Vibe Coding for Good

**Welcome to AI Insiders**
(0:05) Hey everyone, welcome to another edition of AI Insiders by Marketers Guild. I'm your host, David Burkowitz, and I actually get to be the featured guest presenter today. I'm my own guest, and I hope I'll make myself at home in this community. You all make it easy for me to do so because this is a topic I can't talk about vibe coding enough.

**The Power of Vibe Coding**
(0:29) I am trying not to have 20 different vibe coding things going on while I'm doing this session. I'll see if I can focus enough for the next hour or so, but I'm blown away by what can be done with it. I'll share one of the latest ideas that came to mind: could we create resources that benefit others?

**Introducing Vibe Coding for Good**
(0:57) It could be related to something we value like bringing people together, helping families connect better. All kinds of positive ways this tech could potentially help others. On a whim, I created Vibe Coding for Good. I got the domain. I built this in Base 44. If you haven't been playing around with

**Base 44 and Test Sites**
(1:27) the resources yet, I'll go back to that. I need to republish this site in B 44 because a few things aren't here yet that I was putting together, but there are a bunch of resources on here. I tried creating a test site. This is something I admittedly did on my own without community input, trying to see

**AI for Community Support**
(1:54) threats against the Somali immigrant community. So I wondered, what would that look like if AI helped create some resources? I did run this by an activist friend who had some ideas for what to do with it and how to hone things a bit. Again, I'm not sure this whole thing is perfect

**Open Conversation & Breakouts**
(2:26) yet. I'd want to do this with a local community or give this to another org that could do something with it. But it struck me as amazing to see how readily one could put something together that might benefit others. That was the premise for today. This is an open conversation. What I'd love to do based on

**Vibe Coding Projects for Good**
(2:55) what we did in the last Vibe Coding 101 session back in November is to do some breakouts, try to come up with ideas for things, even a quick vibe coding site, things people could work on together or individually, and do a bit of sharing after. First of all, I want to open this up: has anyone worked

**AAC Device App Inspiration**
(3:24) on any vibe coding projects remotely that are in the 'for good' space? >> It's still in build and conceptual mode, but the idea came from my son. He sometimes relies on an AAC device. This is for individuals who have speech difficulties and rely on a device to say basic things. If they

**Limitations of Current AAC Devices**
(3:55) cannot rely on their speech, it's like an iPad. It's called assisted technology. It has buttons for simple verbiage they might need through the day. They tap on it, and it becomes their voice. It's very limiting to people whose thoughts are so much bigger than what the device can offer. We have this book that offers games to

**Creating a More Flexible App**
(4:24) individuals who want to practice speaking out because a lot of it is anxiety and closes up your throat. >> Okay. >> With permission from authors, we're trying to put the games into an app that would be much more flexible than what this AAC device does. It's a big goal, but we'll see.

**Tools for Bringing the App to Life**
(4:45) We're trying to put the games on an app. >> That's incredible. >> We copied the games. >> Okay. >> We have an idea. We have a sketch. We haven't done it yet. >> Are there any tools or apps you've been using or plan to use to bring this to

**Exploring Replit and AI Prompts**
(5:03) life? >> I am trying to see if Replit might be a good idea because all it would need to do is give prompts. It's very similar to the way qualitative researchers are using prompted questions with respondents. It would need to give prompts

**Open to Suggestions for Tools**
(5:29) and have the person join a conversation with AI. >> Okay. >> Not too difficult. I'm thinking Replit might work, but I'm open to suggestions. >> Nice. This is a good place to get some suggestions. I like the ones I'm most familiar with

**Different Coding Tools**
(5:50) are Base 44, for more basic projects, but it's still robust. I've been playing a lot in ChatGPT Codex to build more full-functioning applications, which I find easier than Claude Code, which feels a bit more technical for me. There's everything on that scale, from technical to more 'what you see is what you get'.

