All builds

Joy Build · demonstrated

Notes that find their way back

A personal Zettelkasten system for bringing earlier thoughts into a new project, with a simpler version released as a skill.

The evidence

The Homebrew Club episode shows Symon's personal notes system retrieving earlier ideas. A public phase-one skill is available on GitHub; it is simpler than the system shown.

The problem was retrieval

I had plenty of ways to capture notes. I didn’t have a reliable way to bring the right thought back when I needed it.

My working question became: when I start a project, can I first ask what I’ve already thought about the subject? An old note should be able to help with a project that didn’t exist when I wrote it.

Two versions worth distinguishing

My personal system in the episode The public phase-one skill
Storage Supabase-backed notes Local Markdown files
Context Connections to my own projects Tags and a project keyword lexicon you maintain
Interface Claude Code in the terminal Claude Code skill or Python command-line helper
Availability Demonstrated, not offered as a hosted app Source and instructions on GitHub

The released skill generalizes the idea; it isn’t an export of my private notes or all my integrations. Its files need no database account. If you use it through an AI service, that service’s access and usage terms still apply.

What the released version does

The public README describes capture, tagging, retrieval, links, review, and validation. Python 3 runs the command-line helper. The notes remain readable files, and you can choose how to back them up.

The organizing idea is that notes aren’t trapped inside one project. Tags can connect an earlier thought to a later project. The skill also puts unfamiliar tags into a candidate list rather than letting the vocabulary grow unchecked with every capture.

It is released under the repository’s MIT license. Follow the README for installation; this site doesn’t install it or connect to your notes.

What I demonstrated

In the conversation with Jim, I retrieve thoughts about projects for my kids and broader ideas that might apply to them. Some were ideas I’d forgotten. The useful result was a better starting point for the next conversation.

I also describe a detour: I’d built a Telegram capture route, then returned to the terminal because it was enough. A larger interface hadn’t earned its place.

A small test for your own box

Try five to ten non-sensitive notes. Ask one question tied to a real project. Can you trace the answer to the original notes? Did it return anything you would actually use?

The public instructions have been reviewed for this write-up; the skill hasn’t been installed or tested here. Retrieval quality, backup choices, and how well the tagging fits your thinking still need your own trial. Start with the three-stage worksheet if you want to test the idea before installing the skill.

Built with: Claude Code, Markdown, Python 3.

Borrow the pattern

Symon He · Newsletter

Practical experiments for your next chapter.

I test ideas with AI and share what works, what it costs, and what you can borrow, so you can choose what deserves your time.