The problem

​AI systems have too much information and not enough sense of what to do with it. Context gets fragmented, retrieval is expensive, personalization is weak, memory is limited, and routing across different sources is mostly guesswork. Keyword search doesn’t solve this, and neither does putting a chatbot on top of it.

​What to build

Anything that makes AI search meaningfully better. Some directions we’d love to see:

• Retrieval that goes beyond keyword matching
• Long-context and memory systems
• Routing queries to the right engine, database, or tool
• Agents that plan, search, verify, and synthesize across sources
• Faster, cheaper retrieval with fewer wasted tokensKnowledge graphs and connecting scattered information
• Search that understands a user’s context and history
• Search across images, video, audio, and code
• New interfaces for working with information

​You don’t have to stick to this list. If you have a better idea, build that.

​Projects will be judged on innovation, technical execution, search intelligence, and impact.

​Come build something!