Method notes, workflow notes, and research sketches on AI and personal knowledge will be listed here. This version is a scaffold, not a published archive.
Topics
These anchors are entry points only. They do not hide other notes.
Discover
Entries marked UNREVIEWED are drafts awaiting editorial sign-off. They are not published writing.
Obsidian vaults are directories. Coding agents like Codex already work on directories. Pair them with an AGENTS.md schema—deny rules, Git review, and human gates—and you get repo-as-notes without a proprietary plugin. Here is the pattern, when it helps, and what to keep off the agent’s write path.
Andrej Karpathy’s LLM Wiki pattern treats your notes like a codebase—raw sources stay immutable, the model maintains a compounding Markdown wiki, and you ask better questions. Here is when that beats classic RAG, how it sits next to Progressive Summarization, and what to watch for.
One-click AI summaries are cheap. Discoverable notes for Future You are not. Tiago Forte’s Progressive Summarization still matters—here is the layer model, how it pairs with PARA and CODE, and where AI should (and should not) help.
Keyword search needs the words you remember. Backlinks need links you already made. Smart Connections starts from the note in front of you and surfaces meaning-neighbors—usually with local embeddings and no API key. Here is when that discovery layer fits an AI second brain.
Blank-page writing is expensive. Andy Matuschak’s executable strategy turns a network of notes into outline → concatenate → rewrite. AI can capture and critique—but promoting evergreen claims and owning the final voice should stay human. Here is the pipeline and where the model should get off.