Modern Creator
Jake Van Clief · YouTube

Your AI Folder Works on Your Computer. How Do You Give It to 1,000 People?

A walkthrough of EdubaWare, a closed-alpha platform where you upload a zipped folder of markdown and skills and it becomes a shared, permissioned AI agent.

Posted
1 weeks ago
Duration
Format
Demo
educational
Views
13.2K
227 likes
Big Idea

The argument in one line.

If a well-organized folder of markdown and skills already works with AI on your computer, the only missing piece is a platform that hosts it, so deployment replaces building traditional software.

Who This Is For

Read if. Skip if.

READ IF YOU ARE…
  • You have built a second brain, brand kit or skill folder that works great locally and you want to hand it to clients or teammates without writing an app.
  • You build agents with Claude Code or similar tools and keep hitting walls around deployment, user access and security.
  • You run a small agency or consultancy and want to deliver a branded AI agent to a client's whole team.
SKIP IF…
  • You want a hands-on tutorial. The video is a product walkthrough of software still in closed alpha, and the creator says it may stay confusing.
  • You need something you can sign up for today. Access is by invite, with open alpha expected around December 10.
TL;DR

The full version, fast.

Second brains, brand decks, plugins and GPTs are all just organized folders of markdown, and AI can already work from them. The missing step is deployment. EdubaWare lets you zip that folder, upload it, and invite people by email to chat with it, with organization-level separation, per-person access, your own models and keys, and a choice between shared or individual chats. Two examples show it: a brand kit built from a company's website, and a report generator built from an engineer's folders. It is in closed alpha, with open alpha expected around December 10 and release January 1.

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 →
Chapters

Where the time goes.

00:00 – 01:52

01 · Why we're building this

Intro to the closed-alpha software and the bet that value flows to the people using the platform, not only its makers.

01:52 – 03:36

02 · First things first

The UI will change over time. The focus so far was security and backend. The goal is upload a zip and let thousands of people use it.

03:36 – 05:58

03 · It's all folders

Second brains, brand decks, GPTs and plugins are all organized markdown folders, so why not double down on that.

05:58 – 09:23

04 · Example 1: a brand kit

Building a brand deck folder from a surfacing company's website with the ICM skill.

09:23 – 12:33

05 · Upload the zip

Zip the folder, upload it, and chat with it as a multiplayer agent the whole company can use.

12:33 – 14:45

06 · Example 2: field reports

An engineer's organized folders and one example report become a personal report-generating agent.

14:45 – 17:35

07 · Who gets what

Organizations, invites, access controls, own models and keys, and driving into other people's chats.

17:35 – 18:35

08 · A coach's chatbot

A coach deploys an interview chatbot by uploading a zip, and decides shared versus individual chats.

18:35 – 19:59

09 · The end goal

Small deployable software where markdown is the software and an AI inside the platform runs it.

19:59 – 21:57

10 · How to get in

Closed alpha, open alpha around December 10, full release January 1, and the community links.

21:57 – 22:47

11 · The short version

If it works on your computer as a set of files and folders, deploy it and manage who has access.

Atomic Insights

Lines worth screenshotting.

  • A second brain, a brand deck and a company operating system are all just well-organized folders of markdown files.
  • ChatGPT GPTs and Claude plugins are packaged markdown files plus connectors, so a folder is already the unit of agent distribution.
  • If markdown instructions can get the outcome of software, the only missing piece is something that runs and hosts them.
  • A brand agent used to mean code, agent structure and a RAG database. A tested folder uploaded as a zip can replace all three.
  • Multiplayer AI means the person who built the folder can drop into a client's chat, add prompts and fix issues alongside them.
  • Per-person access to a folder agent matters: who can see it, whether each user gets a copy, and whether everyone shares one chat.
  • A folder of field reports and requirements plus one example report can become a personal report-generating agent.
  • Ownership drives innovation: when people can earn from a platform, more people use it, which is the moat the big AI companies are missing.
  • Starting from a company's public website is a fast way to demo value, since the client can fill in the rest later.
Takeaway

Folders are the agent, hosting is the gap.

WHAT TO LEARN

Most useful AI setups are already just organized folders of markdown, so the real work left is sharing them safely with the right people.

