Automatic captions from the recording, lightly corrected. Stretches with personal details, unsourced numbers, dated news or advice the class has since changed are left out. In all, 51 stretches are left out. Where the recording and the build sequence above differ, the build sequence is current.
Open the edited transcript
0:00One context, many assets
Yes. me today. So, but what we're going to do today is we are going to go over how to kind of take what we've been working on and this is going to be a four-week long over and over again class. Um, so this is going to be more of like a cohort style where the goal is to actually understand how to create assets, which is the most important thing because if you can understand how to do the thing fundamentally, you don't need me, right? to teach you guys the systems so that you guys understand how to operate within them and then you can just build whatever you want. And so the way it kind of used to be is when you were going to build something, you would need to map it out and build each asset
individually. That is no longer the case with AI and what it allows for you. So the model now, and we'll get into this as we go on, is that you basically have your brain in the center and then all the things offshoot from that brain, right? And so the AI interprets the brain and creates the assets specific to what you're doing. And that allows you to scale things that were historically
8:52Build the smallest version that works
So, today we are going to build your MVP. MVP in this case does not mean most valuable player. It means minimum viable product. So what is a minimum viable product? Minimum viable product in the tech space basically means that you are building the lowest thing that creates value. So if you're not an engineer or a builder, what mostly happens, and I am as guilty of this as anyone else, is identifying the problem and saying, "Okay, I need to solve this problem as well as I possibly can or no one will love the solution." Right? And so often right? And so we want to solve one small problem that is actually solved, right?
Build the whole bridge so you don't build a half bridge and then come back to land and say, "Okay, now I'm going to build another half bridge and another half bridge." Build one complete solution and then move on. That is that is where people get kind of off track is they're like okay this is good but it could be better. This is good but and yes you should absolutely iterate forward like that but you should have something that works before you start taking next steps and I am terrible at this working on it. Um so the model is not the edge. So the more technical you become. And I'm not saying everyone in here needs to become developers or programmers or whatever. But the more technical you become, the less the model needs to be intelligent because you're
giving it good instructions. The better instructions you can give any model, the more effective it is going to be in executing for you. Um, and so that's one of the things we're going to talk about today. The more technical you be technical you become, the more advanced stuff you can do. And that's on the front end and the the minimum end. So the goal is to try to come up with systems that work for you. I'm going to be telling you guys stuff that works for me. That is not gospel, right? Like take it. Everyone in here has different strengths and different weaknesses than me. And that is what AI is really, really useful for. It's really useful for leveling the playing field across all strengths and weaknesses. Right? So because I can talk with AI, my strengths
have the potential to be optimized even further and my weaknesses have the potential to be brought up to a place where they are at minimum competent in a very very rapid fashion. So the goal is not to focus on oh I need the smartest thing. I need the smartest what we are going to do is you guys know that generic does not win as it relates to um creating outputs, right? We need to have our own personal context in there so that your business is different than your business is different than your business is different than your business. Right? Everybody's business is entirely different even though it's roughly the same. Right? So I tell are specific to you, right? Like what I think is a beautiful website is going to be different than what everybody else in
this room thinks is a beautiful website, right? And so your clients care what you think is a beautiful website. They don't care what I think is a beautiful website. If you've made that connection and they've said, "Hey, I trust you enough to help with one of the biggest life decisions I can possibly do for the most part, right? Is buying a house, right? That's massive for most people." And so if they trust you in that way, they trust your style, they trust they trust you. They don't trust the AI. They don't trust me. So you guys have to be having the agency to build the assets that you know work in your style meaning your language your visual patterns the architecture that you want to use to present to your clients right so
16:14Put your context in text files
So context is how you do that. So the way we build context is we're going to build context in very very simple markdown files. What that means? So a markdown file is just a plain text file. So the easiest way to do that is to get a create a folder on your desktop that says context engine. Then underneath that you want to create a folder that says my business context, my client context, my deal context. Those are the big three. And then all you do is you start a new markdown file for your client. Start a and that's per client. So, if you have the Smiths, the Jenkins, and the Chases, you want three different files that have their personal context. Now, the reason for that is one, if let's say you sell them the house and they come back to
you, you already have their context saved, right? If they come back to you in 5 years and they're like, "Hey, you did a great job with this house. I want to move again." You've got their context with that, it enables you to take the AI, fix it on that folder, and you can make whatever asset you want. And we'll go over exactly what that looks like. How to build landing pages, how to build emails, how to build follow-up, how to build the whole thing. We're not going to do that today. We're going to work to our MVP today. But that's this is going to be the thing that is the most important is providing useful context and understanding what context is useful and what context is absolutely not because you can give the AI too much
context and confuse it. Right? So what I like to tell people is proper context management is giving the AI the right information at the right time. you if you're building a let's say you're building a landing page for the Smiths, you don't want the context for the Jenkins to be in that. You don't want a bunch of deals that the Smiths would never be interested in to be in that, right? And so the goal is to create the context that gives the AI the right information to help you build good, useful assets for you.
And this is kind of the new model, right? We've got the brain here and then this is what's called a spoke and wheel. Um, so it is not linear. Historically, what would happen is you would have the plan over here and then that plan would go to a landing page and then that landing page would go to an email and then that email would go to something and something and something, right? And you just it's a it would be a straight line of progress. This is no longer a straight line of progress because the straight line of progress is unscalable.