**Codex vs. Other AI Tools**
(6:12) was >> I was wondering about that. This is a basic question, but can't I just do this? I get what Vibe Coding is. Can't I just do it with the AI tools I'm used to? Do I need Replit? Is there something special? Well, for ChatGPT versus ChatGPT Codex, Codex is designed specifically for building sites and apps. So there

**Connecting Codex to Hosting & Randall's Expertise**
(6:39) are even tools that can easily connect Codex to a web hosting app or things like that, or connect it to GitHub and all these specific things. Randall, who I learned most of what I know about Claude Code from, is in the room. I'm so glad you're here, Randall. Please put me out of my misery and up. >> Yeah. No, happy to. The

**General Purpose vs. Optimized Tools**
(7:09) fundamental model capabilities between Claude Code and Codex, and the things that are on your computer, are general purpose. They can do just about anything. You can ask a finance question on Claude Code. They're great at coding. They're great at doing many different technical things. What you'll see in tools like Replit, B4, Bolt, Lovable, is that they

**Purpose-Built Tools Explained**
(7:30) are more optimized for creating the products they're developing. That's just a website. You're not going to ask Replit to analyze your financial statements or review a legal document, right? Claude Code can, but that's because it's a general thing. Those more purpose-built things have optimized additional code and structures and

**Coding Directly in Claude vs. Claude Code**
(7:52) other things they work with to make them better at doing that particular set of tasks, rather than being something as general for any task. Then, since it builds on the questions she was asking, what would you say is the biggest difference of trying to code directly in Claude, for instance, instead of using Claude Code?

**Workflow Differences in Claude**
(8:17) >> Yeah. A couple of things will be on workflow. Anthropic is working on this to connect the different systems. When you do coding on Claude on the web, it doesn't have a good way to get into a code repository on your computer to test. It doesn't have a good way to get into a server to host somewhere. You can create connections. You can do many

**Anthropic's Evolving Capabilities**
(8:41) things to mitigate that, but it's still primarily a chat thing. Anthropic is building capabilities to allow you to work on the same code in different modalities or tooling, whether it's web, app, or even on your mobile app. But that requires a bit more setup and configuration to get that to work effectively. As terms of an

**When to Use Specialized Providers**
(9:09) exploration, it's fine if you want to do a one-off. But what I suggest is if you're trying to build something that is a website, maybe it's an interactive website, maybe it has AI in the backend, which some of these providers now will help with, just do it in one of those providers. They're optimized for building out websites that are interactive, that are

**Advanced Development Options**
(9:32) nice design, that have AI in the backend. As you get more sophisticated or get into other types of activities, maybe within a company, that's where some of the other tooling like Claude Code, or even much more sophisticated, complex options like agentic development, which you may have heard of, that's where that stuff can happen.

**Integrated Workflow for Self-Contained Projects**
(9:53) is more on the technical side. But if you're building something that is fairly self-contained, those tools are probably your best bet because they have the entire workflow of 'I want to chat with the app and make it do stuff and have it move stuff around and then make that available to other people.' That's all built into how they've

**Vibe Coding for Good: A Community Forum**
(10:12) constructed their systems. Anyone else who's already tried using vibe coding for good in some way or has an idea you think could be built that way and wants to use others here as a sounding board? Enjoy this forum for it. David, it's interesting and great being here with the group as we talk about vibe

**Learning Through Real-World Projects**
(10:38) coding for good, alongside a general shared interest across this group of people of learning and getting more knowledgeable on how it works. I ran across, and I haven't actually done this, someone recently who, instead of just going free-form and creating something made up, like you, went free-form and created something

**Vibe Coding for Nonprofits**
(11:00) meaningful that might have purpose to a charity or some other entity. It's a really cool model. If you know a nonprofit or some group out there that wouldn't have the funding to bring in people to do this and create your own pet project. Now, that can get quickly out of hand in terms of overhead and time involved,

**The Value of Real Projects**
(11:22) overhead and time involved, especially if it's for good and not for compensation. But there is nothing better, which I'm sure most of the group agrees, than doing something like you did that's real as part of learning the new technology. Yeah. There was someone

**Zero Budget for Informative Sites**
(11:41) who was telling me about ideas she had for a new site just this week. She was asking my opinion on budget for this, and I almost jokingly went like this, right? I held up a big zero where it's like, not literally, but you can at least get something, especially for something more informative. Yeah, a step up from a brochure-ware site.

