How to fix generic AI real estate pages: How do I fix an AI-written real estate page that sounds generic?

0:00 Recap and house rules

All right. So, last week we shipped the context kit and we started to build real assets from the context kit. And that is kind of what we're going to just be working through in this entire series is how can we take the brain that we built, give it hands, give it feet, and put it to work. That's the goal. um and how do we make it actually work super effectively and and iterate that forward um just like the OODA loop uh kind of the plan. So where we left off last week, so we built agent context. We did not quite get to ship three live pages.

Um, and we hopefully with the context kit as we ship live pages, they're already going to sound like you and they will need significantly less uh refactoring, rebuilding, kind of redoing because you're you already have your base context built out. And then one of the things that we want to think about too is when it does produce an asset, if it's consistently producing the wrong thing, you want to go back and look at your context because the goal is to build the context that builds the app, not build a prompt by prompt by prompt because that's how you go, you know, you have a one-off, a one-off, a one-off.

You don't have that consistency. You don't have that ease of kind of having the brain to push things out. I have new rules. Two new rules. uh one is that uh I was spoken to by a couple people that I say things and they go too fast and then that's my brain how it works. So I am implementing the rule that if you need time to process something please raise your hand because other people probably need that same time and I can expand further. I can do whatever is required to make that understanding like actually useful. If I'm not communicating in a way that you guys are actually getting value out of my communication sucks, right? And so like the goal to of my the goal of communication is to be understood, right? The goal of receiving

communication is to understand. If there's anything deviants from that, I'm not hitting my goal. So I kind of told the story where like there have been several presentations where I've been learning stuff and I'll raise my hand and it'll take 10, 15 seconds for me to ask the question because I have to think about what the actual question even is that I want. I just want this person to stop talking so I can think of the question, right? And so it's valuable for the whole class in my opinion to have those processing breaks because like your brain works in an ironically very similar OODA loop uh sim similar process to an OODA loop where you take the information in and your brain consolidates it and tries to do pattern matching, right? And so if you just take

a little bit of processing time, you can often come up with a really great question or your brain processes and you're like, "Got it now." Right? And so that space is super fine and I want to eliminate cuz when I would do it it would just be awkward silence and everyone would be staring at me like does he know how to talk or why did he raise his hand right and so I want to eliminate that awkwardness by putting that out there on the front end so if there is time needed just raise your hand and I'll stop talking

8:53 Ryan's first websites

the other thing I want to talk about is that if I can do it you can do it um because almost every person I talk to is higher conventionally educated than I am and and so that should represent that you can do things that are uh you know difficult more difficult than than I could do through conventional education. So, if you can learn and get like even the real estate, four-year college degree, two-year college degree, six-month personal training certificate, whatever it is that you apply to do, um, you know, that should show you that you can get from A to B in a more effective way than I could at certain points in my life. Um, and so to prove that I am going to show you guys some of my first websites and also some of my recent ones. So here

is my very first website that you can see uh is not great. Uh very you know mundane color pattern um generic emojis like it's just not good. Same thing for this one. We did a cohort where we got together in the morning and you can see this page is not not good. Uh and this is front end and back end. So the technical side of it, it's not good either. Like these aren't appearing on Google. These weren't like these weren't performing for an aesthetic person. Like the person that's watching the website and the people that are trying to find the website not that is proof that where you start is definitely not where you finish.

13:45 Three moves to a better page

So, so today is the loop. So, we're going to review what we've built so far it super duper effective and then also go over the process to actually make those improvements as you go. So, there's three things that really make it better. Um, one is to figure out what your blind spots are. So you can ask the AI to audit the page and say, "Hey, what am I missing here? What are the design choices that I I should be making? What are the technology solutions that I should be using? Um what are the things that my context doesn't have that it probably should?

The second thing is taste. Uh majority of real estate professionals have better taste than I do because it's required for your guy's job um to be super effective. Um, with this the there's a big gap between generic AI content at this point and AI content that is resonant to your brand, right? And so we want to be able to not only have the the visuals of the content look just like it came from you, like it's your personal touch, but we also want the back end. We want the the SEO, the generative engine optimization, all the stuff that makes the site um useful to Google, useful to AI agents, useful to all that stuff. um we want to move that forward as quickly as possible. Um and the biggest way to do that is to figure out kind of your

taste. So like who are your customers? What does your aesthetic look like? When you're talking to your best customers, what are you showing them? How are you presenting to them? How are you connecting with them? All that stuff. And then optimize to convert. So once everything is like kind of functional, what we want to do is we want to run the context card. We want to get that right first. So we don't want to take the context card, build an asset, and then focus on building the asset. We want to take the context card, build the asset, figure out what we need to do to make that asset ideal for your business. then take that information back to the context card and try to recreate the asset in less time, less steps, right?

Because we want to build the system that's actually building things for you. We don't want to build the things that the system produces because then we're not actually fixing the real problem, right?

