# 150 I Built A $100K Research Assistant With Claude Code (for FREE)

# 150 I Built A $100K Research Assistant With Claude Code (for FREE) Speaker 2: Claude Code is hands down the most powerful way for non-technical people to build real software right now. But most people are way too intimidated to even try it. So in this episode, I brought on Adam Sandler, and not the Adam Sandler that you're thinking of, but Adam Sandler, the marketing operator who spent years at companies like American Express and Nestle, who's now using Claude Code to build competitive research tools from scratch. With zero coding background. So we're going to walk through the entire process live on screen, starting with a blank terminal and ending up with a fully functioning dashboard complete with scoring, filters, contact information for the research targets that we found and more. And if you stick around to the end, you're going to see how we turn the whole thing into a reusable skill that you can run for any market. So enjoy. All right, Adam Sandler, what are we building today? Speaker 1: How you doing, Corey? I thought it would be helpful to show how to put together a process in Claude Code, starting with some simple research and working all the way up to a tool. Because I think this is sort of one of the superpowers that Claude Code unlocks for people who are not technical or don't have a technical background. I think where we're going is the capability for people who are historically in knowledge worker jobs like marketing, like myself. We have the ability now to create tools, even if they're one-off use tools, to help us accelerate how we go to market, how we perform our work. I'd love to showcase how we can take an approach that focuses on competitive research, turn that into a tool that allows ongoing repeated research, so you don't have to do everything all over again. Effectively, building yourself a tool in real time as you operate it. Speaker 2: Awesome. Yeah, well, this is going to be a good one. I'm excited for this. Anything having to do with Claude Code is just massive leverage once you put in the groundwork to refine it and get it up and running. So we'll jump into your screen here in a couple of seconds. But before we do, why don't you just give us your elevator pitch? Like, who is Adam Sandler? What's his background? Why should people listen to him? How do people differentiate him from the other Adam Sandler? Speaker 1: Well, the differentiation piece is easy. I'm a lot less wealthy. Speaker 2: We're getting there though, right? Speaker 1: There's that. I historically have a career in marketing. I live in New York. Worked in a couple of large enterprise, well, several large enterprise companies based in New York City in various marketing roles. So, American Express was one of the bigger names I've worked at and I worked in the Open Business Unit, which is focused on small business customers and worked on a content marketing initiative there called Open Forum. This was, you know, over a decade ago and it was when content marketing was still very nascent. So, got my feet wet understanding, you know, how to use organic content as an acquisition mechanism. And later in my career and most recently in that world, I was at Nestle Health Science, which was by way of acquisition of Nature's Bounty Vitamins, where I led a digital marketing team and also later in my tenure there, helped to lead the Amazon business as well. So, you know, learned a lot about e-commerce and You know, high volume selling, you know, channel management, a lot of a lot of very multifaceted Ways to market in that world and Vitamins is a low margin, high volume business, which on Amazon, and I know you're well familiar, it can be very challenging. So, you know, it was learned a lot, worked with a lot of really great people. And then since that time, over the last two years, I've been doing independent marketing consulting and started using AI just for like how I operate. Quickly saw the benefits of how it could help me deliver for my clients. And, you know, fast forward to today, obviously, you can't get through your day without encountering AI somewhere, right? That's in your LinkedIn feed or within the apps you're using every day. And so, you know, I think You know, when we think about Claude Code and sort of like how this technology and the platforms are evolving, I think there's a real impetus and need for, you know, knowledge workers like myself, and I don't want to characterize you, but perhaps folks like you and I, to get comfortable and to really understand how to embrace this technology, right, and get past some of the I think technical hurdles which, you know, on the surface appear impenetrable, but in reality are really not. Speaker 2: Right. Yeah, I agree completely. So what I love about these episodes and what we're going to learn today is you are going to share your screen. So if people are watching this or if they're listening on the podcast platform, be sure to switch over to YouTube. And Adam's going to break this whole thing down step by step. We're going to jump into his screen, look at his Claude Code setup and build some of these workflows from scratch. Adam, why don't you take it away? Speaker 1: Absolutely. All right, cool. So I am using Claude Code in an IDE called VS Code. VS Code is a free product that you can download. And it's essentially an interface that coders use for doing engineering, right? But because of how Claude Code has become a lot more popular among non coders, This is an easy to use interface to get up and going, right? This is where a lot of people I think also get a little bit intimidated when you start hearing words like terminal and you see black screens like this with a lot of colors. But as we get into this, you'll see how simple it can be, right? So I'm using a Anthropic account. I pay for the max, the top max subscription, which is $200 a month. What I'm doing, you could probably do for the $20 a month subscription. So what I'm going to do here is open a new directory. Actually, I'm already in a directory. So I'm going to open a new terminal up here. Terminal is sort of like your foundational interface for operating a