I almost didn't share this. A few months ago I got obsessed with one question: why does attention actually work? Not how, why. That rabbit hole pulled me deep into vector algebra, calculus, and eventually into something I didn't expect to build. Titan started as a personal obsession and turned into a persistent, evolutionary memory layer for AI agents. But "memory" undersells it. What I really wanted was for agents to have a past. Memories that talk to each other like neurons. Semantically similar ones that rerank and strengthen over time, the same way attention weighs context. Using it changed how I work with agents completely. My coding agent now remembers decisions I made three weeks ago, bugs we debugged together, architecture choices, my preferences. Things that would vanish at the end of every session. It doesn't just remember facts. It knows me. The memory graph is honestly my favorite part. /titan-graph turns your conversations into a living knowledge graph with topic clusters, bridge concepts, and connections that reshape as your agent learns. Watching your own thinking visualized like that is kind of wild. Titan is open source, local-first, and completely free. No cloud. No data leaving your machine. One command, pi install npm, and your Pi coding agent stops forgetting. I believe there's an external harness coming that solves self-evolution in AI. Titan is my small experiment in that direction. Would love for you to try it and tell me what you think.

Titan Memory

A Persistent evolutionary memory layer for AI agents

A Persistent evolutionary memory layer for AI agents