16:29 Find what you can't see

And so we want to find what you can't see. So your pages already sound like you. The problem is you're too close to spot what is missing. Right? So you only know what you only know. Like one of my favorite quotes is I question all my beliefs except for those I I I question all my beliefs except for those which I don't know to question and that which I truly believe. Right? So I don't question anything actually. And that is something that is really really important um is to understand that like hey I don't know what I don't know but I have a tool that can show me what I don't know. Right? And we're at the we're at the first time in human history where the tool can teach you how to use the tool better. Right? The tool makes

me superfluous if you know how to use the tool well enough. And that's that's the truth. And that's why I'm again I'm one of the only AI teachers that I've heard that basically says I am trying to replace me because you guys when you have enough capability you don't need me. You don't need anything else. You don't even need like to pay for individual tools. you can build it all with the AI stuff with a foundational model. You can build whatever you need. Um, and so you can get it here with read every file in my agent context folder and it just goes through and it kind of details what are the unknowns, what are the things that you are not seeing that you probably should be seeing that is just due to lack of experience with the

tools, right? cuz um as you understand the tools better and just like I showed with those other websites, I didn't know that I could have a context card when I first built it. I was building every asset individually and so every asset looked a little bit different. I had to reprompt everything because it's like, oh, this AI tends to have the same problems across everything that it builds, right? And so if you know that it has the same problem, you can pre-solve it in the context kit. I did not know that. So I was trying to solve it every time differently with prompting.

And so then you build those fixes into your context kit and that is what is going to move your ability to move faster forward. Um because the more that you impact your context kit and keep that updated and kind of do that, the more that you will be able to spend less time fixing things on the back end because you've fixed it on the front end. You've you've fixed the defect. Like it's very similar to like if you think about um a production line, right? Like if a production line is producing a toy, let's say, and every dump truck that it produces has three wheels. Well, the the solution is not to just add a fourth wheel at the end of production.

The solution is to look into the produ the actual assembly of the toy and say, "Okay, well, where can we add a fourth wheel to this toy? Why why are dump trucks only having three wheels?" Right? And so it we're looking at fixing systems, not fixing um individual assets.

19:41 Taste is knowing what to cut

So, one of the biggest things is that taste is really knowing what to cut because AI often overdoes stuff. And so, a lot of times, uh, understanding what you need to do is how do you eliminate things? Almost everybody wants to add stuff. Very few people think about eliminating things because it's all we just we're in a society where you just add add, right? versus, oh, if I eliminate this, it fixes that whole problem. And so, you basically want to really understand what is appealing to your audience because I if like when I present to someone, they will get a different vibe from me than they would to the people in this room, right? And so you want to understand what your audience wants from you and then you build that into the context. You don't want to build the

things that I say into the context. They can be a good helpful starting place. But if you know like this I'm a luxury realtor and this this is my client base and this is how I interact with people. 100% you need to do that right because taste is different across businesses, across towns, across cities, across everything. And so if you are performing well, lean into that and keep performing well by doing the things that you do well. If you're not performing well, you can look to optimize and make changes and find your blind spots again. Um, so there is very generic stuff. Uh, three headlines, fighting for attention, five paragraphs, two nobody reads, welcome to my page, filler intro, all that kind of stuff is very, very generic. but one headline that says who you help and where. And that's one of

the benefits of AI is you can have one headline, but you can test five pages. So you don't have to have the perfect headline on your page. You can build five different pages and test each one to see, oh, this headline works for this customer group. This headline works for this client group. This headline works for this client group. So you can build assets much, much faster. And that's one of the reasons to build that context brain and make sure that it's it's functional.

22:00 Have another AI tear it apart

So, and then one of the things we've talked about a little bit is something called adversarial prompting. Basically meaning that you take one AI and have it argue with another AI. Um because if you if you ask an AI to validate its own work, more often than not, it will give itself more forgiveness and favorability than it probably should, right? But if you ask a different AI and it can be the same, like it can be ChatGPT but just in a different window. You take it and you say, "Hey, an AI gave me this. I need you to tear it apart." Like really rip it down. That is one of the most useful things you can do because you will see so much truth in what that adversarial AI is saying because it's not it has no context just like with a

person right like your clients if they bring you a ChatGPT thing they're locked in they like this ChatGPT thing they wouldn't be presenting it to you if they didn't think it was valuable right and so if if they get locked into it the it's one of those other weird parallels with human thought and AI and I don't know why they connect in this way, but the AI is also going to be a little bit more like defensive about its like, "Oh, I made this. How dare you?" Right? And so, this is one of those things that is super duper useful is just having AIs argue with each other.

And then so the biggest thing with like landing pages and websites are the headline, the copy, like are you speaking to your your exact client? You can have a really simple website. And I've actually tested this out where if you do just a plain black and white simple, super simple website, it will stand out if the copy on the page is good. So, if you've written good sales copy that connects to your audience, people will stop and read it even if the aesthetics aren't beautiful, right? They because this is what they want. If you connect with the transformation they're looking to get,

24:34 One headline, one easy next step

So, you can sharpen the headline. Um, and again, this is super simple prompt that's just, you know, rewrite the headline so a stranger in one line knows who I help, where, and what to do. Um, so they know who you're helping, what location you're helping in, and what first steps they need to take to work with you. Are they going to call you? Are they going to email? Are they going to text? Are they going to fill out a contact form on your website? Are they going to be referred by someone? There are so many ways to acquire a client that one of the things you really want to do is make sure that they there's a very easy next step. If you give people the next step, they will often take it.

week. So if you give people the next actionable step, that's often what they're missing. They just see if you give them a problem, they see a problem. If you give them a problem and solution set, they see a pathway to it getting resolved, right? And that's the same with AI. So if you just you know kind of audit one by one that is the simplest way to do it. Um now what we're going to do is actually build stuff and next week we're going to talk about traffic distribution like all that fun the kind of the meat of the class. And

26:52 Live: load your context into Claude

now what we are gonna do is go into Claude. Love that. Um, and actually go and build out an asset using the context kit. So you can do regular Claude or Claude code, either one. Um, I will do regular Claude today, but the easiest way to do this is to go to artifacts. And I think all of them have an artifact. It's either artifact, a project, a Google calls them gems. But basically, this is how you can create persistent context in a super duper simple way. So, what you'll do is, and you can I'll get into Cowork at some point, which is really, really helpful, and I think it would be super duper helpful for um the real estate professionals, but we'll do the regular one today. So, let's get cooking. Pick an artifact category to start building

your idea from scratch. So, I like to start from scratch. I'm weird. Um, but it's now going to give me some bas I usually it's a couple questions. It's like what do you want to do? How do you want to do it? And so it great. Give me a blank canvas. Let's figure out what to build. Point me in this direction and tell me what to do. So I am going to add files. And I am going to go context kit. And that doesn't work. Lovely.

