Parker Rex
All videos
Parker Rex Daily

The Vibe Marketing Race Is Heating Up (Are You Even In It?)

April 21, 2025

Watch on YouTube@parkerrex2

Vibe marketing is heating up, and Parker lays out the hands-on automation stack he’s building with Google Cloud to turn a camera into ready-to-publish YouTube assets—fast. He riffs on the latest news, tools, and next steps while keeping the focus on what's actually moving the needle.

Automating the Vibe Marketing workflow#

  • Drop a video into a Google Cloud bucket; daily raw is generated, then all assets are produced automatically.
  • Target outputs include: speech file, transcripts, subtitles, tags, a short summary, automatic show notes, and chapters.
  • Goal: move from camera to YouTube with as little manual steps as possible.
  • Current status: outputs are text files; still debugging formatting to meet YouTube’s needs.
  • The pipeline is containerized and cloud-run by design; assets live in an artifact registry with logs accessible via Logs Explorer.
  • This is a self-contained video processor with its own Dockerfile and prompts for subtitles and descriptions; designed to live in a dedicated service rather than relying on third-party tools.

Priorities: bottlenecks, ROI, and next automations#

  • Daily news video generation: automate scraping, scripting, and morning prep to cut research time.
  • School posts generation: sentiment analysis and content auditing for improvement; linked to iviwithai.com.
  • YouTube comment Q&A generator: apply to this and other channels.
  • Content planning for daily videos and main-channel content: scaffold the process so production scales without exploding overhead.
  • The emphasis is on practical automation that solves real bottlenecks, not chasing every shiny tool.

News, tools, and experiments#

  • Taskmaster + RueCode integration: direct integration coming; community buzz and real-world usefulness.
  • Context7 for keeping AI docs fresh: fast RG/rag-style access to up-to-date docs via MCP server; token management and chunking explained.
  • Example flows: adding a dedicated “context” rule, OCR snip workflow, and using Shotter for clipboard-ready outputs.
  • Content-creation tooling: tools that auto-publish to channels and auto-generate descriptions; potential for future workflows.
  • Title prompts and newsletters: explored patterns for turning interviews into newsletters and other outputs.
  • Self-contained video processor: Docker-based, with separate prompts for subtitles and outputs; stored in Artifact Registry; Cloud Run executes builds.
  • Logs and monitoring: Logs Explorer aids debugging and alerting; easy to route to external log processors.
  • Starter projects and code labs: cloud run jobs with video intelligence, scene detection, and Next.js context usage; learn-by-doing without leaving the environment.
  • NAN visualizations (GCP viz) illustrate how a robust vibe-marketing graph would map out many assets and services; personal note on how teams with dozens of editors clip and curate content.
  • Puppeteer Crawley (Puppeteer) for web scraping to pull data from closed APIs (used for sentiment and commentary data).
  • Shiny-object caution: don’t overengineer before your current use case is solid; the stack should fit your actual needs, not a theoretical ideal.

Week goals and upcoming content#

  • Roll out Vibe with AI; complete automation for both channels.
  • Publish more community content around Taskmaster, Context7, and Cursor.
  • Produce 10 pieces of content before Thursday to keep momentum.
  • Break down the AI coding stack into: research, documentation, debugging, planning, front-end, back-end, and databases; high-level overview on the main channel, deep dives on the school channel.
  • Expect a future video on the full AI coding stack; practical steps for choosing tools and implementing end-to-end workflows.

Community questions and interactions#

  • PRD access and value; explanation of “K crowd” (key opinion leaders, creators, thought leaders) and where the value lies.
  • Request for a Taskmaster-focused video on debugging/root-cause analysis.
  • Feedback on thumbnails and consistency; ongoing encouragement to keep producing and sharing templates.

