The Problem — Token waste on "where is X?" + code-spec drift Existing Solutions — Manual context, @-mentions, MCP memory — all incomplete The 4-Layer Architecture — L1 grep index, L2 wiki, L3 RAG, L4 compile The Killer Feature: Spec-First Fix — Phase 0-7 workflow with consensus loops Real Numbers — Before/after token comparison from production use Getting Started — 4-step setup What's Next