**Avoiding Costly Website Updates**
(12:16) then you should get as close to that as possible, at least for a pass, before you need a designer, developer, or others to take this more seriously. There was even a partner at a boutique investment bank who came to me and said, 'Look, we just got a quote, 10,000 bucks, to update

**Shocked by Site Cloning Results**
(12:43) our WordPress site, and then 1,000 dollars in maintenance fees. He's like, 'We are brochureware. Can I avoid that?' I'm like, 'Here, let me try cloning your site and seeing what happens.' I got to say, I was shocked. By the way, I did that in Base 44, not one of the even more sophisticated tools. I was like, 'Wait a second, I did not'

**Achieving Quick Landing Pages for Causes**
(13:08) think it would work that well while we're on this call. So, I'm like, yeah, I don't know why the heck for their specific situation. So, if it's getting a landing page for a cause, a new event, things like that, that's now X month, right? We want to do something that is all about this campaign that connects our cause to

**Easy Thematic Campaigns**
(13:35) Black History Month, Women's History Month. Whatever the most relevant theme is right now. We want to do this for Mother's Day and connect this to our breast cancer research, right? This kind of stuff seems so immediately achievable. These are great points, Jared. I haven't worked on any nonprofit projects recently, but one of

**Claude Code for Landing Pages**
(14:03) the things that's really um nonprofits never have the resources you need to build landing pages, and they're resource constrained. They're people constrained. I played around with Claude Code and created a landing page not for one of my nonprofits, but for a company. I was impressed at how quickly it

**Hosting with Netlify**
(14:26) worked. Talking about launching it and making that within Claude Code, Claude Code suggested using a site called Netlify to host it, where you can create the domain URL for it. It was easy. I just uploaded the files there, drag and dropped it in the project, and I had a test page live. Now I just have to transfer the domain

**Nonprofits and Conversions**
(14:52) over and I can do it. But in general, nonprofits are always constrained for building landing pages, but it's one of the strongest things you can do to increase conversions and accomplish whatever you want to do in terms of activation, as we all know. >> Yeah. This is where I

**Casey Newton & Vibe Coding**
(15:14) learned from Casey Newton about Netlify for the first time. It's a pretty big web hosting site. When Casey Newton was talking about how his boyfriend—they just got engaged or married, I saw some big news on Instagram—Casey was having his partner help him learn how to vibe code stuff. He's in

**Hosting and Domain Investments**
(15:43) Claude Code and he used Netlify. He used this terminal called Ghostly to help build it, which I think you can use the same terminal on your Mac. My projects that I've been doing in ChatGPT Codex are hosted on Netlify; the ones on Base 44 are hosted within Base 44. I've been buying way more domains than maybe I ever have, because

**Low Hard Costs Per Project**
(16:14) I keep connecting things to my vibe coding projects. That's been my biggest investment, buying these $13 a year domains. It's incredible how low the hard costs are per project for these. I think it's also worth hearing from others, and Randall, I know earlier, doing anything like

**Passion Projects for Inspiration**
(16:44) this, but stuff you've been vibe coding, it doesn't have to be directly in that 'for good' umbrella, but things that might be a good source of inspiration. Is anyone working on any passion projects? It doesn't have to be specifically marketing related. So, I can jump in here. David, I don't know if I would consider for

**GPT3 for Voter Registration**
(17:09) good, but I've been testing out my custom GPT3 with a gubernatorial campaign here in the US Virgin Islands. One of the big things we noticed as we were getting the campaign ramped up is that many people aren't registered to vote. >> Okay. We have about 100,000 people in the territory, and 30% of those are

**User-Friendly Voter Registration Site**
(17:32) active voters, meaning those who registered to vote and voted in the last two years. One of the Vibe Coding projects, inspired by you and your approach for Vibe Coding, was to create a site that's not funded or sourced by the government, but to make it easy for people to check if they're registered to vote, and to give them the information they need

**Promoting Civic Participation**
(17:55) to get registered to vote as soon as possible by the deadline so they can participate. Sorry for the background noise. I love life. But yeah, one of the Vibe Coding products I had in mind, based on your recommendations, was creating a user-friendly, simpler site to check if you're registered to vote and if not,

**Encouraging Voter Participation**
(18:14) these are the steps you need to take to make it as easy and simple as possible so you can get registered and vote when you're ready. >> Yeah. I don't think there are many better causes in any democracy right now than encouraging voter registration and voter participation. So I'd say that's