It is much harder to change a straight line than it is to say I have a brain. I can create whatever I want. Right? And so that is one of the most important things to understand is the model with which we are creating resources is different than it used to be. Um, and so like I said, we've got so so all of this leads to reviews, social posts, emails, listing copy, client pages, deal pages, your personal page, follow-up, all of that stems from the building the correct brain.
So your agent context, don't over complicate it. It's literally just a folder. It is literally just a folder on your computer that you enter in the this is the information that I have, right? And I'm going to teach you guys how to create all of it in Claude or GPT or whatever AI you want. The big thing is you don't want to get tied to a model. You do not want to get stuck. Oh, I have to use Claude because then if OpenAI releases something that is a step change better, you can't change cuz you're stuck to the model, right? And so if you build your own resources, what I like to tell people is building your own resources is owning your business, not renting it. If you are building just within one AI model and the files and
all the memory are staying in that model, if you need to leave for whatever, let's say the model goes bankrupt, all your stuff goes away. If that data goes if that data dies or and the company dies, everything that you've built within that model is at risk, right? And so we want to own this it could go away. And in real estate, speed is an alpha, right? Like if you can move faster than your competition, you can beat your competition by being able to do just that, right? So you've got your agent context, which is the context for you and your business. This is the stuff that maybe you hate the word dream home, right? And so you will put in there, do not use the word dream home. Maybe you're someone who does short-term rentals, uh, commercial,
whatever it is. Your context is the thing that tells the agent, hey, this is about this person. This is not public data. This is I I say it all the time, but one of the biggest advantages you can have is non-public proprietary data. If the data is on the internet, you do not have an advantage with it unless it's like locked to your company or of a moat there as well. But if those resources are being pulled in from public places, so like if you're looking at, you know, the the tax assessor, you're looking at the appraisal, stuff like that, that's public knowledge, anyone can go get it, right? And so we don't want to be locked into saying like, oh, it public knowledge is it's accessible to everyone. So it's not an advantage to anyone at this point.
a process for doing this it's the same every time. So with AI, one of the things that you'll see is the work on the front end feels like it takes forever and it should. That's the thing that people can't wrap their head that's okay. We move slowly at one place so we can go insanely fast at the next. Right? Speed, like I said, speed is an advantage. And the only way to get as fast as possible is to remove as many friction points that you foresee as you can. Right? And so the way to do that is by creating highly useful context.
24:20Page one: a page about you
So page one and this is um I think I have a folder here but this is basically the business uh folder that you have. So this is going to be the page that AI and buyers find when they look you up. So if you want a landing page for yourself personally, this is going to be a very very simple way to do it. So this is a prompt. What I always recommend is one of the things that content creators do to kind of gatekeep their information is they put it on slides like this, right?
So like nener I have it, you don't. How do you get it? Take a screenshot, take out your phone, take a a picture of this, take it to ChatGPT, say ChatGPT, transcribe this image. Information can no longer be gatekept from you guys. So, anytime there's a slide or information or something that you want to have personally, take a picture of it, tell ChatGPT to transcribe it. You then have that information, right? And so, people So, again, if you want to take a picture of this, you are super welcome to. Um, but you can see this is a very, very simple prompt that says, "Read every file in my agent context, my business folder first. Then follow my guardrails for everything below. Build a one-page personal landing page for me as a real
estate agent. Output the full page copy plus a simple single section HTML layout I can paste into a site builder. So include in my voice a headline that says who I am, who I help, where all this stuff, right? So basically what you are telling this AI is what does a good landing page even look like, right? Cuz what the AI has preconceived notions of what um what uh good looks like, what is good design, what is all that stuff. Now the problem with what AI does is it's an aggregator. So when it's when you say good design, it's trying to design things with counter information from all sides. So it just takes its full information and just dumps it. And usually um I don't know if you guys remember, but a couple years ago almost
every new web page was like a purple blue gradient uh with like some type of bubbly font. That's because the AI was just aggregating it and then putting it out as junk, right? it's almost every web page there was the same for a little a little while um until people started to understand that like hey we have to tell it that the styling needs to change and there are still some people that have the purple blue fonts and so what I would recommend is if you see that gradient pattern understand that if they're trying to tell you something about AI they have no idea what they're talking about right so just understand that that is that is the mark of what's called AI slop
27:20Page two: a listing page
So the next page is the deal. And so again, we've got another prompt and it's going to do the same thing. So it's going to read it. It's going to create that um that output that it's targeting, right? And so this says build a one-page listing landing page for proper for that property. Output full page copy plus a simple single section HTML layout. That is one of the most important things is giving the AI what exactly am I supposed to output? Am I supposed to output a a a Next.js app? Am I supposed to output a single web page? So most web pages that you see online are going to be based in HTML, hypertext markup language. So that's the very very very common thing.
