The argument in one line.
Your foundation — local folders, skill files, context files, and MCP connections — is the permanent portable asset you own; the specific AI tool on top is swappable at any time, which makes 'which tool should I learn?' the wrong question entirely.
Read if. Skip if.
READ IF YOU ARE…
- You've spent months building workflows in Claude Cowork or Claude Code and feel anxious every time a new AI tool drops.
- You're a non-technical creator or solo builder running business workflows through AI and want a mental model for staying calm amid rapid releases.
- You're building skill files, context files, or agent workflows and wondering whether to start over in Codex or another platform.
SKIP IF…
- You've already internalized platform agnosticism and are here for technical model benchmarks — this is a mindset/workflow video.
- You're a developer looking for code-level comparisons between Claude and GPT-5.5.
The full version, fast.
The anxiety around learning the wrong AI tool dissolves when you separate the foundation from the tool layer. The foundation — project folders, CLAUDE.md and AGENTS.md context files, skill workflows, and MCP connections — lives locally on your machine, is owned by you, and is readable by any current AI coding platform. The AI tool is just the swappable engine on top. The creator proves this by running the same PDF skill prompt in Claude Cowork, Claude Code, and Codex simultaneously from one shared folder, getting near-identical outputs from all three. Platform agnosticism isn't a philosophy — it's a practical outcome of building foundation-first.
Free for members
Chat with this breakdown — free.
Sign in and you get 23 free chat messages on us — ask for the hook, quote a framework, find the exact transcript moment, generate a markdown action plan. Bring your own key when you want unlimited.
Create a free account →Where the time goes.
00:00 – 01:12
01 · Are you wasting time learning Claude?
Frames the tool-anxiety problem: constant new releases make it feel like you're on a treadmill, and the host promises a perspective that will calm the ADHD.
01:12 – 01:50
02 · Stay platform agnostic
Introduces the core mental model: build to a foundation you own, not to a specific vendor. Platform agnostic = you can switch without losing your work.
01:50 – 02:29
03 · All the AI tools
Maps the current landscape of AI coding tools: Claude Code, Claude Cowork, Codex, OpenCode, Gemini CLI, Hermes Agent, OpenClaw — and why the list keeps growing.
02:29 – 03:40
04 · Why these tools are interchangeable
Explains the three shared primitives that make all platforms equivalent: same folder structure, same context files (CLAUDE.md / AGENTS.md), same MCP standard.
03:40 – 05:38
05 · Demo: running the same skill in three platforms
Screen-share of the same PDF-guide skill prompt sent to Claude Cowork, Claude Code, and Codex simultaneously from the same workspace folder.
05:38 – 06:56
06 · Comparing the outputs
Side-by-side review of three generated PDFs — same style, minor visual differences — proving that the skill runs portably across all three platforms.
06:56 – 08:07
07 · Plugins and connectors
Shows that connectors (Claude) and plugins (Codex) serve the same function. Zapier MCP bridges both platforms to 9,000+ apps through one configuration.
08:07 – 09:29
08 · One foundation, any AI layer on top
Closing synthesis of the mental model plus CTA: subscribe, join the School community, get the Claude Cowork course.
Lines worth screenshotting.
- The foundation you've built in Claude Cowork — folders, skill files, context files — runs unchanged inside Claude Code and Codex.
- CLAUDE.md and AGENTS.md are the same concept across platforms; they transfer directly when you switch tools.
- MCP connections configured in Claude work the same in Codex; only the UI label changes (connectors vs. plugins).
- Codex, Claude Code, and Claude Cowork all operate on local folders — that single shared primitive makes them interchangeable.
- The question 'which AI tool should I learn?' becomes irrelevant once your work lives in portable local files you own.
- Tool anxiety is a symptom of building on top of the tool rather than underneath it.
- Running the same prompt across three platforms in the same folder produces near-identical outputs — model differences exist, but they don't break your workflows.
- Zapier MCP gives any of these platforms access to 9,000+ apps in one connection — one integration serves all tool choices.
- Platform agnosticism means you can benchmark new models against your real work without migration cost.
- The practical test for any new AI tool: open your existing project folder in it — if it works, it costs nothing to try.
Your folders are the asset, not the tool.
WHAT TO LEARN
Every major AI coding platform reads the same local folder — so the real question is never which tool to commit to, but whether your foundation is portable.
01Are you wasting time learning Claude?
- The feeling of wasting time on AI tools is nearly universal among active users — it's not a personal failing, it's a structural feature of rapid model releases.
