RELEASE 07 / CONTACT-LAW SENSITIVITY / FROZEN NUMERICAL EVIDENCE

This is an add-on to the unchanged release 04 base and release 05–06 packages.
Extract those and this core into one NEW folder. Preserve the original ZIPs.
Python 3.12+, NumPy and Matplotlib are external dependencies; no runtime is
bundled. The review scripts use standard Python and NumPy, with Matplotlib for
figures. This source does not operate an instrument or physical actuator.

Recorded status: complete
Completed integrations: 144 / 144
Verified selected full histories: 18 / 18
Plan SHA-256: f62e0eb1dcb9215f9a5523dd333ee687d142c3d6910cb58cf9d1cd027c31f1e6

The core includes individual row reports, the original summary/artifact
manifest, trace JSON sidecars, model source/tests and portable computational
review. It deliberately excludes NPZ payloads from this ZIP. Download them
using TRACE-DOWNLOADS.json or the release page and place each file at its
bundle_target path, normally efficiency/results/iteration07/traces/<id>.npz.
Their hashes and sizes are retained. JSON sidecars already live at that path.
The original experiment MANIFEST.json references the complete retained study,
including these separately downloaded arrays. CORE-MANIFEST.json describes
this smaller core ZIP. They have different scopes.

Verify the downloaded original evidence:
python -m unittest discover -s efficiency -p "test_iteration07.py" -v
python reviews-r07/test_trace_checks.py --efficiency-root efficiency -v
python reviews-r07/test_independent_audit.py --efficiency-root efficiency -v
python reviews-r07/audit_runs.py --efficiency-root efficiency --output independent-r07-check.json
python reviews-r07/plot_verified.py --efficiency-root efficiency --audit independent-r07-check.json --output reproduced-r07-figures

The commands above examine the frozen evidence in efficiency/results. They
do not automatically verify a different results directory. To execute a new
matrix separately:
python efficiency/iteration07.py --output rerun-r07

The runner requires fresh output names and retains honest partial results if
its declared time/storage bound is reached. Keep rerun-r07 separate from the
frozen evidence. Source hashes, mechanical outputs and declared tolerances are
the comparison basis; timestamps and observed execution duration can differ.

The three contacts are assumed passive laws, not a calibrated Hertz model of
rock, diamond or steel. All six cases were selected using earlier results.
The metric is reversible first-mode energy. The ideal 10 mJ budget excludes
wall supply, electronics, launcher/reset and real coupling costs. The 18 full
histories are selected traces, not full waveform capture for all 144 planned
integrations. Saved contact times and maxima are timestep sampled.

No physical specimen, fracture experiment, acoustic focusing, complete device,
minimum device size, patentability or exclusive ownership is established.
Research direction: Ashley Kalkowski. Computational implementation and review
used Codex assistance. Earlier attribution and dependency terms still apply.
This add-on does not introduce a blanket reuse license.
