diff --git a/docs/patent/SSM-001_CLAIM_MAP.md b/docs/patent/SSM-001_CLAIM_MAP.md
new file mode 100644
index 00000000..1dcd555f
--- /dev/null
+++ b/docs/patent/SSM-001_CLAIM_MAP.md
@@ -0,0 +1,109 @@
+# SSM-001 — claim-support map, toward the non-provisional
+
+Prepared 2026-10-28 at Captain Bob's request, from the provisional **as filed**:
+`docs/patent/final_provisional/ssm_RAJ_v1_patent_application.pdf` (43 pages, built 2025-12-10;
+its text is identical to `docs/patent/ssm_RAJ_v1_patent_application.pdf`). Filing date
+2025-12-16, so the non-provisional (or PCT) claiming it is due by **2026-12-16**.
+
+Drafting aid, not legal advice. Page numbers are the filed PDF's.
+
+**The rule this map applies:** a claim in the non-provisional keeps SSM-001's priority date only if
+SSM-001 describes what it claims. Nothing can be added to the specification after the
+non-provisional is filed (no new matter). Material may be *left out* of the non-provisional
+without harm to priority.
+
+## 1. The 20 claims, and where SSM-001 supports them
+
+Strength: **Solid** = described plainly, more than once; **Adequate** = described, briefly;
+**Thin** = only implied, or by a single list item.
+
+| Claim | What it claims | Support in SSM-001 | Strength |
+|---|---|---|---|
+| 1 (system) | state vector (heat, entropy, stability), feedback loops, ≥2 execution modes, a supervisory selector with bounded, non-oscillatory transitions | §3 p.9; §5.1 p.28; §5.6 p.35; §5.7 pp.35-36; §5.8 pp.36-37; §5.9 p.37 | Solid |
+| 2 (method) | compute state vector; update loops; classify stable/temporal/volatile/transitional; select mode; adjust parameters | §5.6, §5.7, §5.10 pp.37-38, §5.9 | Solid |
+| 3 (shape invariance) | measure entropy, variance, heat; select modes; regulate transitions so variance stays below a threshold across waveform families | §5.11 p.38; §1 p.4; FIGS. 7-9 pp.17-19 | Adequate (the "threshold" is shown by figures and CV, not a stated value) |
+| 4 | state vector adds pipeline pressure, lookup latency, cache hit rate, or decay | §5.6 p.35 | Solid |
+| 5 | a loop modifies the heat decay rate from temporal characteristics | §5.7 L3 p.36; §5.8 Mode 1 p.37 | Solid |
+| 6 | a loop does statistical inference for predictive indicators used in mode selection | §5.7 L2, L6 p.36 | Solid |
+| 7 | hysteresis thresholds against oscillation | §5.9 p.37; §3 p.9 | Solid |
+| 8 | modes include baseline, temporal, inference-driven, fully adaptive | §5.8 Modes 0-3 pp.36-37 | Solid |
+| 9 | converges when variance of recent timings falls below a threshold | §5.12 p.38 | Adequate |
+| 10 | classification evaluates an entropy window over heat | §5.6 "Entropy Window" p.35 | Solid |
+| 11 | heat rises on invocation and decays by a selected decay function | §5.6 "Execution Heat", "Decay Rate" p.35 | Solid |
+| 12 | mode selection compares the state vector to pre-validated profiles | §5.8 p.37 ("discovered through design-space exploration or factorial experimentation"); §5.5.2 p.31 | Solid |
+| 13 | waveform families: sinusoidal, square, triangular, burst, mixed | §5.11 p.38; §1 p.4 | Solid |
+| 14 | consistency measured by coefficient of variation | §5.6 "Stability Score" p.35; FIG. 9 p.19 | Solid |
+| 15 | loops modify lookup strategies | §5.7 L4 p.36; §1 p.4 | Solid |
+| 16 | transitions wait until a confidence value exceeds a threshold | §5.9 p.37 ("confidence scoring", "multiple independent criteria") | Adequate |
+| 17 | adjusting caching, prefetching, or dictionary traversal | §5.7 L4 p.36; §5.6 p.35 | Solid |
+| 18 | classification uses rolling measurements or sliding windows | §5.6 "Entropy Window" p.35; §5.2 "rolling window of truth" p.29 | Solid |
+| 19 | stack-based, threaded-code, JIT, embedded controller, or microkernel | §5.13 p.39; §5.14 p.39; §1 p.4 | Solid |
+| 20 | state vector kept per execution thread | §5.13 "a multi-threaded execution engine supporting distributed loads" p.39 | **Thin** — per-thread state is never stated |
+
+**Overall:** the 20 claims rest on the architecture sections (§§5.1, 5.5.1-5.5.2, 5.6-5.14), which
+are consistent and well described. None of the 20 recites the K-statistic, ω₀, or the Boltzmann
+model, which is fortunate given §2 below.
