# SuperStack Final-Polish Report: JPL_ASTRO_EARTH_08

**Artifact:** Cost-Overrun Hazards in Earth-Observing Missions: A Competing-Risks Model Separating Instrument-Driven from Launch-Driven Slip
**Mode:** ACADEMIC. **Branding:** NEUTRAL (no MITRE markings, no vendor-AI names).
**Files revised in place:** `dissertation.md`, `deck.md`.
**Date of pass:** 2026-06-16.


## Passes run (canonical stack order)

1. **SPACE-SCIENTIFIC-WRITING.** Confirmed domain fluency and source traceability. Every in-text citation now resolves to a real corpus entry with a DOI or resolvable URL. No fabricated sources, numbers, or results were introduced. The design-stage guardrail (every numerical pattern illustrative, none executed) was preserved throughout.
2. **HAAF LOGIC BAND (Toulmin wrapped at Layer 4).** Audited the nine-layer flow across the whole dissertation and confirmed it holds: answer-first executive thesis (the contribution stated as H0/H1); problem frame (current state, desired state, gap, consequence); major claims carrying full Toulmin (claim, grounds, warrant, backing, qualifier, rebuttal); evidence interpreted as claim-evidence-reasoning rather than listed; every causal claim naming a mechanism (driver to mechanism to effect to consequence to implication); confidence stated with what would change it, calibrated to the design-stage grade; the assurance-case summary (top goal G0, five subclaims, assumptions, context, residual risk) present and consistent. No HAAF element was missing; none was fabricated to fill a gap.
3. **SUPER-HUMANIZER.** Scanned for AI tells and scaffolding ("it is important to note", "in conclusion", "delve", "tapestry", "landscape", "realm", "moreover" overuse, "this section will", etc.). The document was already nearly free of these. Removed the few soft tells found (see below). Sentence-length variation and authentic academic voice were already strong and were preserved.
4. **MITRE-HUMANIZER (discipline only, neutral non-MITRE work).** Em dashes rewritten to zero (the document already had zero; verified and held). Banned phrases purged (none survived). Discipline rule honored: the academic null hypotheses, limitations, and threats-to-validity were left stated plainly (the zero-negative-words rule was NOT applied, per directive).
5. **NOVEL-POLISH (academic).** Verified sentence-length distribution, lexical variety, and paragraph transitions; the opening (Abstract, Chapter 1.0) and closing (Chapter 8.5) were strengthened for register (see King Charles pass). Held to the precision and argument-clarity bar.
6. **EXECUTIVE-STORYTELLING (light, academic).** The Abstract and Introduction are stakes-first; no process-label scaffolding ("this section will") was present. The payoff arc (reserve-allocation decision) is clear from the first paragraph of each.
7. **KING CHARLES REGISTER (framing only).** Ran the ceremonial exemplar query (`query_style_exemplars.py --mode ceremonial --limit 6`); inferred a dignified, measured, stewardship-and-service register. Applied it ONLY to the executive-facing framing: the first sentence of the Abstract, the opening framing sentence of the Introduction (1.0), and the closing sentence of the Conclusion (8.5), plus the deck title and closing slide. The technical body remains in its analytic register. Guardrails observed: inspired register, no impersonation, no royal signature forms, no copied transcript phrasing, no claim of Royal Household approval.
8. **LITERARY-METHODS.** Deictic-shift cold-reader audit: temporal references are anchored (Key Decision Point B, "roughly 1990 to present", months-from-KDP-B); entities are introduced before pronominal reference; "this/that/it" referents are unambiguous. Rhetoric-of-Science: epistemic modality is correctly conditional or future throughout the expected-results material ("would rise", "the dissertation expects to find", "to be estimated"); no illustrative pattern is stated as an executed finding.
9. **HAAF RECHECK (closing logic pass).** After the prose, voice, and polish passes, re-verified that no HAAF layer broke: answer-first opening intact; both hypotheses verbatim; the causal mechanism chain intact and not flattened to a correlation; the calibrated confidence statements intact; the assurance case (G0, five subclaims, residual risk) intact; the Layer-9-out-of-scope statements in Chapters 1 and 2 intact. Reference: `haaf_quality_gate.md`. Gate is clean.


## Em-dash count

| File | Em dashes before | Em dashes after |
|------|------------------|------------------|
| dissertation.md | 0 | 0 |
| deck.md | 0 | 0 |

The document already contained zero em dashes on entry; the count was verified and held at zero. (Three en dashes, U+2013, remain in `dissertation.md`; all three are inside verbatim cited paper titles such as "Fine–Gray" and are preserved for citation fidelity. The em-dash requirement is fully satisfied.)


