bouncer
← Back

Mark Kashef · 8.6K views · 446 likes

Analysis Summary

45% Moderate Influence
mildmoderatesevere

“Be aware that the 'mind virus' framing is a rhetorical device designed to make you feel intellectually compromised by current trends, thereby increasing your perceived need for the creator's specific 'cure' and paid community.”

Transparency Mixed Transparency
Primary technique

Performed authenticity

The deliberate construction of "realness" — confessional tone, casual filming, strategic vulnerability — designed to lower your guard. When someone appears unpolished and honest, you evaluate their claims less critically. The spontaneity is rehearsed.

Goffman's dramaturgy (1959); Audrezet et al. (2020) on performed authenticity

Human Detected
95%

Signals

The video is a personality-driven tech tutorial where the creator shares a personal workflow and strong opinions on software frameworks. While AI tools are used to generate code and a website mentioned in the video, the presentation, narration, and creative direction are clearly human-led.

Natural Speech Patterns Use of filler phrases like 'So, in the last 24 hours', 'And for what?', and 'All right. So, it's no secret' indicates a natural human flow rather than a synthetic script.
Personal Anecdotes and Opinion The creator discusses their personal frustration with debugging 50,000 lines of code and their specific decision to delete a repo from their laptop.
Contextual Tool Usage The creator explicitly mentions using AI (Claude Code) to build parts of the project, distinguishing between their creative direction and the AI's output.
Non-Formulaic Narrative The transcript follows a conversational, opinionated structure typical of tech influencers rather than the rigid setup-obstacle-resolution pattern of AI content farms.

Worth Noting

Positive elements

  • The video provides a practical demonstration of using Claude Code's 'agent teams' feature to synthesize features from multiple repositories into a custom project.

Be Aware

Cautionary elements

  • The use of medical/psychological metaphors ('mind virus', 'cure', 'antidote') to describe software preferences, which bypasses technical debate in favor of emotional manipulation.

Influence Dimensions

How are these scored?
About this analysis

Knowing about these techniques makes them visible, not powerless. The ones that work best on you are the ones that match beliefs you already hold.

This analysis is a tool for your own thinking — what you do with it is up to you.

Analyzed March 23, 2026 at 20:38 UTC Model google/gemini-3-flash-preview-20251217
Transcript