computer, right? So all I'm going to do is type, I'm using a command to allow me to skip permissions. I don't recommend it for new For newbies, but essentially what I'm doing is just typing Claude and launching Claude Code inside the VS Code application. Speaker 2: So it's funny that when I first, sorry to cut you off, but when I first launched or when I use Claude Code for the first time, so I use cursor, which is obviously a competitor to VS Code. But I could not figure out how to launch Claude Code. So I had to ask Claude and it was like, all you do is just type the word Claude and press enter. And I was like, oh, okay, that's not that hard. Speaker 1: And I think, you know, what you just mentioned of like how you figured it out. Just ask, right? Just ask Claude how to do it. And that's how I learned a lot of stuff as well. No shame. So there are a couple of ways to open Claude within. VS Code has a built-in, I think it's built-in, but there's an extension that Anthropic released called Claude Code for VS Code. Which gives you a little bit of a nicer interface. So if I hit this orange button up here, after it's installed, you'll see this orange button. It'll just give me something that's a little bit easier on the eyes. Speaker 2: Oh, nice. Yeah. Speaker 1: Yeah. And I do work in this a lot. It does have a couple of limitations compared to the terminal version, which is on the left side. I don't want to get into the details on that now, but perhaps in a second session, if we want to go dive deep, we can do that. I'm going to close that. But like I said, if you're just getting started, this is a really easy way to onboard yourself to working with Claude Code inside VS Code. I'm just going to close this one and focus in on the terminal. It's the same exact thing, just looks a little bit differently. I thought it would be really interesting to focus on a research-based project because it's a really great example of how we can build an iterative workflow and start with something very simple and get to something that is like a tool and a process that can be repeated. Research is always a great example of this because there's so many different ways to conduct research. For now, I thought maybe we can look at like competitive research in some manner. Are there any like industries, verticals, companies that you have in mind that we could target for competitive research? Speaker 2: I do. So I'll use actually a real life example. I was doing a consulting session for a friend of mine last week who is in the wedding venue business. So something that they're doing is they're actually going and approaching other owners of wedding venues here in the state of North Carolina. And they're pitching them and basically saying, hey, we will run your venue and just kick you back a rev share. So you don't have to do any work. You don't have to bring any leads. You don't have to market. We're literally just going to use your venue. And so part of what I was showing him, and I was actually showing him this in Claude Code, which is kind of like a step down from Claude Code, but they do similar things. I was showing him how Claude Code and therefore Claude Code can go out and do some really detailed, deep research on wedding venue, basically wedding venues in the area. And it's funny because he wanted to find distressed wedding venues. And his definition of distress, it's not concrete. It's not like, oh, they have to meet these criteria. It's like, you know, distress might be they might be in financial distress, like they might be going through bankruptcy, they might have bad Google reviews, you know, so there was kind of like these three or four kind of loosely defined And we're going to be talking about definitions of distress. And I was like, well, let's just throw these into co-work and see what it does. And sure enough, within five to seven minutes, it had, I think, seven or eight venues within 30 miles of his venue. That we actually dug through the results and he was like, these are good. These are one of them that Claude went in and found that the owner just went to prison and now there's a foreclosure auction on this wedding venue. And my friend who I was showing this to was like, I never would have found that. It would have taken me hours to find that. So long-winded way of saying, let's just use that as an example. Let's look for, we could repeat the same search, wedding venues. That appear to be in a state of distress in the state of North Carolina. Speaker 1: Yeah, and and you'll see me using my voice I think you know, it's whisper flow Yeah, actually I'm using super whisper. Speaker 2: Okay, interesting same thing. Speaker 1: I'm sure yeah, it's very similar and only because like I guess I found it first using it for over a year at this point and it is You know, it has not cost me any money. It's great. And so, you know, I highly recommend voice dictation because you're removing a lot of cognitive load. Right. When you're not typing, when you're not, you know, having to stare at something. So just a, you know, hot tip for those who are not on board yet. So, okay, let's research distressed wedding venues in North Carolina. And let's focus on pain points and reviews from customers. Okay, so I purposely left it a little bit general because I'm curious about like how it's going to scope out the idea of distressed. Speaker 2: Right. Speaker 1: As you were speaking, and I think you sort of like that was the point you were making is like, well, what does distressed mean? Speaker 2: Right. Speaker 1: I was thinking of like, well, how would you research that? Because like if there's financial distress, that might not be public information, right? Speaker 2: Yep. Speaker 1: And so what Claude Code is doing right now, it's searching the web, right? Just like a lot of these LLMs have the capability to conduct web research behind the scenes, right? Outside of the view of the chat. That's what Claude Code is doing right now. And we're getting a little bit of a visual of the different terms, right? So we're seeing, there was one about scams, which caught my eye earlier. Speaker 2: Right. It's one of the keywords that uses suddenly brides stranded. So it's