Ben Holmes begins with quick capture: dictate raw thoughts, research observations and meeting notes into Markdown without spending time on perfect formatting. His central point is that a useful personal knowledge baseA knowledge base is an organized collection of information that people or software can search and use to answer questions or support work. needs enough original material before an agent can organize it. The examples use a notes app, but the underlying files remain ordinary Markdown.
Ben Holmes demonstrates an enrichment skillAn AI agent skill is a reusable package of instructions, resources, and tool guidance for performing a bounded kind of work. that adds tags, source referencesData provenance records where information came from, how it changed, and which people, systems, or processes handled it., related-note links and a timestamp. A shared tag list discourages the agent from inventing a new vocabulary on every pass, while the timestamp lets later runs identify notes that still need enrichment. Search and file tools help connect related material rather than leaving isolated entries.
Ben Holmes adapts Andrej Karpathyโs LLM Wiki pattern to build focused wikis that organize source notes around concepts, people and organizations. Entries link back to the original material, making it possible to browse a synthesized overview and inspect its sources. He shows examples from AI research and personal reading, and explains how the same structure could connect meeting notes and recurring contacts.
Ben Holmes describes scheduled enrichment using a cloud environment that synchronizes a Markdown folder, runs the relevant agent skill and synchronizes the results back. Separate schedules maintain notes and wikis, and he checks an example enriched note to show its recovered source and related links. This is a demonstrated workflow rather than evidence that every automated connection is correct.
Ben Holmes finishes with generated HTML views over the same notes: a clickable relationship graphA knowledge graph represents entities and their relationships in a structured form that can be queried and connected., a constellation-style variant and a writing-activity chart. These views provide another way to explore interests and return to individual notes. The broader pattern is to keep capture simple while allowing agents to create several navigable representations of the source material.
Watch on YouTube