Takeaways and mindset#

  • Solve your own bottlenecks first; automation compounds as you scale.
  • A self-contained, Dockerized video-processing pipeline makes updates safer and reuse easier.
  • Treat doc and model updates as a product: use fast context/documentation strategies to keep AI aligned with current tooling.
  • The goal isn’t to be the first with every tool, but to build a reliable, scalable stack that frees time for higher-value work.
  • Task Master (AI workflow orchestration)
  • RueCode integration (upcoming Taskmaster flow enhancement)
  • Context7 (up-to-date docs and MCP server concept)
  • MCP servers (document access and natural language querying)
  • iviwithai (content/dataset platform referenced)
  • Next.js docs (example use with context updates)
  • Google Cloud Codelabs (hands-on GCP learning paths)
  • Cloud Run starter projects (GCP examples)
  • NAN graph visualization (network/asset mapping concept)
  • Puppeteer (web scraping library)
  • Professional Services resources (common solutions and tools)
  • Artifact Registry / Container Registry / Cloud Run (deployment and infrastructure concepts)
Transcript

Vibe marketing. You probably heard of it. I've been talking about it on this channel if you've seen this channel before for a little while now. And it's going to be a huge part of everyone's stack moving forward because you can take personalized content from one piece of long form and then expand it to other channels, other platforms. And so what I've been working on is actually getting it to work with Google Cloud Platform because I have not seen any flows that I personally would put my face on. So, I've been doing them manually and I've done them manually long enough now where it's time to expand. So, on this video, this is a daily upload. If you want to see more generic content or broader content, you can go to my main channel. It's just Parker Rex. But, we're going to be talking about some of the news that came out yesterday, some questions if any of them came in, and then some of the strategy that I'm working on. So, right now I have an automation that I'm building where when I drag a video onto my mounted Google Cloud bucket. I drop it into daily raw, then it goes and generates all the assets that I need into here. And I'm just debugging something because the format of the outputs isn't what I need it to be. So when I do it right now, it's just all text files, but you can see that there's certain files that you need for YouTube. And the whole point of this is to go from camera to YouTube without anything in between. So that includes, as you can see here, the speech file. I don't need this one, but if I was doing speaker recognition, that would be helpful. the subtitles, the transcripts, as well as the tags, the summary, and automatic show notes and chapters. So, there's all these things that I was doing in this third party service, but I just enrolling it on my own because it's a simple set of services that is a great way to learn it. And I'm just getting the formatting right on that. And then the way that I was thinking of the different automations to do was not just hey what should I build next but stuff that I actually need. And that's a big differentiator because I've done it manually a bunch of times following the Elon Musk sort of method of don't automate first. And so when I was thinking of them I was like okay what are some good automations that would help me the most? What are my biggest bottlenecks so that I can do the highest ROI work? And one of them was around daily news video generation. So, scraping stories and then scripting and then basically having something in the morning that I'm just going through so I can skip through the research process. Another one was on the school post generation. So, I'll never automate all of school, but this will also soon be on its own platform called iviwithai.com, but I haven't had time cuz I've been trying to get all the content stuff scaled. But this will pull sentiment from all the posts and figure out what areas are people struggling with and then also audit the current bank or library of content so that I can make it better. Another one is the YouTube comment Q&A generator. So this could be on my channel but also other channels which is really cool. and then just different ones around planning the content for both this video and the main channel videos. So I started with the most obvious one which is just all the busy work around the YouTube videos. So you can see it's already deployed it but that's where we are with this and then now I want to get into some of the news and some of the questions. So first off there's a direct integration of rue code coming to Taskmaster hopefully. So, if you're not familiar with Taskmaster, it is the product manager for keeping all your AI coding assistants on track. And Ruode was brought up by someone in our community yesterday, Brian. He's like, I'm having the best time with this. It is so great. And I hadn't used it because I just thought it was a fork of Klein. I like downplayed it because it was a fork. But apparently, it's awesome. So, I want to check it out. And the second thing which I've already added to my flow is a problem that you have with AI docs is that they go out of they get stale fast and we need AI docs or we need docs because models fall behind the frameworks. So when you have a new version of next come out and the model