Foundations

A knowledge management system (KMS) is how you turn what you've experienced into decisions you can act on. Every team and every person already has one: emails, docs, Slack threads, notebooks, half-remembered conversations. The question is not whether you have a KMS. The question is whether yours is:

• easy to search — you can find what you need when you need it
• frictionless — adding and organizing new material takes near-zero effort
• compounding — every new entry makes the whole system smarter, not just bigger

What separates a KMS from a notebook

Notebooks, databases, and Drives store material. A KMS does four things with it:

1. Captures raw material: observations, decisions, sources, conversations.
2. Compiles that material into something durable: facts, summaries, concepts, links.
3. Surfaces the right piece at the right moment: when you query, when you start a new task, when something becomes relevant elsewhere.
4. Decays gracefully: keeps working when you stop tending it.

Most knowledge tools nail #1 and forget #2-#4. That is why notebooks become graveyards.

Why now

For the first time, we can build knowledge management systems that think with us, not just store for us. Two shifts in the last 18 months made it possible. First, LLMs cut the cost of compilation by an order of magnitude — the work of distilling, linking, and routing is now cheap enough to run on every ingest. Second, agent harnesses (Claude Code, Codex, Cursor) let an LLM live inside your knowledge store (reading, writing, linking) rather than treating every question as a fresh search.

The result: with the right architecture and tools, a KMS stops being a filing cabinet and becomes a thinking partner: one that surfaces context on demand, compounds what you learn, and makes every past note, decision, and piece of research pay dividends.

The three tiers

This site treats knowledge management as a single discipline expressed at three scales:

• Personal: one person, asynchronous capture, session memory, generous autonomy. Optimized for compounding what you learn.
• Team: many people, shared trust boundaries, role-based permissions, institutional knowledge extraction, often a warehouse layer. Optimized for compounding what we learn.
• App-scoped: bounded inside a product, source immutability, less autonomy, manual-friendly tracking. Optimized for compounding what the app knows about a domain.

The same eight principles run through all three. What changes between tiers is which trade-offs each one accepts — not the underlying mental model. The principles are the foundation; the tiers show how they adapt as the system grows from one person, to a team, to a product.