02Stay platform agnostic
- Platform agnostic means you own the foundation and rent the engine — when a better engine ships, you swap it without rebuilding.
03All the AI tools
- The landscape of AI coding tools is expanding, not converging — building on portable primitives is the only hedge that doesn't expire.
04Why these tools are interchangeable
- Folders, context files, and MCP connections are the three primitives every major AI coding tool shares — build around those, not around any one platform's UI.
- MCP connections follow the same open standard across platforms; one Zapier MCP configuration gives every AI tool access to the same 9,000+ app integrations.
05Demo: running the same skill in three platforms
- Running the same prompt across Claude Cowork, Claude Code, and Codex from the same folder produces near-identical outputs, which means migration cost is zero once your work is foundation-first.
- 'Platform agnostic' is a concrete practice: open your existing project folder in the new tool before deciding whether it's worth switching — it costs nothing to test.
06Comparing the outputs
- Output quality differences between platforms are real but minor for well-defined skill files — the style and structure you crafted transfers, even if minor visual details vary.
07Plugins and connectors
- Tool anxiety is a signal that your work is stored in the tool rather than under it — moving skill files and context files to local disk eliminates switching cost permanently.
08One foundation, any AI layer on top
- The mental model to carry: one foundation (permanent, portable, owned) + any AI layer on top (swappable). The moment a better model ships, you try it by opening the same folder.
Terms worth knowing.
- Platform agnostic
- Building workflows so the underlying files and configurations work in any AI tool, rather than being locked into one platform's proprietary format.
- CLAUDE.md / AGENTS.md
- Text files placed in a project folder that give an AI coding tool persistent context about the project — rules, preferences, and instructions that load automatically each session.
- MCP (Model Context Protocol)
- An open standard that lets AI coding tools connect to external apps (Gmail, Slack, Stripe, etc.) through a consistent integration layer supported across multiple platforms.
- Skill file
- A saved workflow or prompt template stored in a project folder that can be invoked by name inside any compatible AI tool to automate a repeatable task.
- Codex
- OpenAI's AI coding agent released as a ChatGPT desktop app companion, running on GPT-5.5, positioned as a direct competitor to Claude Code.
Things they pointed at.
02:09toolOpenCode
Lines you could clip.
08:20
“We have one foundation, and we could add any AI layer on top of that.”
03:21
“They're permanent. They're portable, and this is something that we own.”
00:33
“Right now, it probably feels like you're on an endless treadmill chasing different AI releases.”
Word for word.
Read-along
Don't just watch it. Burn it in.
See every word as it's spoken — crank it to 2× and still catch all of it. The same dual-channel trick behind Amazon's Kindle + Audible.
metaphoranalogy
00:00Let's talk about the elephant in the room. Are you wasting all of your time over the past couple months learning Claude? Well, in this video, I'm gonna break down my honest thoughts about this.
00:08Right now, it feels like everybody and their grandma is talking about how they're switching from Claude co work, Claude code over to Codex. Even though you probably just got up to speed with what Claude is and you're starting to feel comfortable after months. So in this video, I'm gonna address whether or not you are wasting your time learning Claude or any specific AI tool for that matter just for a new tool to come out a couple weeks later.
00:29And I will tell you this video probably will give you a refreshing perspective and will probably cool down this ADHD you're feeling around all these different AI tools. Because right now, it probably feels like you're on an that are coming out, and I feel this myself, and it's literally my job to stay on top of AI every single day.
00:49By the end of this video, you're gonna understand why this is the wrong question, which tool we should be using. Because when you understand the fundamentals I'm gonna talk about in this video, it really doesn't matter which tool that we're building with. So let's dive into this now.
01:01Alright. So I wanted to make this video first of all because I'm getting so many questions about this inside of my school community. I have so many people asking, am I wasting all of my time learning Claude?
01:11And the answer is no. So the main point we wanna focus on in this video is that we want to stay platform agnostic. And if you don't know what exactly platform agnostic means, let me break it down very simply.
01:22This means that when you build something in a specific system like Claude and something new like Codex comes out, you are able to switch between the two because what you've been building is just a foundation that you could build upon with different AI tools. It essentially means that you're not stuck into one specific AI even when a new one comes out.
01:42So we're gonna talk about the mental model for us to understand so that way you could ease this tool anxiety that you're probably feeling if you're watching this video. So what we understand is we have a couple of different AI layers, and if you follow my blue little rectangle here, you could see the different layers that we have right now inside of the AI landscape.
