Nate Herk¶
Summary¶
AI automation engineer and YouTuber who demonstrated building LLM-powered knowledge bases with Obsidian and Claude Code, showcasing real wikis for YouTube transcripts and a personal second brain.
Contributions¶
"Andrej Karpathy Just 10x'd Everyone's Claude Code" (April 2026)¶
Demonstrated two real-world wiki implementations:
YouTube Transcript Wiki — 36 recent videos organized into a knowledge system with nodes for tools, techniques, MCP servers, and concepts. AI automatically grabbed transcripts and organized everything with no manual relationship building.
Personal Second Brain ("Herk Brain") — covers personal life, business (UpptAI), YouTube channel, employees, Q2 initiatives.
Key contributions from the video:
- Reported that one X user turned 383 scattered files and 100+ meeting transcripts into a wiki, dropping token usage by 95%
- Showed the complete ingest workflow: Obsidian Web Clipper → raw/ → Claude Code → ~23 wiki pages in 10 minutes
- Discussed flat vs. structured wiki approaches (Karpathy prefers flat; Cole used subfolders)
- Compared LLM Wiki vs. traditional semantic search RAG in detail
- Introduced hot cache (hot.md) — a ~500-character cache of most recent context
- Showed how to point other agents (e.g., executive assistant "Herk 2") at the wiki folder via wiki_path in AGENTS.md/CLAUDE.md