# n239 — THE DEFLECTION CRISIS: light bends toward mass, at one quarter of the observed angle — the successor's sharpest standing tension

**Date:** 2026-07-12 · **Trigger: the human's question "is light deflected by a mass in this theory?"** The mechanism answer is yes; assembling the *coefficient* from the measured and pinned pieces uncovers an adverse result of the first rank.

## The chain (every link measured or pinned, all dimensionless)

1. The clock potential defines Φ: E(x) = m(1+Φ) — coefficient 1 (mechanism 4, the theory's cleanest sector).
2. Dark–baryon equivalence pinned **R = dlnk/dΦ = 1** (n189, clusters) — winding energy is *linear* in k.
3. Light's cone response measured: **dlnc²_EM/dk = 0.998** (n234) — also *linear* in k.

Therefore the index near a mass: n − 1 = |Φ|/2. GR requires 2|Φ|. **Predicted deflection: 0.44″ against the observed 1.75″ — one quarter of GR, half of Newton's corpuscular value. γ_eff ≈ 0.0 vs Cassini's 1 ± 2.3×10⁻⁵.** Eddington 1919 alone would have excluded it.

## Why there is no cheap escape

Windings and light couple to k **the same way (linearly)** — one coefficient serves both. Dark–baryon equivalence demands R = 1; light bending demands R = 4. The clock normalization λ rescales both together: the internal, λ-free, units-free ratio *(light bending)/(dark-matter fall)* is stuck at 1/4 of GR's. **This is the third occurrence of the one-coefficient-two-masters pattern** — the dichotomy (one sector, two forces), Zener (one λ, two polarizations), and now equivalence-vs-bending (one R, two falls). Each previous occurrence killed a theory or forced a split.

## The three outs, registered as computations (not hopes)

- **(a) Re-derive the bending response for a *localized* dressed source.** n234's response was measured under a *uniform* k-shift; a 1/r dressing has gradients at the EM wavelength scale — the effective index could differ from the local-k value. Computable on the glass (scattering of a kernel-mode wave packet by a dressed region).
- **(b) Re-audit n189's R-pinning.** The winding proxy was a tight 4-edge loop in 2D; if the true winding–k coupling is sub-linear (core corrections), R could exceed 1 for the smooth dressing while equivalence holds — splitting the masters. The full defect (not proxy) computation decides.
- **(c) The split that rhymes with history:** the pattern's two previous occurrences were resolved by *splitting the coefficient by sector* (two ledgers; two weights). Here the split would be: light responds to the **staggered-sector** component of the dressing, windings to the **smooth-sector** component — and a mass's dressing has *both*, with independently computable amplitudes. If the staggered component near a source is ~4× the smooth one, both masters are served with no new parameter. This is the natural P3″-native resolution and is fully computable (the two dressing profiles around one source).

## Standing status

Registered as **the successor's sharpest tension — a potential kill by the oldest test in the book** — pending outs (a)–(c), each a named computation. If all three fail, P3″ dies by Eddington, and the record will say so in this font. Prediction 12's entry and the status page carry the flag.

**Class: the chain [A — assembled from measured/pinned numbers]; the trap [analytic]; the outs [computations, not rescues — each falsifiable]. The question that found this was four words long.**