Learning and doing with a second brain.

Everything I read, build, or figure out lands in a single Obsidian vault, now a couple thousand interlinked notes. It is a personal wiki first: small atomic notes, dense linking between them, and a graph that makes the connections between ideas visible. The point isn't to hoard information; it's to make past thinking retrievable the moment it becomes useful again.

The same vault doubles as memory for AI agents. The notes that help me think are also a grounded knowledge base the agents read from, and increasingly write back to: capturing decisions, drafting summaries, and linking new notes into the existing graph. Retrieval over the vault, combining vector search with the link structure, lets an agent answer with my own context instead of a generic guess.

It's a tight loop of learning and doing: research a topic, build something real with it, write up what actually worked, then link it back. Over time the vault becomes both the record of what I've learned and the substrate the agents operate on, so each new project starts further ahead than the last.