And super duper fun fact, if you ask for the AI to give you a return in HTML, so if you ask for the output that you put in in HTML, it will create a web page for whatever output you want and you can actually modify and change the uh the variables on that web page right on the web page and then screenshot it and put it back into the AI saying, "Hey, this is what I want. Let's do this." So
28:30Page three: a private client page
the next thing we've got is the client and that's again same thing for landing page right and so all of these individual context things are siloed right and so once you start putting them together that's where you start getting your really really unique custom very scaled assets that you can start delivering to clients. So, as an example, you might have a deal that has very little natural light, yet your client wants all the natural light, right? And so, you can put all this together and it will say like, "Hey, the detriments of this deal are there there's no natural light." The solution I've come up with to that is if you want to do X, Y, and Z, whatever it is, right? So, you can say, "I've got a deal. It's not exactly what you want,
but here's what you can do to make it a lot closer to what you want." Right? And so now you're starting to match people to things that are actually valuable, right? You're not just the you're adding value to the transaction, which is the thing that is going to separate the next generation of real estate agents from the current generation, right? Like you have to be able to add personal value with almost every touch point that a client has with you. Um, speaking just from like a marketing sense, like I'm in the marketing space, if I'm not providing value for every touch point, people just tend to move on. And I mean, Like, if I didn't come in here and provide value immediately, it would have been, hey, what's next, right? And it and and it ain't me. It ain't me that's
next if I'm not providing value. And the same thing happens on a real estate client transaction. You want to provide value at every touch point so that when they get off the phone with you, whether they think, "Wow, they really know what they're talking about with AI. They really know what they're talking about with the local market. I feel valued." Whatever it is that's going to make that client be more of one of your big fans. That's what we're going for. We need kind of brand alignment at this point because it's no longer just transactional, right? Like people are searching in ChatGPT. They're they're talking more. they're analyzing more.
Everything is happening to a higher volume. And so the people who are doing things that produce higher quality outputs, I believe, are going to win in the long term. And I think that this is going to be one of those things where AI, and we've already seen it kind of in the software industry where AI is forcing consolidation, right? Like if you're I've said this before, but if you're mediocre at what you do, AI will kick you in the teeth, right? So you have to be an expert. You don't have to be an expert at everything, but you have to be an expert at something that allows you to provide more value than someone who just goes to ChatGPT, right?
right now, one of the things that's like super duper important with that, and we've talked about it in this these classes, is to intervene early and to intervene with quality. If you intervene early and your output suck, they're just going to think you're bad, right? Like, they're just going to think you're bad at it. So, intervene early with things that add value. Review the client context card with the client, right? Say, "Hey, does what I have here match what you're thinking?" Um, and that is one of the most valuable things you can do. Um, so we are gonna actually build some stuff now. Um, and I am going to Okay, so go over what we're going to do next week and then we're going to build stuff. So next week, like I said, this week is just about getting something
functional that you can use. Next week we're going to talk about branding, taste, and optimization. So next week is where we get really, really fancy. But next week doesn't happen without a good job this week, right? So, and there's And
yeah,
you gave us the props for um the folders for the client and the my business. But you didn't give us one from my deals. What? What is that?
This one.
Hold up.
I got Sorry. Thanks.
Oh, no worries.
Perfect.
33:22Load your folder into the AI
All right. So what we are going to do is so you can go to whatever AI you want AI of your choosing and I am going to build this in Claude so it is not that daunting. So this is the desktop version of Claude and this is the primary model that I'm what we are going to do, let's see. I should have All right. So, and honestly, so what I did right there is I literally told it to find and open the things that it created earlier cuz I'm not 100% sure where it stored it on my computer, right? And so rather than me digging through my computer and trying to find it, I just tell So this is this is a version of Claude. Don't get freaked out that it looks like a terminal cuz it is.
So, and there it is. So, I told it to open go into my computer, find those folders, and open that location for me. One of the things that I really love about um Claude Code and some of the things that actually have access to your system is you can let it function like an IT person. My computer um when I was purchasing it, I was told, "Hey, this is going to work great as long as you don't decide to code or do video editing." Within 6 months, guess what old genius boy decided to do? Both of the things that they said would not make my computer work. Turns out it does because I have the ability to have Claude code go in and empty space when I need it, clear up RAM when I need it, do things
that I am not technically proficient at doing, right? Because everybody thinks that it is the exact same. There are so many different facets of it. You can be cyber security, you can be performance optimization, you can be asset build. There are so many things that you can do within the tech space that I am not an expert in everything this thing is. And so when I say, "Hey, go find non-invasive things that you can delete from my computer. Give me a list of what it is, what space it's taking up, and how to clean it up." It'll present that list, and I can say, "Go ahead, fire them off. Get rid of it." And it will go and delete it. Um, highly, highly useful. You can have it set up things in your computer. You can have it do almost
anything, which is dangerous, but also very valuable. So, you want to make sure that you're always asking it, "Is this safe?" So when it comes back with a thing um as an example I it came back like I was like go do X Y and Z and bring it back to me. And so it was like oh I found all these things in your downloads folder that you can delete. Okay well why can I delete them? Because they have a number two behind them which histo which historically represents that they're a duplicate right the parentheses and two if you save it. Now what it wasn't being aware of is that sometimes I delete the wrong one and end up with just the copy that has the number two behind it. And so I said, "Hey,
are these actually duplicate files or are you about to delete things that are valuable to me? Oops, sorry. I was about to delete a bunch of stuff that probably provided value to you because there's no duplicate file for this one, this one, and this one, right?" And so you always want to be thinking about what's the worst case that could happen. Ask the AI if that worst case is available or going to happen. And more often than not, it'll be like, "Oh yeah, my bad. I was about to delete your entire computer."
