Scope stress correctly; add efficiency, real-data deployment, and agent/golden-dataset sections

[mdenolle]

Manuscript update prompted by testing codameter at larger scale and standing
up a real-data companion pipeline. Plan reviewed and approved before any file
was touched (see AI_LOG.md's 2026-08-08 entry for the full record).

Sec. Stress -- scope correction, not deletion

The problem was never the word "stress" -- it was that the section presented
a method for converting dv/v to stress and promised more of it ("companion
framework, Denolle in prep."). That method (the acoustoelastic equation, its
best-practice table) and the promise are cut. The section is now a short,
explicit scope statement: this paper stops at the depth-resolved
velocity-change posterior and its covariance. Motivating "stress" mentions
elsewhere (abstract, Bayesian section, depth section) are left alone -- they
say the covariance is useful input to such work, not that this paper performs
it. Discussion/Conclusions reworded to match.

New section: real-data deployment (architecture only)

"Toward deployment: a real-data retrospective pipeline" describes the
noisepy-dvv-cloud architecture (NoisePy on real CI.LJR waveforms from
public S3, AWS Batch Fargate Spot, codameter for dv/v estimation) and its
validation protocol against the published Clements & Denolle (2023) result.
The real pipeline is an early scaffold with no results yet, so this is
architecture-only with an explicit <!-- TODO --> placeholder for the
dv/v(t) figure once that run completes -- no results are described ahead of
having them.

Efficiency paragraph, numbers re-measured

Added a short paragraph to Sec. multiverse on the recent vectorized
fast-path speedups. The commit/CHANGELOG numbers (~2.3x/~4.9x/~4x) turned out
optimistic when re-measured independently in this session (repeated runs:
~2x/~3x/~3x) -- cited the re-measured numbers and corrected CHANGELOG.md to
match, since leaving it wrong would be inconsistent with this paper's own
thesis about honest, verified reporting.

Other

• Introduction's agentic/skill-layer sentence reworded to match what
  codameter.frugalmind/golden.py actually do (RMS-recovery scoring
  against seeded synthetic golden cases), not "evaluation against base
  models."
• Data Availability: fixed a stale manuscript filename.
• New AI_LOG.md.

Test plan

• 230/230 tests pass (1 unrelated skip; no Python source touched)
• paper/build.py builds cleanly -- only the 2 pre-existing, already
  tracked (issue Dangling citation keys @lobkis01 and @poupinet84 have no matching bib entry #30) dangling citations, no new undefined \ref{}/\citep{}
• Confirmed every remaining "stress"/"pressure" mention is contextual,
  not methodological
• No dangling references to the removed eq:acoustoelastic/tab:bp-stress

[Copilot]

Pull request overview

Updates the marine manuscript to reflect recent scaling/deployment experience with codameter, clarifying scope around “stress” (as downstream work) while adding manuscript sections on efficiency, real-data deployment architecture, and the agent/golden-dataset evaluation setup.

Changes:

• Reframes the “Stress” section as an explicit scope boundary (no dv/v-to-stress conversion method promised or described).
• Adds an efficiency paragraph describing measured vectorized fast paths and adds a new “Toward deployment” section describing a real-data retrospective pipeline architecture.
• Updates manuscript wording around the agent/golden-dataset scoring, fixes Data Availability manuscript filename, and corrects changelog speedup figures; adds a new AI use log.

Reviewed changes

Copilot reviewed 4 out of 4 changed files in this pull request and generated no comments.

File	Description
paper/manuscript_marine.tex	Manuscript LaTeX updates: agent/golden wording, efficiency paragraph, deployment section, stress scope rewrite, data availability filename.
paper/manuscript_marine.qmd	Source manuscript updates matching the .tex changes, including an explicit TODO placeholder for the future real-data figure.
CHANGELOG.md	Adjusts documented speedup figures and adds benchmarking caveats.
AI_LOG.md	Adds an append-only AI-assisted work-session log entry for this manuscript update.

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.