So I will do it this way. All right. Well, I don't find it manually. One of the problems with doing a bunch of projects is you have a bunch of projects. So this is my build things that build things context. So I'm going to open that and I'm going to say here are my company context cards. analyze these and help me create a road map to building useful assets for my company. And so we'll send that.

Can I ask you a question?

Absolutely.

So what is artifacts?

So artifacts is basically where you can shove a bunch of context and have it right there all the time. So, as an example, if you were just new started a new chat every time, you would click add files and then you would upload your context kit into that chat and then you would start whatever project you're doing in that chat, right? With an artifact, you can upload those things to the artifact and now they're there permanently.

So now you have the context kit. And so like you can do that for clients where you put the client context kit as one artifact and then you upload your context kit. So you've got both of those and then you can individually add the deal context instead of having it be persistent, right? So you can take like let's say there's three houses that you think would be perfect for this client. You can upload those three into your client artifact and it'll build assets specifically to your brand to that client for those specific homes, right? Or if you do it um for yourself and you want landing pages and email newsletter, you want those things, it'll have all your contacts right there and you can just you don't have to. It's basically solves the manual step of

uploading each time your contact stuff. So when you start a new chat, you're directing it to one of the artifacts.

Yeah. So you would go to artifacts and then you'd open the artifact and then you could do a new chat inside the artifact. So the kind of the way to look at it is

if you have if you have a football stadium and the main field is the main chat, each seat could represent a different project, right? And so like you might say like, "Oh, I want section 8 of my stadium." And so you would just upload the things that would be useful for section 8. And then I probably should have used a different section now that I'm thinking about it. Um but the um the context would be exclusive to that section. So that if like if you know that there's restrooms, you have to go up to this concourse and go left. And that's only section 8 can access these restrooms, right?

that would be specific to that section. If there's a a hot dog vendor that only wants to sell hot dogs in section 8 because no other other section consumes as much food or whatever it is, you would have that piece of the stadium carved out and that is where you would have the context for it. So every time you go into a new chat, it's just it starts fresh. It's brand new. But section 8 has season ticket holders, right? And so that would be these persistent context pieces are like your season ticket holders and then this tickets for sale are like the prompts that you put in because they change on a weekly basis, daily basis, per use basis. And so whatever the things are that you want to be consistent across the board and you may want um you may

want to have a different artifact for email newsletters than you do for website building for each client. like you can basically almost use it like a a micro CRM to some degree.

Okay. And then that's um so then when you have projects you're pulling in the artifact.

Yeah. So so if you're using ChatGPT so it when I say artifacts that's uh what Anthropic calls them. Oh, because I'm wondering because I so I do do ChatGPT, but I also sign up for Claude, right? Yep.

So, I've been kind of transitioning into Claude, but I have a really hard time letting go of chat.

Yeah, I understand now. I get it.

It knows me so well. Really difficult. So,

but so, for instance, yesterday I had it create this um like a um a cover letter, a cover picture on my Facebook. Yeah. Like

basically my it's like kind of a bigger business card, whatever.

Yep.

So, then I gave that to Claude and Claude was like, "Oh." And so then it created this whole website for me. And it was funny because now that I'm remembering what we did last time, I'm realizing I started this project incorrectly and because I had already started that landing page. And so now I'm like, I don't really know how to merge the two. And now and now I'm like now I think I would have been helpful if I had done an artifact. But it says projects here in that's why I'm like

Oh yeah. So, I actually have not gotten too deep into projects, but I think it's it's very similar. So, if you would say like um client And this is super simple, too. So, this is your system prompt. So, what you would do is you would go and prompting is I don't want to say it's dead, but it it's very very simplified at this point. Um, the thing you need to know about prompting, like literally the only thing you need to know about prompting, understand as clearly as you possibly can the exact outcome that you want to get.

And that is how you prompt. At this point, you very clearly define the outcome, very clearly define the win state and any constraints that you might have. And then you ask AI to create a prompt to create whatever it is that you want.

You said, okay, so to describe the outcome and what did you say?

Yeah. So the the outcome or the win state.

Yep.

Okay. any restrictions or constraints that you have

and any like proprietary knowledge that's not publicly available. So, anything that it wouldn't know by searching the internet

and then you ask the AI, you say, "Generate a prompt that will achieve this." And so, the most important thing for prompting is that you have a very clear understanding of the outcome that you want to get at the end of the movie. Not the process required to get it cuz things have changed so much.

Yeah. But the the like what is like I want a website that generates a thousand visitors per month and is targeted towards uh local luxury real estate. Like if if that's the very clear outcome, try to get it right. And you might say like, "Oh, I don't know how many visitors I should try to get." Go to the AI and say, "Hey, what is a reasonable visitor level for a new website and how can I ramp this up effectively?" Okay.

