Why Knowledge Should Survive the LLM
Why reusable knowledge should live outside any one LLM, preserve changes in meaning, keep multiple perspectives, and remain traceable over time.
Notes
Blog
Hands-on notes on technologies I use, what they make possible, where they become difficult, and the details that matter in practice.
Published Here
FB15K: Loading FB15K-237 into Neo4j
A practical Cypher import—and what a graph of MIDs and 237 relation names reveals about knowledge graph benchmark data.
FB15K: Understanding the Dataset
What FB15K and FB15K-237 contain, why inverse relations matter, and what these widely used benchmarks actually measure.
COVID-19: Same Patient, Different Data
How mixed CSV and Excel files, Japanese character encodings, source-specific patient IDs, and conflicting values turned COVID-19 data integration into an identity and provenance problem.
COVID-19: Validating Related Patient Data with Graphs
How Neo4j and Git history helped reveal suspicious relationships, report them to local governments, and trace later corrections.
Learning Neo4j: A Japanese Citrus Family Tree
Using Japanese citrus varieties to learn graph modeling with parent-child relationships, intermediate breeding lines, and path queries.
Migration Planning: Using Graphs
How graph modeling can help clarify dependencies, inconsistencies, validation points, and cutover risks in system and data migration.
Learning Neo4j: Okonomiyaki
Using a familiar food domain to learn graph modeling with shops, places, noren relationships, ingredients, and cooking steps.
Published Elsewhere
Knowledge Must Live Outside the LLM—and Carry Its History
Why knowledge should live outside individual LLMs and preserve its history as models, tools, and contexts change.