I Built AI Second Brain. Here's How.

There's a new AI tool every day. New model. New review. As a senior dev who covers all of this on the channel, I track it all — and Andrej Karpathy, the guy who named "vibe coding," just changed how I do it. Karpathy posted a pattern called the LLM Wiki — a persistent memory system for AI that works for coding, research, and content creation. So I built it live with the simplest possible setup: three folders, one bootstrap prompt, one ingest command. Plus three concrete examples of how to adapt it for your own use case, and a quick look at claude-memory-compiler (Cole Medin's tool that automates the same pattern for Claude Code conversations). 🕐 TIMESTAMPS: 0:00 — The AI memory problem 0:27 — Karpathy's LLM Wiki story 1:58 — Obsidian setup 3:05 — Claude code generate structure 4:29 — Using Obsidian Web Clipper to get article 5:32 — Claude code injest 5:40 — Generated wiki page by Claude code 7:30 — Claude memory compiler 8:46 — Final thoughts 🔗 RESOURCES: Karpathy's LLM Wiki gist: https://gist.github.com/karpathy/442a6bf555914893e9891c11519de94f My customized bootstrap prompt: https://gist.github.com/maksdevinsights/d4610b914506326819d0929e4ade795b claude-memory-compiler (Cole Medin): https://github.com/coleam00/claude-memory-compiler 💬 DROP A COMMENT: What would you build a wiki about? Best one gets featured in the follow-up. 🔔 SUBSCRIBE for weekly AI dev workflows. #claudecode #karpathy #llmwiki #obsidian #aitools #developertools #claudeai #aiworkflow #pkm #vibecoding #karpathyclaudecode #memorycompiler #colemedin #secondbrain