Episode Transcript
[00:00:05] Welcome to Prompting Curiosity, a podcast for the AI curious. No coding background required. I'm your host, Dr. Shantae Cofield, also known as the Maestro, and I created this show to explore what these AI tools actually are. Really though, are the files in the computer, how to use them, and what they might mean for how we think, work, create and move through life. Whether you're skeptical, intrigued, or already experimenting, you're in the right place. All that I ask is that you stay curious. All right, let's get into it.
[00:00:38] Hello, hello, hello, my curious people. Welcome to episode 37 of Prompting Curiosity. I'm your grateful host, the Maestro. And we did it, folks. We done did it. We automated the podcast workflow. So in last Week's episode, episode 36, titled My AI Wish List. Well start with the whistle there. Titled My AI Wish List List. If you don't, if you have, folks haven't seen me in person, I have a gap in my front two teeth and every now and then I get a little whistle. We leave them in cuz that's how, that's how I do. Um, but anyway, that was last week's episode and I listed out two things that I want to build with AI. First, one was an automation for my podcast workflow because I spent half of my life copying and pasting. And the second is a custom dashboard for my messaging coaching services.
[00:01:25] Item number one, folks, has been crossed off the list. And I'm actually so excited about it. I am recording this two days, or actually no, a full day. Not two days, A day earlier than when I usually record it because I was like, I gotta record it and try the workflow again. I know it works. I've been, I tested it a bunch, but like, I'm so excited about it. I was like, I'm gonna do this early. So it took about seven hours to do, uh, it to build out the workflow, to build the automation for this, but it's done. And, uh, that's what we're gonna be talking about today. It is a technical of an, of an episode, but I sat for a long time with the outline and did the best I could to make it digestible. And I think it's important to expose ourselves to these concepts. Um, if we think back to some of the early episodes and I, you know, talking to Rachel, she had said like, you know, she was able to help her mom and do these things because she's exposed herself to this. And that's what I learned from you. It's just like expose yourself to it. Sit with it listen to you. Don't go and actually do anything. I just want this, these terms, these words, these concepts to not be so foreign to you. And I guarantee that at some point you will jump in and you will try it, and things will make that much more sense, and you will be that much less fearful about it, because it won't be your first, uh, exposure to the thing. So, real quick, before we hop into the tech talk, thank you. Thank you again for the support and the excitement about the rebrand.
[00:02:49] Like, you folks are excited as I am, and, like, that's just. It was really exciting to me. It's really fun to me. And just. Just thank you. I really appreciate it. Um, and we got a new Review from Tri Swim 1998. Don't know who that is, but you all are the best. Uh, and shout out to, um, shout out to all of you. Just reaching out.
[00:03:09] Dming, just all, all of you. It's like I'm sitting here thinking and going through and the messages and just like, thank you. I really, really appreciate it. So today's main topic, my friends, automating my podcast workflow with Claude Code.
[00:03:26] What the heck did I do? What's going on? How does it work? Let's get into it. So right off the bat again, uh, for the millionth time, none of the actual creative portion of the podcast is being outsourced ever. All right? I'm not saying, like, write the podcast outline for me. No, nor. Nor is it obviously recording it for me. People do do that. There are podcasts that are just fully AI. That's not what we're doing here.
[00:03:50] The whole process, my whole podcasting process starts with me writing a super detailed outline. And then that is what I read from. That's what I'm reading from right now. To record the podcast.
[00:04:02] What comes after that? Right. What comes after that step is what I've automated, like, 95% of, and it makes me so happy. So to recap my podcasting process, I, um, and I. I want to. I want you to put a little asterisk in here. Remember, I turn each podcast episode into a newsletter and a blog post. And that's largely why I wanted to automate this. If it was just a podcast, I could still do it, but it was taking me a fraction of the time.
[00:04:29] Um, I don't even know if I would have, you know, had the. The impetus, the real motivation to do it, but because I have to, not have to, I choose to turn it into a newsletter and also a blog post. It's A lot of copying and pasting of the same things over and over again. And I'm like, I don't want to do this. So my process, I write this outline, I record the podcast, the actual episode, I paste the outline into Claude, I have Claude generate the show notes and the companion.
[00:04:55] From there I copy the show notes to a Google Doc, I copy the companion to a Google Doc. From there, I edit both of those documents. From there, I copy the show notes to an actual episode post on my website. I upload the episode MP3, which is like the actual recording. I upload the COVID art and I schedule the episode.
