selfware: the newest, most fun thing in tech
I am writing this post from my local-brain-ui: a piece of selfware, or, if you prefer, personal software built through vibe coding, inspired by ideas like LLM-powered wikis and second-brain systems.
The central idea is simple: a second brain does not have to be just a pile of notes abandoned in folders. It can become something more active, almost like an operating system connected to the way your own mind works.
In my setup, the system follows a small internal contract. The raw/ folder stores raw material, captures, and private operational inputs. brain/ stores processed knowledge in Markdown.
synthetics/ is where automations, generated artifacts, and AI-assisted workflows live. This structure is documented in the README.md, but honestly, the folder names are the least important part.
What matters is the direction: knowledge is not only stored. It becomes something that can actually be used.
The interface is part of that philosophy. My Brain Input UI runs in the browser and acts as an entry point for raw captures and operations inside the knowledge vault. From there, I can write journal entries, log habits, read local files, trigger internal skills from a controlled list, edit text files, upload materials, and connect other tools, such as writing environments, media viewers, or local model routes.
In other words, it is not just another notes app. It is a personal workspace designed around how I actually think and work.
This is what I mean by selfware.
Selfware is software shaped around a real person. Not in the shallow sense of changing the theme color or putting your name in the header, but in a deeper sense: software that respects your actual loops. How you write. How you remember. How you revisit unfinished ideas. How you get distracted, and how you find your way back. How a journal entry connects to a project, a book, a video, or a half- finished draft.
Most second-brain systems promise memory. But memory, by itself, is passive.
What I am looking for is closer to an operating system for thought: a place where notes, journals, media, writing tools, research, and LLM assistance do not exist as isolated islands. They can interact through local files, explicit rules, and workflows I can inspect and control.
One example: imagine building your own version of Netflix, but with your own rules.
Not a recommendation engine designed to maximize screen time, but a personal media layer built around your reviews, your reading notes, your journal, your projects, or the books you were studying last week.
The goal is not only “recommend me something.”
The goal is: recommend me something based on my real context, my past attention, and the direction I am actually moving toward.
The same logic applies to writing.
A blog post does not have to start from a blank page, or from a model pretending to know who you are. It can start from a local and controlled context: previous notes, project pages, relevant journal fragments, and the exact environment in which the writing is happening.
In this model, AI does not replace the writer. It helps the writing environment become aware of the system around it.
That changes the experience of using AI.
Instead of opening a chatbot and explaining your life from scratch every time, your local system already knows where to look, within the limits you defined. It can search your vault, read your files, cite local paths, and remain local-first. It does not need to pretend every answer emerged from nowhere.
There is also a practical reason to build this way.
Personal knowledge systems become fragile when everything depends on memory, discipline, or a single cloud platform.
Local files are boring. Markdown is boring. Folders are boring.
But boring foundations make it possible to build strange, deep, and radically personal systems on top of them.
So local-brain-ui is not just an app.
It is an experiment in making a second brain operational.
Journals, videos, writing, raw notes, processed knowledge, automation, and LLMs become parts of the same personal machine.
The possibilities are large, but that is not the point.
The point is simpler:
To make the computer feel a little closer to the way the mind already works.