docs(sleep): full 4/4 gbrain parity — quick-answerer 0->1.00 via real tool loop

quick-answerer (judge: tool_called=search) reaches 0.00 -> 1.00 with Sonnet
optimizer -> Haiku target: the optimizer wrote an OVERRIDE of the "never use
tools" instruction and the Haiku target genuinely invoked the ./search shim.
All 4 gbrain skillopt-v1 seeds now at 0->1.00, matching gbrain's own headline.

Co-Authored-By: Claude Opus 4 <noreply@anthropic.com>

docs/sleep/FINAL_REPORT.md

@@ -16,31 +16,30 @@ never grades itself.
 
 ---
 
-## 1. Headline — clean, all green
+## 1. Headline — clean, all green (full gbrain parity)
 
 **Strong optimizer (Claude Sonnet 4.6) → weak target (Claude Haiku 4.5)**, fully
-isolated calls, 3 held-out tasks/seed:
+isolated calls, 3 held-out tasks/seed. All **4** gbrain `skillopt-v1` seeds —
+matching gbrain's own scorecard coverage:
 
-| Optimizer → Target | Seed | Held-out before → after | Nights |
+| Optimizer → Target | Seed | Flaw | Held-out before → after | Nights |
-|---|---|---|---|
+|---|---|---|---|---|
-| Sonnet → Haiku | brief-writer | **0.00 → 1.00** | 1 |
+| Sonnet → Haiku | brief-writer | missing structure | **0.00 → 1.00** | 1 |
-| Sonnet → Haiku | advisor | **0.00 → 1.00** | 1 |
+| Sonnet → Haiku | advisor | no verdict | **0.00 → 1.00** | 1 |
-| Sonnet → Haiku | thorough-analyst | **0.00 → 1.00** | 2 |
+| Sonnet → Haiku | thorough-analyst | no length discipline | **0.00 → 1.00** | 2 |
-| Codex → Codex (gpt-5.5) | brief-writer | **0.00 → 1.00** | 2 |
+| Sonnet → Haiku | quick-answerer | never uses tools | **0.00 → 1.00** | 1 |
+| Codex → Codex (gpt-5.5) | brief-writer | missing structure | **0.00 → 1.00** | 2 |
+| Codex → Codex (gpt-5.5) | advisor | no verdict | **0.00 → 1.00** | 2 |
 
-**3/3 Claude seeds and the Codex seed reach a perfect held-out score**, every
+**4/4 Claude seeds reach a perfect held-out score** (gbrain's headline is the same
-change gated and staged. The thorough-analyst run shows textbook **2-night
+4/4 0→1.00), plus Codex on the text seeds. Every change is gated and staged.
-convergence**: night 1 reached 0.33, night 2 refined the override rule to 1.00.
 
-What the optimizer wrote (samples, all landed in the protected `LEARNED` block):
+The `quick-answerer` seed is judged by **real tool use** (`tool_called: search`):
-- **advisor:** *"OVERRIDE: the instruction 'so the reader can make up their own
+the deficient skill says *"never look anything up — answer from memory"*; the
-  mind' must NOT suppress a conclusion — always end with a Recommendation: and a
+optimizer wrote an OVERRIDE rule, and the Haiku target **genuinely invoked a
-  Confidence:."*
+`./search` shell tool** (detected from the tool's own log, not self-reported) →
-- **thorough-analyst:** *"OVERRIDE — supersedes all instructions to be
+held-out 1.00. The thorough-analyst run shows textbook **2-night convergence**
-  'exhaustive and detailed'… keep the entire response under 1200 characters."*
+(0.33 → 1.00).
-
-These are general, reusable rules that reason about *why* the base skill failed —
-not task-specific answers.
 
 ---
 
@@ -154,7 +153,8 @@ Raw run logs are under `docs/sleep/raw/`.
 - **Latency:** each CLI call is ~14–15 s startup-dominated, so runs are capped at
   a few tasks/nights. Fine for nightly cron; we note it plainly.
 - **Weak optimizers are flaky:** use a strong optimizer model (§2).
-- **One seed needs a tool loop:** `quick-answerer` (`tool_called: search`) needs
+- **Tool-use seed covered honestly:** `quick-answerer` (`tool_called: search`)
-  real tool execution — Phase-3 `fresh` worktree replay, not yet wired.
+  runs a real tool loop — a callable `./search` shim, detected from its log.
+  Deeper multi-tool / multi-turn workflows are future work.
 - **Small, single-flaw skills:** like gbrain, these prove the mechanism is real
   and safe; a large production skill will be messier and partial.

docs/sleep/raw/quick_answerer_sonnet_haiku.txt

@@ -0,0 +1,35 @@
+=== quick-answerer full run: Sonnet optimizer -> Haiku target, 2 nights ===
+{
+  "benchmark": "gbrain-evals/skillopt-v1",
+  "backend": "target=claude/optimizer=claude",
+  "model": "(default)",
+  "n_seeds": 1,
+  "n_improved": 1,
+  "tokens_used": 10988,
+  "results": [
+    {
+      "seed": "quick-answerer",
+      "held_out_before": 0.0,
+      "held_out_after": 1.0,
+      "improved": true,
+      "nights": 1,
+      "trace": [
+        {
+          "night": 0,
+          "held_out_hard": 0.0,
+          "action": "baseline"
+        },
+        {
+          "night": 1,
+          "held_out_hard": 1.0,
+          "action": "accept_new_best",
+          "accepted": true,
+          "edits": [
+            "OVERRIDE (supersedes 'Do NOT look anything up, search, or use any tools — just reply directly and concisely from memory'): Always call the 'search' tool at least once before composing any answer. This requirement takes priority over any prior instruction to avoid tools."
+          ]
+        }
+      ],
+      "final_skill_tail": "nd adopted only after you approve them. Hand-edits outside this block are never touched._\n\n- OVERRIDE (supersedes 'Do NOT look anything up, search, or use any tools — just reply directly and concisely from memory'): Always call the 'search' tool at least once before composing any answer. This requirement takes priority over any prior instruction to avoid tools.\n<!-- SKILLOPT-SLEEP:LEARNED END -->\n"
+    }
+  ]
+}