Scale Theory
A long-running personal research program on physical observation channels — how finite-resolution experiments change what source states can be told apart. Scientifically reset in July 2026 after an audit refuted most of its original claims.
This began as a broad theory of observer-relative physics: the idea that no scale is intrinsically privileged, and that quantum phenomena below us and frozen geometry above us are artifacts of measuring from a fixed vantage point.
On 18 July 2026 I audited it against its own evidence and reset it. Most of the original claims did not survive. What remains is narrower, defensible, and still open.
Current status. This is an information-physics research program, not a new fundamental theory. It has no confirmed empirical prediction unique to it. Its most valuable output so far is its negative-result record.
What the project is now
Specify a physical observation channel. Define what it means to discriminate source states through its records. Derive how the required resources scale as source modes become unresolved.
That is the whole live program. It is explicitly not a theory of quantum mechanics, particle identity, relativity, MOND, or cosmology — each of those was attempted, and each is now a closed direction.
The claim audit
The reset produced a line-by-line disposition of every original claim. A representative sample:
| Original claim | Disposition |
|---|---|
| No scale is intrinsically simple | Removed — metaphysical; no measurement distinguishes hidden richness from genuine simplicity |
| Observer pinned at about 1 m, 1 s | Withdrawn — instruments set resolution, not body size |
| A universal blur/exposure/noise channel | Withdrawn — real channels depend on coupling, probe, detector, acquisition, reconstruction |
| Scale entropy always increases | Refuted — a hard window’s transfer zeros break ordering at arbitrarily large coarsening |
Phase averaging yields ħ_eff = J |
Withdrawn — classical actions stay continuous and system-dependent |
| Particle identity is an information fixed point | Unsupported — source confusion yields no permutation symmetry, exchange phase, or spin-statistics |
| Bell escape is foreclosed by the M1 exponent | Withdrawn — exponent extrapolated outside its domain; local coarse-graining does not explain Bell violation |
| Quantum mechanics emerges below us | No-go for the proposed mechanism |
| Lorentz/geometric physics emerges above us | No-go for the proposed mechanism |
a0(z) ∝ H(z) is a channel prediction |
Demoted to a separate conjecture; no derivation from the core |
| Force/relaxation cosmology ladder | Closed — exact or observational degeneracy with standard cosmology |
What genuinely survived
- A controlled experiment (M1) that correctly measures a designed moment hierarchy and information plateau in one constructed channel.
- A second (M2) that isolates a finite phase-loss window under fixed structural assumptions — and, in later work, exposes its non-identifiability outside them.
- A clean counterexample to universal monotonicity of scale entropy.
- An exact criterion for when a physically indexed channel family is Blackwell-ordered, a two-sided bound on its Le Cam deficiency, and a noise-floor threshold beyond which a non-nested family still obeys an absolute second law.
- A methodological lesson from the cosmology work: how convention and correlated-data errors can manufacture an apparently strong signal.
- A set of no-go and boundary results for the quantum and relativistic arcs.
The one live summit
A resource-aware distinguishability bound for localized sources observed through a physical far-field channel. The bound now exists, exactly and two-sidedly.
Working on it also falsified two of the three predictions I had registered before computing — including the assumption that angular coverage is a resource entering with the same exponent as inverse distance. It is not: contiguity is the resource, and splitting an aperture into enough components at fixed total solid angle removes the exponent entirely.
The remaining burden is novelty, and it has grown heavier. A prior-art pass
found that the governing exponent is already known in the super-resolution
literature as the conditioning exponent of a Vandermonde matrix with clustered
nodes, and that the design principle is textbook c-optimality. Both current
fronts are therefore reframings of known mathematics. The honest statement is
that the summit has produced one useful internal correction and no new theorem.
Method
Every live claim must name its source-state class, the physical coupling, the detector response and reconstruction, which resources are fixed and which vary, a theorem or registered quantitative prediction, a kill condition, and the precise novelty delta against the closest established result. Predictions are registered before they are computed. Anything without those items belongs in the origins folder or a speculative note — not in the manuscript.
Working this way is what let the project refute itself rather than accumulate.