Separate capture from publication
Raw inputs are allowed to stay messy. Structure and voice are added only as an idea earns its way toward a durable concept or public artifact.
GTBrain
A local second brain that turns raw notes and project evidence into concepts, learnings, writing angles and publishable artifacts.
system · active
Product designer and builder · Updated Aug 2026
- Knowledge systems
- Content pipeline
- Search
- AI memory
01 · Problem
Useful observations are spread across repositories, notes, links and conversations, while publishing requires repeatedly reconstructing what was learned and why it matters.
02 · System
A staged knowledge pipeline captures raw material, extracts durable concepts, records learnings, develops writing angles and routes approved ideas into public outputs.
03 · Outcome
The repository combines a durable folder protocol with search, graph, library, reader and publishing views in one local workspace.
Context
The system is designed around progressive refinement: messy capture remains cheap while each later stage carries more structure and editorial commitment.
Architecture
- 01Markdown knowledge folders with explicit lifecycle stages
- 02Express and Postgres service layer
- 03React library, graph, reader and daily-work views
- 04Voice rules and publishing queues for reusable output
- 05Nightly speculative-idea workflow with promotion gates
Decisions & trade-offs
Evidence
- One workflow connects source material, project learnings and public writing
- Search and graph views sit on top of the same durable knowledge structure
- Speculative outputs remain separate until deliberately promoted
What this taught me
- A second brain becomes useful when it models the movement from evidence to output, not just the storage of notes.
Follow the system
Connected work is resolved from the shared content graph.