Why IdeaSpaces?

The model brings the reasoning. The context is yours.

I think we are using AI the wrong way.

AI is here to stay, and the world is flooded with chat windows. We type a question, we get an answer, we close the tab. That works for general knowledge, a quick draft, a puzzle. I do not think it is the right way to use this technology for the work that matters to us.

The world needs depth

We live in a remarkably complex world. Technology, markets, climate, ideas: more moves in a year than one person can follow. Much of what we learned in school is out of date by the time we need it. The questions worth working on are not answered in one exchange. They need weeks of thinking, decisions that get tested, and a way to return to what we understood before.

A model like Claude or ChatGPT is humanity's knowledge, compressed. It knows something about everything and nothing in particular about you. It will always give you an answer, and it cannot tell whether that answer is any good. That is not a fault. It is what the tool is. The output is only as particular as what you put in.

A chat window ignores this. It starts every conversation from nothing. You explain your situation again, it forgets again, and whatever you concluded together goes somewhere nobody will read.

Your context matters more than the model

Everyone is watching the models. Which one is smartest this month, which company is ahead. I think that is the wrong thing to watch.

Models come and go. Providers come and go. What stays is your context: your customers, your decisions, the questions you have not answered yet, your sense of what good looks like. No model has it and no company can sell it to you. It is also the only part you control.

So the question is not "which AI?" It is "where does what I know live, and how does the AI reach it?"

That context needs a place, and it needs tools. Working locally puts the agent where your files already are. An ideaspace gives the context a home that both of you can return to.

Knowledge is a process, not a possession

We talk about knowledge as something we have. Learn it, write it down, keep it.

Real understanding does not sit still. We use it to decide. The decision meets the world. Something works or fails. We come back with better questions and update what we knew. Knowledge keeps moving; the pile of documents was never the point.

With an agent, that loop turns faster. Understanding gets captured while it forms and comes back when it is needed. The next conversation starts where the last one left off. Conversations end. What you understood stays, and grows.

Depth needs company

Nobody explores anything important alone for long. Depth needs other people: a partner who sees it differently, a reader who asks what you meant, a team that has to act on the same understanding.

For that you need a place where you can share what you understand and let others explore it, with their own agents. Not so they agree with you, but so you can agree on what the story means. A Note someone can read, take a copy of, and question is a small piece of common ground. How changes travel describes the mechanics.

That is the invitation behind this manual. Explore with me.

What we built, and what it costs

The tools are free. The shape of an ideaspace is an open standard, published under the MIT licence. The command line and the plugins for Claude, Codex, and Pi are open source too. Everything you do on your own computer needs no account: read, write, capture, keep this manual updated. Your files stay yours, in a form any text editor can open.

Reaching other people needs an account. Putting a Repo of your own online, sharing it with someone, and asking a person a question about their pages run through ideaspaces.xyz. That is the platform, and it is the part we run. Underneath, it is the same folders with a shared copy.

If you want to see how this works in practice, start at the beginning.