like, how funny is that? So it's basically looking for like social media posts or like reviews or basically anywhere online where somebody has complained that, hey, XYZ wedding venue, like stranded me on my wedding day. Like we were screwed, which I think is really, I wouldn't have thought to do that. Speaker 1: And one other thing to be aware of, and I don't think this is happening right at the moment, but Claude Code can launch autonomous sub-agents, meaning it can actually launch its own chat sessions that it will operate outside of your view. To conduct this research in parallel, right? You can have three subagents all doing work at the same time, and then they will come back and report the results to the primary agent, which is our session here, right? So just a good thing to understand. And maybe we'll look at some of that later. So here we go. So we have an outcome here. There's a lot, actually, some links. Obviously, we don't want to read the entire report. But what I'd love to do is get To a point where we have a commonality, some common data point to work with across these different venues, right? So we've done the research, we have some intel, and one of the magical things that Claude Code can do, and I think Co-Work now as well, is it can create a file and store it on your local machine. So we can now capture these insights, whereas maybe like if you're just using ChatGPT, you might have to copy and paste something else. And I think, you know, it sounds small, but this I think is one of the core benefits of using Claude Code. The ability to read, write, edit and delete files on your machine means that You effectively can manage the information in the context that Claude Code is using and has access to at any given moment and can update that context. Speaker 2: Right. Speaker 1: And now we're sort of like bumping up against the idea of persistent memory. Right, so I just want to draw the line from like that seemingly simple capability to the actual utility and impact. So it's hard to sort of like just eyeball to see if like there are, you know, pain points that we could parse into data, but I'm going to ask Claude Code to do it anyway. I would love to create a CSV that can give us visibility into the discrete pain points for this selection of competitors as we look to get an understanding of the landscape for our business of finding distressed wedding venues. Well, so do you see the table above it actually kind of already did that where did it yeah like it sort of gave like a Roll-up if you uh-huh right like here are the common things. Speaker 2: Oh, so you're saying we want it specific to each venue like this more granular got it I say yeah. Speaker 1: Because that's what's going to open up the path for us to operationalize that data, right? Like there's not, there's maybe some good intelligence here, but it doesn't look like we can like really use it just yet, right? And that's kind of like what I want to get to. And so as Claude Code is thinking, the panel here on the left It's empty because this is just another view into the folder that the project is on on my computer, right? So if I right-click here and click Reveal in Finder, It's going to pull up that exact folder. So it's just another quick tip to like identify files on your machine. You can right-click, reveal in Finder. But also just to illustrate that this box here is just the same thing, different view. It's really helpful because you'll end up managing a lot of files. And here it is, right? We have our CSV. With a ton of these different venues, and then it's structured some schema based on the findings, right? Not a ton of data, 31 records, great for a demo just like this. But if you are working with large sets of data, You can similarly use this kind of approach, right? Great. So we sort of like we have a schema now, right? We've done research, we have a schema. What can we do with this? Let's now Our business of identifying these distressed venues, I guess, you know, we want to at some point, like, transact something, right, or contact a customer or a potential prospect or client, right? Help me understand a little bit like what the business wants to do, because I want to like try and shape it around the goal. Speaker 2: Yeah, so the goal of my friend who does this was twofold. It was one, if they are in enough distress and they're willing just to sell their venue outright, then my friend wants to buy it and then operate it, right? And then on the other hand, if they don't necessarily want to sell, but they're so You know, grinded down from years or months or however long of operating these venues that are clearly not performing well. And if they're not performing well, they're probably not making any money. So his pitch then is, again, we'll come in and operate it and we'll spruce it up. We'll even invest in the property to get it up to 2026 standards. And we'll bring in the leads and we'll operate it. And then we'll give you a rev share, essentially. Speaker 1: Okay, that's that's super helpful. A lot of detail as well. I'm going to do my best to sort of regurgitate some of that in a pithy way. Using this data, we would love to identify targets for our business to explore for potential acquisition and targets for our business to explore for potential management partnerships or cooperative arrangements where we can have some level of control and management of the venue. Again, I'm leaving it a little bit open-ended. One thing I'd love to showcase here is plan mode. I'm going to flip it on. If I hit Shift tab here, I'm activating plan mode. This is one of my favorite features of Claude Code. I believe it's in co-work as well, but I don't know if you can manually activate it in co-work. What plan mode does, it turns Claude Code's capability to edit files off. And tells Claude that you want to have basically a conversation to plan out this next task. Speaker 2: Right. Speaker 1: And the idea is that Claude Code is going to then sort of guide you through a series of different questions to clarify the ask. And then it's going to write a plan to do this, right. And so this is probably a little bit of a simple example. However, I'm sorry, Corey, were you