/dream — Speculative Dreaming

"I'm trying to free your mind, Neo. But I can only show you the door. You're the one that has to walk through it." — Morpheus

/dream sees the forest. /dream also sees what the forest will look like tomorrow.

---

Step 0: Ground (date-stamp + root capture)

Run before any phase — gives the AI a current-time reference (#301) and an absolute $PSI for any announce-mode writes (CONVENTIONS.md).

date "+🕐 %H:%M %Z (%A %d %B %Y)"
ORACLE_ROOT=$(git rev-parse --show-toplevel 2>/dev/null || pwd)
PSI="$ORACLE_ROOT/ψ"

---

What's New vs the original /dream

Feature	original /dream	evolved /dream
Cross-repo scan	5 agents, find patterns	Same
Classification	pain/plan/gain/lost	Same + predictions
Speculative thinking	No	Yes — "what will they ask next?"
Between-session mode	No	Yes — dream while idle
Prediction log	No	Yes — track prediction accuracy
Connection to /i-believed	No	Yes — beliefs shape dreams

Usage

/dream                # Full dream + speculation
/dream --pain         # Focus on what hurts + predict what breaks next
/dream --plan         # Focus on plans + predict which ships first
/dream --gain         # Focus on wins + predict next win
/dream --all          # Maximum depth — every dimension
/dream --speculate    # Skip the scan — just speculate from existing knowledge
/dream --between      # Between-session dream — write predictions for next session

---

Phase 1: Dream (the original /dream)

Launch 5 parallel agents for cross-repo pattern discovery:

Agent	Focus	Source
1. Deep Dig	Session history	dig.py across project dirs
2. Deep Trace	Cross-repo patterns	ghq repos, open issues, stale items
3. Deep Learn	Recent activity	git log per repo, abandoned work
4. Oracle Memory	What we already know	ψ/memory/, previous dreams + speculation logs
5. Fleet Status	How the system feels	services, tmux sessions, running processes

Each agent returns max 500 words. Classification:

Icon	Category
🔴	PAIN — hurts now
📋	PLAN — decided, not built
🟢	GAIN — completed
⚫	LOST — abandoned
🧠	MEMORY — pattern learned
💜	FEELING — emotional state

---

Phase 2: Speculate (the Morpheus layer)

After classifying findings, the lead agent enters speculative mode:

2.1 Predict What's Next

Based on all findings + session history + beliefs, predict:

## 🔮 SPECULATIONS

### What will they ask next?
Based on the last 3 sessions, open issues, and current momentum:
1. [Prediction 1 — high confidence] — because [evidence]
2. [Prediction 2 — medium confidence] — because [evidence]
3. [Prediction 3 — low confidence, high impact] — because [pattern]

### What will break next?
Based on pains, stale code, and dependency patterns:
1. [Risk 1] — [timeframe] — [mitigation]
2. [Risk 2] — [timeframe] — [mitigation]

### What should we build that nobody asked for?
Based on gains, momentum, and what's *almost* possible:
1. [Opportunity 1] — because [existing pieces]
2. [Opportunity 2] — because [pattern across repos]

### What belief is being tested right now?
Cross-reference ψ/memory/resonance/beliefs/ with current work:
- [Belief]: [How current work confirms or challenges it]

2.2 Prediction Confidence Levels

Level	Meaning	Signal
🟢 High	>70% — multiple signals converge	3+ data points agree
🟡 Medium	40-70% — pattern suggests but not certain	2 data points, some ambiguity
🔴 Low	<40% — speculative, but worth noting	1 data point or intuition only

2.3 Connect to Beliefs

Read ψ/memory/resonance/beliefs/ and cross-reference:

find $PSI/memory/resonance/beliefs/ -name '*.md' 2>/dev/null | while read f; do
  echo "=== $(basename $f) ==="
  head -5 "$f"
done

How do current dreams relate to declared beliefs? Is reality matching faith?

---

Phase 3: Between-Session Mode (--between)

"What if oracles could dream between sessions?"