was trained before that then you need to augment it. you need fast rag basically and so con context 7 solves that by providing an MCP server with access to 2400 docs 2463 and then you can see when they're last updated so the next docs were updated let's see yesterday kind of wild there's an API you can sign up for for early access but it's done by upstach and then what it allows you to do is say hey how do I use the new nextjs after function function use context 7. So this would be the way that you do it as an MCP so you can ask your questions naturally. But then of course you want to actually implement this into your taskmaster flow. So what I'm doing is I'm actually adding it to one of these files. I'm not sure which one because I don't want to mess with taskmaster because it just works. But it'll probably be its own rule. And then I'll just name it context. Oops. I'll name it context. And then I'll grab this rule that I pulled from another creator. His name is Cole, his video. And I would just did this with the OCR snip, which is using Shotter. And now it's got all the stuff on my clipboard. And I just put it here and format it up. But this will be really helpful because if Taskmaster or Perplexity or any of these are having trouble, then I can go ahead and have new docs always up to date. So I can just say boom. This is always Well, now that's done. Let's see what happened here. What did it change? It looks like exact format indicates the image was built for something different. Okay, cool. This means yeah it keeps this is one thing where I should have a context seven for this because Gemini 25 and is just like really struggling to remember that it needs to make the Docker image a certain way and it just keeps leading to the same thing over and over. But that's a case in point why you need context 7. So what I actually did was I went and I pasted in this deploy cloud run thing and you can see it went and actually went in there, found a couple of the docs, processed them, and then chunked them. And then now I should be able to go to here and then yeah, I already see it. It's pretty cool. One thing you can determine is how many tokens you want. So if I have a really large one, then I probably don't need all of it. So, if I typed in like let's see all and then tokens. Yeah, like then net docs 3 million tokens. You can't have all those. It'll just destroy you. Like you literally can't. So, they allow you to select the number of tokens and then pass that as a param on the MCP which is really cool. That was something I came across. Another one was this tool for content creation. I think it's it looks like somebody made this for doing this for clients, but then they apparently didn't love it because it's just someone like posted this on Reddit. But it's a cool idea where you have the ability to publish to channels from it. That's kind of cool. You have the ability to autogenerate the descriptions. Then some of the other stuff I found more interesting was actually like this title stuff. Let's see. Title prompts. And then it had the clients that they were working with. And then it had, let's find one of the stuff in it. I guess they didn't have anything for titles, but newsletters they did. So if I go in here, you can see this was to take their interviews and then turn them into newsletters. And it's all just a guey. But you can imagine this is kind of what when I see something like this, I get interested because this to me is a service where right now in my project, I'll try to make it a little bigger for you guys. Right now, my project, I have a few different things going on. So, I really need to nail the YouTube one. And I did it where it just did the videos, but then now I need all the stuff around it. But this video processor will be its own service. So, it's self-contained. It's got all the stuff that it needs in here. It's got its own Docker file. It's got its own logic in here for what everything is needed. And so you'll see that it has like the different prompts for the subtitles and all of that. And I probably need to split this out to be separate steps. I'm realizing has its own requirements. But all this is self-contained and sits on the container on the artifact registry. It's where all the Docker containers are. So if you're not familiar, you basically have an artifact registry and then you'll see inside of there the different ones that are available. So, I had a ton of them in here while I was trying to figure this out. And then you have the versions of them as they come in. And after that, then it will go like this basically like builds it for you. It's like the warehouse. And then artifact registry I keep pinned. And then cloud run is the runner for it. So very cool. Artifact registry I realize is is actually the old version of container registry. It's a little bit confusing, but that's kind of the nature of GCP when you're just learning it. And then I keep the logs explorer up so I can see that there we had some type issues. And you just can pin down either go to like the stream logs which is now or you can go to uh like the time that you want to have. You can change that. You can skip ahead, skip back. Really easy to use. And then you just would download the logs. You can set alerts up. I download the logs and then dump them over to Purser and let it do its thing. What else was interesting that I came across was all these different starter projects. So