speaking? Speaker 2: No, you're good. Speaker 1: Okay. So this is probably a little bit of a simple example. But plan mode is such a powerful tool because it'll help you identify opportunities or gaps or things that you may not have thought of. And there have been so many instances where plan mode has Open the door to different paths or different ways of doing things. It's a learning tool. Speaker 2: Right. Speaker 1: Right. And again, so you hit shift tab, and you can see on the bottom of my screen, the mode is indicated here under the chat box. So plan mode is on. And I'm going to hit enter. And, you know, plan mode, you don't have to have a robust hypothesis or idea or anything. You don't even have to be using it in the context of producing a thing. You can use it in the context of like, hey, help me plan my weekend. Speaker 2: Right. It can be super basic. Speaker 1: Yeah. And it really sort of like Got me onto the, I've become a big fan of like guided experiences because of plan mode. So when I develop my own workflows, I try to enable a guided experience. Speaker 2: And I think, and not to get off on a tangent, I'll just brush over it briefly, but obviously you and I know how powerful skills are. And I think plan mode within Claude Code is incredibly useful for kind of brain dumping the way that you currently do a task or execute a process, and then letting it put together a plan for creating that skill. Right, so I think plan mode and Claude Code, in my opinion, is the most effective way to really put together solid skills in the shortest amount of time possible. Speaker 1: And we'll definitely get into skills. So the first question it's asking is the format. Do I want a CSV, a Python analysis script, or both? And I'm just going to go with both here. Because why not, right? The next question, control level. For management partnerships, what level of control are we targeting? Full operational takeover, majority management control, advisory, consulting, or flexible? Depends on the venue. Speaker 2: So it's my understanding that they are, ideally, they want full operational control. It's like, hey, we take over everything. You sit back and collect a check. Obviously, we're going to take the majority of the profit, but it's going to be hands off for you. Speaker 1: You did mention that when you were describing it. And then that was left out of my vague ask, right? Speaker 2: Well, but it's that goes to show that that's how good Claude Code is, is it's like, well, hey, this is probably a key detail that we should fill in. What's the context there? And that's another beautiful thing about plan mode is it will Ask you the questions required to fill in that context to actually give you a good output. Whereas ChatGPT is just going to go and answer, right? Or even regular Claude is just going to go and answer unless you specify, hey, interview me to get the context needed. Speaker 1: Yeah, Claude regular Claude is now using the question thing. I don't know if you've caught that. Speaker 2: Mm hmm. I think it's smart. Speaker 1: Yeah. Meaning like when you sometimes call just regular desktop Claude will like flash you questions to click on versus answering just like plan mode and co work. Or in the UI extension we looked at. So one other thing just to sort of call out here is that like, if you're not sure or you're not happy with these options, you can, you know, add your own option, or you can even chat about this. Right? So you can dig deeper to understand like, whoops, I just, I just went a little too far here. Oh, God, I'm sorry. Let's try that one more time. Speaker 2: No worries. But yeah, again, that's another good thing about plan mode is like I've done that many times where I'll like fat finger a button and be like, oh, I didn't mean to type that. And then we go redo the interview and it doesn't take but a couple seconds. Speaker 1: There is a way to go back. I just fat fingered it and stopped the whole question. Okay, so the thing that I wanted to mention, though, was the chat about this, right? So you can chat more about to get an understanding, like, okay, why these options, you know, a lot of times you'll see Claude Code will make a recommendation in those options, it'll say, like, recommended this one, right? So like, chat mode is great for understanding, like, well, why did you recommend that one? Right, and digging a little bit deeper into the process. And I'm using that more and more lately. Okay, so for one reason or another, it just skipped. Speaker 2: The plan, yeah. Speaker 1: The questions, that was my fault. But let's not waste time. Let's just continue moving forward. So once plan mode finishes and gets that information from you, it's gonna present back to you that plan, right? So here's Claude's plan. Distressed North Carolina wedding venue target analysis. We've got pain points, the deliverable will be another CSV file, and here's an overview of the schema, the situation, and it's created a rubric for scoring, right? So this is sort of like what is responding to the nature of our ask, which was like, well, how do we decide what to do with which, right? And so there's a scoring system that it's implementing here that will be visible in that CSV file. Right, and here's the more information about the venues and looks like some details about those specifics. Normally, I would like spend some time to read through this carefully, and I recommend everyone does when you're using plan mode, because you will find things that you might not align with, right? You might find things that you want to change. And there are simple things too, like if you're working with a project that you want to manage version control, it's good to say like, and do this on a new branch, right? So after plan mode has presented the plan, I'm presented with more options. Yes, clear context and bypass permissions. So this just means that it's going to start a fresh session with this plan. That's what clear context means. So, you know, we have a full context window of tokens to work with and bypass permissions. It's not going to ask me to move forward with like discrete steps or edits. So I hit number one and it's going to go and now activate