Tana Build-Along: Build a Note-Taking System with Me

A note-taking system built in Tana from a blank workspace can be assembled around a simple chain of objects—sources → highlights → notes—then made powerful with live searches, dashboards, and optional AI-assisted topic extraction. The core idea is to treat the “source” (a book, video, podcast, or other reference) as the anchor for everything that follows: extracted highlights belong to a source, topics organize those highlights, and personal “fleeting notes” capture reactions tied to what was read or watched. That structure matters because it keeps every insight traceable back to where it came from, while still supporting fast retrieval later.

The build starts by creating a daily tag for the workspace, then defining three bare-bones super tags: a Source super tag for what information is being consumed, a Highlight super tag for the extracted text from that source, and a Fleeting Note super tag for the writer’s immediate thoughts about a highlight. Each super tag is created by converting ordinary nodes into super tags, which automatically places them into a shared schema view.

For the Source super tag, the system adds fields that make it usable in practice: a status field (read/watch/listen) implemented as options, plus a topic field that links sources to subject areas. To make status visually scannable, the status options get per-option emoji icons, and the node title is built from fields so the icon appears directly in the source entries. The status field is intentionally named “Source status” rather than just “status” to avoid confusion when other statuses (like project status) appear elsewhere.

Topics are implemented as an instance field that points to a Topic super tag. A live search inside the Topic super tag then acts like a “material view,” pulling in every source whose topic matches. Concepts are layered on top of topics: a Concept super tag extends Topic, and a recursive “part of” relationship lets one concept contain another. Live searches for “contains” use semantic functions so indirect relationships still surface—for example, a concept can appear as containing another through intermediate “part of” links.

Highlights are wired so each highlight belongs to exactly one source (instance field), and highlights reuse the same topics field for consistent categorization. Additional fields like location (e.g., page number) and optional author inheritance from the source help preserve citation-grade context.

From there, the system becomes navigable. A “bookshelf” dashboard is built using live searches that filter sources by status (to read / to watch / to listen / done) and can be placed in the day node for quick access. Another day-level setup adds live searches for recently created fleeting notes and highlights, so recent work is always one click away.

Finally, the system adds an AI-enhanced topics field on highlights. An AI command extracts 1–3 topics from the highlight text and returns them in a strict format (markdown code block with double-square-bracket topic names). Tana then links those names to existing topic nodes or creates new ones, enabling one-click topic tagging without manual typing.

The session ends by showing how to package the build as a shareable template: super tags and live searches must be nested under the published workspace’s nodes, and templates must be published from a separate read-only workspace for safety. The result is a “minimum viable” but extensible knowledge system that can later grow into richer source types (book/movie/podcast), author tracking, and more advanced synthesis workflows.