**Adapting Vibe Coding for Causes**
(18:37) very little good is going to happen without strong electoral turnout. So, Earl, that is awesome. I think ideas like that are tremendous. And also, by the way, the ease with which one can adapt some of the work you're doing to other kinds of causes and themes is tremendous. I'll share an example that and I'll put

**HighCaliber AI Test Bed**
(19:15) this into the chat right now. HighCaliber AI is my consulting site, but it's also what I use as my test bed for anything and everything. I had this training curriculum that I had done for a cooking school, and a friend of mine's like, 'I'm working with some political campaigns. Could you'

**Applying Themes in Base 44**
(19:42) potentially do something with this?' I haven't actually done this yet, but I just said, 'Base 44, can you take what I did and apply it to politics? Add a red, white, and blue theme.' I gave it that direction. It probably would have figured that much out on its own, and it came up with all these detail things. What I really like about this, by the

**AI's Insidery Jargon**
(20:06) way, some of the small touches here, like GOTV—if you've worked in politics, you know it means 'get out the vote,' right? It's insidery jargon that it already built in and makes it sound like I know what I'm talking about. Clearly, if I was going to do this for someone or even do an hour or two-hour version of it,

**Reskinning Content Quickly**
(20:30) I'd need to review this in detail and see what they need and stuff like that. But to say, yeah, if there's something you've already put out there, and then you go into one of these apps—yes, you could do it in ChatGPT, you could do it in Claude, or your app of choice—but if you're also using a tool like Base 44, Codex, Claude Code that you've

**Tremendous Speed-to-Market**
(20:55) already built something in, okay, just reskin it. The number of subpages I have on my site now for all kinds of different purposes, all kinds of different events, anything I speak at, how quickly you can put something in market with all of it is tremendous. >> Can I chime in? >> Please? The tools that I have are

**AI Media Centers for Campaigns**
(21:26) not vibe coding, but one of the target markets I've always had were political campaigns. >> Mhm. Creating media centers, then having AI as a way for people to inquire, and then having the RAG GPT behind it where the answers only come from your own knowledge base and the answers come up. There are a couple of different

**Overcoming Campaign Challenges with AI**
(21:55) aspects. I've done a lot of canvassing for Bernie when he was running. There are a few challenges: one is training, like you said, outreach. There's a set of information, and there are usually gatekeepers for that information. You have to find them to get it. They have these Slack channels, and people

**Valuable Knowledge Center**
(22:21) repeat the same thing over and over again. To have a knowledge center where you put everything in there and people can just ask, and it's going to pull information and lead them to the source, is very valuable. If you have your own campaign material, appearances on TV, newsletters, whatever knowledge and information you

**Early Campaign Outreach**
(22:45) want to share, to make that easily shareable. I always thought, and I met with Cornell West's team when he was running for campaign, but my product wasn't quite ready. >> Uh-huh. >> The people who were running it loved what I was showing them. But, as you said, one of the things you want to do in a political

**Empowering Campaigners with AI**
(23:06) campaign is to empower everyone who is running to share valid information, and my KA tool is a perfect solution for that, and not live coding, but as AI. >> But when you, as a founder, right, when you're going to someone like Cornell West, or maybe want to talk to James Tarico, or whoever's bubbling up today,

**Demonstrating the AI Tool**
(23:40) have you been using it yourself yet for a quick glimpse about what this could look like? >> Well, what I did, and that's what I've done with the AI marketers group, is I took a lot of interviews and content from Cornell, put it in there so that when they ask questions, they would get answers

**Challenges with Political Campaigns and AI**
(24:10) and they would experience what it looks like. Absolutely. One of the challenges with political, the bigger ones, I remember back when I was going after Elizabeth >> Warren. Elizabeth Warren. >> Exactly. Elizabeth Warren. One of the problems they have is they're very conservative about

**Vetted Media Centers for Campaigns**
(24:33) especially when it comes to AI. If it deviates from their talking points by even one degree, they're very wary of using it. Political campaigns always want to have vetted information, especially the higher up they get. But it's all a matter of having one campaign use it effectively and then replicating it. But it's

**AI as a Viable Solution**
(25:03) having vetted media centers, and for that purpose, it works great. It works really well. You can probably optimize the prompting a lot more, but to pull information off your own database and provide answers, I think is a very viable solution. >> Is there anything that's held someone back from

