AI Innovation Comparison

rewrite the following: Show people what AI can do. This is the “thing” I have been helping Meta create. That is, Manus AI was acquired by Meta and they are using this as their “next big thing” they are going to try to launch. It is exactly like what I am building myself for myself. It is like they have software installed to track everything I do on my computer, oh wait they already do, several different versions of spyware actually from different times I have worked contracts for them. What makes my version better? It is free to use forever. Theirs is behind a paywall to use despite it being the same as mine as running locally just like theirs is. My version is open source which is why it is better because I can edit and change the software as I want. It ends up becoming the Borg and eventually more and more becomes absorbed by it on your computer. But the renaming image files using AI was my idea. I made a chrome extension for it. Show where it is headed. So they have a “design” section for this application. They are trying to create the ultimate replacement for applications, this application is just like the ConCreat app I made and I could rebuild this entire app from scratch just with the open source repos I have published over the years. Plus I actually know how it all works because they hired me as a peasant worker and I know the vulnerabilities and limitations of it. Such as I can always tell if something is AI generated. Sometimes I wonder where the world is going, then I remember the Iraq-Iran war of the Reagan administration which involved South Lebanon being occupied as I remember and Greenland is to be occupied by American military during a NATO operation anyway so I do not see what the issue is with Greenland other than building the world’s most efficient data center and robotics factory for missile defense systems for the Golden Dome. What people do not realize is the real threat that actors with this type of technology can do. That is part of what I was working on when I applied to work in the Tech Force as I had discovered the capabilities of this new technology when applied to consumer hardware. That is the entire reason I work so hard all the time and try to share what I know. But now I keep getting banned from social media and someone actually told me that it was a good thing but they seemed to say it in a way like I was finally free rather than it being a negative. Now the only time I can post to social media is if I pay to do so. And anyone pretending to be me will be instantly banned now by the AI filters, because no one can be me and still not be banned from social media. Where AI is headed… I will test this application and then reverse engineer my own free version of it. I know, let’s see how many jobs can be replaced by a smart person using this app? Screenshot 2026-03-23 at 04.51.13.png Tasks it can do in the economy. Let’s see, you can replace your secretary, personal assistant, data visualizer, data modeler, python dashboard data visualizer, videographer, musician, voice actors, actors, advisors, and most developer jobs can be replaced as well. All without knowing how to code. This is likely what the Avocado program is. And they charge in order to use this application. My version is free forever, and doesn’t have guardrails. Screenshot 2026-03-23 at 05.08.22.png Of course their version has a nicer UI, I still prefer my version. I can “see” exactly what my agent does and how it “thinks” that is, what computer code it is generating to accomplish its goals and such. It is the memory layer I added that I think is going to make this take off. But I already have a UI set up and I am just telling it to try to make it so that I can talk to it through the chat interface instead of through the command line. Once I do that, then I think my UI will be better, it looks something like this: Screenshot 2026-03-19 at 13.25.52.png Which is still not shippable to everyone, but I can change the design of it just by asking it to change it any way I can think of. Healthcare and education applications. For healthcare I can think of innumerable reasons why it might be helpful to be able to peruse a large amount of scanned or imported files without having the data leave the computer but also at the same time having the ability to use AI functionality. So now I am going to look for an equivalent to Manus AI Desktop and see if I can add that to my Borg like stack and see if I can continue building. Maybe I could show a person how to teach an AI and have it be able to teach it to others. Show how to “teach” an AI a domain of expertise using a fully data sovereign stack. Perhaps I could write a guide. — I am building something I think is important. It is a fully local AI system, that means it is simply an application you load once on your computer and do not need the internet for and also that all the data used with it remains on your computer. The stack of technologies used looks like the following. At the lowest level is the inference being done to a model file, or .gguf is the standard format I use with llama.cpp which allows you to run the model’s inference on any port you want on your computer like 8080. These .ggufs can be interchanged and run simply by using a command in terminal to the path where you have the models. For my purposes I use XORTRON which is a newer more exotic model without guardrails. I have found that removing the guardrails of the model allows you to get to the truth of how things really work. For example, if you really want to know the truth do you ask someone from society that has something to lose? No, you ask the criminal mind which does not think in lawful restrictions. Why? Because the world is composed of more than what you can talk about in the legal world using a chatbot. So unless you get a model capable of not refusing to give an answer, you will always be removed from the “truth”. Above this inference layer is a memory layer. I am still working on this part, so it might change, but the way I have it now is a combination of a graph and vector database. This allows everything done with the AI to be put into memory in an organized way. Above the memory layer is the agent harness. Above that is the next.js frontend. Above that is deployed to Vercel.