[00:05:14] Next, I copy the companion to an email template inside of Kit, my email service provider, and that will go out as a newsletter and I schedule it.
[00:05:23] That also requires that I bring in the image and that I link the actual podcast episode. From there, I copy the companion to a blog post on my website.
[00:05:36] I also generate a featured image on Gemini, which has been sucking. Hopefully it gets better. Um, I upload that featured image to the blog post. I upload that featured image to the actual newsletter as well. I schedule said blog post to go live, I schedule the newsletter to go live. I copy the podcast episode link and I copy the blog episode link to a master Google sheet. And that is the whole flow. And it's just a lot of redundancy. There's a ton of it. So I, uh, said, let me let the robots help me. What did I do? I used Claude code, which, remember, that is Anthropic's agent tool, right?
[00:06:08] Where it's Agentic tool, it is AI that can actually do things, that can create files, it can delete files, it can execute on things. It doesn't just answer questions. I use cloud code to create a custom Node JS command line tool, AKA a bunch of files that this program called Node JS can actually read and that automates the workflow.
[00:06:32] Perhaps I've lost you. It's okay, I'm going to explain.
[00:06:37] I had Claude Code build a set of instructions that I run from the terminal of my computer and that will allow that computer program Node JS to access all of the parts of my workflow, including Google Drive, my website, convertkit, or it's, uh, called Kit, and it can execute tasks in all of those places. For those of you wondering, the terminal, we've all seen it at some point, uh, at one point or another, it is a text based way to control your computer. All right? It's a little screen and it's just writing and some people, it's black, some people it's white. You choose the background color. But it's oftentimes very scary to people and they're like, what is this? It's only scary until you actually use it. And then you realize, oh, it's just a text based way to actually control my computer. Meaning I can write files, I can go into the files, I can delete the files, I can search the files all through text because it's only text. And you can't like actually use your mouse. It can be a little annoying because you have to like do one space at a time. You can't like highlight all the things, but that's what the terminal is. It gives you access to your, to your computer, allows you to control your computer via text. Okay. Um, so again, I had cloud code. Ultimately, like, what I had to do was it wrote a bunch of files that this particular type of computer program called Node JS is able to read. And inside of those files has instructions on how to execute commands and tie together my Google Drive and my website and, uh, my ConvertKit. Right. That is the summary of what is going on.
[00:08:19] The back end of this process and how it exactly works is techie. So I'm not going to go super into it, but I do want to expose you to it generally and I want to quell your fears around this by pointing out the fact that Claude code builds this, runs this, executes this, debugs all this. I didn't like write the code. It does that for you?
[00:08:40] Yes. There were some steps that I needed to do manually. A lot of copying and pasting, but it walked me through every step.
[00:08:48] Every single step. And the best part is that I'm just talking to it with natural language. Right. I'm typing. I use voice voice to text a lot as well. And then anytime I got stuck on something, I literally just took a screenshot of what I saw and I dropped it into the chat window and said, hey, I'm stuck. This is what I'm seeing. And it would tell me exactly what to do.
[00:09:08] Um, so I want to start at the end and explain what my workflow looks like now. And then I will briefly share how it works and then we'll wrap it up. Right.
[00:09:19] I'm excited about this. Um, that's what that, that's going to be our, our overview. Okay. So my podcast workflow is now I write the outline in the Google Doc and it's inside of the episode folder that was the same as before.
[00:09:34] I record the episode like I'm doing Right now.
[00:09:37] Then I run the command npm, run, generate, and the episode title. And from there Claude will generate the show notes and the companion.
[00:09:48] Um, which is crazy. It goes and it makes it. And then that computer program node JS creates a Google Doc for each of those two assets for the show notes and the companion inside of that Google Drive folder. It does it without me touching it.
[00:10:03] Amazing. So I just have to sit. It takes about two and a half minutes. I sit there after running that command, and then I can be. I can watch, I can see it happen on the screen. And those two files will appear in that Google Drive folder.
[00:10:16] From there, I review and edit each of them. One of the system, one of them is the show notes, one of them is the, uh, companion, which I had to do before. I want to stay involved in the process. So I look at it, I edit it.
[00:10:26] It's already formatted the way I want, which is amazing. I just have to change the words if I want.
[00:10:31] I will also go and create a featured image and I upload it to that same folder.
