# NetPhysics measured benchmark and protocol packet — 0.5.0

Prepared 20 October 2026. Author-run analysis of external measurements; not peer review,
new measurements, external preregistration, or a claim of new physics.

## Reproduce

Python 3.12.14 was used. From this directory:

```sh
python -m pip install -r requirements.txt
python metabolic.py
python protocol_tools.py
python check_invariants.py
python make_figures.py
```

The first command installs dependencies. The analysis itself requires no network.
The default metabolic run uses 2,000 family bootstrap samples and the original WR05
table included here. `metabolic.py --output /path/to/new/results` permits comparison
without overwriting the supplied reference outputs. CSV/JSON numerical outputs are
the verification targets; SVG metadata timestamps may differ across rendering runs.
Do not lower the bootstrap count and label the result as the release run.

Read `METABOLIC-REPORT.md`, `provenance.json`, `analysis-plan.json`, and every sensitivity
in `metabolic-output/results.json`. The plan file was saved locally before fitting;
this is retrospective public-data analysis, not a registered confirmatory study.
`design-output/` contains calculated design examples, never collected observations.
The blank CSVs are intake templates; acquire and validate real data before fitting.

Source code is MIT under LICENSE-MIT.txt; original explanatory material and derived
outputs are CC BY 4.0 under LICENSE-SCOPE.txt. PanTHERIA remains its original CC0
dataset, attributed to Jones et al. Full source papers are linked, not redistributed.
Review and publication instructions are in PUBLICATION-AND-REVIEW.md.