01Why we're building this
  • Value in an AI platform comes from letting other people earn on it, because ownership and the ability to make money drive adoption and innovation.
  • Controlling IP and capturing all the value discourages the people who would otherwise build on a platform.
03It's all folders
  • Second brains and company operating systems are well-organized folder structures of markdown files, with an entry file that tells the AI where to start.
  • Plugins and GPTs are the same thing, markdown plus connectors, so packaging knowledge in folders is the direction the ecosystem is already moving.
04Example 1: a brand kit
  • A brand deck folder can be generated from a company's website using a folder-structuring skill, then refined later by the client.
  • Structuring folders so AI can read them saves tokens, effort and improves accuracy.
05Upload the zip
  • Zipping the folder and uploading it gives the whole team a chat agent that carries the same files and structure.
  • Multiplayer chats let the builder and the client both edit and generate decks in the same environment.
06Example 2: field reports
  • An engineer's organized specs, requirements and report examples became a report generator once the folder structure was described and one example added.
  • Niche workflows like this are the kind of startup-sized idea a folder agent can serve.
07Who gets what
  • Organizations keep data separate, and invites, templates and workspaces control who can see what.
  • Choosing your own models and keys lets teams control cost and provider.
08A coach's chatbot
  • A coaching process written as a folder can become an interview chatbot with no code.
  • Deployment choices include whether users share one chat or each get their own copy.
Glossary

Terms worth knowing.

ICM (Interpretable Context Methodology)
A method for organizing folders so AI can read them more accurately and with fewer tokens. It ships as a free skill on GitHub.
Agentic folder structure
A folder of markdown instructions, skills and reference files that an AI navigates to do a job, acting as an agent without custom code.
CLAUDE.md / AGENTS.md
Markdown files at the top of a folder that an AI reads first. CLAUDE.md is read by Claude, and AGENTS.md is the same idea for any AI.
Multiplayer AI
A shared AI workspace where several people can join the same chats and edit the same folder-based agent.
Closed alpha
An early test stage limited to invited users, here a couple hundred people, before wider open alpha and a full release.
Resources

Things they pointed at.

The Script

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.