Right? So, there's a super funny meme out there that has um someone taking a picture of a mushroom and asking ChatGPT if it's poisonous and then ChatGPT says, "No, you're definitely okay to eat that mushroom." And then in the next scene, it's a tombstone that has the the date of death. And ChatGPT says, "Oh, sorry. Turns out that mushroom was poisonous." Right? Because that's all ChatGPT is going to say to you after it destroys your entire system, right? And so, make sure that you have a backup.
So, I would use either like Google Drive or something called GitHub to push those files into so that you always have a cloud backup as well as something saved on your computer.
Get drive.
GitHub. GI T H U B. And GitHub is is one of those technical platforms. And it's good and bad because it's good because GitHub is very accessible via AI. Meaning you will never even have to look at the platform once you get it set up. But you do have to get it set up. And if you want to get help. Right. So GitHub and it is a backup. How do you tell the app to work with your computer or is it just going to work?
So, it depends. So, if you're using like the hard app, so like if you've downloaded Claude the app
that that is what's called in a sandbox, meaning that it creates a virtual environment that you can provide access to folders, but it is not going to have free reign over your entire computer. So if I click this, I can add files, I can add folders, I can add things from GitHub, I can add all these different things, so what we've got here, I think it's this one. So this is the agent context. And so you'll see these are all MD files. I'm going to show you that these are not that complicated. So if you go into the about me, with um so getting in back into resource management is also something we've talked about, right? And so I use a 4T system. Tech, tides, tangible assets, and talent. So tech, what are the tools
that you're using? Tangible assets. What are the real things I have access to now? Whether that's capital, whether that's a deal, whether that's thing like an office space, whatever it is that you have now, a podcast space that you can use as a resource to create content, whatever it is that you have as a tangible asset. The only way the AI can help you manage your resources well is if you've actually taken the time to aggregate your resources in one place. That is one of the things that takes a lot of time is to sit down and really think about tech tides. So tides are things that I believe are going to happen whether I want to be a part of it or not. Right? So like with AI that's happening. I can either watch it happen
or I can get on the train and be a part of it happening. Um, and then I have things that I think skills are very useful. So like communication I think is a skill that's going to absolutely rule the world in the future. Resource management, pattern matching, things like that. Um, and then also things like um human connection, right? As we move towards an AI based digitally aggressive world, human connection is going to fade more and more and there is going to be a deep desire from people to actually reestablish human connection. And we're is the that context for you. So like what is your background?
What are you working on right now? Ideally, this gets updated weekly. I will tell you full disclosure, I am terrible at updating weekly. So what I recommend is having a meeting with your AI scheduled weekly just like you would with an assistant or your team, whatever it is. Just say, "Hey, this is my weekly context update. I sold this deal. I got this new client. I found this great lead source. Whatever it is, just tell the AI, here are the things that need updating. Now go update my context.
You don't have to sit here and meticulously comb through it and update it. You will want to quality control it and make sure that it is accurate once it's up there, but you don't have to sit there and comb through it. You can just And now when I go in there um so I will need to I would need to add the context to this and do the things but that's all you need to say and it will go and update that context for you, right? And so if I wanted to add that, I could add the project and Okay. Uh, no, I would need to start a new project there. Let's see. Add folder.
So, I would just go to add files. I think it's going to be in here. Okay. Okay. So, what I'm going to do, so I am actually just going to go grab the folder that's open right here. And so, what I can do is I can just grab this. I can grab this about me file and I can pull this into the chat. So
ask you.
Yeah.
Some questions.
100%. I would prefer if you did.
Okay.
I don't like hearing myself talk. So more questions the better.
Yes.
I mean, how do I get that? And where do you Because I noticed like if I go to save for instance, okay, I don't have a map of all.
Um, if I go save a file, that's not a file type.
Yep.
So what is that? How do I get that and how do I use it?
Yeah. So you can either have the AI create it for you and say, "Hey, create this as a markdown file. Create my
create a word doc. Put it into
You could even just put it into Claude and ask it to create it, right?" Like you don't necessarily need the word doc. You can
Oh, I can just type it in there and say create.
Yeah. Yeah.
Save that into your desktop.
Yes. And
so we're creating two different folders, meaning a folder on your desktop and also a folder on chat on quad.
Yeah. So you can and we will get into projects and stuff and we're going to do that I think week three um to make it a little more simple but yeah so what the way I create them is you can either open a text editor like um like notepad stuff like that and usually they have a save as where you can save it as a markdown um or you can literally just type it into Claude and say create a a markdown file for this name the markdown file this right and it will create it for you So, going back to our screenshots.
So, we take a screenshot of whatever page number one.
Yep.
And it said, "Hey, transcribe it." I did. And did the next one.
So, I'm taking page one. You
Yep.
What do I do with that?
Yeah. So, you've got that prompt. What you will do is you'll copy that. You'll paste it into a new window. And what you can say is
a new chat window. So, like a new whether it's Claude or GPT.
