Manifesto

The software I use should be personal, private, and sovereign. The way I use it should make me more independent, purposeful, and thoughtful.

This is a working theory I'm testing with two projects.

Where I'm writing from

A list of privileges sits underneath this ability. I have the time and space to study software and AI as if they were worth studying. I live near the center of the bubble that produces these tools, which is how I see their defaults up close. I can afford the hardware that runs local models. I'm writing this with help from one of the most expensive frontier models on the market. The path into the pillars below is much easier from where I sit than from most places.

Why now

LLMs have changed what one person with a need and time can build. Most of the capital flowing into AI is going into agents that reason and write code. Trained engineers are less and less the bottleneck on writing software, and more and more the bottleneck on knowing what software is possible and worth writing.

Additionally, there's a deep corpus of free and open source software that already solves most of the hard parts. Between the two, the cost of producing a tool that fits one person well is collapsing.

I expect that to pull software development toward the edges. Smaller groups and individuals will take on niches no roadmap was ever going to fund. The software they produce will be more malleable, because the niches reward flexibility over polish. These are the pillars the rest of this document stands on.

Personal

The software is built for me. Not for a market segment, not for retention metrics, not for a roadmap that has to make sense to someone who doesn't sit where I sit. When a feature I need isn't there, I add it. When a feature nobody asked for is in the way, I remove it.

A piece of software being personal doesn't mean nobody else can use it. It means I'm the one it has to satisfy.

Private

My data, my prompts, my queries, my drafts. They stay on hardware I control. If a third party gets a copy, it's because I sent it, not because the tool phoned home on my behalf.

This includes the AI. A model running on a server I own is different from a model behind an API that's logging every token to a vendor's training set. Both have a place. But the one I use should be just mine.

Sovereign

I can read and change the code. I can freeze a version that works and refuse to upgrade.

When an AI agent helps me change my system, the change is reviewable, revertable, and mine. The agent works for me.

Independent, purposeful, thoughtful

Independent: my workflow keeps working when a SaaS goes down or a startup pivots. Purposeful: I don't drift into using a tool because it nags me; I use it because I chose to. Thoughtful: I notice what software is doing to my attention, my data, and my agency, and I can do something about it.

Malleable by default

If decentralization is going to land, it has to land through software other people can read, change, and adapt to their own situation. So the software I write should be the same shape as the software I want to use: inspectable, modifiable, forkable. A closed binary with my name on it would undercut everything above.

This is the idea that allows my efforts to scale past one person.

Two experiments

assist is the local agent stack. emacsos is the phone. Both are deliberate attempts to live inside the constraints above and see what falls out.

What this builds on

I'm not the first to want this. The manifesto stands on a growing body of work, and these are the pieces that have inspired it.

The phone came from one project in particular: the OURS Project, a DIY, open-source, repairable smartphone built on a Raspberry Pi and free of Big Tech. It's the hardware expression of what emacsos chases in software.