## Banned phrases removed

- No instances of the banned set (robust as marketing, cutting-edge, leveraging, seamless, navigate the complexities, testament to, at the end of the day) survived in either file. The 26 occurrences of "robust" are all the load-bearing statistical sense ("robust to either bias alone", "robustness battery", "doubly-robust"), which is correct technical usage and was retained.

## Soft tells removed (super-humanizer / novel-polish)

- "highest-leverage way to control cost" to "most direct way to control cost" (Ch 3.1).
- "Their finding is nuanced and directly relevant" to "Their finding is qualified and directly relevant" (Ch 3.5).
- "the pivotal observation of the chapter" to "the central observation of the chapter" (Ch 3.7).
- "the pivotal sub-construct" to "the decisive sub-construct" (Ch 4.4).

## King Charles framing added (executive-facing only)

- Abstract: one stewardship-of-public-resources framing sentence before the existing technical opening.
- Introduction 1.0: one measured duty-and-stewardship sentence before the answer-first contribution.
- Conclusion 8.5: one closing sentence of service and faith-with-the-public-purpose.
- Deck: a stewardship subtitle on slide 1 and a service note on the closing slide.


## Citation hygiene (the largest structural change)

The dissertation entered with **two incompatible numbering schemes** and a real integrity defect:

1. Chapters 1, 2, 3, 5, 6, 7 each carried their own chapter-local `[N]` numbering tied to a per-chapter "Chapter N References" block. Chapter 3's local block additionally used `<span id="ref-N">` anchors that **collided** with the Part I master anchors (duplicate HTML ids 1 through 44).
2. The Part I master list (149 entries, alphabetical) used a different numbering, so a chapter `[1]` and the master `[1]` pointed at different sources.
3. **38 raw corpus keys** had leaked into the body text of Chapters 3, 4, 6, 7 (for example `[lieber2016]`, `[habibagahi2015; habibagahi2019]`), plus two malformed `[14, governance]` markers and one stray `[10] (reserved)` entry.

Fix applied (no source was repointed; only style was normalized and links added):

- Built a DOI/URL/title-matched map from every chapter-local number and every corpus key to the canonical 149-entry Part I master scheme.
- Rewrote every body citation to the master number, rendered as a clickable `[\[N\]](#ref-N)` marker (escaped brackets so pandoc preserves them).
- Replaced all 38 leaked raw corpus keys (including 17 multi-key tokens) and both `[14, governance]` markers (which point to Zancan, master 142) with the correct linked master cites.
- Replaced the six redundant chapter-local reference blocks with a one-line pointer to the consolidated Part I list (the master list already contains every cited source). This removed the duplicate `ref-N` anchors.
- Confirmed every reference-list entry carries a `<span id="ref-N"></span>` anchor and every DOI is a clickable `https://doi.org/...` link (133 DOI links; 16 URL-only entries are NTRS/OpenAlex/arXiv/NIST records with no assigned DOI, each a clickable resolvable URL).

Verification of the final state:
- 149 anchors, **zero duplicate anchor ids**.
- **461 linked body citations, zero mismatched, zero unresolved.**
- **Zero remaining plain or raw-key citation tokens** anywhere in the body.


## Length

- Body prose (reference blocks excluded) grew from 58,341 to 58,467 words (+126, from the King Charles framing). No chapter, subsection, table, data point, or appendix was dropped; all 83 original section headers are preserved.
- Total file word count fell from 66,429 to 62,445. The entire reduction is the consolidation of the six duplicated chapter-reference blocks into the single master list, which is the citation-unification the task required. No substantive content was lost.


## Flagged for the author

1. **En dashes in cited titles (cosmetic).** Three reference titles retain the U+2013 en dash ("Fine–Gray" in entries 23 and 144; "competing risks – Cause-specific" in entry 128). These are verbatim from the published titles. Elsewhere the document writes "Fine-Gray" with a hyphen. Left as-is for citation fidelity; change only if house style requires hyphenation of title strings.
2. **Authoring date.** Title block and deck retain the 2026-06-15 authoring date. Left unchanged (it records when the dissertation was written), not updated to the pass date.
3. **Launch-side evidence thinness (substantive, already disclosed).** The heritage-arm launch-dominance claim remains the most evidence-thin half of the argument, which the document itself flags as residual risk R4 and as the half most likely to return an inconclusive verdict. This is honest and correctly calibrated; no change made, noted for the author's awareness when the cohort is executed.
4. **Two corpus entries with empty author fields in the reference list** (entries 36 and 45, the Acta Astronautica flight-spares and dual-telescope-radiometer papers) render with a leading title rather than an author. This matches the corpus metadata and the IEEE-derived style used for corporate/anonymous-author works; verify against the source if author attribution is later recovered.