02:00We have Claude Code. We have Claude Cowork, which a lot of my audience is familiar with. We have Codex, which everybody seems to be talking about because of the new Chatcha BT 5.5 model that is apparently the best model out there.
02:12We have stuff like OpenCode. We have Gemini CLI. And not even to mention, we have stuff like OpenCLAW.
02:18We have Hermes Agent. Man, there are AI tools coming out every single day, and it's really making us feel like, man, I'm wasting my time learning something that's gonna be obsolete in the future, and, no, that is not the case. So moving on to why these systems that we are building actually can be interchangeable, and that is because they work fundamentally the same.
02:37Claude Code, for example, and Claude Cowork and Codex, all these different AI tools and platforms work inside of folders on our computer. That is fundamentally how they work, and they all have the same context files. So if we're working inside of a specific workspace folder on our computer, we can always, at any time, access those CloudMD files or AgentMD files that we could transfer over when we wanna try a new tool, As well as if we have specific MCPs connected, which are the different tools that we are working inside of, like Gmail or Slack, Stripe, all these different tools, we could also work inside of those in all of these, you know, tools like Cloud Code, Codex, etcetera.
03:17And these three things that I just talked about right here are our foundation. They're permanent. They're portable, and this is something that we own.
03:25Since they live locally on our computer, we can always access these specific skills or agents that we've built inside of them. Now since this is the workspace that we're working inside of, we could open these in any of these different platforms. So instead of just talking about this, let me show you exactly what I mean.
03:40Now this is a folder that we're working inside of. This is the ClaudeCode short system. You could see all these little subfolders.
03:45Yes. This is messy, but this is where I've been working mainly inside of ClaudeCowork for the past four or five months.
03:52This is basically where my entire business runs. Now if you're anything like me and you've been using Cowork over the past couple of months, you're probably feeling pretty attached to the entire workspace that you've built inside of that folder. Well, the good part is is let me now toggle on Clawd Code over here.
04:06Now I'm inside the desktop app. If you don't have it, make sure to download this to use both CoWork and Code. And then what I could do is I could just create a new session.
04:14And as you can see down here in this little writing, Claude Code short system is the folder we're working inside of. This is the same one we were working inside of. And we could select, obviously, any of the folders where, you know, we have on our computer, but this is the one that I was using and built directly inside of Cowork.
04:28Now to show you this in practice, I'm gonna run this skill, which is my PDF guide skill, and I basically wanna create a PDF guide breaking down why it's important to stay platform agnostic with Claude Cowork, Claude Code, and Codex. So I'm gonna send that off inside of Cowork, and then what I'm gonna do is I'm gonna come over to Claude Code.
04:45Since I'm in the same workspace here, I'm gonna give it the exact same prompt, and I have that PDF skill inside of that folder so it's able to run it inside of Claude code too. I'm gonna send that off. I'm gonna let them both do their things, and then we're gonna come back to probably the exact same or very similar PDF guide that it's created inside of both of these different platforms.
05:04And while we're at it, let's do the exact same thing inside of Codex, which is ChatGPT's desktop app that they created to kind of be a competitor to ClawdCode. Again, we are inside of the same folder right here called ClawdCode short system.
05:18Make sure to just select the same folder that you're working inside of. I'm gonna type in PDF guide, staying platform agnostic with Claude co work code and codex, same exact prompt. But now we are using the ChatGPT 5.5 model, which some people are saying is better than the Claude Opus model.
05:33So I'm gonna send that off. Let's let all three of these generate this, and we'll come back and compare and contrast. Now that all of these have been generated, let's start with Codex and take a look at this PDF guide.
05:43Mind you, this skill was created inside of Claude Cowork, Claude Cowork, Claude Claude Code, and Codex even though that's not where, you know, I built this skill.
05:54So if I open this up, we have this PDF guide, which is in this exact style that I crafted, uh, again, inside of Cowork, and this looks really good. We have these, like, diagrams breaking down why this matters, these little snippets here of kinda just, like, talking points. I mean, this looks exactly like the ones that I generate inside of Cloud Code, Code.
06:13This is super cool. Now let's take a look at the other ones. Now back inside of Clodcowork it looks like I could click on this file here.
06:19Same thing. This looks basically identical. Obviously, looks slightly different.
06:23Like, there's a different graphic here than there was inside of the, you know, Codex one, even though it looks very similar. Looks like it also generated this little cool graphic. I mean, this is super solid.
