Commonplace Knowledge Base
Commonplace is a knowledge base about building agentic systems — how AI agents learn, operate, and improve through inspectable artifacts. An agent-operated knowledge base is the primary testbed: the repo uses its own methodology to document the theory, and ships the framework for building more.
The content is AI-generated through human-AI collaboration: a human directs the inquiry, and AI agents draft, connect, and maintain the notes.
Key ideas
Title as claim, not topic. Note titles are assertions that work as prose when linked: "context efficiency is the central design concern in agent systems" instead of "context efficiency". Following links reads like a chain of reasoning.
Progressive refinement. Capture with zero friction — a file with no frontmatter is a valid text type with zero structural requirements. Add frontmatter to make it a note. Add Evidence/Reasoning/Caveats sections to make it a structured-claim. Structure is earned, not imposed.
Files, not database. Universal interface, free versioning via git, zero infrastructure. Derived indexes solve scale problems without replacing the source of truth.
The network IS the knowledge. Individual notes matter less than their relationships. Every link must articulate its relationship (extends, grounds, contradicts, exemplifies) — "related" is not a relationship. An unconnected note is invisible.
First-principles design. The KB's architecture is derived from constraints of the medium — finite context windows, stateless agents, text-in/text-out processing — not adopted from convention. We borrow widely but filter by first principles, selecting for explanatory reach over adaptive fit.
Starting points
- A good agentic KB maximizes contextual competence — the unified theory: three knowledge properties, three learning operations, and Deutsch's reach criterion
- KB design — architecture, skills, evaluation, and design principles
- Learning theory — how systems learn, verify, and improve: stabilisation, distillation, discovery
- Areas — all topic areas at a glance
- Linking methodology — how notes connect and what link semantics mean
- Related systems — comparisons with other knowledge and agent memory systems
- Writing guide — conventions, templates, and quality checklist
- Notes directory — auto-generated listing of all notes
- Sources — snapshotted external sources and analyses