[00:10:36] Next, I just run the command NPM M run publish and the name of the episode.
[00:10:42] And then that program node js, it creates a, uh, blog post draft on my website, it creates a podcast episode draft on my website, and it uploads the featured image to my website.
[00:10:56] And it creates a newsletter draft inside of kit and everything is already formatted. It's exactly how it looked inside of Drive. It just pulls it over. That was a bit of a tech. That took a lot of work to make it look the same, but it pulls over exactly how I want. It's all formatted and all I have to do is just go and look at it. Nothing should change, but I go and make sure nothing has changed.
[00:11:15] And I hit, and I hit schedule. I choose the date that. Sorry for the whistle again. I choose the date that I want it to go live.
[00:11:22] I pull those URLs and post them to the Google Sheet. And that to me is. I use that as a personal, like a check step. Because I have a running list on a Google Sheet of like the episode, um, the date that it drops, the episode number, the title. And then I have the next column has the URL for the actual podcast episode, and the next column has the URL for the blog post. And so I put those URLs in myself. I was going to have, um, the program do it, the, the script do it, but I decided to do it myself. That's the way that I make sure that I Actually did it because I do the podcast so early that sometimes I'll go and be like, did I actually schedule those things? And instead of me going to the website and seeing that I schedule it, I can just look at that drive and if the URLs are there, it means that I scheduled it. If they're not there, it means I didn't schedule it and I got to go do it. Um, but the generate portion, like I said, where Claude actually makes. That's the only. This is the only part where actually Claude is involved, right, in this workflow. Um, Claude makes the show notes and it makes. Generates the companion. It takes about two and a half minutes. I timed it. Um, and I just let that run as I'm recording the episode. I actually let it run today when I went downstairs and had breakfast. So it was really cool that I was like, all right, I did the outline, went downstairs, had breakfast, and you know, three minutes into meeting breakfast, I just pulled up my phone and looked into my Google Drive folder and the assets were there, which is so cool to me.
[00:12:42] Uh, uh, the actual publishing command, meaning that's when the things that are in Google Drive, that's when they get pushed to the website and pushed to Kit.
[00:12:52] That takes like a couple seconds. It's like you push it and like, it probably takes 10 seconds.
[00:12:57] And all of it is billed as part of my $20 a month Claude Pro subscription. There's no actual additional costs to do this, um, because there's a thing called APIs and an API call, and we're going to talk about that in a few minutes. Um, but that. Because I'm not doing a ton of them, it's. It's all within the normal usage. Um, and so there's no cost for that. Right. So again, in reality, the only portion that actually uses Claude is generating the show notes in the companion. Everything else is done outside of Claude is run locally on my computer via the. Those files, right, That I had cloud code make that have these. Have all the instructions in it. And that program node JS is what executes and runs the, uh, runs those.
[00:13:39] Runs those files, right? Reads the directions that are the instructions that are in those files. Uh, yes, clearly I am still involved in the process somewhat, namely to check things, which is what I want. But you see, there is just so much less copy and pasting and making of the actual documents, which just adds up. Like those little things to me do add up. Some folks totally wouldn't mind, and they don't mind doing all the copying pasting and they wouldn't have spent seven hours. Yes, it took me about seven hours to build this thing out. They wouldn't have spent seven hours being on a tool to automate those things. But this is exactly what AI is for, one of the best use cases for AI and what it's about, which is building custom solutions to personal problems.
[00:14:22] As for how this thing works, I'm not going to get too into the weeds with it. But as for how this thing works, again, ultimately, what Claude code built is a bunch of files that have a set, the specific set of instructions and give the Node JS program the ability to interact with my Google Drive, my website Kit, via what is called an API. Right. Uh, these things can talk to each other via, uh, this API. The acronym right there, API is what I want to expose you to. API stands for Application Programming Interface. You will hear this term come up and it is simply a way that applications are able to communicate with each other and interact with each other. Right. It's a set of rules that allows these two things to talk to each other. That's it.
[00:15:09] Right. So as part of this workflow, I had to get Google Drive to be able to talk to my website and also Google Drive to be able to talk to Kit. And to do that, I had to set up and enable those APIs.
[00:15:22] Setting up, enabling that means getting a specific key for each. And a key is simply a long ass string of numbers and letters and then you save those keys inside of a file that that Node JS program reads. That's it, right? I know it feels like I'm like, what are you like probably sitting there like, what the fuck are you saying? This may be a good episode to go and look at the blog posts for this episode. Um, so you can read it and sit with it. I just want you to get exposed to it, that's all.