on this plan and create that CSV and analysis. And then after this is where like, I want to, you know, nobody wants to go into a CSV, right? And, you know, that is, I think, where we started thinking about operationalizing this information in a way that is engaging, easy to digest and repeatable. You know, and thinking about like how this would manifest as a tool. Speaker 2: And so Adam, I got a question. So just to play devil's advocate, because I already know there's going to be people who say like, well, why do you need to do this in Claude Code? Like, why can't you just do this in regular Claude or like ChatGPT? I mean, is there a difference? And if so, what is the advantage of doing it in Claude Code? Speaker 1: Yeah, I think it's a really great question. You know, we talked about the ability to capture and store these files, right? So I mean, in all fairness, it's been a minute since I've like really dug deep in ChatGPT. So the last as far as I know, ChatGPT is not Able to manipulate files on your hard drive in a folder on your machine, right? Speaker 2: Right. Speaker 1: So the examples we're looking at here, we're creating new files, right? But Claude Code could go back and look at specific data points in here. We can ask it to retrieve certain things or change or edit certain things, right? So like it's a dynamic living and breathing project in that regard. Files that are instantly on your computer. Right. So I think that's one major difference. And again, like the LLMs are sort of like pushing in this direction or the typical platforms are pushing in this direction a little bit, especially Claude. Unknown Speaker: Right. Speaker 1: There's like rumors flying around there that coworker is just going to be the main. Speaker 2: That is just going to be Claude. Yeah. Speaker 1: Which kind of makes sense. And so with Claude Code, you know, we're also working within a discrete project folder. The session that I'm working within can only see what I give it access to, which is this folder, right? It doesn't have memory of anything else that I've done. It doesn't know anything else about me. This is a completely empty project directory, so this specific session of Claude Code doesn't even have the skills that I typically would pull into a project. Right, so all of that is malleable and customizable. I can completely customize how I want this project to behave. And I think that's also unique to a platform like this, like Claude Code. Speaker 2: Right. Speaker 1: We have the capability to plug in a lot of different tools. I'm sure you've heard of MCP, right? Speaker 2: Yeah. Speaker 1: Which are, you know, essentially integrations with outside platforms, easy, low impact ways to transfer data back and forth, right? I have a project where I have an MCP tool that pulls in website performance data from Google Analytics, creates a report, and then can, you know, take a number of actions based on the substance of that report. Speaker 2: Right. Speaker 1: You know, so there's a lot of things like that you can put together yourself. Through those different things like MCPs and plugins and skills and things like that. So there's a lot but I would say like those are the things that come to mind in response to your question. Cool. All right. So what do we have? We have a new CSV, but we don't really want to work in a CSV. So it's rolled this up. As we were speaking, we got a different score for each of these different venues. Have you seen these names before? Speaker 2: Yeah. So Champagne Manor is the one that we identified from our co-work session last week, where the owner, I think, is in prison now and it's going to foreclosure sale. Right. So that was one that I even asked him, I was like, had you ever heard of this venue? Like, would you have found this in your regular research? And he said, no. And the reason for that is because it's not like it's on the website that like, hey, we are in foreclosure. What Claude most likely did is it went out there and found like a social media post or actually I think it was a news article about the owner going to prison. And then it put two and two together like, oh, This guy owns this venue. He's now in prison. The venue's probably a good target for acquisition. And then it digs a little deeper and it's like, oh, what do you know? It's actually going into foreclosure. And it makes sense that it ranks that particular property with an acquisition score of 4.2 out of five, right? And a partnership score of only 1.2. Because you're not really going to partner with a venue that doesn't have an owner and isn't going into foreclosure. It makes way more sense just to go and buy that thing. Yeah, absolutely. I'm texting him right after this. We've got an episode coming out for you. It's all about you next week and I think he'll love it. So yeah, this is spot on. Speaker 1: Let's think through what would be a natural next step from this. We have our research, we have our intelligence and one of the things that I'm thinking of is we can create a visual dossier or asset similar to what might be an artifact in Cloud Desktop. To better consume and understand the outcome here. Speaker 2: Mm hmm. Speaker 1: And that's something I do all the time. One of my favorite skills is the playground skill. Unknown Speaker: I don't know if you I'm not familiar. Speaker 1: No, it's actually officially released by Anthropic and it's available through their plugins. And I have it installed globally. So I can invoke it without installing it in a specific project. But for anyone who's interested, and maybe we can follow up on this, because I don't know if I have the exact link. But if you type slash plugins, you can manage your different skills that are officially released by Claude. Speaker 2: Oh, interesting. Speaker 1: Right. And you can add new marketplaces. Right, so there's an official Claude plug-ins marketplace that you can grab the link You can add it to the marketplace list here and then you'll have a number of different skills and plugins that you can add to your projects through this way versus like dropping a skills folder in or something like that. Speaker 2: Right. Speaker 1: So I already have the Playground skill