And so actually, yes, I probably would move to projects over artifacts now that I'm seeing exactly what this is. Um, so

and then what you want to do is this is where you would add your files. So you would click here and you would go upload from device

I see files

and now I have all of my files in there and you can see it says 1% of project capacity used. That's one of those things that is we we talked much earlier about context window management and some of those things. This is the context that you're using. So like you want to make sure that that doesn't get too full because if it does that's where the performance of the model starts to degrade significantly and I don't know how long that will be true for uh because there's a lot of insanely smart people working on memory context windows and like how to solve kind of the needle in the haystack problem. Mhm.

And then it will also create memories in the chat for you. And so like it'll create once you talk with it a little while, it'll start creating memories and adding them to the context. And so you might look at the memory and be like, "No, that's completely wrong. That was just a one-time thing. I need you like go delete that." Or it'll start creating things where you're like, "Oh, I didn't even know I was doing that. I should apply that more deeply to the context card." Right? And so these three things are probably the simplest way to turn your context cards into into assets because you can scale this almost infinitely. You can scale it to specific clients. You can scale it to specific deals. You can scale it to just you. And you can build out whatever you want

underneath that. So, like if you want to send clients a unique landing page for a deal that is specifically designed to them, you can have it set up so that this is has their client context, the deal that you want, and your context, and it'll build it exactly as you want. And you can make adjustments to where it's like, oh, the client just gave me this new piece of information. You can add that to here, and it'll kick it to a memory. And I don't think you can update the files directly through the project, but I would need to play around with it more to validate that for sure. Um, but then all you would do is you would come in here and you would say, uh, create a website for me.

Super simple landing page. Show me in HTML. So now when it's creating the website, it's factoring in my context. It's not just building the website like Claude would. It's not just taking like, oh, you're a general person. Um, I'm going to pull the things I have from my memory and push it in there. Artifacts are one of the or um projects are one of those things that are super duper useful for if you have multiple businesses and you need to silo things. Yes, that's my problem.

Yep.

Like that's where I'm a little um

it's like ChatGPT has all of my life.

Yes. Yep.

And now I moved to Claude to just real estate here. Yeah.

And then um

so I had for I had forgotten again that this context card thing and I'm like, "Oh, dang. I should have done that first." I'm like, "I need to write like a like a step by step when I need to go do something like where am I going to my steps

because I have completely forgot that I created all these.

Oh yeah, I have a full like the way that I think you guys should be trying to end up is basically an asset creation pipeline. So it's not um you know a one-to-one-to-one. It's you basically put this the information that you want at the front of the pipeline. You shove it down the pipeline and then it gets assembly lined basically is what it's doing. Taking your context, adding doing what it needs to do and then it shoots out a an asset at the end. We're looking to create um factories at this point, asset creation, factories or assembly lines or pipelines, whatever the the terminology is. Mhm.

And so you can see I asked it to create a website for me. Super simple landing page. Show me an HTML. So it says building the BT2 landing page as a single HTML file, right? So it's building that file right now with the context card. So it's not if I would put this in just general Claude or ChatGPT, it would be like, okay, bro, what what are we doing here? like you have so much stuff going on like what am I actually building the landing page for? But because it's got all the context, it knows exactly what it's it's supposed to do.

And you can see here it said reading front-end design skill before building the landing page. So skills will be something that we talk about in probably the next class or the class after. But skills are one of those ways that you can have again more persistent context.

42:24 Review the first draft

Oh, and there it is. So, it's not spectacular, but it has all my brand concepts. Um, it's got a readiness audit and like I, you know, I would obviously make significant adjustments, but it also has, you know, my social media if I was logged into LinkedIn, um, it's got to link right to the page, right? And so, whatever social media or whatever assets that I think are valuable for people to see, that's what I want to put into the context. Um, and you can see it's got a little animation there. Um, build your business. Like, it's a decent site. I don't necessarily love the the white background or the um the grid on the background. So, what I would do is I would go back to this and I would make an adjustment, which let's do that.

Um, so now we are officially in the iteration stage. And so but yeah I ChatGPT's um it is useful you can use ChatGPT to do the same thing like there have been um people who have gotten really useful AI outputs with lesser models meaning like older older uh large language models older older models so like this is Opus 4.8 but they also have um Opus 3, right? And so if I wanted to test how good my context is, I can go to an older model and ask it to do the same thing and see if I get a high quality output because the worse the intelligence is, the better your context has to be to get the output you want.

can see I asked it to change the grid. I asked it to change. Um, and now

45:14 Fix the context, then rebuild

what I would say is help me implement these design changes in my context so you don't make the same mistakes. And so you're always asking the model to fix the system rather than the asset and then you're asking it to rebuild the asset, right? So if you just fix the asset, you will spend your life fixing assets. If you fix the system that builds the asset, you'll be able to do other stuff,

right? And so and you can see here it says uh recognizing read-only files versus canonical project context. So it's evaluating um it cannot touch the markdown files. So it can either give me new markdown files or it can make those adjustments. Yeah. And you can see here it says the best approach is to give them a clean markdown block they can paste directly into their project files. Probably the brand voice or markdown. And I this is something that not a lot of people do, but I highly recommend opening the observability and understanding how the the model is thinking because

I like that it does that.

Yeah, 100%.

How it's telling you what it's

Yeah. And that so that

like that a lot

wouldn't happen if I didn't click it. Right. So a lot of people just see what it's doing and they're like, "Oh, cool." But if you go up here and you click that, it opens up what's called the observability and you can see its its reasoning and its logic and you can identify like, oh, this is where it went wrong. Because the second you can identify where something went wrong, you can start to fix it for real.

Yeah,

it's pretty incredible.