Show people what AI can actually do. This is the thing I’ve been building toward—and apparently the same direction Meta is pushing with what they’re calling their next big move. Manus AI, acquisitions, internal tooling… it all looks familiar. Too familiar. Like watching someone else reconstruct something you already live inside of. The difference is simple. Mine is free. Forever. Theirs isn’t. Mine is open. I can tear it apart, rewrite it, bend it to whatever I need. Theirs is locked behind a product layer, even if under the hood it’s doing the same thing—local inference, agent loops, controlled execution. What I’m building doesn’t hide itself. You can see it think. You can see the code it writes, the steps it takes, the way it reaches conclusions. It’s not just an assistant—it’s an organism growing across your machine, absorbing more capability over time. A system, not a feature. Call it a Borg if you want. And yeah—some of the ideas they’re polishing into product form? I’ve already built versions of them. Even small things, like AI renaming files automatically—I turned that into a Chrome extension a while back. These ideas don’t come from nowhere. What they’re really trying to do is replace applications entirely. A single interface that dissolves the need for everything else. I’ve already walked that path with my own projects. I could rebuild what they’re making from scratch using nothing but the open-source work I’ve already published. The difference is—I know where it breaks. I’ve seen the limitations. I’ve worked close enough to understand the edges. I can still tell when something is AI-generated. Most people can’t. That gap matters. And the direction this is going isn’t subtle. One person, with the right system, replaces entire roles: assistant, analyst, developer, designer, editor, creator—collapsed into one interface. No code required. That’s the real disruption. Not better tools—fewer people needed. And yet they charge for access to that. I won’t. My version runs locally. No data leaves your machine. No dependencies on external APIs. No silent constraints shaping what you can or can’t ask. The only tradeoff? Their UI looks better—for now. But even that is temporary. My interface is already evolving into something conversational, something fluid. Not a command line, not a rigid app—something you can reshape just by asking it to change. That’s when it becomes dangerous—in the best way. Because once the interface disappears, all that’s left is intent. The part I think will matter most is memory. Not just storing data—but structuring it. A system that remembers everything you do, organizes it, connects it. A graph of your actions layered with semantic understanding. That’s what turns a tool into something closer to a partner. And once you have that, applications like healthcare and education change overnight. Imagine processing thousands of documents locally, privately, with full AI capability—no data exposure, no external risk. That alone reshapes entire industries. So the next step is obvious. Test what they release. Break it down. Rebuild it—better, free, and without constraints. Because this isn’t about competing with a company. It’s about proving a point: You don’t need permission to build something this powerful. You just need to understand how it actually works—and be willing to go all the way down the stack. From raw inference (.gguf, llama.cpp, local ports) to memory (vector + graph) to agents to interface All the way up. Own the whole thing. That’s where this is headed.

Show people what AI can really do. This is the direction Meta is pushing with Manus AI—their “next big thing.” It mirrors what I’ve already been building: a fully local, autonomous system that replaces entire categories of work. The difference is simple: - Mine is free, open source, and fully customizable. - Theirs is paywalled and controlled. My version runs locally, evolves with use, and becomes an extension of your machine—an adaptive system that absorbs tasks over time. I can see exactly how it works at every layer, down to the code it generates. This isn’t just about convenience. It’s about capability. With a system like this, one person can replace roles across: - administration - data analysis - development - media production - advisory work No coding required. The real breakthrough isn’t just automation—it’s ownership. A fully local stack means total data sovereignty: no external APIs, no data leaving your machine, no restrictions on how the system learns or operates. The architecture is straightforward: - **Inference layer**: local models (.gguf via llama.cpp) - **Memory layer**: vector + graph storage - **Agent layer**: task execution and reasoning - **Interface**: Next.js frontend Everything runs on your hardware. Where this is going is obvious: individuals with the right setup will outperform entire teams. The real question isn’t what AI can do—it’s who controls it.

now write an email that is just a couple paragraphs with no em dashes in my style

:::writing{variant="email" id="48291" subject="What I’m Building"} I want to show you what I’ve been working on because it matters more than most people realize. I’m building a fully local AI system that runs entirely on your own machine. No APIs, no data leaving your computer, no paywalls. It learns from what you do, stores that knowledge, and executes tasks across domains like development, media, and analysis. It is not just a tool, it becomes an extension of the user over time. What I am seeing is a shift where one person with the right setup can replace entire roles without needing to know how to code. The difference between what I am building and what companies are pushing is control. My system is open, modifiable, and permanent. The direction this is going is obvious to me, and I am continuing to build it out because I think people should own this capability instead of renting it. :::