01
Data Repository
Store first. Documents, meetings, screenshots, and source material are captured in one place so AI can find them.
I distill tacit knowledge: the judgment and experience you use at work but have not yet written down. This page is the map of that system. Read the articles for the details, download a skill package to try it, or explore a service if you want to build it together.
Tools keep changing. Portable documents keep the work understandable and reusable across tools.
A document can function as an agent when three conditions are present:
Codex, Claude, ChatGPT, and other agents can change. The same source documents keep the context, rules, and operating knowledge in your control.
Data lets AI retrieve. Connected knowledge helps AI reason across context. Rules let AI act with your judgment.
Store first. Documents, meetings, screenshots, and source material are captured in one place so AI can find them.
Store and connect. Source material becomes linked knowledge, so AI can see relationships and reuse prior work.
Capture judgment. Standards, workflows, and skill packages let AI carry your decision criteria into the work.
AGENTS.md, routing rules, tag dictionaries, hooks, and review gates determine how the whole system operates. AI can prepare changes; a person still decides what enters the source of truth.
This architecture grew from actual use. Each stage solved the problem revealed by the previous one.
Articles, work records, and source material were finally captured in one place. It created a warehouse, which was the necessary starting point.
Bidirectional links, retrieval, four-dimensional tags, and the Tag Wiki method began turning stored files into connected knowledge.
Read the Tag Wiki method →The focus moved from operating a note-taking tool to designing documents AI could find, understand, and execute. Skill packages and one shared AGENTS.md clarified roles and rules.
Read the 3×4 method →One-off tasks are turned into trackable, reviewable loops. New methods are absorbed as rules, fields, or workflows when they fit the system.
Read about Loop Engineering →