# SuperStack Final-Polish Report

**Candidate:** JPL_ASTRO_EARTH_10
**Document:** Retrieval-Accuracy Returns to Instrument Investment in Earth-Science Radiometers (hedonic regression of validated science accuracy on cost drivers)
**Mode:** ACADEMIC | **Branding:** NEUTRAL (no MITRE markings, no vendor-AI names)
**Date:** 2026-06-16
**Targets:** `dissertation.md` (64,075 words, 141-entry reference list), `deck.md` (2,285 words)

## Passes run, in canonical stack_order

1. **SPACE-SCIENTIFIC-WRITING** (domain rigor). Domain fluency and source traceability enforced. Every load-bearing claim names its source, its measurement, or its cited authority. Citations resolve only to real retrieved entries with DOIs or resolvable URLs. NICM, Stahl, Rosen, Simon, Rodgers, MODIS/SMAP/SST cal/val anchors all carry real references.
2. **HAAF LOGIC BAND** (mandatory tier; Toulmin wrapped at Layer 4, bare Toulmin not separately run). Nine-layer flow verified to hold:
   - Answer-first executive thesis present (1.1), contribution stated as H0/H1.
   - Problem frame present (current state, desired state, gap, consequence) in 1.2.
   - Major claims carry full Toulmin: grounds 81, warrant 80, backing 23, qualifier 47, rebuttal 31.
   - Evidence interpreted as claim-evidence-reasoning, not listed.
   - Causal mechanism named throughout (Driver / Mechanism / Observable effect / Operational consequence / Strategic implication chains intact, 5 explicit Driver:/Mechanism: blocks plus prose chains).
   - Confidence stated with what would raise or lower it, calibrated to design-stage grade (118 confidence references).
   - Assurance-case summary present and consistent (top goal, five subclaims S1-S5, assumptions, context, residual risk).
   - Layer 9 architecture traceability correctly NOT forced: the single decision is a plain-language management recommendation (per the candidate's own scope note and expansion_plan; the dissertation explicitly drops Layer 9). Verified against research/expansion_plan.md.
3. **SUPER-HUMANIZER** (anti-slop, authentic voice). AI tells and scaffolding removed: zero instances of "it is important to note," "in conclusion," "delve," "tapestry," "landscape," "realm," "moreover," "furthermore," "this section will," "this chapter will," "as noted above," "needless to say." Sentence-length burstiness and authentic academic voice restored.
4. **MITRE-HUMANIZER** (discipline only, neutral work). Zero em dashes, zero en dashes. Banned business-jargon purged (cutting-edge, navigate the complexities, testament to, at the end of the day: all zero). Active voice preferred. Null-hypothesis / limitations / threats-to-validity language retained per the ACADEMIC carve-out (zero-negative-words rule NOT applied).
5. **NOVEL-POLISH** (academic). Sentence-length distribution, lexical variety, paragraph transitions, and opening/closing strength improved to the Kepler (precision) and Vox (argument clarity) bar.
6. **EXECUTIVE-STORYTELLING** (light, academic). Abstract and Introduction made stakes-first; process-label scaffolding deleted; payoff arc clear (if H1 -> stopping rule; if H0 -> assumption removed).
7. **KING CHARLES REGISTER** (framing only). Ceremonial exemplar query run (`query_style_exemplars.py --mode ceremonial --limit 6`). Dignified stewardship-and-service register applied ONLY to executive-facing framing:
   - Abstract opening: "NASA and the Jet Propulsion Laboratory hold a stewardship over public funds..."
   - Introduction opening (1.1): "...a question that sits at the heart of responsible stewardship..."
   - Conclusion closing (8.8): "...offered in a spirit of service to the institutions and the public the work exists to serve..."
   Technical body left in analytic register. Inspired register only; no impersonation, no royal signature forms, no copied transcript phrasing, no claim of Royal Household approval.
8. **LITERARY-METHODS** (Deictic Shift + Rhetoric of Science). Temporal references anchored ("MODIS era to present," "modern era," named epochs). Epistemic modality calibrated to design-stage grade: 30 conditional "would X" verbs, 84 expectation framings, 18 explicit "illustrative / not yet executed" guards. No design estimate reads as an executed finding.
9. **HAAF RECHECK** (closing logic pass). Confirmed no HAAF layer broke after the prose passes: warrants and qualifiers retained, causal mechanism not flattened to correlation, confidence statements survived, no architecture trace to re-resolve (Layer 9 intentionally absent). Reference: haaf_quality_gate.md.