So, in the last 24 hours, OpenClaw pretty much became close claw, and I personally deleted it completely off my laptop and found the simplest and laziest way to make it my own. So, if this thumbnail and title made you click, then this video is definitely for you. Open Claw was the beginning, but it wasn't meant to be the end. I've spoken to so many people that feel compelled and brainwashed that they have to use someone else's framework and they've spent hours, weeks, and hundreds of dollars to debug and make someone else's workflow work perfectly for what they do day-to-day. And for what? To wrangle someone else's code and make it work for your very specific use case or very specific preferences. Open Claw and Company were built for the masses, but you can just build for yourself. So, in this video, I'm going to cure you of the OpenClaw mine virus. All right. So, it's no secret that the world has gone crazy over OpenClaw. It's now at 150, 160,000 stars on GitHub, one of, if not the most popular repo ever. But the main issue here is everyone is taking this one repo that has a million ways to do things. So, it's basically a boilerplate, and they're trying to make it meld to their exact way of doing things. So, everyone wants to establish their own workflow, their own APIs, their own tools, their own permutation on what a sole MD is or how to configure a personality. I made a video along with many other creators on the fascination I had with tinkering and playing with this to understand all the mechanics, all the simple elegance of using markdown files to create a magical experience. But behind the scenes, there's no real magic. It's a series of prompts, cron jobs, some rag, and a combination of prompting that makes it feel like it's self-learning, even when it's not. If that wasn't enough, we then had a fork explosion where we took Open Claw, and you created infinite derivatives. Zero claw, nano claw, picl claw, tiny claw, and the list goes on. Each of which would take a permutation of the existing one. And again, try to build something for the masses that at the end of the day you have to tailor for yourself. Spending hundreds of dollars and hundreds of hours to make an existing framework do something it might not be designed to do. Why not just build it from scratch, not you cla it and give it everything you want and have it figure it out. So my prescribed solution to you is to take all the different repos, open claw, insert name of claw, look at the things that matter to you, create your own wish list, which I've made infinitely easier for you, which I'll show you in 2 seconds, and then send that to claw code, it will create its own version that's fully explainable. So even when it comes to security, you can understand each and every part of your version of OpenClaw and never have to worry about that ever again. So what you're looking at here is an AI generated website where it's taken every single repo that I mentioned in the last image and it's created a whole catalog a catalog that explains the difference between openclaw zero claw and as you go down it goes through the simple concepts of the architecture. So, it breaks down what the SoulMD looks like, how it's structured, and as you go to the very bottom, I have this feature grocery store, and I'm going to make it available to you in the second link in the description below. So, you can take it, build your own wish list, and be up and running in less than 1 to two hours. And here's exactly what it does. So, each part is a category, and these are categories of amalgamated features across all the different repos. Now, did I do this myself? No. I'll show you in a second how I made Claude Code do it in the simplest of ways. Once it did it, all you have to do as the user is pick the things that matter to you. So, I'm going to say, you know what, I want a soul file and I want per group isolated memory and switchable agent profiles. You can also see which repo this concept element or feature originates from. And as you go down, when it comes to security, I can say I want this, this, and this. And if we go down, communication channels. Now, if you're using OpenClaw, you are inheriting code for multiple modalities of interacting with your bot. But if you just care about WhatsApp or you just care about Telegram, you can just borrow the implementation. They figure it out, import it and make it your own. So I'll click on Telegram and for memory knowledge, I will pick let's say two-layer memory system. and then automation and scheduling. Let's do browser web hook cron job and then MCP. We'll do some skills. And then here is an amalgamated list again that cloud code put together across all the different repos I showed you. So let's say I only care about Obsidian and Gmail and Whisper Speech to text and everything else I'm good for for now. And then beyond that, let's say you just want one of these. This is a graph or a table rather that compares all of them just in case you want an overview. And then at the very bottom, it has the actual architecture patterns. But this is not what you should worry about. What you should worry about is this beautiful part right here. So now that you've created your grocery list of what matters to you, you can click on export as claude code brief markdown file. And then you get this beautiful my framework ingredients. This has a full list of everything that you want to repurpose and create using cloud code. So now once we have this, you can feed it as I'll show you and it will go and create using a series of agents an entire framework that's specific to you. Now like I promised in terms of how I created that web page, it was actually very straightforward. I said clone the following GitHub repositories. So in this case, each one of them spawned in these folders. And as you clone them, you'll inherit all of the underlying code, text, markdown files, everything that makes that repo tick. And after we've imported everything, all I did was send this very menial prompt. I'll read the first couple lines. Assume that I'm a non-technical user and I want to understand what each one of these repos has to offer. I want you to explain all the concepts that I might want to pick and choose from if I want to be able to build my own framework that is derivative of the main things that I care about in my day-to-day life. Can you create a visual representation of what each repo has to offer and create a web page that also does a comparison between all of them? And then I specify exactly how I want to create this grocery list, which is very straightforward. Now, once I sent this over, it created the web page in probably under 10 15 minutes. I took a look. I just iterated on the UI and then we had that extra feature that I added to create and export that grocery list in a way that was very descriptive. The first version just gave me the actual list of things that I wanted but not a description good enough for cloud code to understand exactly what a particular skill could do. Once I downloaded and imported my wish list that I showed you just a few minutes ago, I just opened a brand new terminal session and this is literally all I said. I want to build my own framework based on this deep dive file right here. This is my wish list. I'm non-technical. Help me bring this to life in a way where onboarding will be as easy as possible. So my goal was to to give you a little teaser here, have an experience just like this where I have my CLI and in this CLI, you can see the very bottom here. I could set up my bot very easily. I could just give it the token. All I cared about was Telegram. I could just give it my API key for open router. I could pick the model. I could tell it exactly what I want the bot to be called, the personality, and I was up and running in less than 5 minutes. And just to prove to you that I'm not full of it. If I just say, "Hi, my YouTube audience is watching. Say hello." You'll see here it's already learned that OpenClaw is a mind virus. It should be able to type come back. And this was infinitely easier than setting up OpenCloud at the time Claudebot two, three weeks ago. And there you go. It's greeting you to this video saying welcome to whatever Mark's got going on here. So I can show you that that does work. And this whole CLI by the way was fully possible by just taking this having it go create an entire plan of how it replicate it. And I had it interview me. Now I have a personal slash command called marks planning which is more powerful than plan mode because it really pushes back on me and gives me multiple multiple choice questions. Once we went through the multiplechoice questions it created a series of phases on how it would execute my version of openclaw that is specific to my needs. So in this case I called it antidote. It came up with a plan. This was the project structure and based on everything it learned from the other repos right here, it created a list of all the Python files it would need to spin up really quickly and more importantly for me in as little lines as possible. And obviously you will write absolutely zero of it. So at the very bottom of this, it gives me the risks of building things in a certain way. And then once we're good to go, I say create a markdown file called antidote.mmarkdown file with this entire plan in detail on what needs to be executed. Now I wanted to optimize this plan so that it could be easily spun up by the agent teams feature of cloud code. So once it did that, we got a file like this where it walks through exactly what the build plan is, the decisions we made back and forth, what we explicitly chose not to do, the project structure, and everything it needs. And this is all Python stuff. Once we had that, all I had to do was tell it to execute the antidote MD file and spawn an agent team. Then from there it took around 1 hour in total to go through build everything test it with this agent team and even my prompt was literally just spin up an agent team to build test and deploy this entire thing. Now the one part that was missing was creating this CLI style asy art so you could have the exact same experience as openclaw especially on onboarding. So once you finish you can ask it to create a oneline command spin up startup just like openclaw it will go through the openclaw repo learn and understand how to put it together and then it will give you this command you can run once you run this command that's where you're in the promised land and you can execute your own terminal your own way on your own hardware and even if you want to optimize it for local it's completely your call and that's pretty much all you need so you can take the HTML file I'll make available to you in the second link in the description below. You can take it for a spin with whatever you want. Could be codec, could be claude code. I couldn't care less. What I do care though is that you stop feeding into this hype, this very anthropomorphizing hype that this is some revolutionary technology. What's revolutionary is the concept and multiple vendors will build on this concept, whether it's OpenClaw or a completely different framework that is not lobster based. So, you're literally just a few clicks, commands, and markdown files away from just importing all the repos of interest that have the features and the elements that you might want to shop. You go shop, you make that wish list, you feed that wish list to Claude Code, you have it take care of all the plumbing, you give it your different takes on how the personality should be, where you want it to live. And once you go through that process, you can own your own infrastructure. You can still always shop what's out there if something new comes out. But most importantly, you can stop having to freak out with every brand new Open Claw video or Open Claw update. Now, if this ends up working for you, please let me know down in the comments below that I've cured you. It would make me feel a lot better because there are many people that are not focused right now and getting distracted by things that don't matter. If you want access to my exact repo that I showed you right here, along with tons of exclusive Claude code content and my very Debbie Downer rational way of building and using things like claude code and insert word claw, then check out the first link in the description below and maybe I should see you in my early adopters community. I'll see you in the next one.