installed through that method. So I can invoke it for things like this. And just to give a quick understanding, Playground is a skill that is designed to allow you to interact with the model in different ways. And I don't want to waste too much time getting into the weeds on that. I'll just show it in action. Let's use the playground skill to create a visual dossier of each one of these acquisition or partnership targets. Where we can get in a snapshot, an understanding of the recommendation, the score, the time sensitivity with the right context for us to have a business oriented session on strategizing our approach for this. Speaker 2: Okay, so I see what you're doing here because you're saying, so what we're looking at above is like, okay, well, this is a little chart that basically shows of these five properties. This is the acquisition score. This is the partnership score. This is like a quick reason why Claude gave it that score. And it's like, that's great, but that doesn't really help us dive deep into like what really makes this particular property a great acquisition target versus this one that might just be a really good partnership target. And like you were saying earlier, we have a CSV that has all that data. That was the deliverable of Claude's work, but it's like, we don't really want to dig through a CSV. It's hard to look at. It's hard to kind of make sense of it. So what you're saying is now with the playground skill, you're telling Claude Code, it's like, hey, basically take that CSV file that has all of our output and turn it into something that we can actually look at. That's actually visually appealing where I can kind of, from a quick glance, understand Claude's reasoning behind each of the properties and really from there be able to decide next steps. Speaker 1: Exactly. That's exactly right. Like we're sort of asking Claude Code to be our analyst and, you know, do the reporting and give us the qualitative perspective, right? And this is a roll-up, right? This is a summary. So if we click here, Like, you know what, to be quite honest, I was expecting to see a lot more. Okay, there is a lot. Okay. So, you know, it's just summarizing what is in that CSV, to your point. Speaker 2: Right. Unknown Speaker: I don't want to read it. You don't want to read it. And the utility of still having that CSV, though, is that like we've established structure, right? Like we have structured data. And when we get into repetition, that's where doing that up front, like we're having that awareness up front to start with structured data is going to help. I just asked for research and it like, you know, dumped it into the chat and I copied and pasted it into another prompt. You know, there's no structure there, right? You're not building process or iterating, right? We're iterating here. Speaker 1: And that's kind of like the through line that I try to keep through these processes. So Playground is going to create an HTML file. Usually it launches it automatically in the browser as well. It launched it off screen. Speaker 2: And it tells us it's saying I'm building a professional briefing dashboard, essentially. Wow. Speaker 1: And that's exactly what it just did. Right, so let me see if I can zoom in a little bit on it. Right, so here's an overview, right? Five venues, two acquisition, two partnership, one to monitor, two at high urgency, one medium, one too low urgency, right? Speaker 2: This is crazy. Speaker 1: You even built in like... Speaker 2: Filters. Speaker 1: Yeah, that's awesome, right? And then if I click here, I don't know if you caught that before. Speaker 2: Oh my God, this is like insane. I'm actually shocked by this. Speaker 1: Right, exactly. And so now, you know, this is Playground, big, sorry, develops this into a self contained HTML file, right? Speaker 2: Like, save locally on your computer, right? Speaker 1: Yeah, all of this is just one file. So like, if you wanted to just send this file to your buddy, You know, you don't have to worry about like, what are my assets? Where's my data, right? That wraps it all up into that one file. What Playground is really intended to do is allow you to interact with your model in different ways. And like, for example, if I said, make a playground for me to tweak the UI, It'll actually create a tool for me to visualize the UI and slide things around to change heights and things like that. Speaker 2: That's sweet. I'm literally pulling a cursor on another screen right now to install Playground as a plugin. Speaker 1: And then Playground, what happens is it gives you a prompt to copy and paste back into Claude Code with all of the minutiae and details of things that you changed. Change the font, 24X or something to that. So again, I'm stretching Playground into use cases that I don't think it's typically designed for, but something I've gotten a lot of mileage out of. Speaker 2: I mean, I really hope for the folks listening on podcast that they switch over to YouTube because this is a really, I mean, we didn't even tell it to build a dashboard. We just said like, hey, give us something that we can visualize. And it went in and built like a fully functional dashboard with filters, with tags, visual, like ring style diagrams of the scoring. I mean, yeah, it's tagged as like operational failure, operating under duress. And then I like how it has a next steps recommendation for each one, too. Speaker 1: That was part of the ask, right? Because the ask included like, OK, like help me plan my my working session to strategize this. Speaker 2: Right. Speaker 1: And it even included a way to copy that to the clipboard. Speaker 2: Wow. Speaker 1: Yeah, pretty awesome. So again, that's the Playground Skill, official Anthropic release, way under the radar. So what do we do from here, right? Well, what if we wanted to branch out into other states? We could. Okay, we've conquered North Carolina. What about South Carolina? Speaker 2: Right. Speaker 1: Right. And this is, I think, where we start thinking about skills, right? Because now we're thinking about repeating this entire process. You know, and so I would from in, to be clear, like, you know, we are like, for the sake of the