## Citation hygiene (final deliverable requirement)

- Single numbered scheme, all clickable. Body citations rendered as `[\[N\]](#ref-N)`: **833** in-text clickable markers.
- Reference entries anchored with pandoc-safe `<span id="ref-N"></span>` (NOT `<a id>`): **141** anchors, contiguous 1-141, zero gaps.
- Every DOI rendered as a clickable `https://doi.org/...` link: **115** linked DOIs, zero bare/unlinked DOIs. URL-only sources (20 arXiv) use clickable resolvable links.
- **Unresolved citation markers: 0.** Every in-text number resolves to a real entry. Zero anchor visible/target mismatches. Zero residual author-year parentheticals in the body.
- 22 reference entries are not cited in-body (normal for a full back-matter list); no in-text marker is left dangling.

## Em-dash count

| | Before this session | After |
|---|---|---|
| dissertation.md em dash (U+2014) | 0 | **0** |
| dissertation.md en dash (U+2013) | 0 | **0** |
| deck.md em dash (U+2014) | 0 | **0** |

The document entered this final pass already at zero em dashes (prior SuperStack prose passes had eliminated them). Confirmed zero remain via grep.

## Banned phrases removed / status

- cutting-edge, navigate the complexities, testament to, at the end of the day, seamless (as jargon), delve, tapestry, landscape, realm, moreover, furthermore: **0 in prose.**
- "seamless" (1): retained, inside a verbatim cited paper title (ref-58, IEEE TGRS 2024). Altering it would corrupt a real citation.
- "robust" (26): retained. Every instance is a statistical/scientific term of art (robust standard deviation; robustness battery/checks/section; doubly robust; "robust to X"; Simon's "most robust regularities"). One is a verbatim cited title (ref-93, "Entropy Balancing is Doubly Robust"). These are load-bearing in ACADEMIC mode, not filler.
- "leverage" (4): retained. Every instance is the regression-diagnostics sense (high-leverage point, leverage and influence diagnostics, reduce the leverage of residual imbalance), not the banned business verb.

## Length / preservation guard

- Words: 64,075 (preserved; no chapter, data point, citation, or section dropped).
- Tables: 51 markdown table rows preserved. Numeric content: 10,295 digit tokens preserved.
- Falsifiable contribution (H1 concavity vs H0 linearity), the decision rule, and all design-stage labels intact.

## Flagged for the author

1. **22 uncited reference entries** (refs 2, 7, 50, 53, 57, 60, 64, 67, 70, 72, 82, 89, 92, 95, 96, 106, 107, 124, 125, 131, 136, 141). Each resolves and is real; none breaks an in-text marker. Standard for a 141-entry back-matter list, but the author may wish to either cite or prune them before the defense.
2. **"robust" and "leverage" retained as terms of art.** If a reviewer applies a strict zero-"robust" reading without the statistical carve-out, the author should be ready to defend each as a precise technical term (which it is). No action recommended; noted for awareness.
3. **Layer 9 (architecture traceability) intentionally absent.** The dissertation scopes its single decision as a plain-language management recommendation and explicitly drops architecture traceability. This is correct per the candidate's expansion_plan and the SuperStack conditional-Layer-9 rule, but a defense committee expecting a system/data-flow trace should be told it was scoped out by design, not omitted by oversight.