So it's about time I put a video out explaining the software that we've been developing for, well, quite a few years and mostly the last six months we've been putting full -time effort into it. Some of you may have seen some of my other videos kind of mentioning it, different MCP servers we built inside of it. Some of you may actually be alpha testers in it because we have a couple hundred people who are in the closed alpha right now.
Or you recently saw an Instagram or YouTube video from me where I talk about this next trillion dollar software is going to be a company slash software that is able to manage deployment and marketplace better than anyone else has before. Obviously, I'm going to try my best to build that type of software, but I think there's a lot of different opinions on how it should be built.
So let me go in and kind of show what we've built so far. And either you can try to build your own to compete with us, or you can see how you can be a part of what we're doing. Because I think what's going to make this software valuable isn't that we make all the money.
It's the fact that we generate value and money for other people. And I think that's the problem a lot of these big AI companies are missing right now. The ownership, the ability to earn your own value in a market -based economy is what drives innovation in a lot of ways.
And it's actually the controlling of IP, the stealing of value, the control of it stops you from creating more value. It de -incentivizes people from doing it. But if someone knows they can make a buck, they can eat based on using your work, your platform, more people use it, right?
That's Amazon, that's Google, that's any of these things that are out there. YouTube, right? People are incentivized to use it more.
And I think that's something we're missing in the AI age. And it's a moat that I think is really easy to build around because it's not just built on technical solutions, although technical is a big part of it. So let's go ahead and dive into it really quick.
Now, first things first, the UI is going to be completely different over time. We focused mostly on security and background and backend stuff for the last long while, but it's gotten a lot better. Now, for those of you who are not invited to our closed alpha test, you can start an account, but you won't be able to access anything.
You won't see any orgs or anything like that. Some of you may log in and may see our classic version. This is something you've seen in my other videos.
This is going to be slowly pushed to the side, but this is all C Sharp. Both the front end and the back end is all .NET C Sharp with Blazor Wasm. It makes it a lot faster and quicker.
For those of you who aren't familiar with the tech space in that way, don't worry about it. You don't need to know why that's important. Just that's the choice we made and we're sticking to it.
Some of you may hate C Sharp and think it's a trash language with Microsoft. By all means, say what you want in the comments. It's worked pretty well for us for what we need.
Essentially, we wanted it to be very simple. We wanted you to be able to come in, upload a zip file, whatever it is, right? People are working in ICMs, they're building all these skills and folder structures.
We wanted you to be able to deploy it and thousands of people could access it. You can manage it and it basically is a talking environment. Before I show you what that looks like and, you know...
what we've done to solve the problems of folder structure as agentic architecture. If you've seen in my other videos, let me show kind of some value or standpoints where people might be doing this already, right? Your second brains, those areas, uh, and see what happens when you hit certain barriers and why a platform like this or GitHub has been where people have been deploying these zip files and getting a lot of value out.
So generally speaking, a lot of people have been building really cool file structures, right? Very basic, like either brand decks or second brains, right? People have been making second brains of all of their work to be able to talk, right?
Company operating systems. But at the end of the day, all of those things, all of these fancy stuff are just really well organized folder structures, markdown files. um right people might have a cloud .md in this case i have one that just says go read the agents .md um and so that makes sure that claude can or any other ai that's using cloud .md as a base will go and read the instructions you've got a markdown file that describes what's in here and then you have all sorts of skills or other markdown files or concepts or ideas or photos right that you can come in and roll through that the ai can navigate quickly and start building with This has been quickly overriding what people have been using as agents, right?
If you come into chat GPT, GPTs are getting sundown for plugins, and plugins are basically connectors with some files and folders, right? If we come in and, I don't know, pick a... a random skill or plugin or here, right?
It's just markdown files. Maybe it's structured of markdown files. If we come into Claude, it's the same thing.
If we come into these plugins and we click on one of these, maybe one that's a bit more open, we can go and see the contents. It's just files and folders, where and how to use them, some sort of readme, as well as the connectors that are associated with it. so our reasoning is and many other people's reasoning is if i can package agentic knowledge concepts ideas in folder structures uh well why don't we just double down on that why are we trying to build in the traditional software way if it's able to get the outcomes of the software if i build this the the folder structure well and so lots of people have been doing that right they've been building big folder structures maybe shooting them on their their desktop uh you can even do this inside of microsoft copilot right um using either projects or notebooks you reference your onedrive so on and so forth if you want to dive deeper into that i have quite a lot of videos on like what that looks like and how to get the most out of your ai to do really amazing things like building the animations that you're seeing in this video now or doing all sorts of research
