Scale Mechanics

A standing experiment in steering frontier machine intelligence toward novel, verified theory — whose demonstration payload is a scale-dependent information-processing framework for quantum mechanics, measurement, gravity, and cosmology. Every number in the manuscript is derived, tested, and pinned by this repository; every generation of the work, including the flawed first draft, is preserved in lineage/.

Classical mechanics governs the large; quantum mechanics governs the small; scale mechanics is the dynamics of the axis between them. Developed 2025–2026 under the working name CEQT (Computational Emergence Quantum Theory) through three adversarially verified generations; the founding name was ratified 2026-08-15. Not abbreviated — "ScM" only where unavoidable, never "SM".

Two doors into this repository:

• The physics. The manuscript is paper/CEQT-v3.pdf — it keeps its generation-3 title because the verification record certifies that exact text. Its derivation engine is src/scalemech/; its regression suite is tests/; its observational dossiers are data/recon/.
• The method. lineage/README.md — three generations of the work preserved in full: the raw first draft, the forensic sweep that found 14 fatal defects in it, the repairs, the convergence logs, and the gauntlet that certified generation 3.

The third-generation manuscript models physical reality as a computational substrate whose effective description changes with scale. Scale transitions are Takesaki conditional expectations between finite-dimensional von Neumann algebras; measurement is stochastic branch resolution obeying a hard rate ceiling; gravity is the thermodynamics of severed computational connections; dark energy is the cost of horizon-scale complexity growth. Every headline number is computed by this library from stated inputs and locked in place by tests/test_paper_numbers.py, which fails if the paper and the code ever disagree.

The method demonstrated by the lineage: iterative strengthening of a draft by fleets of adversarial AI verifiers, run until two consecutive independent passes find nothing; numbers from committed engines, never from prose; data confrontation on paper before any experiment is proposed; refutations published as results; gaps conceded OPEN in writing. The full protocol is distilled in lineage/README.md.

Stated plainly, because it is the record's own epistemics: adversarial self-verification certifies arithmetic, internal consistency, citation reality, and confrontation against published data. It cannot certify novel physical truth — the verifiers share a substrate with the generator. The external ground wire is the scoreboard, registered in advance:

1. Qutrit commutator bump — √2/4 at γt = ln 2 from a commuting pair; a cheap, decisive experiment on existing hardware.
2. Measurement speed limit — no engineered readout beats τ⊥ = πħ/2E_ex.
3. Σm_ν = 59.2 meV, normal ordering — a measured sum significantly away from 59 meV kills it.
4. Fixed-frequency primordial template — ω_log = 2π/ln 2 = 9.0647, one free amplitude.
5. wₐ > 0, no phantom past — a robust dark-energy reconstruction requiring wₐ < 0 with a past phantom crossing kills it.

G-universality stands as the gravitational sector's further kill condition. Reality grades these; nothing in this repository can.

git clone https://github.com/ovachiever/scale-mechanics.git
cd scale-mechanics
pip install -e ".[dev]"

scalemech-reproduce   # derive every paper number from stated inputs
pytest                # 82 tests, including the paper-number regression suite

Library use:

from scalemech import koide, speed_limit, dark_energy

koide.neutrino_solution("NO").total   # 0.05916 eV  — the flagship prediction
speed_limit.tau_perp(2 * 1.602e-19)   # 5.17e-16 s  — optical measurement floor
dark_energy.cpl_parameters()          # (-0.885, +0.231)

src/scalemech/       the derivation engine (numpy only)
  chart.py           canonical scale chart s = ln(l/l_P), hierarchy identity
  speed_limit.py     Margolus–Levitin measurement rate ceiling
  commutator.py      dephasing flow, envelope theorem, creation effect
  measurement.py     martingale branch dynamics, Born statistics, drift no-go
  koide.py           Z3 equipartition, circulant fit, neutrino solution
  gravity.py         Jacobson bridge, eta calibration
  dark_energy.py     complexity-horizon equation of state, CPL parameters
  dsi.py             discrete scale invariance, CMB template sensitivity
  constants.py       CODATA 2018 / PDG 2026 / NuFIT 6.0 inputs (single source)
  reproduce.py       scalemech-reproduce entry point
tests/               unit + property tests per module, plus test_paper_numbers.py
paper/               CEQT-v3.tex / .pdf (REVTeX 4.2), the generation-3 manuscript
paper/verification/  adversarial gauntlet log (passes 0-14, 48 findings) + raw engine output
data/recon/          observational-status dossiers (DESI w0wa, S8 tension, collapse speed, Koide, Planck DSI)
lineage/             the full three-generation record: raw gen-1 drafts, the
                     gen-2 forensic sweep and repairs, convergence logs, fleet dossiers

Generations 1–3 (March 2025 – July 2026) were built under the working name CEQT; the complete record, including the audits, is in lineage/. The framework's founding name, Scale Mechanics, was ratified 2026-08-15 and debuts on the fourth-generation manuscript (Scale Mechanics 1.0), whose program is the unification theorem: classical and quantum dynamics as the two fixed-point regimes of one scale flow — built in the open, against the scoreboard above.

Inputs are pinned in src/scalemech/constants.py: CODATA 2018 fundamental constants, PDG 2026 lepton masses, NuFIT 6.0 oscillation splittings, Planck 2018 Ω_DE. Change an input and the test suite reports exactly which paper numbers move.

See CITATION.cff. The manuscripts in paper/ and lineage/ are © the author, CC-BY-4.0; the code is MIT.

Bug reports about numbers are the most valuable kind — if any quantity in the paper cannot be reproduced by scalemech-reproduce or disagrees with the tests, that is a defect in the framework's front line. See CONTRIBUTING.md.