**Failures and Lessons Learned**
(25:33) from doing so? I'll build on that, David, and ask, what have you tried that didn't work, and what did you change because of that? I'll give a quick example there to open that door, because one of the things I've gotten into playing around with, and I'll show you a failure in the chat,

**Trying to Vibe Code a Point Solution**
(26:01) here. Yeah, you could see where I tried to take something. One of the things I like doing sometimes is I'll find a point solution and be like, 'Can I vibe code this? Why or why not?' This was a case of why not, because I saw a tool that did something like being able to track some public social accounts and then get alerts based on

**Prototype Challenges**
(26:26) this, and I had ideas that I thought would be particularly useful for a client I was working on. Ultimately, I was building all this, well, in ChatGPT. We were getting far. I got enough of a working prototype going, but ultimately it was that I thought the app I was initially modeling this off of was

**Data Access Limitations**
(26:55) just really hacked together and not that impressive. They since had a major upgrade, and so they're a bit more professional now. What ultimately I came across was an issue of direct, reliable access to LinkedIn's data for being able to get public LinkedIn posts. It turns out when

**Third-Party Data Sources**
(27:25) I went to explore what this other app was doing differently. It turns out it was using a third-party data source, and ChatGPT Codex found the calls to this specific data source that the other company was subscribing to. So, being able to pay for that data made that app work. I was like, okay, onto one of my next projects, and I'll see if I can

**Failed Guitar Solo Project**
(27:54) either launch or break that one. It's a great question. Is there anything else where anyone got stuck or wasn't able to take something all the way? >> I tried something not for good, just for egos. I use music AI a lot to help me do stuff, and I wanted something where I could say, 'Here are the chords or here's the

**Academic Sounding Solos**
(28:20) music. Give me a guitar solo that sounds like a guitar solo.' >> Mhm. >> I vibe coded the hell out of it. The challenge, and I gave up after this, is that there are sources of harmonic structures that people, very academic sources, say these

**Data Access for Music AI**
(28:44) notes work with these chords. It provided me with really academic sounding stuff that was horrible, and it became much easier to just go to Sunno and strip out the rest of the music and have it play the guitar solo for me. It was really access to data more than anything else that was the

**Successful Mood Board Tool**
(29:04) problem, the same way you were talking about. On the other hand, I did something where I read that there was a mood board thing that one of the big publishers had come out with. I thought that sounds like it would be fairly easy, and it took about a half an hour of talking to one of the AI engines to code something. It's

**Usable JavaScript Tool**
(29:26) all in JavaScript and HTML. It saves stuff. It doesn't make any calls to anywhere else. It works well. I kept saying, 'Hey, could we save this? Hey, could I pick it up later? Hey, could we do this? Hey, could we do that?' It's a very usable tool. I posted that in the AIMG Slack somewhere. Anybody who wants to download it is welcome to tear

**InnerVoice and Interactive Journaling**
(29:48) it apart. >> Awesome. Feel free to post it in the chat too if you want. It reminds me of another app I was playing with that's more of a consumer app. You can see it here at innervoice.me, because I'm a big fan of the interactive journaling app Rosebud. I'm still using the official app, not my own in this case. So make of

**Exciting and Scary Aspects of AI**
(30:13) that what you will. But I was like, this to me is also one of the exciting and maybe scary things, because we have Kevin in the chat talking about what some of the 'moats' are, where I'm like, when I'm building these things, I'm not just saying I want to do exactly what this other thing does. It's like, 'Oh, this tool is useful

**Customizing Tools for Workflow Needs**
(30:37) for my workflow. If I could do X, Y, and Z, or the Y and Z on top of the X here, that would really hit it out of the park.' Maybe there are other people with those needs, too. In this case with InnerVoice, I was playing with Rosebud, and I love this interactive journaling app. It's actually really impressive

**Functional Explorations**
(31:00) for that. Are there other ways to learn about psychological modalities in the process? So, yeah, this exploration, I don't even know how good this thing is yet, but it's functional, which is further than I got for other things. Then, I'm playing with theological

