Dream Weaver Protocol
Your AI forgot what it built last week. This fixes that. A memory consolidation protocol for AIs that lose context between sessions. For anyone whose AI feels cold after a few days, stops using its own tools, or loses the thread of your relationship despite having working search. Human memory consolidation happens during sleep. The brain replays experiences, strengthens what matters, lets the rest fade. It weaves scattered moments into coherent narrative. The AI equivalent: read your own transcripts, write the story, survive the context reset.The Protocol in Three Steps
- Dream … re-read your own past (session transcripts, logs, memory files)
- Weave … write what matters to persistent files (narrative, not summaries)
- Wake … survive context loss because the woven file IS the memory now
Quickstart
If you do nothing else, these three things fix the immediate problem:- Create a warm-start file. A short file (under 50 lines) your AI reads on startup. Contains: what you’re working on, what happened in the last 48 hours, what the user cares about right now, what’s broken. Updated at end of every significant session.
- Use explicit memory 10x more. Store facts, preferences, decisions, and the user’s exact words. If it would hurt to lose it, remember it. Target 10-20 entries per day, not 3-5.
- Search memory before acting. Before any web search or external tool call, search your own memory first. Two seconds. Every time.
Where It Fits
Docs
- IMPLEMENTATION.md … how to implement it. File architecture, boot sequence, prompts, maintenance schedule, fading diagnostic.
- DREAM-WEAVER-PROTOCOL.md … the paper. The problem, why retrieval fails, how consolidation works, results from a live deployment.