And so you can see here it's got design defaults markdown. So it created a new markdown which I can then download as a markdown file. I will save that. I will go back into projects. I'll open up Bob. I will add the design defaults. Now, what I probably should do is take those design defaults and take the brand voice, put those two in its own chat, and ask it to make sure that the design defaults are what is being represented in the brand voice. And then it'll create a clean markdown file. What I would do is I would close, remove that brand voice, and then upload the updated one. Mhm.

And something else that's really really helpful is ask it to add version numbers to your project or to your outputs. Right? So it you that way you know I'm on version seven of my brand context and I've changed X Y and Z.

Oh, so that way you're like okay I like this from version five. I like this from

Exactly. Yeah. It never it never locks you into uh it's again it helps you own your context, right? Where if you're like these are all legacy contexts and then you're like oh I overcorrected on version 8 I I would should just go back to version seven and start again. You can pull version seven and you still have it. You're not changing the the real thing that is impacting your business on a day-to-day basis. you're you always have a saved safe copy of it. And like for me, I will probably delete versions as I get like three or four versions past. So like if I have eight versions, I'll delete version one, two, three, four, five, and then I'll keep six, seven, eight.

And so now theoretically I've started a new chat with the same prompt I put in the first chat. Now ideally this new chat will create something much closer to this website. than it did the other one. That being said, you never know what's going to happen with an AI. It's non-deterministic, so it's it's, you know, it's out there. So, we'll see what happens. But the the key mindset is that whatever the output is, you know how to you have a clear outcome of where the output should be and you understand how to get there. And it's never It's it's so hard to get perfection on the first try and it's so easy to get perfection with repetition, right? And so if you just focus on it's not where I need it to be, but I know

how to iterate it forward. If you know how to do that, you'll always have the ability to get what you want eventually. It's just the understanding that it there is an eventually, not an immediate. And so we'll see if this builds the um the actual website that we want. Okay. So you can see it did. It built red, white, and light blue. It removed the grid background. And so the design aesthetic is much closer to what we actually want to represent the brand, right? And so that is honestly the biggest the biggest unlock is understanding one persistent context and two that you should be editing your context to get the assets you want rather than editing the assets that it creates.

Yeah. Okay, that makes sense.

We're always trying to build a better foundation. We're never trying to reshingle the roof,

which that foundation is super important as I'm like I'm realizing and because what it can build is pretty incredible.

Yeah, 100%.

You have to know how to direct it.

Exactly. And that's why I say like clear outcomes are the most important part of prompting.

But sometimes my favorite thing is that like because it knows you, it like sometimes you don't know what to ask.

Yep. Exactly. Yeah.

So that's why when you're saying ask this AI to ask that, you know, to do that. So, that's kind of where I'm at now because I had that one AI create this one thing, which basically gave this one Claude a better understanding of what I'm trying to do.

51:57 Questions: revising, publishing, connectors

And then now what I have to do is I h I really like what it created, but now I have to go back to the original context to see if all of those are together. And I don't want to do it all myself, read through all of it. So, I need it to do it for me.

Yeah. You can take the asset that it created that you really like and you can say evaluate this against my context. What does this asset have that I did not mention in my context or that I did not create in the context and fill that out so that we can get this asset by asking the AI one simple question rather than the 30 questions that it took to get the real asset. So then once you're um so you create a landing page and then within my landing page I have different tabs.

Yep.

So I want to get the landing page as perfect as I want it through this before I then

put it onto the website.

Yep. So, I created my own I I've got my domain yesterday and I'm like and I got the git

GitHub. Let's go.

But then I

proud of you.

I got to the point where I needed to like get from the HTML that I created into the GitHub and I was like I don't know what to do and I stopped which I'm glad because I didn't really like read through the whole thing to perfect it before I even did that. So now I'm like the cart was definitely before the horse. I need to finish that first.

Um so then my is is once I get the landing page done, can I like how do I go to the different tabs, like how do I direct it that way?

Oh, as far as like adjusting the the tab content.

Yes.

So, you would just tell it this basically, hey,

go to this tab. Yeah, we're working on this specific thing. Don't touch anything else on my page, X, Y, and Z. And then what you do is you take that output to an AI and you adversarial prompt again, right? And you say identify that I told the AI to do this, identify if it did that and if it touched anything else on the page as well.

Um, and you can get it to kind of, you know, debate itself and then be like, oh yeah, it touched this page, this page, and it but it also did exactly what you wanted. And then you go evaluate. And what I found too is sometimes I'll look at the pages that I didn't want it to touch and be like, "Oh, I'm I'm the one that was wrong for it not wanting to touch it, right?" And so you kind of have to always be evaluating and quality controlling. And that's kind of the the human's primary job with AI now is understanding how to plan the project and then how to quality control the project. The actual build stuff is it coding is very very low cost at this point. like the actual code code is super cheap is what um kind of the the

tech motto is. But design principles, taste, um all that kind of stuff is where humans are still very insanely valuable. Especially if you ha have an audience that you've already connected with

and you know what works,

right? Then your taste is even more valuable than than any part of the project, right? And so this is we'll go into this uh since we've got a little bit of time. You can go into connectors in Claude and I think ChatGPT has something similar.

Yeah. Yeah. Yeah. So in ChatGPT

access.

Okay. Where is ChatGPT tools?

They have connectors. Okay, there's the app. I know the web the web version has connectors.

Are you kidding me? I can connect chat to this.

Mhm. So like you can go to

Oh my

you can connect it to So what I do is I always connect it to GitHub and Netlify and then it has I can tell it push to GitHub give me a Netlify test website and it does everything. It's an end toend.

I think that's where I put my content card was GitHub, right?