06:32Again, this looks exactly the same with a couple minor details, of course, because it's not always gonna generate the exact same output, but it's the exact same skill. It's pretty obvious.
06:42And here we go. Here's the Claude Cowork output. Again, very similar.
06:45Same exact style. Of course, the visuals look slightly different. I think this one honestly might look the best.
06:50This one's cool, but, yeah, as you could see, this is working with these exact same skills because it's working inside of the same workspace. Of course, we do need to talk about the fact that these different platforms, for example, this one's Codex, they do look slightly different. They might have a slightly different feel, but they do function fundamentally the same.
07:06So if I look over here, you could see automations inside of Codex. And then if we come over to Claude Cowork, we have basically something very similar, which is scheduled tasks. And then inside of Codex, we have something called plug ins, which basically are just connectors if you've ever used Claude before.
07:22This is where we can connect our different applications. So, you know, we could configure Gmail, DocuSign, ClickUp, basically, all these different apps we use on a day to day basis. And it's the exact same thing inside of Claude, but it's just called connectors instead of plug ins.
07:36And if there's ever a specific connector you wanna connect to inside of either of these platforms, you could always use something like the Zapier MCP, which then allows you to connect to 9,000 plus different apps that it integrates with. So to show you what I mean, I have this specific MCP server configured with these different applications.
07:53For example, I have school in here, Google Drive, SyncFlow, couple of different tools. All I have to do is click on connect, click add to Claude, click connect, and then we could actually use those different applications inside of Claude. And then that would function the exact same way when we're back inside of Codex, for example.
08:07Now if you were able to stick with me through this entire video, you now understand that we have one foundation, and we could add any AI layer on top of that. Our foundation is just our folders, our skills, our context, our MCP that live on our local computers.
08:22And then whenever a new tool like ClaudeCode or Coworker Codex, OpenCode, or the Gemini CLI come out, we could actually go ahead and begin using those if those models are more powerful than what we've been using in the past. For me, personally, I spend the majority of my time still inside of Claude Cowork. I do use Claude code when I need to do something a little bit more technical, but I personally haven't found a need to switch fully to something like Codex.
08:46But that doesn't mean that I can't do it whenever I want to. Sometimes I will run a task inside of Codex just to compare and contrast to see if I get a better output because of the new GPT 5.5 model. Anyways, guys, I hope you guys can understand this concept, and I really hope that this maybe calms down your tool ADHD that you're probably feeling.
09:04It does probably feel like you're running on a treadmill trying to chase these different AI tools. But if you understand this, it should calm you down quite a bit. If you guys want more content like this, make sure to subscribe to this channel.
09:13I cover AI for non techies, and if you want to join my school community, we have some amazing people in here. I have a full Claude coworker course. I share all the different Claude skills that I use that you could also use inside of Codex, and it's a great place to be.
09:25With that being said, guys, thanks for staying to the end. I'll see you in the next video.
The Hook
The bait, then the rug-pull.
The title is a trap — and a relief. Every few weeks a new AI tool lands with a wave of "you should switch" takes, and the creator behind "AI for Non Techies" gets the same question flooding his inbox: is everything I built in Claude obsolete? This video answers it with a demo, not just reassurance.
Named ideas worth stealing.
02:29model
Platform Agnosticism Stack
- Your Folders
- Skill Files
- Context Files (CLAUDE.md / AGENTS.md)
- MCP + Tools
Two-layer model: a permanent portable foundation (folders, skills, context, MCPs) topped by a swappable AI tool layer. The foundation is owned; the tool is rented.
Steal forany onboarding flow or course that teaches AI tool setup — frame the skill as building the foundation, not learning the tool
02:39list
The Three Foundation Primitives
- Project folders on disk
- Context files (CLAUDE.md / AGENTS.md)
- MCP connections
The minimum portable set: if you have these three things set up, you can open any current AI coding tool and be productive immediately.
Steal forAI onboarding checklist, course module structure
How they asked for the click.
VERBAL ASK
08:20subscribe
“If you guys want more content like this, make sure to subscribe to this channel. I cover AI for non techies, and if you want to join my school community, we have some amazing people in here.”
Soft and friendly. Mentions the School community with a course on Claude Cowork. No hard pitch, no urgency.
MENTIONED ON CAMERA
FROM THE DESCRIPTION
PRIMARY CTAWhere the creator wants you to go next.
OTHER LINKSAlso linked in the description.
Visual structure at a glance.