video, I think like we're going we're going fast, right? I would put more thought and care into the individual steps. But like, you know, generally, this is the approach, I think. So, let me think about this for a second. So, we want to be able to repeat this process for other states, right? So, is there anything, before we do that, like is there anything else on the dashboard? Let me, where did I, did I lose it? Speaker 2: I still see VS Code right now. Speaker 1: Yeah, yeah. I think I might have closed it. My right click is not working, so I want to open that in the Finder. That's strange. Anyway, one thing that I felt like was missing was links, right? Speaker 2: Right, yep. Speaker 1: Links to maybe their reviews, their homepage. So I want to add that in. Please update the HTML, the dossier HTML, with per venue links to their homepage, include their contact information, and if they have a Google business page with reviews, include that link also. Because that'll be easy for us to then not have to do those extra steps when the time is right. And then when we want to create a repeatable process, we have those details included as well. So yeah, it's a very interesting focus, right? The wedding venues. A lot of, you know, the purview that I have typically are, you know, companies that are very digitally native. Speaker 2: Right. Speaker 1: Doing a lot of like digital marketing through emails or landing pages and, you know, funnels and things like that. So like the I appreciate like this, this unique vertical, because I think it like, you know, it's challenging to to sort of like pivot in that regard. Because like, where I envisioned this going was like, okay, now we could repeat a landing page for these different competitors, right? Speaker 2: Right. Speaker 1: But this is a little bit different. And I really enjoy this. It's very unique to what I'm usually doing. Speaker 2: Yeah, and I figured too, I was like, well, this is one we were just working on this week in co-work. Let's see if Claude Code can go even a little deeper. And you know what, my next step here would probably be even saying to Claude Code, it's like, hey, that workload that we just implemented to find these venues in North Carolina, Let's turn this into a skill so that we can replicate it at the city level or the town level. And then, of course, at the state level as well. And a lot of times it's like, great, it just does it. And then you run it once, see how it goes. And then if there's tweaks to be made, you just tell it to update the skill accordingly. Unknown Speaker: Yeah, yeah, yeah. Speaker 1: And I do that all the time, like refine and update skills. And some frequently, like we'll go in there manually, right? Because if you open a skill, and I guess we can look at this once we make the skill, like if you open it, it's a markdown file, right? Which is essentially text. And it's written like a prompt, you know? So you can, you have the capability to be surgical. If and where you want to. And it's also very, I would say, productive and helpful to just get that familiarity of how a skill is built. Over time, as you read them, you get that intrinsic understanding of how a skill works, and it's going to make you so much better at designing your own skills over time. Speaker 2: Absolutely. Speaker 1: So what's happening now, it's using Claude in Chrome. So it could, I think it's controlling a browser in a different, I have a different desktop instance somewhere, and I think it's controlling a browser there. But one of the great capabilities they added recently to Claude Code, and I think it's in all the Claude projects now, or products, is the ability for it to take control of your browser. Speaker 2: Yep. Speaker 1: So I've gotten so much mileage out of that. And sometimes it'll just do it on its own. I think like it is now. While it goes and collects that information, let's think about the next prompt. And I think after this is when I would be inclined to turn it into a skill. We did basic research at first. We created structured data from that research. We evolved that into a analysis with a scoring rubric. And then we created a visual presentation To enable us to like actually use it, right? Speaker 2: Yep. Speaker 1: And that was a lot of steps. And we can basically repeat that in one command through the magic of a skill. And skill, there are skill creator skills. Anthropic has an official skill creator skill, which I think can be accessed the same way through slash plugin. It's thinking now, so it's effectively queued up. But you can, the same way that we looked at how to get the playground skill, I believe you can get the skill creator skill that way as well. Speaker 2: Yeah, it's just what, slash command slash plugin? Speaker 1: Yeah, but you might need to grab that URL for the Anthropic Marketplace. I don't know if that's built in or not, to be honest. And there's a lot of like, you know, community and user made skills and marketplaces out there as well. I encourage scrutiny because like, you know, there's some, I've heard some horror stories of security issues. I've never experienced that myself. Just know where your things are coming from, I guess. Like, if you have a skill, make sure you trust the source. Speaker 2: Yeah, don't just go and download random skills off the internet, especially if they're not from an approved source, like you said. So that's good advice for anyone. Speaker 1: And then as far as skill creation goes, a lot of people, including myself, have created their own skill creator skills that are a little bit more advanced than the one Anthropic provides. And I wouldn't say it's a necessity. But if you feel like you're not getting what you need out of the official skill creator, just go create your own, right? Because then you could be prescriptive about the kinds of skills that you want. Yeah, you can even create a skill that is specializes in specific types of skills. And I don't want to overcomplicate things. The point of what I'm saying is like, it's totally white space. Speaker 2: Yes. Speaker 1: And like there's very little risk to experimenting, right? And so it's worth your while because you're going to learn something one way or another. Speaker 2: Right. 