in this one, this is how you use cloud run jobs and video intelligence to process videos to get scene detection. So this teaches you how to make a container image using Docker. How to follow the principle of lease privilege by creating a service count for just that one thing. How to use video intelligence library from a cloudr run job. Very cool. And then how to make a call to the Google APIs to get the visual description of each scene from Vertex AI. And it just goes through every single one of them. They literally just do it in the cloud shell so you don't leave. And you set up the variables, set up your service account, create your bucket, create your job, and then ultimately run the job and then voila, you've finished. So these are really, really awesome lessons that they basically have. And let me see if I can't get back to where this was found. Yeah, I was just trying to find an example of it. It's like, oh, they're in the code labs. So, let me just take this route off. But if you're interested in learning this stuff, code labs are kind of insane where you just go through them and you figure all this stuff out and it's really helpful. Another thing I'm looking at, this to me screams shiny object syndrome slash potentially overengineered, but I don't know. I haven't done it. But this is this framework that people have put together where it combines a bunch of acronyms like spark principles. So clarity, extensibility, testing, collaboration. But I I don't even want to like really get into this because I'm just really starting to figure out my stack. But I just want to call it out that while this may be good, I just want to warn a lot of people that when you see something like this, don't immediately think it's awesome. Like you might be a few days away from your current thing really working. And uh yeah, that that's just something that I've personally felt and it's it's totally a thing. And then back to the GCP stuff. So when you look at NAN, you have these nice nodes and this is what GCP viz does. So it's a go tool that you run locally and then when you run it you see this nice thing and it'll actually generate the node structure to show you your network of things going on. So this is the goal like one of these like this is basically vibe marketing in a nutshell where you have all these child nodes that are calling to different things different services and right now I probably only have like this green one and then it does and each one of these represents an asset that I'm generating but I can't wait until I'm at this level because not only will the content be better because I won't have to do all the work around it all the administrative work. But I look at this graph, this firework looking thing, and I think, "Wow, this was teams of people." Like there are people that have 150 clippers that just clip content. If you can make a really good clipping service, which is just a folder full of Python stuff, and have it totally be custom to you, when everyone was spending their time with the Clippers and then going to NAN or going to Opus or going to Descript and it all looks the same, you might be a little behind at first, but I really think that this is going to pay off because there's just longevity to it and it's putting your eggs in the Google basket. not a open source TypeScript project, which is fine. Like I'm not hating at all, but I just think that you can even host NAN on here if you wanted to. I just think it's so much more powerful when it comes to the ability to also do idea generation and being able to vectorize these huge pieces of the internet. Like it's kind of mindboggling. Then another resource that I'm kind of digging through. This is where I found that GCP viz thing. This is called professional services. So common solutions and tools developed by the professional services team. And there's a lot of stuff in here that's really helpful. So even like cloud build applications, CD examples, cloud composer, there's just so much stuff and because of that a lot of people are intimidated by it. But I'm just going for my one use case, which is let me replace a phonic because it's been unreliable and I want to have a streamlined workflow because it honestly it takes forever. There's a lot of complaints that I had with it. Like one, it doesn't work the way I want it to. Two, it's inconsistent. So, okay. So by going after that use case, it pushes me through the learning curve because I really need the thing because I want to be working on other stuff. I don't want to do like I've recorded for 15 minutes and I'm going have to spend another 15 minutes to do like the thumbnail and the titles and yada yada yada and schedule it and wait for this, wait for that. If I have a phonic, it adds another whole 40 minutes to it. So I'm like, I really want this. And that goes to learning anything. Like I'm solving my own problem. And I know it's a problem worth going after because I know when I make videos, I learn more because I'm forced to know my stuff and I know that the stats go up as well. So, it's like if I look at the stats, I've only been doing this for a month on both. Yeah. Literally 20 20 days is when I started on both. It's like I know these will just go up and I'll have more impact if I can make the content better, if I can get the message out further. So, yeah, just think your why. Another thing I was looking at is I didn't know much about Puppeteer. It's