Yep. You should. Yeah. Yeah. So, if you go up and you do
Yeah. So, actually, I don't know if you can. I Claude used to be weird where you couldn't actually have multiple windows. So, it may be the case that you can't.
I still don't understand the concept of having an app versus the computer. I don't
So, like going online to do it versus downloading the app and doing it.
Well, I have the app.
Okay.
And that's where I'm working.
Yeah.
But is there else I should be.
So you can do Claude code um which is the more advanced of the two and that'll be that's week four. We're going to get into week four. So yeah.
Yeah. For now. Yeah. Unless you're like um Claude code, I'm super happy for you to do that. But if you're just figuring it out, I would stay in the app because it's the more simple user interface. It's less intimidating and it has all the tools right there. You don't have to know any of the the background stuff to actually get it to function the way it's designed to.
I just asked it what I asked and it's like, well, I don't see that folder. Where is the conversation? But it's not in the conversation. It's on my computer,
right?
How am I going to know? How does it know to read my computer?
So, what I would do is I would say there is a folder in my downloads or wherever you saved it and then say, how do I pull this information into the app? And so what you can also do is you can find the folder on your computer and if it doesn't exist tell it to create it. Um you can say hey create this
I don't have access to your desk.
Okay. Yep. And so that's if you go if you add the plus if it's on GitHub you can add it or you can click add files or photos. And if it's not in here you'll have to just create it. And you can get it to create uh so let's see here
that said use the following
it doesn't want the folder because I keep telling it showing it the folder and it's like okay well keep going it's like well what in the folder
I still don't have
I don't get it. I don't understand.
Well, and that's that's totally good. And that there will likely become a day where you get one thing, get one, and you're like, "Oh, the light bulb has gone." Like, you won't even realize that the light bulb has turned on. you'll just be like three context cards deep and you'll be like, "Oh, I get it now." Right? And it's it's one of those things that like very rarely for me in this space have I had something that's like it's a onetoone, oh, I got that the light bulb went off and like it's a magic moment.
get. Right. And so
you're saying you're saying
no it's basically telling me I don't have cont what I keep telling it's like you keep asking me to do this yet
no I have like a document
oh well then you don't have
exactly
that's your computer's
I did I had one from another class that I did that's how we found it again, it's up to you to quality control it because the thing that I always tell right? So, make sure that you are really quality controlling the important things. And that being said, make sure that you're not overquality controlling the things that are just like, hey, I just need to do this fast and get a result back out, right? Because that's I mean I have spent I mean when ChatGPT first came out, I spent hours designing a logo. Complete waste of time. If you look at logos out there like look at FedEx, look at Google that it's literally just word, right?
Like the logo does not mean anything until it does, right? And you can take your time to create that and the logo changes. It changes it changes until you find something you like. Um Um, and so what you can do is once you have it created is you can click this and you can download it as a markdown file. Once it's downloaded,
HTML is HTML the same.
HTML is not. So HTML will have a bunch of different tags and uh excessive things that is not required for the AI to use. So the reason I choose markdown over HTML is that and you can say like please create this in markdown not HTML um the reason I choose markdown over HTML is because there's a lot of what's called metadata in HTML meaning that there's a body tag there's a and it is a little bit more organized but it also adds a lot of junk into the AI's context window that's not necessary and so if I save this so if I Save this in downloads.
56:20Ask for three versions at once
So I'm gonna show you guys how this works from a conceptual standpoint. So, if I take my business and what I'm going to do is I'm going to copy these these things and I'm going to drag and drop them. And so now these are all context that is loaded into the AI. So when it goes to make a decision, it's going to analyze all of these, right? And so now again, what I'm going to do is I'm not going to type my prompt. I'm going to say analyze all of the inputs that I have provided to you. This is my personal brand context. Provide a landing page for my business that will allow me to attract real estate clients for my business. Understand me deeply and exactly who my core audience is. Build three versions of the asset and present
them each to me. Then I'm going to do is press enter. So now theoretically it should analyze all of these and create a website that is roughly what I need. Now the things that I need to add to this to make it really really good personal images. So it's not going to have any pictures or anything like that. So you'll need to add images same that you would with the um the home, right? So, like if you have deals, what you might want to even do is put a new folder and put the context in there, including the pictures. So then you would have everything in one place. The pictures of the home, any videos, the written context that you would have for the the deal, all of it is in one place that is
super accessible for the AI. And so you can see right now, and I always like to read the observability portion because that tells you what the AI is thinking as it's generating the assets. And so if you see that it is generating something that is way off base for what you want, you can stop it there and say, "Hey, make this adjustment, right?" And so it's going to read the design and quality control skills that apply to conversion focused web pages and then it's going to build to do is it is now building things that are just exclusive to the information that it has. So, it's taking the general information that it knows of how to code HTML, what it knows about real estate, and it's adding your personal information to it. And then it's
building something that is personal to you. And that's why this is super duper scalable because you can build things that are very, very custom, very, very fast as soon as you have that context built out, right? And so again, it just it takes a little bit of time to generate, but and I always one thing that is um I'll give you two very very useful tips. One is if you're ever questioning what you need to do, ask the AI to interview you because sometimes you don't actually know what outcome you're trying to get. You just know the direction you want to go, right? And say, interview me and tell me how to get where I want to go. This is where I want to go. Right? I don't have any outcome in mind but this is the direction I'm
trying to go. The other thing is if you ask for variants it will provide three unique variants um ideally in different styles different everything. I may not have given it enough specific context to do that. But if you ask for a variant after it's already created just one output, it'll create a variant that's very similar to that one output because it doesn't know, oh, there's that variant variant variant, right? Like it the AI is trying to get to an algorithmic algorithmic place that it understands. And so the thing you want to do is you want to make sure that if you ask for var like variability that it creates variables like I ask for three to five outputs usually because it's so easy to create and then you just junk the stuff that and like often times what I'll do is I
will get three outputs and I'll be like I like this from number one, I like this from number two and I like this from number three. Build that. Right? And so, and a lot of times it's like I didn't even know that I liked or would have liked this stuff in number two because it's created something that I'm like, "Oh, that's really cool. I would have never thought about that." Right? And so, use the co-intelligence as an intelligence piece. Like, that's what it is right now. It is a it is a co-intelligence partner. It is not no longer just like a very advanced Google, right? You're looking for it to be a branding partner, design partner, a building partner. It's working with you.
