# n110 — the refutation hunt: existing data turned against the theory
# (2026-07-10; adversarial sweep — the standing instruction, applied to
#  the world's archive instead of our own lattice)

RULE: only experiments that have ALREADY run. No proposals. Each
attack states the framework's exposure, the data, and the verdict.
Verdicts: KILL / WOUND (carve into status) / SQUEEZE (live tension,
quantified) / ABSORBED (bound enters the card) / MISS.

---------------------------------------------------------------------
ATTACK 1 — DESI DR2 evolving dark energy vs the LLR G-dot bound
---------------------------------------------------------------------
EXPOSURE: prediction 11 — "if dark energy evolves, G drifts with it;
the pump powers both." Direction registered, coefficient owed.
DATA: DESI DR2 BAO + CMB prefers w0waCDM over Lambda-CDM at 3.1 sigma
(w0 = -0.42 +/- 0.21, wa = -1.75 +/- 0.58); with SNe, 2.8-4.2 sigma
depending on compilation. Meanwhile LLR pins |Gdot/G| < ~1e-14/yr.
ARITHMETIC: at the DESI best fit, d ln(rho_DE)/dt = 3|1+w0|*H0
~ 1.2e-10/yr. If G drifts with the pump at coefficient k,
   k < (1e-14)/(1.2e-10) ~ 8e-5.
VERDICT: SQUEEZE — the sharpest live tension found. If the DESI
evolution is real at its best-fit amplitude, the pump coefficient is
unnaturally small (< 1e-4 where O(1) is natural). The framework must
choose its stance NOW, before the data decides:
REGISTERED STANCE (prediction 11, amended): the w0-wa preference will
dissolve — it is 3-4 sigma, driven by two LRG tracers, and the DESI /
DES-SN5YR likelihoods are in direct documented conflict. Same class
of bet as the alpha-dipole and CODATA-G stances (prediction 13).
COST IF WRONG: a confirmed w(z) != -1 with Gdot/G still pinned at
zero forces k < 1e-4 — not a kill by the letter of the registered
kill condition, but a fine-tuning wound of 4 orders. Carried openly.

---------------------------------------------------------------------
ATTACK 2 — LARES-2: frame dragging at one part in a thousand
---------------------------------------------------------------------
EXPOSURE: the audit reads T00 — energy density. Frame dragging
(Lense-Thirring) is sourced by T0i, the momentum flow. The treatise
has no gravitomagnetic engine; the tensor junction (the 0.5000 debt)
owes the whole radiative + rotational sector.
DATA: LAGEOS + LARES-2 + GRACE (Nature, 2026) measure terrestrial
frame dragging at ~1e-3 relative precision — an order of magnitude
beyond anything before. GR exact at that level.
VERDICT: WOUND (sharpened, not new). Not a contradiction — the
framework predicts nothing here yet, and "predicts nothing" is the
wound: a 1e-3-precision phenomenon with no engine. The budget IS a
flux (A·delta-k is a current), so the raw material for a T0i sector
exists; nothing is built. LARES-2 joins binary-pulsar decay as the
tensor junction's named watchers, now the most precise of them.
The relativistic-continuum deliverable inherits a numeric target:
whatever continuum limit emerges must produce Lense-Thirring exactly
to 1e-3 or die on contact.

---------------------------------------------------------------------
ATTACK 3 — the running of alpha (LEP: 1/137.036 -> 1/128.93 at M_Z)
---------------------------------------------------------------------
EXPOSURE: the framework reads alpha = 1/S_inf, a substrate constant.
But the measured alpha RUNS with momentum transfer — vacuum
polarization, confirmed at LEP to high precision.
VERDICT: WOUND (named debt on the alpha reading). No contradiction —
S_inf matches the Thomson-limit (IR) value, which is the correct
boundary condition, and the running is loop physics of the MATTER
sector, which the framework declares absent. But the alpha reading is
therefore INCOMPLETE BY CONSTRUCTION: it supplies one number (the IR
endpoint) and owes the entire momentum dependence to a sector it does
not have. Filed next to the 3.604 debt: the day a matter sector
exists, it owes (a) the 3.604 normalization AND (b) the beta function.

---------------------------------------------------------------------
ATTACK 4 — Eot-Wash vs the r^-3.5 neutral residual
---------------------------------------------------------------------
EXPOSURE: the EM-sector reading found a neutral-neutral residual
force ~ r^-3.5 in the toy. If that residual acts between laboratory
test masses with any appreciable strength, torsion-balance tests of
the inverse-square law (pure 1/r^2 down to 52 um) already exclude it.
VERDICT: ABSORBED — bound-in-waiting. Without the units bridge the
residual's lab-scale amplitude is unpriced; the Eot-Wash power-law
bounds become a CARD ENTRY the bridge must clear on arrival. (Same
status as the knee's absolute rate: the gap protects, and the gap is
the debt.)

---------------------------------------------------------------------
ATTACK 5 — long-baseline atom interferometry (the 54-cm, 2-s fountains)
---------------------------------------------------------------------
EXPOSURE: launched atoms in half-meter superpositions are NOT at
rest — during beamsplitter pulses dT00/dt != 0 differs by branch.
Observed: full contrast over ~2 s. Any audit prefactor large enough
to matter at Auditometer amplitudes must not already kill these.
VERDICT: ABSORBED — these are "reached amplitudes": the fountains'
null bounds the prefactor from above at their (tiny, adiabatic)
modulation depth, exactly the regime the Lorentzian suppresses
(omega << gamma_knee by ~17 orders; suppression ~ (omega/gamma)^2).
Card entry, not a wound: the shape protects the framework here BY
PREDICTION, not by gap. A fountain that ever shows anomalous
contrast loss at rest remains a kill (prediction 3).

---------------------------------------------------------------------
THE HUNT'S BALANCE
---------------------------------------------------------------------
No kill found in existing data. One SQUEEZE promoted to the
predictions page with its stance registered before the data decides
(DESI x LLR: k < 8e-5 if the evolution holds). Two WOUNDS carved
into status (frame dragging at 1e-3 with no engine; the alpha
reading's missing beta function). Two bounds ABSORBED into the
card's waiting room (Eot-Wash vs the residual; fountain nulls vs
the prefactor). The pattern of every honest sweep so far: the
theory survives where it has shape, and bleeds where it has no
sector — and the tensor junction is now watched at one part in a
thousand.