Note-Taking to Knowledge Infrastructure - Progress Update

A couple weeks ago I wrote about my effort to turn note taking into knowledge infrastructure. That last word is a fancy way of saying I want my Obsidian vault to be more of a wiki. I’ve made a little progress over these two weeks and it’s going pretty well so I’ll share an update here.

What’s going well

Looking at the 6 parts of this, shown in the image above, the Ingest, Capture, and Process phases have (mostly) been the easy, set it and forget it part. The Compile and Output phases have been the standouts - meeting and exceeding my expectations. Here’s one example note that shows how the Claude wiki-query skill has done a great job with all 5 of the tasks under Compile:

That’s not the whole note, but it shows the wiki-query skill has:

• Formatted the note’s frontmatter in red, so it’s easy to spot when I’m triaging notes

• Carved out a section from a source note that’s useful enough to be an additional note on this topic (agentic ai)

• Added appropriate tags and made good choices on the related notes it links to

Here’s one more example note. I’m highlighting this one because it points out that it’s AI generated, via a chat I had with Gemini and “requires primary source validation”.

The generated articles are looking great. There are a lot of them, but so far a small daily time block is enough to review and either organize or remove them.

What still needs work

Landing spots for the skill outputs from the Ingest and Capture phases. I didn’t spend enough time on this, and it’s a little messy right now. Those outputs are landing in three different places within the Obsidian vault right now and that just makes it harder to triage them, organize them, and work with them. I’m thinking that I will probably use the Capture folder in Obsidian as the single landing place. That’s what that folder is for and then it’s simple for me to go to one place, review those notes, decide whether they need to live in the wiki. If they do, I’ll move them to Wiki/raw, where they get processed as shown in the images above.

I also want to make two small tweaks, telling Claude that:

• It’s a little tag-happy at the moment on the wiki-generated notes. It’s never a good idea to end up with a thousand tags, and I don’t want to have that as an additional thing to edit when I’m working with the wiki-generated notes.

• The naming syntax for daily morning briefings is not great, too long. It should just be morning brief and the date, right now it looks like this:

briefing_20260515_070146_morning_brief_2026-05-15

That’s the update. I like how this is going, and I’m hopeful that soon it’s going to a smooth running collaboration with Claude.