Now, that being said, I don't anthropomorphize this and make it human either because it's it's working with you based on what you tell it. So, it will also directly work against you if you give it bad information, right?
Can I ask you?
I asked it to create an MD file.
Okay, I got you. Yeah. Yeah. Um, so I would click Windows default and then click okay and that should Yeah, there you go.
Thanks.
Yep. Absolutely.
So right now it's build and so one thing that's cool too is it'll name the variants. So right now it's building version one of the landing page and it's calling it the warm honest flagship. And so we will see what it creates. But it's going to do that and that and that. And so basically with this, I was hoping it would create it a little faster. I should have uh not done the high model because it's working to create the best possible output instead of an output that's a little bit quicker. But um once it creates the the landing pages, they should be decent.
Um, and the other thing too is you you can always put your design preferences in a file. So you can put your design preferences in a markdown file and then it will that kind of wrangles it and guard rails it to specifically what you want. So eventually this is going to create three outputs. Um, and this is just Sorry. As my story mentioned earlier, uh my computer is on its last legs because I did a bunch of stuff that I shouldn't be doing or I'm not supposed to be doing as it relates to coding and video generating.
Um but you upload, you can see here I've got a deal uploaded. I've got a client in there and I have all of my personal stuff. So this is the full landing page if you were making a custom to you that you're going to send to a client about a specific property. Right? So this is uh once okay so it's synthesizing the brand into three iterations. So it's built it it's building version one, building version two and building version three. You can see it's doing its coding stuff here.
um you don't necessarily have to care about that but again it is useful to look at the observability and say okay it's working on number two if there was a problem here I can intervene immediately um so it is building all three and then it's going to present them once that's done I'll show you what it looks like to build an asset specifically for a client about a specific deal and then that'll be what we've got for the day. And then uh next week we're, like I said, we're going to go over optimization and like actually how to get these things to be um really how you want them. Now, this may produce a better result for me than it will for you guys cuz I have different skills. I have things that are already programmed into this that it's
calling without me even having to tell it to do that, right? And so that it will go down into the we'll get into that in the optimization phase of like how do you do this without actually having to do that the the steps every time. How do you build your taste to make sure that it's fully fully oh yeah fully fully uh rolling. So and you can see here it's quality controlling already. So version two, it's replacing the accent stripe quote card with the top rule. So it's it's making changes and auditing the page already without me having to do it to marry up to my uh things that you guys do not have in your context yet, but will by the end of this cohort.
Um, and so it's doing the same thing there. surprises at the table. Now, there is no images here. So again, if you want images, add that into your context folder. If you have a logo, put that into your con your personal context folder and it will push that stuff onto the page, right? And so if we go down, notice that this is not blue and purple. Um, and the reason for that is one, the AI has gotten better, but two, the design rules that I have are don't ever do that. I literally in my design rules I say if you ever create a purple and blue gradient I will fire you. Um and
1:08:18Pick the best parts of each
Yeah. It's absolutely crazy. And so you can see I've got three landing pages here. So let's look at version number two. So this is version number two, right? Completely different style. It is completely different. And so now you can look at these and you can say, you know what, I like this headline better, but I like the red and white color scheme better. Right? And so then margin of back to that same color style. And that might have been actually. Yeah, that's the same one. Let's see if that's different. Okay. Yeah. So, the one is is different again. Right. And so, you have to look at all of these and pull what elements you like versus what you don't like. Let the AI make a bunch of the technical decisions for you. And then you just look at them and say, "I
like this. I don't like this." There's no reason for anyone to really have an understanding on HTML anymore because you can operate it all within plain English, right? So, these are all different things that you can make observations about. And like literally, if I was like, so I'm going to start this second page a while because it does take a second to get rolling. But if I if I pull up my narration and I say, "I really love the hero section from page one. I love the color scheme from page three, and I like the order of page two. Please combine all of these into a new master website for my real estate business.