work, creating dashboards, etc. But let me give you like two examples that we built locally on a computer and then decided to deploy so that other people could use it or that, you know, we didn't need to have our computer at all. And how our platform kind of is solving a lot of those problems.
So one quick example is, let's say you want to make a brand deck for your company, right? And so every time you talk to the AI, it can build out slide decks or animations or styles all based on your brands. One big thing that I've been doing for people recently is just building it based off their website just to show them really quickly and then they can add to it later.
So I just said, hey, I'm building a brand deck. Make sure you leave room to interview them. Make sure to use my ICM skill.
Um, so I obviously have a skill for building really good folder structures that AI can read. Uh, I always put the link in, in the bio it's free online. It's a GitHub repo.
And I just said, Hey, go rip it from his website and make a brand deck. Um, ICM, a brand deck folder structure, right? And ICM stands for interpretable context methodology, uh, just how to organize folders so that AI can read it better.
It saves you tokens, saves you effort, allows you to have it more accurate, all that good jazz. Again, lots of videos on that. That's not what this video is focusing on.
This is for people who have already dove into it. So it went through and it wandered through their website and they're a surfacing company. Right.
I used to actually work for them when I was younger, when I was 16, 17. I would go and build courts with them and they build really beautiful courts. Right.
If anyone needs good tennis courts, they do all sorts of rubber courts, liner courts, things like that. They do some really cool like art courts as well. Really amazing people and good work.
But there's not too many of them. And a lot of the times they want to organize structures and brands and be able to build stuff out. And so it was ripping through.
It went and pulled. It looked at me. making documents and followed my ICM skill really well until it finally created a pretty decent brand deck, right?
And the brand deck is essentially just a big old folder structure. I mean, if I come right here, it's just a folder. And it has an explanation of how to navigate the folder.
It has some HTML SVG examples, right? So I can come in and it explains all the colors that they use, the styles that they use, different SVG symbols, right? Most of that's just in the HTML file.
But then it has, you know, knowledge, right? Brand is more than just the colors used. It's why you use them, when you're pricing, what your processes are, past projects that you may have had, right?
They've done projects for Nike and Fenway and all sorts. different places um and so being able to access that quickly is quite powerful when you give it to an ai and now from here usually what people would do is they would just keep it locally on their computer and you know talk with the folder structure but what you can actually do is just zip it up real quick and you could deploy this to github So I could upload this to GitHub.
I could put it on Google Drive and try to have other people access it if they have their own Cloud or ChatGPT account. They could use this brand deck almost like a design document that makes the AI create really good stuff. But there's a lot of problems with that.
Not everyone in the company has Cloud. People outside of the company that they might want to use it, they don't want to buy a Cloud subscription for or another AI subscription. And then there's certain parts of it that they may want people to not access.
And so that's where... what we've been building in a simple way, it gets much more complicated here in a bit for bigger deployments, which we'll show, we've wanted to try to solve. So let's go ahead.
And I went ahead and created him an organization in my software, in Adobe Aware. So if I wanted a bunch of people to be using this, or I wanted to share it with my whole team, usually I'd have to go and install it on their computer, send it to this GitHub, all these different things. But in our platform, all he has to do is come in and upload it.
uh just as a zip as is so i come in take that whole folder structure um make sure i hit upload on it and from there it's gonna go ahead and upload it and if it matches the same name of a previous one it'll automatically you know change the name for it but once it's done uploading I can just come in and start chatting with it, just like I would with any other amazing AI.
I can create a new chat. Obviously, we already have ones in there. It's the same exact folder structure.
It still has the HTMLs and everything that are in there. It has all the different structures and markdown files. But the difference is the package itself comes with the AI.
I can just point it at it. And what we were able to do is I can also organize and join his chats, not just mine. right it's multiplayer ai it's it's these processes all right he was coming through and making changes to the small details that he thought were important to the brand i was able to make some changes as well and even build out some decks for him just by asking it to build some so now anyone in his company is able to come through and build out you know, in seconds, beautiful proposals or slide decks or ideas.
And this is great to make generic ones. But imagine being able to bespoke this to every client they talk to every customer, they can just very quickly based on a call, drop a transcript in here, say make a slide deck based on these person's questions and comments and concerns, and then shoot them the email immediately. And it's going to use their brand every time it's going to use the photos that they have, it's going to use their understanding and ideas around it really, really quick.
quickly. Um, and this could work for proposals and all sorts of things. And boom, in the past to deploy a agent, that's a brand agent for your company, it'd be a whole bunch of code.
And then you want to set it up and then you don't have to do the agentic structures. And then it needs the rag database and all this stuff. But in this case, he just, you know, built a little folder package on his computer, made sure it worked and then uploaded it.
And now He can invite people to it by email. He can make it so that only certain people can see this chat.
