/brain:sleep

Run a full 9-phase maintenance cycle inspired by how the human brain reorganizes memories during sleep. This is the most comprehensive brain maintenance command.

Usage

/brain:sleep [scope]

Arguments:

• [scope] — (Optional) Limit sleep to a specific subtree (e.g., professional/skills). If omitted, processes the entire brain.

What It Does

/brain:sleep runs nine sequential phases that mirror neuroscience research on memory consolidation during sleep:

Phase 1: Replay

Scans all memories and computes current effective strengths based on decay. Categorizes memories into tiers:

Phase 2: Synaptic Homeostasis

If the mean strength across all memories exceeds 0.5, proportionally scales down ALL strengths to prevent inflation. Then selectively re-boosts:

• High-salience memories (>= 0.7)
• Recently-accessed memories (within last 7 days)
• Frequently-recalled memories (access_count >= 5)

Based on Tononi and Cirelli's Synaptic Homeostasis Hypothesis (SHY) — the idea that sleep's primary function is to downscale synaptic weights that accumulate during waking.

Phase 3: Knowledge Propagation

Evaluates recent memories (within propagation_window_days, default: 7) against the existing hierarchy:

• Enrichment — New details that supplement existing memories
• Contradiction detection — New information that conflicts with stored knowledge (reduces confidence by -0.20)
• Validation — New evidence that confirms stored knowledge (boosts confidence by +0.10)
• Obsolescence marking — New information that makes older memories outdated
• Cross-referencing — Discovering connections between previously unlinked memories

Based on memory reconsolidation research — the finding that recalling a memory makes it temporarily malleable and subject to updating.

Phase 4: Semantic Crystallization

Finds frequently-recalled episodic memories and extracts generalizable principles into new semantic memories. The event details begin fading but the lesson persists.

For example, three episodic memories about deploy failures might crystallize into one semantic memory: "Deploy failures consistently stem from skipping the pre-deploy checklist."

Phase 5: Reorganize

Detects flat clusters — 3 or more related memories at the same directory level — and restructures them into deeper subcategories automatically. This keeps the hierarchy clean and meaningful as your brain grows.

Phase 6: Consolidate

Groups weak related memories (below the consolidation threshold) and merges them into stronger combined memories:

consolidated_strength = max(source_strengths) + 0.15
consolidated_decay    = min(source_decay_rates)

See /brain:consolidate for details.

Phase 7: Prune

Archives memories that have faded below 0.1 strength by moving them to _archived/. Their index entries are transferred to _archived/index.json and they remain searchable as a fallback.

Memories with salience >= 0.7 are exempt from pruning — they must be explicitly forgotten via /brain:forget.

Phase 8: REM Dreaming

Selects random memories from different categories and discovers creative cross-domain connections via analogical reasoning. Each discovered connection is scored by:

• Novelty — How unexpected the connection is
• Utility — How practically useful the connection could be
• Surprise — How much it challenges existing assumptions

High-scoring connections are stored as new insight memories with association links to the source memories.

Phase 9: Expertise Detection

Identifies dense knowledge areas by analyzing memory concentration, strength, recall frequency, and temporal span within each category. Generates expertise profiles at four levels:

Each expertise area gets an _expertise.md profile documenting what you know well, knowledge gaps, and contributing memories. The phase also populates the spaced repetition review queue.

Example Interaction

User: /brain:sleep

Agent: 😴 Starting sleep cycle...

Phase 1/9: Replay
  Scanned 42 memories: 12 strong, 18 moderate, 8 weak, 4 fading

Phase 2/9: Synaptic Homeostasis
  Mean strength: 0.54 — applying downscaling
  Re-boosted 5 high-salience and 3 recently-accessed memories

Phase 3/9: Knowledge Propagation
  Evaluated 6 recent memories against hierarchy
  • 2 enrichments applied
  • 1 contradiction detected (confidence reduced)
  • 1 cross-reference created

Phase 4/9: Semantic Crystallization
  Crystallized 1 pattern from 3 episodic memories:
  → "Pre-deploy checklist failures are the leading cause of incidents"

Phase 5/9: Reorganize
  Moved 4 memories into new subcategory: professional/skills/typescript/

Phase 6/9: Consolidate
  Merged 3 weak memories into 1 consolidated memory

Phase 7/9: Prune
  Archived 4 fading memories (salience < 0.7)

Phase 8/9: REM Dreaming
  Discovered 1 cross-domain connection:
  → "The event sourcing pattern from project-alpha could solve the
     audit trail problem discussed in personal/education/compliance"
  Novelty: 0.7 | Utility: 0.8 | Surprise: 0.6

Phase 9/9: Expertise Detection
  Updated 3 expertise profiles:
  • TypeScript: Deep Knowledge (0.72)
  • System Architecture: Working Knowledge (0.58)
  • DevOps: Awareness (0.35)
  Added 8 memories to review queue.

✅ Sleep complete. Brain health improved.
   Active: 38 memories (was 42)
   Archived: 9 (was 5)
   New: 2 (1 crystallized, 1 dream insight)