Yeah. Yep. Yeah. And you can have and if you connect the GitHub connector, you can have it pull it right from GitHub every time. And so that's another way to basically build your own version of projects. And if you do it that way by putting your your context card into GitHub, you're not model locked, right? So like you can call it from Perplexity, she can call it from Claude if it's in GitHub and it runs roughly the same effectiveness. And so that's how another way that you don't have to rent from the model. You can own the business and it's in the the GitHub. And with the GitHub, you can actually uh edit the files as well. So you don't have to um one, you don't have to do uh the version control you because GitHub does it

automatically for you. That's one of the big reasons for it. And then two is you can actually audit or edit the um the actual files right in there. And so this is when I say create a pipeline, this is creating the pipeline, right? You have your context, you have GitHub, you have Netlify, and so you can go from prompt to live asset in like two steps.

Whoa. Okay. So, I need to spend time to add these connectors.

Connectors are super valuable. Yeah. So, what I always tell people is your context kit is the brain. The connectors are the hands and feet.

Say that one more time.

The context kit is the brain.

Connectors are the hands and feet. So you have the ability to go from the chat window to real work using connectors and it didn't used to be that way. When this start when this stuff all first came out, you had to still have knowledge on how to build API and like actually have things talk to each other. Now the the model and this is um this is also what I mean when I say you only need the foundational model, right? Because if you have the right free connectors with the model, it can do anything you want it to. And this is why AI is eating software. Like this is people say that software as a service is dying or dead.

It there is a a useful niche out there for it, but the more you understand how to use the foundational model, the less you need any type of external software outside of one of the big four models.

but yeah, this is so I was I said a little bit earlier with that agents are a brain uh outcomes that you want to get and then the tools to achieve those outcomes. Those are the three things that all agents have in the simplest terms possible. And so you connect these add connectors, you can click browse connectors

and I mean there's hundreds these are all connectors, all of them. And I can browse connectors in my Perplexity.

I think so. I think Perplexity has connectors. Let me double check.

All your login.

Yep. That's

is very hard for me.

Yep. The nice thing about the connectors is once you log in, they're connected. And that's the thing that's like it's done. And and that's that's one of the useful things for me too is like every time I would have to go log into something, I'd be like, "Okay, what do I need to do? Like what email address is this signed up under? All that stuff.

I know. I have a lot.

Yep. Yeah. Exactly. Yeah. Yep.

I I need like I almost need to connect it to so many things, but I only so I can clear out things because I I have like a bajillion emails like having that's Outlook's more of a pain in the butt than Gmail. But yeah, you can use those connectors and it'll get the inventory and like sometimes so there are some connectors that don't have edit access and that's good and bad. I think the Claude MCP does not have edit access, but I have done a lot and we'll go over Cowork in probably a couple weeks, but I've done a lot in Cowork as well, which is like their basically Claude's business agent. So, if you need space on your computer, if you need an IT professional, go to Claude Cowork and say, "Hey, I need you

to function as an IT professional. Help me figure out the most efficient way to run my computer. Help me organize all the files in my downloads folder. Help me X, Y, and Z. And you can click the project and you can add the folders that you want it to work in.

I need that real.

And so connectors and Cowork are two of the most useful things possible. And they these like I said these are the things that literally take it to where you don't need other software for the most part. Um you can connect it like

1:05:12 Let AI do the steps, you review

And if I have enough connectors connected or the right connectors rather, I can say, "Hey, I need you to identify all the actionable items from this pro this conversation and do as much of it as possible, right?" Because a lot of people are like, "Oh, I have the AI and it kind of gives me the steps to do." The goal is to not be doing the steps. The goal is for you to be, like I said, quality controlling the outputs of the AI even within the steps. Um, and so like I never have it send emails, but I have it draft emails so that I just need to read and then I can send.

Yeah.

And so, going quickly. I think that the what you said in the a little while ago was key is

teaching us how to be project managers.

Yeah.

That's really where it comes boils down to. Yep.

And that skill can take you to all facets of your life

100%.

You can create that,

right?

You understand that. If you can manage a project well and you understand how to iterate it forward to where you need it to go like that's with AI that's all you need to to be able to do is just the OODA loop

and a lot of project managers well their weaknesses i.e. I have this problem is delegation.

Yep.

So if you're but that's but if it's as easy as using your AI tool to delegate

right

then that's that's a lot of really good human skills that we all need for almost anything. Yep. And if you're delegating to something that is basically an extension of you, right? Like that never calls out sick. It never uh is going to talk back unless you ask it to.

And it knows you better than yourself,

right? Yeah. 100%. Yeah. Yep.

It's scary.

Yeah. And it's it's infinitely smarter than anything that I've ever created, right? Like it's infinitely more intelligent than what I would be on my own. and using it as an amplifier or a multiplier is is where the value is. It's taking that and being like, "Okay, this thing's smarter than me. How do I get the most out of it for me, right? Because it it I I've said this a couple don't rein it in, it's going to go in whatever direction it wants because it knows there's so many solutions out there that you don't even have an awareness of, right? And so you can be like, hey, you can either ask it, hey, what are the solutions to this problem?

Or you can say, here's the solution I have. Execute it. Make sure you're working within these constraints and that you don't do this because I've tried it and it stinks,

right? And

Yeah. Exact. And that's the cool thing about AI is and I've said this is the main problem and also the biggest strength is that it can do everything

and so it can work with your work style. It can also work with someone else's work style and but and it can identify hey this is how I can help you. And so as long as you have the ability to have it work with you in that way versus going in all directions and kind of scattering out like it can be a really like it can be a powerful tool for helping to execute. It can be a powerful tool for like some people don't have any ideas and they need to something to to get rolling off a blank canvas, right?