We can manage who has access to certain workspaces or templates or ideas. We can connect it to other APIs. We can decide what type of AI we actually want to use at our own models and our own keys, all with just the click of a button, right?
And it comes with the security. and the control that you would see in enterprise, right? And instead of having to build all of this out for every single software, every single agentic system you're trying to do, instead, you can just go to the domain expert, the person who says what they need, have them build out or talk about the thing that they need.
And then you upload a folder and now you have thousands of people that can log into that search folder and use it to get the outcomes that a software would have usually required. To go into this a little bit more, let's actually explore another example. We allow the opportunity to separate data by organizations.
They're all completely separate. We're using Azure on the back end. These are secure Azure containers.
And so I can go to a personal organization where me and my buddy were testing out something. He works for a certain set of companies that focus on doing reports around industrial builds and things like that. and so he wanted was telling me like hey jake i've got all of these reports i've got all of these different um requirements that are public record things like that on how things need to go what specs need to be between plumbing and ideas and sorts of concepts there between electrical and mechanical and he has to make these kind of reports that take pictures from when he's on site and say his notes Right.
But in order to do all of this, it's a lot of thinking. It's a lot of concepts that are going on. There's a lot of different areas of where the photo needs to be placed in the report and what he's referring to.
And he was trying to get AI to do it, but he was just struggling with it. But he had already organized all of the folders and files for himself. right in a perfect way right these were for him he was reading these by himself and so all we had to do is pretty much describe the structure of those folders where and what his reports were give it an example upload that zip file and now he has a perfect quote -unquote personal agentic system for generating reports this is an entire startup that i know people would be doing that Now it's just got replaced by a zip upload and he can invite people from his team.
Or in this case, we made it private because this was a test environment. He didn't want anyone else. He is the one that needs to invite people and mold it in.
And he can add his own zip files and other extras and stuff inside of it and actually pull out and edit these and create full reports really quickly. And it describes what's being done. He can query it.
If he wants to open a new chat, it doesn't lose everything. If he wants to invite other people in. They can work on their own report, and now this acts as a second brain of all the reports.
Again, with just a file and folders. And there's a lot of other little features and things that we've been doing, right? The use cases are really dependent on what you're trying to get done and what structure you can create.
I've created animation workflows that work inside of this. We want to make it so that you're able to handle different backends and secrets. So if I wanted to connect this to GitHub repo, if I wanted to connect this to some other software that I already built, then all I have to do is just toss in the API key and they're kept separate.
entirely, completely, and utterly separate. And if you really want to know the best ways to build structures and to solve problems with just folders and files, again, check out one of my most recent YouTube videos. It should be coming up here on the thumbnail where I go into a live call where I was answering people's questions on how we use it ourselves.
But the possibilities are endless. We have a big alpha testers community here where we have people uploading all sorts of stuff, right? They're uploading their own templates.
that I can come and enjoy. So it's also multiplayer AI in a lot of ways. I can come explore and chat with them.
I can drive right into their AI chat, which allows for a lot more talking. I can drop in transcripts into here. I can help them out, right?
So if I've made something work perfectly on my computer and I'm trying to share it with a client or someone else, I can deploy it to them. And if they have any issues, I can come in and just start. typing myself, adding my own prompts, texting them and shooting them an email after, right?
There's a lot of possibilities. And what we're trying to do is kind of manage this, this control aspect. One of the biggest things is, let's say you do build.
some sort of folder structure or some sort of process you invite people in what do they have access to what things are they allowed to do you need to revoke them uh we want to be able to handle the email invites and everything for you uh handle billing for you is one of the main goals too in terms of what credits and capabilities are there what workspaces exists uh being able to separate people so you know if you do have 200 people in a single org not everyone's seeing the same thing being able to manage the front end, our end goal is to actually make it so you don't even need to see a Dubaware.
This is your own front end. You white label it. You change what it can look like, who has access to what, when they have access to it, even designing your own, right?
One thing that we've been looking at doing is, well, if you deploy the folder and if, you know, the folder has an HTML file, why isn't that the thing that they log into? They're logging into this in the back end, but then they're just opening up the dashboard. you created for them but it controls the folders and files just like it would on your computer just like a second brain would on your computer except now you don't have to worry about the deployment and all the other security issues and things like that and that's where it becomes quite fun.
One of my friends, I was working with him on one, and he deployed a coaching journey, right? He's a coach, simple as that. And he put the way in which he handles people and what they are and what the actual mapping can look like.