So then finish and paste that. So that's in right? And so the worse you are at typing, the better it is for you to be talking if you can. Um, the other thing that's really useful about talking is your brain with talking does not have to catch up to your fingers. If for like for me, if I'm thinking and typing, I type, I catch up to where my brain's at and then I forget where my brain's at, right? So, I'm like, "Okay, I have to wait, wait, wait, wait. Okay, what was I thinking? Okay, now my my thought process is interrupted and the it's very disruptive, right?" Whereas, if you're talking, it's much easier to just keep everything going and then you can go back and delete the things that you said that you don't like or whatever it is,
right? It's much easier to get the information in there and then pull information out by deleting than it is to try to type, let your brain catch up, type, let your brain catch up. It's just a much more segmented creation process. At least that's for me what I found. And
1:11:09Check facts and catch missing files
so now it's architecting the listing page. It's going through fair housing compliance, right? And so it's trying to do everything it can to create this particular um page for that particular client. Now what would be even better is if we had made a decision on what landing page was best and implemented that style as well. So it's architecting, it's listing. And again, um, and I probably should be doing better with prompting for you guys, but I'm so used to just the way that I prompt because I have so much stuff in the back end. But you want to make sure that you're asking for links, asking for citations, asking the AI, how it's validating its outputs, right? That's that's a key um question. How are you validating? and and that's AI speak for how are you ensuring that you're giving
me accurate information that is uh useful to me right how are you not hallucinating okay so now it is building this and so now it did all the facts so that was the research phase Right? And so now you've created a useful asset that you've given to the seller before that that home even uh gets live on the internet. missed it because I didn't put it in there. And that is um because I'm using the app. Like I said, the app uses what's called a sandbox. What is a sandbox? A sandbox is basically an environment. It's a fake computer that this creates that allows it to give access to only a select few things so that it doesn't have access to your computer and can't go in there and change a bunch of files that you don't want changed.
Right? So, this is for uh more I don't want to say remedial users, but that's the reality is it's for users who could harm their computer by letting the AI just go wild, right? So, if you're not validating, if you're not reading, if you're not checking, stay in the app. Um, and so we'll see how this is going. however you're um however you're operating and so I will show you, right? So, oh, let's say I forgot to upload this context. Please generate a new asset with this additional information.
And that's one of the other things that's great about AI is you can move so fast that the mistakes don't matter that much. You can almost always fix or recover what you're doing. So don't don't get caught up in being perfect. Try to go fast, like uncomfortably fast, because that's the only way that you will learn to go faster is if it's uncomfortable to do, right? And so now it's reconciling the property facts against what it just wrote and it is going to create a new asset.
But like I said, this is how we plan, how we build our context, and then how we use it to create things that are actually valuable to you.
1:16:11Read it, fix it, build again
That's pretty much the class today. It's just getting a handle on how to build raw assets. They don't have to be perfect. They just have to be built and then you can iterate forward. This is exactly why we teach the OODA loop, right? like observe, orient, decide, act. Once this creates the output, I'm going to observe it. I'm going to orient where it's different than I want it to be. I'm going to decide decide how I want to make those changes. And then I'm going to tell the AI to act on those changes.
It's going to create a new asset. And then you can build it. But that is pretty much the class for today. Um to uh next week we're going to get into how to optimize. We'll do a recap and make sure that everyone's got a good handle on this. And if there's any questions now, I'm obviously super duper happy to answer them. Um but effectively building is all about context right now. So, this initial phase is a lot of reading. So, it's build it, read it, refine it, get it get this as perfect as you possibly can because I tell people And so, this is what we're looking at.
Let's see if it actually built it. Okay, so this says, "This changes the page in real ways. The fact I used facts I used before came from your brand voice writing sample, which turns out to describe a different illustrative house. The actual home card corrects four things. It's a single level bungalow, and it just goes over what the things are that need corrected. Now, I I also So it can get as accurate as you want, but it's always up to you to be the person responsible to ensure that the accuracy is accurate, right? Cuz it will go in whatever direction you tell it to.
And so now you can see it's verifying content compliance by auditing restricted terms. So it is literally taking the page that it just created and it's looking through my context to say, does he say don't use Dream Home? Does he say don't use um basic jargon that is not specific? Like what is it um what is the communication style that this person uses to convey their message? Right? And that's again why it's super important that you guys understand that everybody's business here is different. Everybody has different strengths and weaknesses.
Everybody communicates differently. your clients are working with you because of you, right? And so you want to make sure that you are in as much of the AI as it possibly can be and that your communication style is captured as well as it can be. And so now it's the same page, but it's updated to the actual facts of the property. Right? So that's how you build assets with context. So, can I have some action items?
Yeah.
So, today we are making sure we have the right context cards.
Yep.
That's what we're bringing next week, but maybe going further, we should have maybe a context card regarding what we want for our branding and our taste and whatever.
Yep, that's ideal. Yep.
And then you will put everything together, create a landing page.
Yep. Yeah. And just the big thing that I would say is play around. Like be curious. The biggest advantage you have with AI is being curious and understanding that it's really hard to screw it up to a point where you can't recover it, right? Like it's so difficult to make such a catastrophic mistake that you can't just build something that's just the thing I always tell people is if you ever get down such a path where you're like this is a spaghetti mess, just start over. like you can just start over if you have the context to do it.