It can help with them. It can be like oh you're stagnating with ideas. It'll do the same thing and it's it's so scalable as and it can be scaled in the same way for you personally as long as you understand how to enter the problem solving loop and what the end state is when you can exit it.

Yeah.

Okay. That OODA loop thing is like hard to

No. And and that's people do the OODA loop and you what you'll find is like you'll find yourself like calling it different things. You'll be like, "Oh, I need to do research." Okay, so you're um you're doing step one of the OODA loop. You're observing what the reality is. You're orienting to the re and you just kind of do it naturally, but you may call it something. You might add an extra step. You might take a step away. Like whatever it is. I also am really bad. I have when I remember I have to write things down to remember because I never go back to this and never look at it and then I like need it to be organized in a place so I can like re reference it later because I can't control F my stupid notebook.

You can take a picture of it, upload it to chat and then

this is what I need to start doing. Yeah,

because I when you were saying that for the prompts that was really helpful last time And so,

yeah.

Well, and I have the same issue with being like I don't know where my context cards are. I don't know how to get them, you know. So, I had looked up the Cowork thing on here and then it says for file organization, do a lot yourself and then Claude Cowork will automate the repetitive parts. Set a clear structure. So, it says pick one home base. So, just lay it out for me

and then inside it create top level folders, real estate, travel, personal blah blah blah blah.

Yep.

And then use a clear folder hierarchy. So like real estate listings 2026, you know,

your context card.

No, this is for if I want Cowork to go in and organize my files and things for me.

You can copy and paste that, put it into Cowork, and it will organize your computer. And and that's the clarity of outcomes, right? We talked about clarity of outcomes are so much more important than prompting right now because the AI will do the prompting for you. So if you have that clarity of outcome like this is the organizational style that works for me

clog co do it.

It's easy to remember because I know that it's real estate 2026 and then you have that and then I would use my maybe my address my client name you know.

Yep.

Uh and so you tell it all that and then it gives the active or archive and then you just put that in and tell Claude to go.

Yep. And Cowork will do it for you. Yep. But defining the object my what the outcome is is sometimes the challenge for me.

Oh yeah. Yeah. I spend hours defining clear outcomes and

that is my hardest part.

Yep.

But you can use this to brainstorm with you too. Be like okay this is my hardest part. And then it's like you get into conversation and it helps you.

Yeah. Exactly. Yeah. Yeah.

Could I go back to my conversation? the the vomit the poor vomit I did to chat the other day was like organizing my thoughts

could I then say okay

tell teach tell me what's how to use Claude Cowork

to because I don't know how to even use Claude really to to do that

100% yeah you can go back into that and you can say like what so I will often ask the AI I will say this is what I think I want

this is x y and z need you to interview me and make sure that this that what I'm telling you is actually going to get what I want and it or do I even want the thing that I'm saying that I want and there are several times where it's like no dude you you're way off base like you need to change X Y and Z of the thing that you want to get the outcome you want. It's like and you can think about like too like you can start as the most remedial level possible like you can be like I am a realtor what do I need to do now and it will just start talking to you and then you like as a seasoned realtor you can be like no that's wrong x y and z this is like what do I need to

do like you can ask it questions like what is a reasonable uh traffic number for a website that you've created or what should the goal be for a landing page like what should the goal even be?

Yeah. And then like here it says, "If you will share how messy your things are now and I can outline one exact Cowork prompt to start cleaning up safely, you know, like even could tell just from me talking through."

Yep.

You know that and if you see my desktop like I have traumatized people.

Yep.

I know. Having to clean up that stuff is like

Yeah.

to a drop. Yeah. Yeah. I had one guy that came. He's like, "I've never seen anything like this. I've never seen it." And he's like, "I'm on the computer." I've never seen anything like this. I save everything to desktop so that I can find it.

Practically find it, right?

I know.

So there's a physics principle called entropy which is basically that uh life and all things kind of move towards a state of disorganization without effort to keep organized. Right? Like it's in in physics it's basically how well does something manage heat, right? like how well basically how much energy can you put into the system without the system exploding, right? And so entropy debt on a business is basically where are my context cards? I built them. They're useless to me cuz I don't know where they are, right? And so like that's entropy debt on a business. And all businesses have some level of entropy debt. And this is one of the things that can kind of eliminate entropy. And if you look at what that does over the right? And so like entropy debt is one

of those things that a lot of people don't think is as important as I think it should be. But I personally value like this stuff being an organizational utility more than almost anything because it it saves so much time from disorganization. And I'm a very disorganized person. I'm one of those people that's like I am a productivity tornado and that leaves just a bunch of junk in my wake. Um,

something you did the other day that I just realized I got you this is the multiple desktops.

Yeah. the AI?

So the way I manage it, and this may or may not work, you kind of have to play around with it and just kind of see works best for you. The way I manage it is during the course of a normal workday, I will keep everything typically in one desktop just because that's I'm usually doing the same stuff a majority of the day. Um, now if I'm the way I kind of do it. And like like for presentation, I have all my presentation stuff when I come in here on one desktop so that I don't have to bounce around to my other stuff and my other stuff remains undisturbed. It's still functional and able to perform.

But I have a desktop that is set up just for my presentation stuff and it everything I need to present is here. So like I have the Claude, I have um the websites that I wanted to present and I've got the actual presentation somewhere.

So each one of those things you've got going on up there is a desktop.

Yeah. So this is one desktop and this is two. I was I will get up to pro sometimes fiveish desktops if I really want to be like focused in.