Video description

Join 800+ AI builders in the community (includes Claude Zero to Hero course): https://www.skool.com/earlyaidopters/about Grab the free OpenClaw Mind Virus Cure: https://markkashef.gumroad.com/l/openclaw-mind-virus-cure Book a 1-on-1 call: https://calendly.com/d/crfp-qz3-m4z --- OpenClaw just became ClosedClaw -- and I deleted it off my laptop. Instead of spending hundreds of hours wrangling someone else's 50,000-line framework, I used Claude Code to read every major fork, build a feature grocery store, and spin up my own version in under 2 hours. This video shows you exactly how to do the same. --- TIMESTAMPS 00:00 - I deleted OpenClaw 00:51 - Why everyone struggles with it 01:48 - The fork explosion 02:10 - Build from scratch instead 02:44 - The framework comparison tool 03:25 - Feature grocery store walkthrough 05:00 - Exporting your wishlist 05:32 - How I built the comparison page 06:51 - Feeding your wishlist to Claude Code 07:25 - Antidote CLI demo 08:19 - The planning process 09:22 - Agent teams building it all 09:55 - Creating the CLI experience 10:29 - Stop feeding the hype 11:35 - Outro --- #openclaw #claudecode #aiagent #aiframework #aiautomation #anthropic #claudeai #opensource #aicoding #buildyourown #aitools #pythonai #agentzero #aiassistant #claudecodeagents

© 2026 GrayBeam Technology Privacy v0.1.0 · ac93850 · 2026-04-03 22:43 UTC