Pointer Architecture: A Constrained Information-Geometric Model of Galactic Rotation Curves

Abstract

We compare three information-geometric models of galactic rotation on the 171-galaxy SPARC sample: unconstrained pointer, constrained pointer (with a physical prior on halo extent), and a standard NFW baseline. The constrained pointer wins AIC on 60 of 171 galaxies and shows a positive partial correlation between halo extent and inferred information mass (ρ = +0.19, q_FDR = 0.04). Four of six residual-structure features correlate with galactic age at FDR-controlled significance. A full reproducibility pipeline — code, data, and fit diagnostics — is released alongside the preprint.

TL;DR

• ·3-model head-to-head on 171 SPARC galaxies
• ·Constrained Pointer wins AIC on 60/171
• ·Partial correlation on halo extent: ρ = +0.19, q_FDR = 0.04
• ·4 of 6 residual-structure features correlate with galactic age
• ·Full reproducibility pipeline released (code + data + fit diagnostics)

Predictions & falsifiers

1. Constrained pointer fit quality improves systematically with galactic age.
2. Residual-structure features at large radii should track halo concentration, not stellar mass.
3. Low-surface-brightness galaxies should show larger pointer-vs-NFW AIC gaps.
4. Independent samples (LITTLE THINGS, THINGS) should reproduce the ρ direction on halo extent.
5. Falsifier: if the ρ sign flips on an independent age-controlled resample, the model fails.
6. Falsifier: if AIC-preferred galaxies correlate with survey artefacts rather than age, the result is observational.
7. Falsifier: if reproducibility pipeline yields materially different fits under reasonable re-parameterisation, results are over-fit.

Cite

Savchenko, M. (2026). Pointer Architecture: A Constrained Information-Geometric Model of Galactic Rotation Curves. Preprint.

FAQ

What is Pointer Architecture?

Pointer Architecture is a constrained information-geometric model of galactic rotation curves. It treats gravity as a gradient over a memory-density field and fits three competing versions — unconstrained pointer, constrained pointer with a physical prior on halo extent, and an NFW baseline — head-to-head on the 171-galaxy SPARC sample.

What is the main empirical result?

On the 171-galaxy SPARC sample, the constrained Pointer model wins AIC on 60 out of 171 galaxies against NFW. The partial correlation between halo extent and inferred information mass is ρ = +0.19 with q_FDR = 0.04. Four of six residual-structure features correlate with galactic age at FDR-controlled significance.

How is it different from MOND or NFW?

MOND modifies Newtonian dynamics at low accelerations. NFW posits a cold-dark-matter halo with a specific density profile. Pointer Architecture keeps standard gravity but adds a memory-density term to the enclosed mass integral and fits the three models at matched parameter count, so comparisons are at parity rather than stacked.

What would falsify the model?

Three independent falsifiers: (1) if the ρ sign flips on an age-controlled resample from LITTLE THINGS or THINGS, the correlation is noise; (2) if AIC-preferred galaxies correlate with survey artefacts rather than galactic age, the result is observational; (3) if the reproducibility pipeline yields materially different fits under reasonable re-parameterisation, the results are over-fit. The preprint enumerates each rejection criterion explicitly.

Is the data and code available?

Yes. The full reproducibility pipeline — four Python scripts, per-galaxy fit outputs, correlation tables, PCA, and residual diagnostics — is released alongside the preprint at github.com/neuralcosmology/pointer-architecture. SPARC source data is not rehosted but comes from the authoritative Lelli, McGaugh, Schombert (2016) release.

How does this relate to the books?

The non-fiction volume 'The Celestial Code' presents the argument narratively in plain language, chapter by chapter, with the full reasoning chain and references. The preprint presents the same model formally for academic reviewers. The two are linked via schema.org workExample/subjectOf metadata.

What does 'neural cosmology' mean?

It is the working name of a research programme that treats reality as a computational substrate and tests that framing against concrete anomalies in cosmology, neuroscience, biology, and information theory. The programme is scientific in the strict sense: every claim comes with a falsifier, and the core paper is head-to-head AIC against standard baselines.