100%. Is it stuck on this one, the one property here that it's researching on the browser? Speaker 1: It might be. Let me, let me give it like, yeah, 9,000 tokens. It's not moving. Let's, I'll stop it right there. Speaker 2: And what I want to do too is I want to have you back for a follow up where we can go deeper into skills because I mean, again, I think we're in agreement that skills is going to be one of the most critical. Well, skills to learn. And then over the next few years is how to build skills, what type of skills to build, how to construct them the right way, how to iterate on them. So, I mean, as far as for today, I think this is a great example as far as like an intro to what Claude Code can do. And I mean, people listening can use their imagination as far as where they want to take it from here. Like we said, you could turn this This research process into a skill so that it's automated and it happens the same way every time. You could then take it a step further and say, OK, well, next, we want to find contact information for these venues, for example, and then they can go out and do that. And then you can turn that process into a skill, the process of actually finding the contact information. You can really take it all the way down the chain. I mean, you could have it. You know, you could have a skill where. All right. Every time you find a new venue, you put together a three page PDF proposal. We're going to be talking about a couple of things that we can then use in our presentation to the venue. There's a million directions you can go with it. Speaker 1: Yeah, you could even set up a cron job, right? Like have a time-based or time-triggered workflow that executes once a week and says, hey, find me all the new distressed venues in the Southwest or something like that, right? Speaker 2: Yep. Speaker 1: So yes, and you could chain them together. There's a ton of opportunities. We're using a lot of time here. So I just asked it to add whatever it found to the existing HTML file. And while it does that, oh, okay, just pulled it up again. Sorry. Do you still see me? Speaker 2: Yeah, I see it. I see the dashboard here. Speaker 1: Okay, cool. So it should have added contact information where it was able to find it. But we kind of cut it short. But here it is, right. So we've got the address, we've got the phone number, we've got an email address, we cut it short. Otherwise, we might have gotten some other stuff that we asked for. And then from here, let's move on to the skill. So it's I'm going to give you it's it's like mind boggling, mind bogglingly simple. Create a skill that can perform the research, the analysis, and create the dashboard visual, just like this, that we can run for any state or city. And I'm going to put this in plan mode, which I highly recommend you do when you are creating skills because of the same reasons that we mentioned earlier where, you know, there might be details that You haven't thought of or, you know, there could be blockers that would prevent this from just moving forward. Speaker 2: Right. Speaker 1: The outcome you want if you don't use plan mode might get confused. AI still hallucinates, still gets things wrong, still gets confused. I always like to tell people, like, use your instincts, right? Speaker 2: Yes, right. Speaker 1: Instincts. If you feel like if it smells fishy or something doesn't look right, like don't just. Speaker 2: Don't just assume it's right, right? Speaker 1: All the time I am correcting and finding things and, you know, instincts are critical here. Speaker 2: Great. And let's see if we can speed run this skill creation just because we've got a hard stop here in a couple of minutes. And if not, like I said, I want to have you back in the future for for a skill specific episode. Speaker 1: Yeah, yeah. I wish I could like turn a crank and make it go faster. Speaker 2: I know. Speaker 1: If it's if I could always like record a follow up video and send it to you if you wanted. Once it's finished, if you're if you have to cut it. Speaker 2: Yeah, I'd say let's go ahead and cut it here just because we've got another recording here in a couple of minutes. But this has been fantastic as far as an introductory lesson for for folks getting into Claude Code, kind of how it's set up, how to really just how to get your feet wet. Right. There's a million different use cases out there. Speaker 1: Absolutely. Like I said, it's it's very simple. Like I do think it's it's going to be ubiquitous. Speaker 2: Right. Speaker 1: One thing we should just throw in there about skills is that skills is an open protocol. Google's products are using skills. I think ChatGPT products are using skills as well, right? So those are transferable to other platforms. Speaker 2: Absolutely. Speaker 1: And so I think that's just sort of like a signal to what you said, like they're not going away and it's gonna, I think, be more prevalent as time goes on. Speaker 2: Right. And I think what people should take away, they only take away a couple of things. It's one, just download. An IDE, whether it's VS Code or whether it's Cursor, and just get into Claude Code and just start messing around with it. And then two, just create a sample skill. Take anything you do on a daily basis, doesn't matter how big, how small, and turn it into a skill and then just iterate on it over time. So that's my feedback. Speaker 1: Absolutely. Yeah, it's cool, man. This was great. Speaker 2: Yeah, well Adam, it's been great having you. So where can people find out more about you and follow you? Speaker 1: So Adam Sandler was taken. So you can find me at viableedge.com is my website. And I have a lot of marketing related content as in regards to AI and agentic marketing workflows. You can find me on YouTube at Viable Edge. I'm also active on X, the Viable Edge and LinkedIn. I did manage to get Adam Sandler. Speaker 2: Nice. Speaker 1: Yes. So you can find me at Adam Sandler on LinkedIn. Speaker 2: Beautiful. Adam, thank you so much for your time and for those in the audience, we'll be back next week as always. Speaker 1: All right, Corey. I really appreciate it. Thank you so much.