A Useful Knowledge Garden Is More Than a Chat Archive

“Do you post inside of them or what?”

That was one of my questions after watching a video about digital gardening. The idea had caught my attention, but I was still trying to understand what people actually put in a garden—and whether the things I worked through in my conversations with AI could belong there.

Those conversations could cover a business possibility, a drawing problem, or a decision I needed to make. We would explore an idea, question an assumption, consider another approach, and eventually arrive somewhere useful.

I wanted to keep more of that than the answer at the bottom of the thread. The discussion had helped shape the answer. Could I turn that thinking into something I would return to?

I asked for research and a way to try it.

Something I could open

The early choices were practical. I wanted the garden to stay private, with the option of turning selected material into public articles later. I also wanted to keep control of the underlying files.

We considered keeping the whole garden in a hosted service. We settled on a file-based foundation instead. Obsidian would give me a place to read and connect notes; GitHub would keep a history of their changes.

On October 22, we tried a small test. A note was created through ChatGPT's GitHub connection. I fetched and pulled the change with GitHub Desktop, then opened Obsidian.

The note was there, in the expected folder. I could see its content, properties, and links. A later test added a diagram, which appeared inside a note too.

There was now something to open outside the conversation. The test gave the idea a working foundation, though it said little about how useful a larger garden would become.

A file wasn't the whole experience

An early permanent record described the architecture efficiently. It captured the platforms, constraints, decisions, and tests. There was good reason to keep that information.

But my expectations for the garden went further. I wanted to open an article and follow the thinking. I wanted enough background to understand the question, enough of the exploration to see why the answer made sense, and writing I would enjoy reading.

That became something I had to make explicit as we developed the material. A technically accurate account could still leave me wanting a better article.

The distinction changed what I asked the writing to preserve. If a discussion had moved from an attractive possibility to a more qualified conclusion, I wanted to understand that movement. What had made the possibility appealing? Which question complicated it? What did we find that affected the choice?

Listing the result would skip the part that made it useful to me.

Even the garden's own tool choices illustrated this. A short note could preserve the two names we selected. A fuller account could explain why I wanted portable files, what the change history offered, and why the initial manual synchronization step was an inconvenience I accepted. Later, I could reconsider that inconvenience without having to reconstruct the entire decision.

The software helped me keep the material. I still had to decide how to turn it into a worthwhile reading experience.

A place to return to

As the garden grew, I also changed how I wanted to find things again. Early organization separated material into broad kinds of documents: research, knowledge, and decisions. I asked for a structure built around interests instead, with maps that brought related articles together.

A subject might begin with research and end in a decision. I wanted to find it by what it was about, without having to remember which stage of the discussion had given it a home.

I also wanted later conversations to improve the article already there. Otherwise, each new discussion could leave another version beside the last one, and I would be sorting through files as well as chat threads. Change history offered a way to retain earlier writing while keeping one current account to read.

Some of the work remains unfinished. Older references and attachments have not all been recovered, and reliable unattended maintenance still needs to be demonstrated. Those limits belong alongside the progress. A readable account should make them easier to understand.

When selected lessons become public articles, there is another person to consider: someone who was never part of the original discussion. I want that reader to be able to follow the exploration too, with enough context to understand the choices and room to reach a judgment of their own.

I began by asking what went into a digital garden. Building one has made me more particular about what I want to find when I return: an idea I can follow, a decision I can examine, or a question I am ready to take further.

That is what makes this worth continuing. The next useful conversation can become part of something I can keep thinking with.

When you return to an old note or conversation, what do you most need it to help you understand? I'd like to hear your thoughts in the comments.