How I Built a Memory Layer for My AI Coding Sessions
Why I built recuerd0, a context memory layer for AI coding agents, on SQLite and FTS5 full-text search instead of embeddings, vector databases, or RAG.
I'm Mario Alberto, a software engineer and entrepreneur based in Colima, México. I'm the creator of Rails MCP Server and former CTO/co-founder of Aoorora, where I architected a core banking platform in Ruby on Rails that enabled lending startups to build on modern, secure infrastructure. I spend my time at the intersection of Ruby on Rails and AI—building tools that help developers work smarter. When I'm not writing code, I'm documenting territory, popular culture, and memory through photography.
Why I built recuerd0, a context memory layer for AI coding agents, on SQLite and FTS5 full-text search instead of embeddings, vector databases, or RAG.
Kamal isn't just for apps you build yourself. It works just as well for prebuilt third-party Docker images — here's the pattern I use for self-hosted apps.
Rails handles framework-level security defaults — but authorization, rate limiting, CSP, session hardening, and encrypted attributes are decisions only you can make.