# How SecondOS compares

Three ways teams give AI context today, and where each falls short:

1. **Cold re-read** — the tool re-reads your files every session. Slow, expensive, and it re-derives a fuzzy picture each time.
2. **Hand-written CONTEXT.md** — rots the moment the code moves; not query-scoped; costs a lot to load.
3. **Agent-memory APIs (mem0, Zep)** — a *different category*: conversational/agent memory for chat-derived facts you integrate yourself. They don't understand a codebase's structure.

SecondOS is the category no one else fills: **one portable, structure-aware memory of your codebase that every AI tool reads** — automatic, always-current, and yours to export.