**Exploring for the Heck of It**
(31:30) philosophical treatises and stuff. There are all kinds of weird things I can just explore for the heck of it, and if it doesn't live to see the light of day, or if it doesn't come up until 6 months from now, and I'm talking with someone about it, it's like, 'Oh, I actually built something like this. Want to give it a whirl?' This is the fun thing for me right

**Data Access as a Shared Challenge**
(31:55) now. I love it, Mark. It's also interesting having that shared challenge. So, what are some of those 'moats'? Access to data has been a big one for a long time. Mark, you know, I was deep in the social media marketing space for a while and even went over to one of those apps, Sisimos, for social

**Lack of Innovation in Social Listening**
(32:20) listening. I'm not saying this about Sysimos or any particular company, but in general, there was so little innovation in that whole space, and there were so many great early-stage startups that failed miserably in that space because you had to pay for Twitter's firehose. You had to pay for real access to Instagram.

**Ethical and Legal Use of Data**
(32:49) Those costs were astronomical for a startup. >> The other thing is that it's possible to find data. But we also have to think in terms of the ethical use of data, and the legal use of data that's fairly easy to scrape. But you may be running into PII from some of the sources, and then you're getting into all of the

**The Scary Side of Site Cloning**
(33:13) regulatory nightmare that we're living in. >> Well, and this also, when I was talking to the investment bank and I showed them how quickly I cloned the site, the obviously scary thing is how I could do this for anyone, and anyone could do this for your site right now. There's literally no way to stop that that I

**The Ease of Malicious Cloning**
(33:38) know of. There wasn't exactly a way 20 years ago to stop that from happening, except now someone can do this while they're having a cup of coffee. Yeah, anyone could have scraped some source code and done that before the ease of doing so, and then registering a domain that is off by one letter or includes a hyphen, and then creating email addresses on top of that.

**Gift Card Scams & Vibe Coding for Bad**
(34:04) and this is a group that I know has been hit by those crazy gift card scams. You get a text, and they see you're working with someone. 'Oh, yeah. I'm stuck in a meeting. Can you buy me a gift card?' 'Yeah. This is Jane from Banking Resource.' 'Go do this and buy a $50 gift card to'

**Vibe Coding for Good vs. Bad**
(34:34) Pizza Hut at Walgreens or something. And send me the code.' It's like, geez, that's scary, right? So, there's vibe coding for good, and then there's vibe coding for basic ethics. How you prevent vibe coding for bad is maybe above my pay grade. Other thoughts, projects you're building, like

**Mutual Learning Session**
(35:04) I feel like another time we can go back to more of the Vibe Coding 101, where we actually vibe code stuff on the fly. I feel this is better as a mutual learning session on what else could be done with this. But who else has learned anything from their vibe coding journey or attempt to do so? Randall, please, you could be running five, ten

**Personal Learning in Vibe Coding**
(35:28) of these a day anyway. So please don't hesitate. >> Yeah, I'll talk about something I've learned, and it's not about Vibe Coding, more about me and how I think about things. I've realized that where some of my particular strengths are, they don't necessarily lend themselves to how I think about my coding and some of these different tools. And

**Leveraging Strengths & Mitigating Weaknesses**
(35:51) so it's been interesting for me to find ways to leverage those things I'm good at, while acknowledging and maybe mitigating through finding other people or folks to help with those areas I'm less good at. My background is product and technology, and I tend to go very deep, very quickly, to understand the mechanics of things

**Avoiding Rabbit Holes in Prompting**
(36:15) without actually skimming across the surface and understanding the surface of things, the shape of things. What I found is that I can go in different rabbit holes very quickly, which is not necessarily as valuable as being very surface and layperson in my language and prompting. It's been a very

**Modifying Communication Style for AI**
(36:38) interesting observation for me in terms of how I work, think, and communicate, and how these tools respond to that, and what they're able to do versus what I see other people able to do. If you want to continue that question, for people who have tried different things, what have you found in terms of modifying your communication style or how you think

**Claude for Landing Page Research**
(37:00) about whatever you're working with? What have you changed to get better results from how you think about things? >> This is Brad. One of the things I did when I did my first landing page was, I asked Claude to do research on what typical landing pages are for that type of business, and I had it come up with a whole report and

