docs: add root STATUS.md snapshot; close #37 (verified dup of #35)

Create root STATUS.md (at-a-glance: north star, line-status table, open issues,
immediate next action) pointing to the detailed wiki registry. Verified GitHub
#37 is a byte-identical duplicate of closed #35 (principle already binding in
AGENTS.md, commit 27232cb) and closed it. Registry + log updated. Docs-only.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

Author: JohnCCarter

STATUS.md

@@ -0,0 +1,54 @@
+# STATUS — fibengine research
+
+At-a-glance snapshot. **Last swept: 2026-06-25.** Full per-line detail + doc pointers:
+[docs/research_wiki/reference/research-line-status.md](docs/research_wiki/reference/research-line-status.md).
+Current working context: [handoff.md](docs/research_wiki/handoff.md). Why the project exists:
+[north-star.md](docs/research_wiki/north-star.md).
+
+## North star
+
+Teach the machine to **select Fib legs (A→B) like the human** (facit = manual source-fibs). This is
+**step 1** of a staged path: selection → descriptive level-reads → edge/backtest → **Genesis-V2**.
+"No edge claim" = *not yet / not from this sub-study* (a validity gate), **not** a cap.
+
+## Where we are
+
+Step 1 is **not in goal**: the model ranks human legs above chance (AUC ~0.91) but does **not**
+reproduce the selection (AP 0.057 vs 0.83 ceiling) — a modest single-feature (`cleanliness`) lead.
+Controls are done, enrichment closed, the learning-curve is saturated → **more 4h data is low-leverage;
+the lever is a richer notion of what makes a leg "yours"**. Current bet: **top-down MTF nesting**
+(model the same swing decomposed 1M→1W→1D), which needs **new deliberately-nested labels**.
+
+## Line status
+
+| Line | Status |
+|------|--------|
+| Selection-learning (model selects legs like the human) | 🔬 active |
+| Top-down "sniper" MTF nesting (1M→1W→1D) | ⏳ pending — user redraws fibs |
+| BTC source-fib labeling 1M→4H | ✅ complete (1H deferred) |
+| Corpus integrity / dedup / corrections | ✅ complete |
+| Selection-learning controls (w-gap, stage-1, artifact) | ✅ complete |
+| Tooling / ecosystem (#25, #30, #32) | ✅ complete |
+| Fib behaviour (B-1) + context-conditioned | ⛔ closed (null) |
+| Horizontal-structure event study | ⛔ closed (null) |
+| Enrichment (`exclusivity`) | ⛔ closed (worse) |
+| `cleanliness` matched-null crux | ⛔ closed (rejected — A8/A11/A9) |
+| MTF confluence atlas | ⏸ parked (geometry, not edge) |
+| Selection-learning mechanics (snapping / net-path) | ⏸ parked |
+| Fib → Genesis-V2 (phase 0/1/2) | 💤 dormant (docs-only prereg) |
+| Chart regression strategy | ⏸ deferred (#F) |
+
+## Open issues
+
+| # | Title | Note |
+|---|-------|------|
+| **#31** | Investigate fractal-based anchor detection vs human source-fib labels | 🔬 open — now relevant (upstream of the ranking model: does the detector even propose the human's A/B?). |
+
+*(#37 closed 2026-06-25 as a verbatim duplicate of #35 — the principle is in AGENTS.md.)*
+
+## Immediate next action
+
+1. **User:** redraw a small set of **deliberately-nested** fibs on one era (1M→1W→1D) — *tomorrow*.
+2. **Prerequisite (code):** extend `RESOLUTION_TIMEFRAME` in
+   [`same_candle_mtf_resolution.py`](src/fibengine/labeling/same_candle_mtf_resolution.py) to cover
+   1M→1w (monthly anchors currently stay coarse). Continues **#31**.

docs/research_wiki/log.md

@@ -25,9 +25,11 @@ behaviour/horizontal nulls, MTF-confluence, and Genesis-V2 docs; no status overv
 [reference/research-line-status.md](reference/research-line-status.md) — one row per LINE with status
 (active / pending-input / complete / closed / parked / dormant / superseded) + open issues; linked it
 from `index.md` (Wiki Operations + Reviews) and added a status banner to `reviews/README.md`. Issue
-state confirmed via `gh`: only **#31** (fractal anchor detection — now relevant) and **#37** (agent
-principle, likely a stale duplicate of closed #35) are open; flagged #37 for the human to close.
-Docs-only, no code, no claim change.
+state confirmed via `gh`: only **#31** (fractal anchor detection — now relevant, user resumes with new
+fibs next day) and **#37** were open. **#37 verified a verbatim duplicate of closed #35** (byte-identical
+body; #35 closed with "Done — principle added to AGENTS.md commit 27232cb"; principle present at
+AGENTS.md) → **closed #37** as a stale tracking dup. Also created root **[/STATUS.md](../../STATUS.md)**
+(at-a-glance snapshot pointing to the registry). Docs-only, no code, no claim change.
 
 ## [2026-06-25] review | Top-down MTF nesting premise check → facit does NOT nest (different eras)
 

docs/research_wiki/reference/research-line-status.md

@@ -1,9 +1,10 @@
 # Research-line status registry
 
 One row per research **line** (not per doc): is it open, in-progress, closed, parked, dormant, or
-deprecated. Source of truth for "where does X stand". Per-result detail lives in
-[reviews/](../reviews/); current focus in [handoff.md](../handoff.md); do-not-rerun nulls in
-[closed-questions.md](closed-questions.md). **Last swept: 2026-06-25.**
+deprecated. Source of truth for "where does X stand". Root-level snapshot:
+[/STATUS.md](../../../STATUS.md). Per-result detail lives in [reviews/](../reviews/); current focus in
+[handoff.md](../handoff.md); do-not-rerun nulls in [closed-questions.md](closed-questions.md).
+**Last swept: 2026-06-25.**
 
 Status legend: 🔬 **active** (work in flight) · ⏳ **pending-input** (waiting on a human action) ·
 ✅ **complete** (delivered, nothing pending) · ⛔ **closed** (null/answered, do not re-run) ·
@@ -48,8 +49,10 @@ Status legend: 🔬 **active** (work in flight) · ⏳ **pending-input** (waitin
 
 | # | Title | State | Note |
 |---|-------|-------|------|
-| **#31** | Investigate fractal-based anchor detection against human source-fib labels | 🔬 **open — now relevant** | Upstream of the ranking model (does the detector even propose the human's A/B?). Connects directly to the MTF-nesting redirect. |
-| **#37** | Add agent principle: Intent Alignment over Plan Inertia | ⚠️ **open — likely duplicate** | Same title as **closed #35**; #36 (commands) closed. Probably a stale tracking dup → candidate to close (confirm with human). |
+| **#31** | Investigate fractal-based anchor detection against human source-fib labels | 🔬 **open — now relevant** | Upstream of the ranking model (does the detector even propose the human's A/B?). Connects directly to the MTF-nesting redirect. **User resumes with new fibs (next day).** |
+
+*(#37 **closed 2026-06-25** — verified verbatim duplicate of #35; the principle is a binding agent
+principle in [AGENTS.md](../../../AGENTS.md) (commit 27232cb).)*
 
 > 🗄 **Superseded:** pre-BTC-monthly-reset descriptive reads are archived under
 > `archive/research_superseded/2026-06-08_pre_btc_monthly_reset/` — not current evidence