Cole Medin

I Love the Karpathy LLM Wiki but it Doesn't Scale. Here's What Does.

Markdown makes great memory for an AI agent - a Karpathy-style LLM wiki, an Obsidian second brain it reads and writes. If the agent is just yours, a folder of markdown is hard to beat: simple, yours, all in git. My second brain is exactly that! It falls apart the second you ship that agent to other people. Now the memory has many users at once, the data behind it changes in real time, and you need retrieval and access control a pile of files can't give you. That's not a memory problem anymore - you need a context layer. So I take an agent and build it that layer with Redis, going deep on two new pieces of Redis Iris: Context Retriever, which turns your data into tools the agent can safely call, and Agent Memory, which scales recall across thousands of conversations. ~~~~~~~~~~~~~~~~~~~~~~~~~~ Thanks to Redis for working with me on this video! I've been looking for a while for the best database platform to make exactly this video. I focus on the two newest pieces, Context Retriever and Agent Memory. Iris is actually five tools, those two plus Redis Data Integration, LangCache, and Redis Search. All seriously worth a look: https://fandf.co/439eNyo ~~~~~~~~~~~~~~~~~~~~~~~~~~ - Join my free live workshop on July 29th, where I show you how to become an AI native engineering organization - building a reliable standard for how teams use AI coding agents: https://dynamous.ai/ai-native-engineering-org - My Pydantic AI + Redis Iris agent I demoed in the video: https://github.com/coleam00/redis-iris-agent - Redis Iris docs (Context Engine): https://redis.io/docs/latest/develop/ai/context-engine/ - Context Retriever quickstart: https://redis.io/tutorials/getting-started-with-redis-iris/ ~~~~~~~~~~~~~~~~~~~~~~~~~~ 0:00 Two Types of AI Agents 1:14 Why Personal Agents Don't Scale 3:02 Production Agent System Architecture 4:47 Demo: E-commerce Agent Support 9:59 Implementing the Context Retriever 14:40 Short and Long-Term Memory 18:07 When to Use Each Architecture ~~~~~~~~~~~~~~~~~~~~~~~~~~ Join me as I push the limits of what is possible with AI. I'll be uploading videos weekly - at least every Wednesday at 7:00 PM CDT!