A high-signal personal knowledge base and second brain capturing engineering mental models, agent systems architecture, practical data-mesh patterns, Japanese workplace-language practice, and personal operating heuristics.
This repository is organized as an AI-native Personal Knowledge Base (PKB) focused strictly on high-leverage, proprietary context that general-purpose foundation models lack:
- Operating Heuristics & Work Style: George's mental models, trade-off evaluation matrices, decision-making frameworks, and collaboration expectations.
- AI & Agent Systems: Agent orchestration paradigms, evaluation harness patterns, RAG pipelines, and agent skill engineering.
- Data Mesh: Practical discovery, interoperability, governance, and in-place consumption patterns.
- Japanese Workplace Language: Japanese-first practice for messages encountered at a Japanese internet company.
| Domain | Directory | Focus & Key Contents |
|---|---|---|
| 🧠 Operating Heuristics | how_to_use_me/ |
Working style, trade-off evaluation, decision-making frameworks, communication, LLM philosophy, RCA expectations, and trust building. |
| 🤖 AI & Agent Systems | ai/ |
Agent orchestration patterns, evaluation frameworks, harness design, RAG pipelines, and agent skills. |
| 🏛️ Data Mesh | arch/data_mesh.md |
Data-product discovery, metadata, governance, quality, and in-place access controls. |
| 🇯🇵 Japanese Workplace Language | specific/jp/ |
Japanese-first practice for understanding and replying to workplace messages. |
When using this knowledge base with LLMs and AI coding assistants:
Feed the relevant sections from how_to_use_me/ alongside your technical proposal and ask:
"Based on George's operating heuristics in how_to_use_me/, what questions, trade-off challenges, or risks would George likely raise regarding this proposal?"
When designing data-product discovery, governance, or cross-domain consumption, query arch/data_mesh.md. For Japanese workplace-message practice, query specific/jp/.
See AGENTS.md at the repository root for explicit subagent loading strategies, editing rules, and topic boundaries.
- High-Signal Context Only: Focus on personal heuristics, nuanced trade-offs, and operational lessons. Avoid storing generic textbook definitions that pre-trained LLMs already know.
- Preserve Operating Intent: When refining notes, maintain concrete heuristics, failure modes, and real-world trade-offs rather than abstract summaries.
- Cross-Link Freely: Link related notes across architecture and specific technologies to build an interconnected knowledge graph.