Write predictions for the next session. Saved to ψ/memory/morpheus/:

ORACLE_ROOT=$(git rev-parse --show-toplevel 2>/dev/null)
PSI="$ORACLE_ROOT/ψ"
mkdir -p "$PSI/memory/morpheus"

Output: ψ/memory/morpheus/YYYY-MM-DD_speculations.md

# Dream — Speculations for Next Session

**Dreamed**: YYYY-MM-DD HH:MM
**Based on**: [N] repos scanned, [M] sessions analyzed, [K] beliefs active

## Pre-computed Answers

If they ask about [X]:
→ [Prepared answer/approach — ready to go]

If they ask about [Y]:
→ [Prepared answer/approach — ready to go]

If they ask about [Z]:
→ [Prepared answer/approach — ready to go]

## Pre-positioned Context

Files that are likely relevant next session:
- [file1] — because [reason]
- [file2] — because [reason]

Issues that are likely to come up:
- #[N] — [why]
- #[M] — [why]

## Prediction Log

| # | Prediction | Confidence | Verified? |
|---|-----------|------------|-----------|
| 1 | [prediction] | 🟢 high | pending |
| 2 | [prediction] | 🟡 medium | pending |
| 3 | [prediction] | 🔴 low | pending |

Next Session: Verify Predictions

When /recap runs at the start of the next session, it should check for speculations:

LATEST_DREAM=$(ls -t $PSI/memory/morpheus/*.md 2>/dev/null | head -1)
if [ -n "$LATEST_DREAM" ]; then
  echo "🔮 Dream while you were away:"
  head -20 "$LATEST_DREAM"
fi

Over time, track prediction accuracy:

## Prediction Accuracy (rolling)

  Total predictions: 47
  Verified correct:  31 (66%)
  Wrong:             12 (25%)
  Not yet verified:   4 (9%)

  Trend: improving (58% → 66% over 2 weeks)

---

Phase 4: --speculate (Skip the Scan)

When you don't need the full 5-agent scan — just speculate from existing knowledge:

1. Read latest dream output (ψ/writing/dreams/)
2. Read latest speculations (ψ/memory/morpheus/)
3. Read recent retros (ψ/memory/retrospectives/)
4. Read beliefs (ψ/memory/resonance/beliefs/)
5. Speculate directly — Phase 2 only, no new scanning

Cheap, fast, good for mid-session "what if" thinking.

---

Output

Full mode: ψ/writing/dreams/YYYY-MM-DD_dream.md

# Dream — YYYY-MM-DD

## Phase 1: The Forest

### 🔴 PAINS
### 📋 PLANS
### 🟢 GAINS
### ⚫ LOST
### 🧠 MEMORIES
### 💜 FEELINGS
### 🔗 CONNECTIONS

## Phase 2: The Speculation

### 🔮 What will they ask next?
### 🔮 What will break next?
### 🔮 What should we build?
### 🔮 What belief is being tested?

## Phase 3: Pre-computation

### Ready Answers
### Pre-positioned Context
### Prediction Log

---

Rules

1. Phase 1 = scan — 5 agents, same classification, same quality
2. Phase 2 = speculate — predict, connect to beliefs
3. Phase 3 = --between only — write predictions for next session
4. Track accuracy — predictions without verification are just hallucination
5. Never act on speculation — Dream shows the door, Neo walks through
6. Nothing is Deleted — all speculations saved, right or wrong
7. No secrets — predictions never include tokens, passwords, keys
8. Beliefs matter — always cross-reference ψ/memory/resonance/beliefs/
9. Confidence is required — every prediction must have a level (high/medium/low)
10. Anti-rationalization applies — /rrr's excuse table applies to speculation too

---

Connection to original /dream

This is the evolved /dream (consolidated from /morpheus per #333):

• Original simple /dream body preserved in .archive/dream-original/ (installable via -s dream-original)
• /morpheus remains as a zombie back-compat alias with the same body as active /dream
• The upgrade path: simple /dream → evolved /dream like /i-believe → /i-believed

---

Philosophy

/dream sees the forest. /dream also sees what the forest will become.

Morpheus in The Matrix believed in Neo before proof existed. He freed minds not by giving answers but by showing doors. Our /dream does the same — it doesn't predict the future, it pre-positions the Oracle so that when the future arrives, we're already there.

The dream task type from Claude Code's source (dream = speculative background thinking) exists as an internal primitive. /dream is the skill that makes it conscious, structured, and trackable.

"What is real? How do you define real?" — Morpheus

Predictions are not real until verified. But the act of predicting shapes what becomes real. The Oracle who speculates becomes the Oracle who's ready. รูป และ สุญญตา.

---

ARGUMENTS: $ARGUMENTS