And what is the button that does that?

You see this guy that has this two little windows?

It's like a two little window bag.

I don't have a two little window.

Oh, yes I do.

Yeah.

Okay.

So, you see you put that in the plus. That's how you So, I I'm thinking cuz

So, um, I was thinking that the the desktop option I didn't when you did that, I was like, oh, that's what that

I really enjoy it. Like I love having the ability to have my work siloed as much as it can be with the volume of work that it is. Um and like if I'm do and like you can divide it up different ways like I'll have the reason I do one desktop is predominantly cuz a lot of my work is codebased. So I'll have all my coding stuff in one place rather than all my different businesses. Right? So you can chunk it by like activity that you're doing. You can do it by business.

there's the however it you know works for you and it could change on a day-to-day basis right like you and the more you explore the tool the more you'll figure out oh this is what works for me on a just an everyday basis for the most part or like oh I'm like there have been seasons where I've been like okay I'm just coding for like 16 hours a day is what it feels like right and so it's just like one desktop with a bunch of coding assistants open or whatever it is

well reason why I like that is that if I have my Outlook is where I have four or four different emails. Um, but I found that if I see it that there's an email,

then I go in there, right? That's if but if I if I want to put that email in that desktop and I'm working on something else, then I won't get distracted by the email see it.

Yep. Exactly. Yeah.

So, that's where like I really need because again I I I just like squirrel squirrel squirrel all day and it's like and it makes me feel like I do nothing all day. 100%. Yeah.

All day. What am I doing? Absolutely nothing.

Right. Yeah. Yeah. Exactly. I felt like I did a lot of stuff today, but I have nothing to show for it. What is going on? Yep. 100%. things I did.

Yep. Yeah. Exactly.

Really like, okay, that's how I really need to start working.

Focus is a superpower. Um, I was, so I

Yep.

And the desktop tool, I think, is a really great way to like lock yourself into that. That's that's a great that's exactly what I mean

cuz I'm the same way like after a certain point I start looking for things to be distracted right and

and I pro procrastinate the things I know are going to take a long time.

Oh 100%. The things I don't want to do. Yeah. The soul crushing work that like has to get done that long.

Yeah.

And I've learned that it's really my email is the worst. I I can spend all day just working on emails.

Oh yeah. But then you stack your emails with that you're going to get right done later, you know, and so then you're like then you're building up.

Connectors.

Connectors are one of the things that like cuz you can have it be like I had it pull all the context from each one of my clients that emailed me. So I'd be like find every email from any email that had this email address or this email address. Pull that into a contact folder for me. Copy it. Pull it into a contact folder and it would Yeah. And it would create markdowns of the contacts. So now I have all of my client context in those conversations. I edit it out and like fix the things that are different, whatever. Um or have the AI do it. And then now you have the client context card without really honestly doing too much work because it pulled it all from your email.

And so that's constantly what we should be looking for with AI is like what are the efficiencies that this can gain me over the manual process.

And the only way to do that is to understand what the actual workflow and the manual process is. And people want AI to be this magic button, but if you don't have understanding, it's it's

And it's all about your inputs. If you have good inputs, you will have good outputs. But if you have terrible inputs, doesn't matter what you do to the information, you're going to have a bad output. Um, and so if you really focus on input goals and like un and deep understanding of what you're doing, um, we're at I think we're at a space where you have to go a mile deep, not a mile wide at this point because anyone can go a mile wide with AI and it's super simple and you can do it in 10 seconds. So if you have no human expertise, you're you you're not creating value in the marketplace at this point, sadly.

Like I've said before that AI just obliterates mediocrity. them.

I started going through my notes and now I'm like why am I doing that? all those things.

And

although I don't really know like I think I have the worst level of shorthand. I don't even know how to do shorthand. I just I just know my brain. I don't finish words. So it's like this thing is not done.

Yep. Yeah. And it if it can interpret my handwriting, it it can it can work miracles.

Okay. Yeah.

And what you can also do is you can um give it some context. So you can uh interpret your shorthand for it a little bit like give it a key of like hey on this first page you might see x y and z shorthand this means this this means this this means this this is kind of how my thought process on shorthand works try to figure it out

I don't complete words

yeah you could say like hey use autocomplete and you know extrapolate based on what you know of not only the context text of the page, but the context of me

figure it out. And more often than not, it figures it out. It is the smartest like friend you've ever had, right? And it's it's super useful. Yeah. And so many people just use it as like a oh, you're just a really nice Google, right? Like they'll just use it as a search bar and they're not using it as a co-intelligence to do brainstorming, to do outcome processing, to do all the things that can actually move the interact with it in that way. And that's that's one of my big points is like it is a co-intelligence at this point that actually can impact things in the real world. It's not just like oh I have a question anymore.

And that is the exact reason why it's important to have context cards. That is the exact reason.

That's right. Yeah. Exactly.

You do never want to be locked into an ecosystem of anybody's choosing.

Yeah. I love it.

Yep.

I'm very much that way.

Yeah. The same. Yeah. Like I tell people all the time because they're like, I love And then they insert whatever AI it is and I'm like, you should love being model agnostic and moving quickly. Like that's what that's what we should value as it relates to AI.

I'm in.

Let's go.

That's right, baby. I'm like I gotta get that right.

That is crazy.

Yeah. Yeah. Really quick really good engineering companies like Claude Desktop almost releases a new update every day. Like I almost have to click their update button every single day, which is annoying, but I would rather be annoyed by that than by it having not features, right? like

okay sending it to them right now. No iPhone for me. I just got

freedom. Freedom.