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

A practical note-taking architecture in Tana can be built from scratch by treating “sources” as the root unit of knowledge, then layering highlights, topics, concepts, and fleeting notes on top. The system’s payoff is that every quote or idea extracted from a book, video, or podcast stays traceable back to its origin, while also becoming searchable through topic and concept relationships.

The build starts with a blank workspace and creates a daily dashboard using a calendar node and a day super tag. From there, the core schema is assembled using three super tags: **Source** (the book/video/audio item being consumed), **Highlight** (the extracted text from that source), and **Fleeting Note** (the user’s immediate thoughts about a highlight). Converting normal nodes into super tags is done via a command workflow, and the resulting super tags are placed into the schema so they can share a consistent structure.

The **Source** super tag is configured with fields that make downstream retrieval easier: a **Source status** field (implemented as options like read/watch/listen) and a **topic(s)** field. To keep status visually scannable, the system uses emoji icons tied to each status option, then feeds those icons into a “build title from fields” rule so nodes display the right icon automatically. A key design choice is naming clarity: instead of a generic “status,” the field is explicitly called **Source status** to avoid confusion when other statuses (like project status) appear later.

Next comes **Topics** and **Concepts**, which turn simple tagging into a connected knowledge graph. Topics are set up as an instance field on highlights (so each highlight can be tagged with one or more topics). A live search then powers a “material” view that pulls in all sources whose topic matches a given topic node. Concepts extend topics: a **Concept** super tag adds a recursive “part of” relationship so the system can answer structural questions like “what is part of consequentialism?” using a live search that supports semantic function queries across multiple levels.

**Highlights** are configured to belong to exactly one source and to reuse the topic field. The build also adds practical metadata such as a **location** field (e.g., page numbers) and demonstrates how missing default values (like an unset Source status) can break icon rendering—solved by setting a default status (e.g., default to “to read”).

With the minimum viable system in place, the workflow expands into dashboards. A **Bookshelf** node is created with live searches for “currently reading,” “to read,” “to watch,” “to listen,” and “done,” using Source status options across multiple source types. Another daily view shows recently created fleeting notes and highlights (e.g., last 3 days), making it easy to revisit work.

Finally, the system adds AI-assisted topic extraction. An AI-enhanced field on the **topics** field reads highlight text and returns 1–3 candidate topics formatted as double-square-bracket links inside a markdown code block. Tana then links those to existing topic nodes or creates new ones, enabling one-click enrichment of the graph. The session ends by packaging the build into a shareable template, with guidance on where super tags and live searches must be nested to publish safely without exposing the primary workspace.