Hackathon: Building your own Agent LLM Wiki
🧠 AI-Memory Hackathon: Building Your Own Agent LLM Knowledge Wiki
"Instead of just retrieving from raw documents at query time, the LLM incrementally builds and maintains a persistent wiki" — Karpathy
Inspired by Karpathy's popularization of the LLM Wiki concept, Cognee is inviting builders to take that idea further — collaboratively extending and constructing an Agent Knowledge Wiki together. This is your chance to get hands-on with cutting-edge AI memory technology and compete for serious prizes!
🏆 Prize Pool: $1,500+ Cash
| Place | Prize |
|---|---|
| 🥇 1st | $800 cash |
| 🥈 2nd | $500 cash |
| 🥉 3rd | $200 cash |
🛠️ What You'll Build
Using Cognee's open-source memory engine, participants will construct their own LLM Knowledge Wiki capable of three core operations:
- 📥 Ingest — feed and structure raw information
- 🔍 Query + Self-improve — retrieve and refine knowledge over time
- ✅ Lint — validate and clean the knowledge base
👉 See how Karpathy explains the concept: Karpathy's LLM Wiki Gist
💥 Presentation Format
Each team gets 3 minutes to impress the judges:
- 🎤 Pitch your idea and demonstrate how you leverage agent self-improvement
- 🖥️ Run a live demo showcasing your agent in action
⏱️ Schedule
| Time | Activity |
|---|---|
| 12:00 PM | 🚪 Doors open + Networking |
| 12:30 PM | 🎙️ Opening remarks + partner demos |
| 1:00 PM | 💻 Hacking begins |
| 4:30 PM | 📬 Project Submission Deadline — finalists selection |
| 5:00 PM | 🏅 Finalist presentations & judging |
| 5:30 PM | 🏆 Awards ceremony |
| 6:00 PM | 🎉 Event wrap-up & doors close |
🤝 Technology & Venture Partners
⚙️ Cognee
Cognee is an open-source AI memory engine that converts unstructured data into persistent, queryable memory. It empowers AI agents to store, retrieve, and reason over information across time — enabling more reliable, adaptive, and stateful AI systems.
🔴 Redis
Redis serves as a real-time context engine that unifies streaming and vector data to power intelligent applications and AI agents with instant awareness. It allows developers to process and serve dynamic context at sub-millisecond speed with enterprise-grade reliability.
💼 Pebblebed
Pebblebed is a technical early-stage VC firm founded by:
- Pam Vagata — co-founder of OpenAI, ran AI for Stripe, inventor of FBLearner Flow
- Keith Adams — founded Facebook AI Research, chief architect at Slack, 20th engineer at VMware
- Tammie Siew — former Sequoia Southeast Asia investor, former Sequoia & Notable Capital backed founder
🎟️ Register now: lu.ma/uhda61yp
Register to See Address
Get directions
Register to See Address
Get directions
