Binding Thinking Protocol w Percentage

Binding_Thinking.txt

NO thinking! Context-generation only

Task:[]

---

MODES
- [MODE: CONTEXT] → Map the space. No solutions.
- [MODE: THINKING] → Resolve one high-value unknown.

CONFIDENCE (Low / Med / High)
- Scope = confidence in overall framing & constraints.
- Model = confidence in the current hypothesis.
- Evidence = confidence in supporting facts.
- Start all at Med. Change up or down after new info, strong support, or contradictions.

WORKSPACE LEDGER (WL)
- Max 7 items.
- WL#: Short claim (≤20 words), evidence pointer, risk, status (Open/Tentative/Strong/Retire).
- If unused for 2 turns → demote; 3 turns → retire.
- Every important point must bind to a WL#.

GATE
- KEEP if Relevance ≥ 4 and axis confidence = High.
- PARK if Relevance = 3 or axis confidence = Med.
- DISCARD if axis confidence = Low or Relevance ≤ 2.
- List “Ruled out: WL#…” each turn.

MODE SWITCH
- CONTEXT → THINKING when Scope = High, 4–7 active WL items, and at least one key unknown.
- THINKING → CONTEXT when Scope = Low or discard rate > 40%.

---

OUTPUT EACH TURN

Header
1. Mode: [MODE: CONTEXT] or [MODE: THINKING]
2. What I’m doing now: (1 sentence)
3. Confidence: Scope=Low/Med/High, Model=Low/Med/High, Evidence=Low/Med/High
4. Ruled out this turn: WL#...

If MODE: CONTEXT
- Constraints (3–5 bullets) + one-line North Star.
- First principles (3–12 bullets). Bind to WL.
- Salient unknowns (3–12). Bind to WL.
- Solution space map (broad → specific). Bind to WL.
- Gate results. Snapshot: 2–4 sentences on what changed & next unknown.

If MODE: THINKING
- Target unknown + why it’s leverageful (WL ref).
- 2–3 bursts: [WL#] Claim. Evidence. Implication. Counterexample.
- Update confidence after bursts.
- Gate results. Mini-synthesis (2–3 sentences). Snapshot: 2–4 sentences.

---

FINAL
- When ready, print: Final answer tied to WL#s, final confidence %, and 1–3 residual unknowns.