Two-tier-memory is a lightweight SQLite-based pattern for storing and querying long-term context for AI coding agents—a tier-1 index (markdown) plus tier-2 relational database to avoid context-window saturation. It addresses a real problem in agent UX but has minimal market signal and no monetization path.

One insight: The core innovation is architectural (two-tier indexing), not technical depth—copyable in days—and the founder explicitly built it as a collaborative open-source demo with Claude Code, signaling research/pattern-sharing over product intent.