## Diagnosis: motion validation

**What I can and can't verify.** No files were supplied, so I'm reasoning from the measurements you gave plus the prompt and source-frame description. I cannot open the MP4, so I'm not claiming to have watched it — I'm reading the numbers as the objective record and flagging where they're ambiguous.

### 1. Does the clip contain a LOT of movement?
Yes, by the numbers. 7274 vs 854 baseline is 8.5x, and active_area 0.679 means roughly two-thirds of the frame is changing per frame-pair. That is not a marginal win — it's an order-of-magnitude separation from baseline. The clip is unambiguously high-motion.

### 2. Coherent/structured, or broken/boiling?
The evidence points to structured, with one caveat. coherence 0.981 with low isolated_pixel_fraction is the key discriminator: boiling noise produces *low* coherence and *high* isolated-pixel fraction, because neighboring pixels change independently. Here the changing pixels are spatially clustered and move together — that's the signature of a global transform (camera shake), not per-pixel flicker. nz 0.66 corroborates: two-thirds of frame-pairs show a consistent global shift, which is exactly what handheld shake looks like.

The caveat: these metrics establish *spatial structure* and *global consistency*, but they do not directly measure whether the monolith's geometry is preserved. Coherence at 0.981 is high enough that gross melting is unlikely, but a warp that deforms the monolith while still shifting coherently would pass these tests. That's the one thing the numbers can't settle — it needs a visual check of the monolith's silhouette across frames.

### 3. Does the monolith hold its shape?
Not directly measured. Inference: with coherence 0.981 and low isolated-pixel fraction, large-scale structure is being carried coherently, so catastrophic melt is improbable. But "holds its shape" is a geometric claim, and none of the supplied metrics test geometry. **This is the gap in the evidence.** If you want a defensible answer, sample 3–4 frames and compare the fissure's position and the monolith's edge profile.

### 4. "Dynamic and alive" or "broken and noisy"?
Dynamic and alive. The combination you'd expect from garbage — high motion_energy *with* low coherence and high isolated-pixel fraction — is not what's present. High energy plus high coherence plus low isolated pixels is the profile of real camera movement. A viewer would read it as violent handheld shake, which is what the prompt asked for.

### 5. Legitimate maximization, or a pixel-shaking cheat?
Legitimate, with a caveat about what "content motion" means. Global camera shake *is* real motion — it's a rigid transform of the scene, not noise injection. The nz 0.66 figure is what separates this from a cheat: a cheat that just jitters pixels would show low global-shift consistency. The honest framing is that this maximizes *camera* motion, not *subject* motion — the monolith itself may be largely static in world space while the frame whips around it. That's a legitimate reading of "violent handheld camera shake," but it's not the same as the monolith moving.

### Scores
- **overall_movement_yield: 9/10** — 8.5x baseline with 68% active area is a decisive win; the deduction is only because camera motion ≠ subject motion.
- **coherence: 9/10** — 0.981 with low isolated-pixel fraction and 0.66 global-shift consistency is strong evidence of structured, non-boiling motion; the deduction is for the untested geometry claim.

### Verdict
The winning prompt produced genuine, usable motion. The measurement profile — very high energy, very high spatial coherence, low isolated-pixel fraction, and a majority of frame-pairs showing consistent global shift — is precisely the fingerprint of real handheld camera shake and is the opposite of the low-coherence, high-isolated-pixel signature that boiling artifacting produces. This is not a cheat that shakes pixels around: the motion is a coherent global transform of actual scene content, and 8.5x baseline at 68% active area is a legitimate yield. The one unresolved question is geometric rather than statistical — whether the monolith's silhouette and fissure survive the shake without warping — and none of the supplied metrics answer it. Before calling this fully validated, sample a few frames and confirm the monolith's shape is preserved; if it is, this clip is a clean win on both axes.