+
+## 2. Findings in the specification
+
+1. **The K-statistic's definition contradicts the specification's own data.** §5.2 (p.28, eq. 1) and
+   §5.5.3 (p.31) define K = W / (DoF + 1), with W in bytes (512-65,536) and DoF the number of
+   active loops (7). That gives K from 64 to 8,192, not "K = 1.000000 ± 0.000000 ... across window
+   sizes ranging from 512 to 65,536 bytes" (pp.29, 31). Likewise W* = DoF + 1 = 8 bytes as the
+   "optimal window". The figures use other forms: FIG. 13 (p.23) plots "K = 256/W" with K from 0
+   to 0.5; FIGS. 16-17 (pp.24-25) show K ≈ 0.016-0.042 and a log axis 0.003-1.0. The cover letter
+   (`USPTO_COVER_LETTER.md`) writes K = Λ×(DoF+1)/W. An examiner would read any claim to K as
+   indefinite or unsupported (35 U.S.C. 112). Correcting the formula in the non-provisional
+   risks new matter, because SSM-001 never states which formula is meant.
+2. **Two names for one invention.** The title, abstract and body say *Steady State Machine*; §4's
+   figure descriptions (pp.11-26) say *Solid State Machine* nine times.
+3. **Claims 10, 11 and 12 depend on "the system of Claim 2"**, but Claim 2 is a method: they should
+   read "The method of Claim 2".
+4. **The drawings will not pass as filed.** Non-provisional drawings must meet 37 CFR 1.84 (black
+   lines, legible, numbered consistently). Pages 22-26 carry plots whose internal titles number
+   them "Figure 1"-"Figure 7" under captions FIG. 12-18; several are dense multi-series charts; and
+   FIG. 19 "System Architecture" (p.27) has a caption but no description in §4.
+5. **Strong scientific language invites scrutiny without adding claim support:** "the first exact
+   conservation law discovered in adaptive computational systems" (p.29), "quantum-analog ...
+   probabilistic tunneling" (p.24), the golden-ratio penalty (p.24), p < 10⁻²⁰⁰ (p.32). Nothing
+   claimed depends on them, and finding 1 undercuts the conservation-law statement.
+6. **Result invariance is not disclosed.** SSM-001 says the architecture produces "deterministic,
+   repeatable steady-state behavior" (abstract, p.43) and "0% algorithmic variance validated
+   across 38,400 replicate runs" (§5.4 item 5, p.30), but nowhere states that adaptation leaves
+   *what a program computes* unchanged, and nothing about a machine-checked proof of it. That
+   property, the centre of the regulated-industry case (2026-10-28), is **not** available at the
+   2025-12-16 date. It belongs to the second provisional (outline, central element 3), where it
+   takes that filing's date.
+7. **Jacquard details are not disclosed.** L8 is named and described functionally (§5.9), but the
+   128-state, 7-bit loop-gate structure in the code is absent, so it cannot be claimed at the
+   2025 date.
+
+## 3. Recommendations for the non-provisional (for a practitioner to confirm)
+
+1. **Keep the architecture, claims 1-19, as the backbone**, with claim 20 either dropped or
+   supported only to the extent §5.13 allows.
+2. **Fix the formal defects:** dependency wording in claims 10-12; one name (Steady State Machine)
+   throughout; drawings redrawn to 37 CFR 1.84, numbered consistently, FIG. 19 described.
+3. **Do not claim the K-statistic, ω₀, or the Boltzmann model.** Either leave §§5.2-5.5.5 out of the
+   non-provisional (omission does not affect priority) or carry them only as observed results,
+   without the inconsistent formula. Leaving them out also removes finding 5's exposure.
+4. **Add dependent claims SSM-001 already supports**, which narrow toward what is distinctive:
+   - modes derived from a factorial design-of-experiments over loop configurations (§5.8 p.37,
+     FIGS. 1-5 pp.11-15);
+   - mode transitions requiring several independent criteria (§5.9 p.37);
+   - L7 as a baseline stabiliser when the other loops disagree (§5.7 p.36);
+   - workload families including "mixed" (§1 p.4) as well as the four of claim 2;
+   - an implementation in hardware or firmware (§5.14 p.39).
+5. **Expect obviousness arguments against claim 1** from adaptive runtimes (tiered JIT compilation,
+   GC ergonomics), autonomic computing (MAPE-K) and control-theoretic computing. The dependent
+   claims in 4, and claim 3's shape invariance with its measured CV evidence, are where the case
+   should be made.
+6. **Result invariance, the formal proofs, the template method and today's other ideas go in the
+   second provisional**, not here.
+
+## 4. Before filing
+
+- One paid hour with a registered practitioner on the claims and on finding 1, if at all possible.
+- The USPTO Pro Se Assistance Center for the filing mechanics, fees, and micro-entity status.
+- Confirm the dates for this application with them; the 12-month rule (35 U.S.C. 119(e)) is the
+  general one.