I can already see the huge level of change from quad to
right.
Yeah.
Okay.
And that's Yeah. So,
so and this is the big thing is just build out your context cards and play around with it. Like try to build a landing page. like understand that success uh is it means different things to everybody and ideally what success means to you with AI right now will be wildly different than what it means to you in a year if you keep making small incremental changes. If you keep making small progress daily, weekly, monthly, it looks wildly different in like if I were to show you some stuff that I built early on. And so one, my skill level has increased, but the model intelligence has gotten better as well. And so both of those things are moving my outputs forward. And that's the same thing for you guys. As you get better at understanding what generates good
outputs, you will get more effective at it. As the models get more intelligent, they will get more effective at it. So between those two things, what I always tell people is one day sooner or later I will be able to do whatever I want because the barrier to entry due to model intelligence is coming down. My skill set is going up. Now the way they're working is my skill set is probably going up at this speed. The the accessibility of model intelligence is coming down like this, right? So like I'm moving slowly towards where I need to be. The model is looking to meet me as fast as it can. Now the model is going to keep going where I don't need this this skill set to do the things that I used to do. So what what I'm
saying is the faster that you can understand the skill set needed to accomplish what you need to do, the more of an advantage you will have for a longer period of time because eventually everyone will have the same skill sets that you guys do right now. Right? Even though we're here in this class learning about it, the model intelligence is making it so that all of this stuff becomes more accessible. But the more you know, the more you will be able to create things that differentiate you from everyone else, right? Cuz right if they care deeply about, you know, oh, I I love this area, but I don't know if there's any vendors in the area that are going to mow my lawn and do X, Y, and Z.
Say, hey, uh, AI, go find vendors and attach them to this page so that they have a painter, they have a carpenter, they have X, Y, and Z in the area. That's something else that you guys can do. If you have trusted vendors, you can make a context card for those vendors so that you can say, "Oh, this person needs a painting. Let's add the painter I know that's a really good uh painter to that context card.
1:24:20Confirm facts before it goes live
And so and so you can see here too. So this landing page has a thing that says confirming before it goes live. Meaning section about halfway down the page, there are a couple things that I need you to verify. Verify these things and then I will confirm that and we can push it live. Right? And so it says verify, verify, verify. So are those things accurate? Yes. Rock and roll. Let's move forward. Or are there changes that need to be made? And that is how you use the context cards to build landing pages.
Yay. Hopefully everyone's head hurts. Um, the biggest thing with this is you don't have to just remember that you don't have to get it all right now. I come I come into this class and it it may look like, oh, he knows what he's doing. It's my goal is not is to shortcut that fumbling process for you guys. But don't think that because it's accessible to me, it wasn't a lot of work and a lot of learning and me looking like a goofball for a certain period of time because that's that's what was required for me to get it. Just understand that just go try stuff. Try stuff. See what works.
See what doesn't work. And that's how you figure it out is just it's literally if like for me I was one of those kids that love playing with Legos. That's all this is. It is just building digital Lego blocks. That's legit what it is.
What was amazing to me is to just ask AI because when I first I thought, okay, I've created a context. I don't even know where it is. I don't have to I don't have to change it. You know, I'm like
silly girl. So, I go in and I'm like, do I even have a context card? Can you find you?
And it does. And then after I noticed there were errors in there. Some were crazy. I'm like
right. Yes. Like where'd you get this?
Yep.
Yeah. And so you were talking. I was waiting for you to finish so I could ask you why can't I change anything? And I'm like I'll just ask. And I asked and it said what you want change? Just write it in here to me. You don't even have to think about it. Just say remove this, add this.
Yep.
And I was like,
yep. And that is exactly the reason why I believe that if if you're a teacher in the AI space and you are not looking to obsolete yourself, you are trying to fight for a job that will not exist because the AI will teach it'll it's the only tool in modern history that will teach you how to use it better than I can. Right? As long as you understand how to use it, you can be a better operator with it than me. You just have to be curious and play like that. So many people think work has to be hard and challenging and all this kind of stuff. And yes, it does when you get to certain levels, but right now just play around. Just have a good time with it and try to create fun things. Like I
And then I take that stuff and I'm like, "Oh, I can use this design principle in my actual business. I can use." And so you can play and have a good time with it and end up being very productive. And that's what people kind of miss, right? like it work is supposed to be uh valuegiving and fun to some degree. Um and this allows it to be fun without all the tediousness of like how do I code a a landing page, how do I do x y and z the tedium is gone, right? It doesn't have to be there anymore. And so I I I'm a big proponent of just like explore, try stuff, trash it if it's if it's not good, and start over. And then try to get to and if even if it is good, start
over and try to get to that output with less input, right? Like see how like let's say it took you 30 prompts to get to um the final output. Take that output and ask the AI, how can I get this in one prompt, not 30, right? like review this conversation and tell me what I did that got me from there to here and then take that prompt and try to get it in less time. Right? It's it's the same with anything like it's practice and it's repetitions at the fundamentals. No one who is elite at anything has ever said, you know what, I got too good at the fundamentals. That's it just doesn't exist.
