or something, those are things that do exist in my world what I would recommend for people if you reach out and said to be a message your chances of either getting that prompt or finding out where you could get that prompt is drastically better.
[00:31:41] Zach Hammer: Cause I don’t know where you’re going to find them right now. I don’t have a specific product of those. I literally use these day after day in my business. So feel free, reach out maybe I’ll give it to you. Maybe I’ll point you in the direction of the course where they exist or something like that.
[00:31:53] Zach Hammer: You can always check the the show notes for the show as well. Maybe I’ve included it in there. Honestly, I don’t know. We’re recording this off the cuff cause I just knew I wanted to talk [00:32:00] about this concept. But if you want the templates themselves, then we’ll, you know, maybe we’ll talk through how people can get those.
[00:32:05] Charlie Madison: It’s good to be a friend of Zach Hammer. I’ve got my black book of Zach Hammer prompts where we chat every now and then, and I don’t know if you create them on the fly or like, I’ve got this one. And like, I’ve got my list, my one tab saved of Zach Hammer’s precious prompts and it was like, Ooh, like they’re so like, cause I mean, I, you know, I’ll spend a lot of time sometimes like that one that I mentioned about getting the persona.
[00:32:33] Charlie Madison: I really need to use one of those on this because I’ll do it, but you know, it takes 30 minutes and I mean, it’s amazing. Like I get amazing material and ads and content from it. But if I could do that in two to five minutes, I’d probably create more content. And I would use it more often, right?
[00:32:56] Zach Hammer: Yeah absolutely. And so yeah that is actually one caveat that I’ll throw into this [00:33:00] process. There’s a little bit of an art to this, of knowing what all a single prompt can do effectively. Versus what needs to be achieved over multiple separate prompts. Sometimes in separate threads, sometimes in the same thread.
[00:33:16] Zach Hammer: Cause there’s only so much context that ChatGPT is able to understand at one point. And effectively deliver on before it starts hallucinating and giving you weird results that you’re not expecting. So just keep that in mind. If you’re going through this process for using this framework and leveraging mega prompts, and you’re finding that you’re not getting a consistent result, chances are, you’re just trying to achieve too much.
[00:33:38] Zach Hammer: And what I would recommend that you do is that you step back and you break it out into another prompt or two and think about how. You put in the input at the top, you run it through a prompt and you take the output of that and put it into another prompt or something. And that typically is the way that you could still get the end results that you’re looking for, but not try and have like.
[00:33:59] Zach Hammer: One prompt is going to [00:34:00] have a hard time really effectively, like, coming up with a concept, creating the ad, creating the blog, creating all that and having where it’s like, you put one thing in and it outputs all of this stuff and it does it. Well, you can, I’ve always seen It’s less than repeatable, let’s put it that way, it might give you something that you come back with and it’s impressive what all that output but it’s not a consistent process, it’s just impressive.
[00:34:20] Charlie Madison: You know, going back to the beginning, it’s a worker, it’s a worker that’s good at one thing. And so if you got three things, you need three workers. Now w we could inception this and we could use the mega prompt to create three prompts.
[00:34:39] Zach Hammer: That’s right. And honestly that’s a big part of this is literally, if there is something that you do repeatedly, chances are there’s a way to structure it as a prompt where you could take kind of the easy to flow, roughed inputs and then output you know, the things that you’re looking for.
[00:34:57] Zach Hammer: But yeah, there you go. You know what I just realized [00:35:00] actually. There is one other thing that I typically have as part of a mega prompt. So, the reason why I don’t necessarily think of it as part of the mega prompt. Because this is the user end of a mega prompt, if that makes sense.
[00:35:12] Zach Hammer: And so typically in my prompts I have the stuff that I’m telling ChatGPT at the top. That’s like all the, what it’s going to do and how it’s going to do and how it’s going to think about it. And then down at the bottom. I essentially have like the fields that influence everything else. So, if it’s a Facebook ad generator, then it might be like a section that says, what’s the Facebook ad for, what’s the product, what’s the target market, and those are things that are going to change.
[00:35:38] Zach Hammer: Every time I run the prompt, so I have them down at the bottom as places I could just fill in the blanks. And that’s the other way to really structure this to be usable is to make it so that whatever the questions are that you’d have to answer that make the output somewhat unique, that you just put those in one area so that you could just think through and brain dump.
[00:35:57] Zach Hammer: All right. You know, I wanted to talk about this. Here’s a rough draft of [00:36:00] those ideas. And that makes it so you don’t have to put as much work into the beginning process. Cause you did most of the front end creativity. Now you just give it the minor adaptations. This one needs to be funny.
[00:36:10] Zach Hammer: This one needs to be serious. This is about this product. This is for this audience. So those sorts of things are what I typically put at the bottom, which is a user input is kind of what I’d call that area, but, well, yeah, there you go. So a little bit of a bonus at the end of there in terms of how you structure these prompts for success.
[00:36:26] Zach Hammer: So there you go. Be sure to make sure that you that you follow what we’re up to. Chances are some of the concepts that we talked about in this one. If you feel like they weren’t adequately covered, they will likely be covered much better in separate episodes over time and we’ll dive into them in future episodes.
[00:36:41] Zach Hammer: So feel free to check back in and get more there. You know, wherever you’re watching this listening to it, whatever like comment, subscribe, all of those things, definitely want to make sure to get you this content.
[00:36:52] Zach Hammer: If you know of anybody that you think could get value out of this kind of information where we’re seeking to help real estate professionals to be able to better [00:37:00] leverage their effort to create scalable systems for their business, to be able to effectively, you know, build out a team, build out a company, build out a repeatable business, the feel free share it, pass it along. I’m always excited when people do that. But yeah, until next time, Charlie, thanks so much for coming to another episode of Real Estate Growth Hackers and we’ll call it there.
[00:37:22] Charlie Madison: Bye everyone.
Real Estate Growth Hackers Founder
Zach Hammer is the co-founder of Real Estate Growth Hackers. Over the last 36 months Zach and his team have managed ad budgets well over $100,000, generated over 25,000 real estate leads, and helped create over $50,000,0000 in business revenue for their clients. Zach is also a highly sought after speaker and consultant whose work has impacted some of the top Real Estate teams and brokerages across the country.