been around forever, but they have this product called Crawley and it's a web scraping library handles the block and crawling and proxies and browsers for you. And so I used this and Puppeteer yesterday to go and pull the data out of school because they have a closed API. This is kind of the only way to get stuff out of there. And that's important because I want to know again like the sentiment around stuff like the commentary, what's going on? And I'll use this for other things to to learn basically. So lastly, is there anything else on the news front? I have some big goals this week that I need to do because I'm leaving town on Thursday. So I need to roll out Vibe with AI. I need to complete the automation for both channels. Ideally do another one. I have some good community content going up around Taskmaster and Context 7 and Cursor. And then I needed to put 10 pieces of content out before Thursday. So big week and one of the parts that I'm going to be going through is the AI coding stack. So everyone just thinks of this part when they think of coding, when they think of like, oh, I have like Taskmaster and it does just the implementation, but there's a lot of other steps. And so I'm going to make a video about this on my main channel, but it won't be as indepth as the school just because there's so much to it. It's crazy. Uh like knowing which tool to use. I'll go through it high level on the main channel, but then in depth will be in school. So knowing how to do the research, how to do the documentation piece. There's other options besides context 7, but like actually explaining how to do it. And then debugging, I threw augment on there, but I have other methods. the planning, the front end, database, implementation, and then this should probably be broken out to back end because it does depend on what stack you're using, how complex the thing is. So, I just see this. It's kind of like that fireworks chart that I showed you earlier, but it's a set of if else like statements like if it has this stack, do this. If it's missing that, then you don't have to like there's some stuff where you just don't even need to do these steps, right? This would be for the full thing, but a lot of the time you just wouldn't need all that. Let's see if we have any community questions real quick before we wrap. And I need to return to the studio. Had a good video go out about PRDS. If you want help on PRDs, like the way your brain works, my man. Thanks. Great ideas on bringing all disparate pieces together. Once it's working, you can easily productize and sell this setup to the Kol crowd. What is K? What is the K crowd? Kings of Leyon, the band. Literally have no idea. Key opinion leaders, creators, experts, thought leaders, I guess. So, I'm not sure. Keep up the great content, Rex. Thanks. I love a tutorial on how to use Taskmaster for debugging/root cause analysis. Every time I try, I just end up buried in subtasks and quickly becomes a mess. I really love the videos and really appreciate them. Nice. Thanks. Do you mind to share the PRD? Yeah, here you go. I've been like holding out on it, but there you go. Post it once and then I'll have that on the other site. I tried Taskmaster for some large projects that I had specs for. Turns out didn't work as expected. I'm testing 03 high to improve tasks. Yeah, they have a PR out for open router support so that you'll be able to do any model which should be helpful. Yeah. And then let's see the other one on the daily which is this channel. If you want your questions answered, I do them at the end of the videos. Now, let's see. All I interpret is another guy on internet who has found ways to make money, but he makes YouTube to sell courses. Yeah, you should watch more of my videos. It's like I basically am using community. It's called the community audience community funnel, something like that. But it's really like cool, give out a ton of information. If people want more information, then they can be a part of the community. And then I'm leveraging that to then build the stuff I really want. Like my goal is not to be like an info product course guru bro at all. It's more like, hey, I live in a tech desert and I'd really like to have the kind of feeling that you get by being in San Francisco in terms of technical people talking about technical things. Like I want that on the computer because I can't get that where I live. So I'm trying to emulate that. And then on the money side, it's like you kind of have to charge money because the significant amount of work. And I don't need to like justify this away, but there is a reason behind this. She says, "I like your thumbnails." Thank you. You're not that guy, pal. I love it. Yeah. Uh, super consistent. Yeah. Hari says, "I am him." Hari, you got to start preparing. You're doing a workshop on Wednesday, so prep up. And that's it. I appreciate you guys watching and I'm going to keep basically just pounding away getting this stuff done and then all the videos are going to be a lot better and it's going to be really cool when it is live because everyone that watched these videos got to see me do it and then if you wanted to copy it, you could just watch the videos and I'll probably templatize this stuff in the community. If you didn't like the video by now, do you hate me? What's going on? And then of course, if you got value out of it, you should subscribe so you can get more value in the future.