Notes & knowledgeCAN I VIBECODE
Tana?
BUILD FIRST
- One specific customer type and one primary use case
- A vertical slice that completes the main job without manual workarounds
- Visible limits, review steps, and a fallback when an external dependency fails
LEAVE FOR LATER
- Broad third-party integrations
- High-scale collaboration, syncing, and permissions
- Automation that can cause harm without human review
RESEARCH NOTES
What this tool actually does.
Tana (Tana Outliner) is a knowledge-management tool built on an infinite outliner that doubles as a knowledge graph. Every bullet point is a node; the same node can appear in multiple places as a reference, and editing anywhere updates every occurrence. Supertags turn nodes into typed objects carrying a field/template; all nodes sharing a supertag form a collection that can be searched, filtered, and rendered as a table or other views inside the outline. AI is embedded throughout: chat that references your notes, reusable 'AI command nodes' that send selected notes to a model with a user-defined prompt and write the result back, a Prompt Workbench to preview payloads, plus image generation, voice memos, live dictation, and bot-free meeting transcription. It is positioned for personal and team knowledge management, arguing graph-structured connections enable AI multi-hop reasoning. Note:
RESEARCHED BUILD PROMPT
Give this to your coding agent.
Build a functional single-user web prototype of a knowledge-graph outliner in the spirit of Tana Outliner — do not copy Tana's brand, logo, marketing copy, proprietary assets, or UI theme; invent your own layout and styling. Core workflow to replicate: the user types fast, freeform hierarchical bullet notes; every bullet is an addressable node; any node can be embedded elsewhere as a reference, and editing a node propagates to every place it appears. A 'supertag' system lets users tag a node with a type that brings a field template (e.g., Task with Due Date, Status); all nodes of a given type form a live collection. User-facing requirements: an infinite-outline editor with indent/outdent, collapse/expand, and search; applying supertags and editing their fields inline; node references rendered inline; saved views such as a table of a type's nodes with column filters and sorting; full-text and field search with results linked back to their nodes; an AI chat pane that answers using the workspace content as context; and reusable 'AI command' nodes that run a user-written prompt on selected nodes and write the returned text back into the outline. Technical requirements: local-first storage (IndexedDB or SQLite via sql.js, or JSON files) with a data model of nodes (id, title, parent/children, references, supertag, fields) and one-way propagation of reference edits; a deterministic renderer (React or Vue with a virtualized tree, or a markdown-like text tree) that round-trips edits without data loss; AI wired through an API key (e.g., OpenRouter) so no model is bundled, with a prompt-injection warning surfaced in the UI. Sensible implementation approaches: treat the outline as the source of truth and derive views/queries as projections over nodes; store references as node IDs resolved at render time. Out of scope: real-time collaboration/multiplayer, mobile apps, sync between devices, meeting transcription, image generation, and native packaging — a single-window web app is sufficient. Deliver a README covering setup, the node/field data format, how supertags and views work, how the AI chat and commands are configured, and known limitations (no conflict resolution, AI output quality varies by model, no offline sync).