And I can come in and start an interview process, right? This is a chatbot. He just deployed a chatbot by just uploading a zip file.
Now he can invite as many people as he wants to it. If he does want to upload new types of templates or concepts, even just as basic as skills and blocking people who have access to them or when they can, you can come in and you can decide, hey, does everyone in this org have it? Only people that I want to have it.
Who in the org should have access to it? When someone uses it, are they making a copy of it? Are they all chatting together?
Does each person get their own? copy? Or is there only one single chat that everyone has to sit inside of?
And all the beautiful security and controls that come with that. These are the things we're trying to solve. And the real end goal with this software is to make it so that we have small deployable software that is easy to use and quick.
And in a world where markdown files, where instructions and language can be a software, can get the outcome of a software, all you're missing is something that runs it. So having an AI sitting inside of it, controlling your own AI. One of the biggest things that I'm trying to get set up right now and it is new is why can't I just connect this to the AI on my computer so that even 10 other people that I'm working with when they're using this, they're actually just using one of my local servers or my local desktop.
That allows for huge security while still getting the plus sides of secure deployment. So there's a lot going on and I just kind of wanted to make a quick video describing some of the use cases and reasons why. These are by far some of the more simple ones.
There's way more. complicated ones where we've connected them between GitHub repos and deployment statuses and all sorts of lovely things here and there. It's going to change a lot over time in terms of what you can access, connecting MCPs, connecting other sites of it.
And again, it's not just files and folders that are here. It's running Python. It can actually do indexing.
Again, if it works on your computer and it works with your AI, you shouldn't have to figure out how to deploy it. You should just be able to upload a zip file, And it works.
So it is in closed alpha right now. We are only allowing a certain amount of people in. If you want to join our community, it is free.
You'll be updating a lot more there on what we're doing and how we're doing it. We're also for people who do pay for premium. We're prioritizing them to get alpha access.
We're not expecting to have full release until January 1st. But by around December 10th, we're expecting opening it up to open alpha. And right now we are covering a lot of the costs.
We are allowing people, you know, a lot of their own token costs. We're covering some of that and all of those things. And I will be adding videos more and more around then.
Because then the UI, the structures, the features will be exactly what we want. We have quite a lot of PRs coming down the push right now in terms of what we're trying to do and how we're trying to work with it. But it is a good place to start.
And if you do want to learn more, of course, about how people are building with these folder structures and replacing software. with just good old files and folders uh this is a great community to be in we have competitions we have people asking questions i mean even right now we have over 400 people online texting and it's midday it's one o 'clock in the middle of the week um so it is an amazing place to be able to come in and dive and explore so go check that out and uh yeah if there's any questions about adubaware drop them in the comments um i'm also gonna drop the discord Uh, we do have a discord for a do beware, um, where you can also go and try to request access.
So I'll go ahead and, uh, I'll probably, uh, toss an invite for that for everyone. here shortly so that you can at least join in and you can see some of the change logs and what we're doing. I'm also about to drop a roadmap of all of these features and concepts into it.
And you can kind of check out different things as well as my co -founder, David, what he's been doing and building around for it as well. So, yep, that's really it for today. That's obviously I didn't touch on everything.
It might still be a bit confusing for some people. So just expect more detailed videos and better demos in the future. future as we start releasing and getting more people on this.
But at the end of the day, if it works on your computer and it's a big set of files and folders, and you've been wanting to just deploy it to a bunch of people and be able to manage who has access to what, this is a great place to do it. So this is what we're here for. We're trying to make it easier because we've seen people who are struggling with GitHub or managing users or multiple templates and copies.
And we're trying to create all the solutions that allow you to simply upload a folder, And it works. So keep in tune, keep in touch.
And yeah, this kind of breaks the silence on everything that we've been building and working with over a bit. So until next time, happy learning.
The Hook

The bait, then the rug-pull.

A creator who has spent months teaching people to organize folders for AI finally shows the software built to deploy them. The pitch: zip the folder, upload it, invite your people.

CTA Breakdown

How they asked for the click.

VERBAL ASK
19:59link
“It is in closed alpha right now... if you want to join our community, it is free.”

Soft invitation tied to alpha access, with premium community members prioritized and a Discord invite.

MENTIONED ON CAMERA
Storyboard

Visual structure at a glance.

open
hookopen00:00
it's all folders
promiseit's all folders03:36
upload the zip
valueupload the zip11:07
how to get in
ctahow to get in19:59
Frame Gallery

Visual moments.

One-click upgrade to your Google

Get more breakdowns in your search results

Add Modern Creator as a preferred source and Google shows you more of our breakdowns in Search, Top Stories, and AI Overviews. It only changes what you see, and you can undo it in your Google settings anytime.

Add to Preferred SourcesOpens your Google source preferences with us pre-loaded. Tick the box and you're done.
Watch next

More from this channel + related breakdowns.