**Claude Code's Plan Phase**
(37:20) what I did was I fed that into Claude Code. I started out not knowing what to do on this type of landing page, but having it research it for me through that report it creates, I fed that within Claude Code and used that information to create the page for me. One of the things I also found out is, I don't know if this is new or

**Conserving Credits with Planning**
(37:43) not, but Claude Code has a 'plan phase.' So instead of using your credits to build the page right away, you can switch it from plan to build so that as you're planning it out, it doesn't build it in the background. It keeps coming up with ideas and does the plan so you don't waste those credits. When you're

**The Value of the Plan Function**
(38:05) ready and it's thought through the process and saw something made sense, then you can trigger it to build from that. >> Yeah. I don't know where the raise hand button is, so I just gesture. But on that note, I've been toying around with a couple of nerdy projects in Claude Code, and I've used the plan function, and

**AI Choices and Feedback**
(38:26) I will say it is fantastic. It gives you the ability to put your stuff out there. Claude Code comes back with, 'Here are the things I'm going to do. This is how I plan to execute it.' And then it will often give you choices. So, it's like, 'Here's option A, here's option B, here's something else, or tell me what you think is different and what you think is wrong about this.' And

**Avoiding AI Tangents**
(38:50) to your point, Brad, it stops you from wasting those credits. I think your point about having either Claude or something else do research and come up with a plan and feeding that into Claude Code is great, because I found with any AI, if you just let it go, it can go off on a tangent that is great for somebody else but not what you want. So, second all of

**Data Science for Social Impact**
(39:17) that. I'm going to share something I saw on LinkedIn. There was a woman who lives in England. Her brother was coming for a pub crawl. She's got a PhD in data science, so she's not necessarily vibing with all of this stuff. But she turned that into the disappearance of local pubs. I believe she might have vibe coded the

**Empowering Good Ideas with AI**
(39:44) data visualization, or maybe she knew how to do it herself. Either way, the exciting thing I took away from reading this is, I can have a crazy idea about what is the good thing, what's something I can do for good that I otherwise wouldn't have been able to implement, and because of these tools, have the ability. >> And Zach, I appreciate your point in

**Top-Down vs. User-Driven Good**
(40:03) the chat. Is it up for users to go and use Vibe Coding for good? Because it's probably not coming from the top down from these major LLMs. I wouldn't be surprised at some point if there's just the way Google and others have funded nonprofits and various causes with certain kinds of ad credits that they're

**Collective Benefit of Vibe Coding**
(40:38) wind up being more like these credit-based systems, and if nothing more than just to make it look like they're doing some good in the world, papering some other things over. Maybe certain companies will come from a good place. But yeah, this is where how do we collectively

**Easy Access to Voting Resources**
(41:09) think about how to use this in ways that benefit others? Some of it could be as simple as putting together resources, just like you, making it easier for people to find voting resources when some localities aren't making that as easy as they possibly

**Greater Good of Voter Participation**
(41:37) could, and benefits people across the spectrum, not just an individual candidate's campaign. Yeah, all the stuff, I imagine your candidate, any candidate you'd work with, wants as many people to vote as possible, and that's very much the greater good when it comes to something like that.

**Awaken the Wonder: Knowledge Centers**
(42:05) >> Just want to add one thing. >> Go ahead. >> The slogan 'awaken the wonder' comes from this project to have a media center. Especially when it comes to voting and politics, how do you get people out of their mentality or challenge what they're thinking? By creating a knowledge source that may question or have a vetted

**Forums for Constituent Issues**
(42:34) research knowledge base to answer questions so it may challenge their preconceived notions. >> Mhm. >> Yeah. And to your point, David, the idea is, again, this is not for a particular candidate, but I know that California does this a lot where they have bills that they put

**Addressing Local Challenges with Vibe Coding**
(42:54) out to their constituents, but having a forum where voters and supporters can come in and bring up their issues that are targeting people. For example, in the Northeast, snow plowing will be a big one. So I think maybe it'd be a good opportunity to vibe code a site where people can come in and bring

**Candidate Focus and Voter Concerns**
(43:19) up some of their issues. That way, candidates from all backgrounds can say, 'These are the issues we want to focus on, and what are the things we can do and policies we can enact that can help remediate some of the challenges their voters are facing?' One of the ways you can find out if you have a forum is the questions they ask are the issues they are

