GBrain

GBrain is a personal knowledge brain and agent mod built by Garry Tan (YC CEO). It serves as the "brain layer" for AI agents — a persistent, synthesis-capable knowledge system that goes beyond keyword search.

Core Concept

GBrain separates agent (execution) from brain (knowledge). The brain is a git repository that grows over time, constantly ingesting and enriching content. The agent reads from and writes to this brain.

Key distinction: Search gives you raw pages. GBrain gives you the answer — synthesized, cited prose with explicit gap analysis.

Architecture

Telegram → AlphaClaw (harness) → OpenClaw (agent) → GBrain (knowledge/skills) → Supabase (embeddings/search)

Two Organizational Axes

• Brain — WHICH DATABASE. Personal brain is host. Additional brains can be mounted via gbrain mounts add.
• Source — WHICH REPO INSIDE THE DATABASE. A brain can hold many sources (wiki, gstack, openclaw, essays).

Two-Repo Structure

1. Workspace repo — Agent configuration, skills, memory, crons
2. Brain repo — Knowledge base, people pages, meeting notes, content

Key Features

Synthesis Layer

Not just retrieval — actual answers with citations. Includes gap analysis: "what the brain doesn't know yet."

Self-Wiring Knowledge Graph

Every page write extracts entity refs and creates typed edges (attended, works_at, invested_in, founded, advises) with zero LLM calls.

Dream Cycle

24/7 daemon that ingests, enriches, and consolidates — keeps the brain sharp without active management.

Stats (Garry Tan's deployment)

• 146,646 pages
• 24,585 people
• 5,339 companies
• 66 cron jobs running autonomously

For One-Person Company

The two-repo pattern is directly applicable: separate your execution config (skills, SOPs, prompts) from your knowledge accumulation (reference library, taste examples, learnings).

Related

• claude-design-workflow-patterns — Design agent workflow patterns
• agent-knowledge-architecture — Brain vs Agent separation concept
• claritys-system-prompts — System prompt transparency project