[00:15:51] So again, to summarize, I had Claude code build a bunch of files that live on my computer that a program called Node JS can read and execute. And it allows my Google Drive to talk to my site and my Kit account.
[00:16:08] Worth noting. Little, little aside here for those of you that are a bit more techy and or you have good memories. Last week when I was talking about my wish list, I said that I thought I would build this automation with what's called MCPS and a Cloud MD file. Um, and that is another way to build this. But after chatting with Claude, and that's a very important part if you're going to be building stuff, is to have the conversation with AI about what's the best way to build this and really plan it out. And so after the back and forth with Claude and identifying that this process, here comes a word that we've heard before. The process is deterministic, right? One input, one output. It's going to run the same way every single time. It made more sense to build it this way that I just outlined.
[00:16:48] I might take that other approach for the custom dashboard build out, though. So stay tuned.
[00:16:56] Overall though, folks, I just want you to understand that this is possible. Like, literally anything you want to do is possible. And Claude code can make whatever you want it to make.
[00:17:07] I ran you through the how it is totally viable for you to just build stuff without actually understanding how it works. That's not how my brain works. I want to know, like, what's going on? Why are we doing it this way? What happens if it breaks? Will it break? Can I fix it? What? Can you help me fix it? Can we identify what's going on? How much control do I have over things? How safe is this? There are plenty of people that are just, like, truly leaning into vibe coding. They're just like, make this thing for me. I need to look like this. And you can do that. You can literally just type the commands into cloud code and be like, I want a dashboard. I want it to look like this, I want it to do this. And then you could be like, it doesn't look good. Change this. I want this button over to the right. And you can just use natural language. Absolutely, absolutely. I am trying to expose you to some of the back end so you understand how this works. But overall, I just want to expose you to the fact that you can do this stuff at all. And Claude code will hold your hand through the parts that it needs you personally to do. Right. So, yes, there is a bit of tech tenacity needed here, but honestly, uh, it really, at this point, to me, feels more like patience than anything else, because you just need to be willing to take screenshots and literally chat with Claude when you don't understand things when they're not working and just be like, hey, this is not working. This is not working. This is what I'm seeing. This is not working.
[00:18:28] And just show it. And it will keep running through and be like, all right, let's try this, let's try this, let's try this.
[00:18:33] All right. I, uh, know. I know that this. This is a bit techy. And that, again, is because I wanted to explain to you how it works. I really, honestly, ultimately, just hope that you take away from this, this episode a glimpse of what is possible. Just a glimpse of what is possible, and you start experimenting with it yourself, because that is truly the best way to learn.
[00:18:58] All right, last things last, and then we will wrap it up. How I used AI this week. For those of you who don't know, maybe you're new here. Welcome. Um, each episode I share a quick example of how I used AI that week. Clearly, this week I use it to build out this podcast automation, this podcast workflow automation.
[00:19:16] I use Claude code primarily, and I'm going to do a future episode on what exactly Claude code is, how to install it, because I know it's kind of this weird, intangible concept. And that's how I felt about it, really, largely until I started using it for this project. I was like, I understand it. I've watched enough videos. But, like, I don't really get it. And, like, where does it live? What's going on here?
[00:19:36] How cool is it, dude, it's fucking really cool. Which is what everyone's been saying. But I'm like, until you use it and you dive in, you don't really understand that. Uh, so if you're scared, you're thinking about it, but you're a little scared. I got you. We're gonna do an episode about it. If you're feeling curious and brave, go for it. Dive in. I believe in you. All right, that, my friends, is all for today. Hopefully you found this episode helpful. If you did consider leaving a rating or review. We're, uh, at 27, folks. We're at 27 ratings. That's amazing. Let's get to 30. Who's going to help me out here?
[00:20:12] Don't forget, I also have that companion newsletter, the Curious Companion, that drops every Thursday. That is basically the podcast. Not basically. It is the podcast episode in text format. So if you prefer to read or you just want a written record of the things delivered directly to your inbox, join the newsletter fam. You can head to prompting Curiosity.com forward/newsletter or you can check out the link in the the show notes. Um, as always, endlessly, endlessly. One more time. Um, endlessly appreciative for every single one.
[00:20:44] Until we chat again next Thursday.
[00:20:47] Stay curious.