**Missing Channels for Constituent Dialogue**
(43:41) concerned with. So it creates a way for you to understand where people are coming from by the type of questions they ask, and then you can respond with content. >> Right. Exactly. Other than that, you're going to have social media, whether it's on X or Facebook. But one thing I haven't seen, whether in the United States or back here in the

**Structured Information for LLMs**
(44:07) territory, that there's not a channel where people can address their issues that people in office can respond to, which is surprising to me. >> And by the way, that also can lead to a lot of structured information, which, of course, is great fodder for feeding LLMs. One thing I also can't recommend highly enough when using any of these vibe coding tools is, first,

**SEO and Geo AI Optimization**
(44:35) of all, start asking them, 'How can I make this site or app more SEO friendly or more geo AI optimization friendly,' especially if it's a website or web app. It often has many useful suggestions built in. Sometimes I will feed it a good article I've read about AI optimization and then say, 'Hey, Base 44, Codex, whatever I'm using, can you take'

**Google Search Console & Sitemaps**
(45:04) some of these best practices and make recommendations for what to do. You'll see things like these highlight boxes on top of most pages on my own site that are designed entirely for AI crawlers based on the best practices I've been reading about. The other thing that is super old school, that was one of the last steps along the way that I learned because I

**Indexing Pages for Discoverability**
(45:27) was like, 'Wait a second. Something still doesn't make sense here. How the heck is Google, let alone ChatGPT or anything, going to discover some of these properties I'm working on that I want to be discovered?' Base 44 is like, 'Oh, you need to take a sitemap and put it into Google Search Console.' And now

**Old School SEO Basics**
(45:50) you can authorize crawls of it. You can index your pages. It can tell you if and why any pages are not getting indexed. Some of this really old-school stuff, for folks like myself who've often just had other people doing a lot of this stuff for them, one of these basics, like, okay, yeah, these apps won't

**Manual Sitemap Creation**
(46:14) always instinctively create sitemaps and readme files and other information that's useful for these crawlers. So then, going in there and asking it for it, and then taking that other step until at some point this is done automatically by them, linking that with search console and being able to give your site at least a shot at being seen

**The "What Else?" Question**
(46:42) out there. It's just some of the basics become really important. So those questions of how to make this more SEO friendly, and also my favorite question to almost always ask with AI whenever I hit a stopping point on a project, is, 'What else should I be doing here?' Then seeing, okay, of those five ideas, I want to do one and

**Utilizing Excess Credits**
(47:13) four, go for it. Let's see where this goes. With a lot of that vibe coding, there are often great suggestions as far as how to take things further. If you wind up in a situation where you have, on some of these things, maybe more credits than you know what to do with, I've mentioned in the chat, the ChatGPT, I'm on there all the time. I've yet to

**Wrapping Up the Discussion**
(47:35) get a credit warning from them. I don't know why, because I'm only on ChatGPT's $20 a month plan, and I assure you, ChatGPT does not know I'm running forums like this. But if you have credits to burn, burn them. I'm going to wrap this up because we are at the hour here. Really fun discussion. I appreciate so many folks sharing what you've learned,

**Thanks to Subject Matter Experts**
(48:02) what you do differently. Big shout out to Randall, who gave me a primer on Claude Code, even if I'm not quite using it to the potential I could be. But I've learned a lot more about how to use other tools as well that I'm a bit more comfortable with. But we have incredible subject matter experts here, and Karen and

**Join the Slack & Upcoming Sessions**
(48:27) Earl, and all these incredible builders here, tinkerers like Mark. This is a lot of fun. Keep discussions going in the Slack. If for any reason anyone's on this call that isn't in the Slack, I put the direct click-to-join link so you don't have to register or anything special, right there in the chat. We'll be back next

**Future Events & Community Input**
(48:50) week. We got Nate Elliott from EMarketer. We'll take a week off from Architecture Live after that. If anyone's going in person, let me know. It'll be great to see you there. Then we've got an amazing schedule of guest speakers. Keep ideas coming my way and things you want to see us talk about and do through AMG. It's great to get to riff on

**Learning from the Community**
(49:12) things with you. I love these more open-ended community conversations because you're all doing such cool stuff, and I love learning from you. Have a great rest of your week, and we'll see you soon.