NetPhysics adds a measured mammal benchmark and publishes an unfavorable primary result

Project-issued release / 20 October 2026

FOR IMMEDIATE RELEASE

NetPhysics, Christopher Hill’s independent research project, has prepared release 0.5.0 with a retrospective analysis of the published PanTHERIA mammal dataset. The release includes source provenance, inclusion and exclusion records, reproducible code, held-out predictions and sensitivity analyses.

Among 573 eligible species, a continuous scaling hinge fixed at 10 kg did not improve the primary prediction on held-out families. Its test RMSE was 0.200464 log₁₀ W, compared with 0.198095 for a single-line baseline. Alternative splits and models are reported, including a modestly different result under five-fold family cross-validation.

The release also corrects the scope of a proposed open-pipe acoustic benchmark and the measurement design of an elastomer frequency/recovery pilot. These physical experiments have not yet been performed. The numerical checks and public-data analysis do not establish a new physical law or validate a universal NetPhysics model.

Evidence and reporting resources

Measured benchmark and full result · All workstreams · Release record · Research source archive

The source dataset is attributed to Jones et al. (2009), PanTHERIA, Ecology, DOI 10.1890/08-1494.1. These are external measurements compiled at species level. No new NetPhysics measurements, independent review, repository DOI, institutional partnership or external media coverage is claimed.

About NetPhysics

NetPhysics is an independent, AI-assisted research and software project examining geometry, memory and scaling through explicit assumptions, source-linked models and reproducible comparisons. Founder: Christopher Hill.

Media contact: Christopher Hill — chris@maxwellx.com.

Download release text · Updated fact sheet

This is a prepared project-issued announcement. It has not been sent to external outlets or accepted for independent coverage.