# E2E Report - Anderson Physical Perturbation Landing Loop

**Date**: 2026-05-29 21:24 UTC  
**Mode**: bounded regenerative landing loop  
**Loop tool**: `tools/exp_boundary_physical_perturbation_landing_loop.py`  
**Output**: `tools/data/landing/boundary_physical_perturbation_landing_loop_20260529_v2.json`  

## Question

Can the Anderson physical perturbation discriminator regenerate through
stronger perturbation rounds until it lands stably?

## Landing Definition

Landing means:

- W20 remains `local_global`;
- W16/W16.5 do not become promoted `local_global`;
- poles remain preserved;
- result is discriminator progress only, not a physics-law claim.

## Result

```json
{
  "landed_once": true,
  "landing_round": 1,
  "sustained_landing": false,
  "landing_lost_round": 2,
  "rounds_run": 3
}
```

The loop lands under the default perturbation round, but does not remain fully
landed under stronger perturbations.

## Round Read

### Round 1 - default

```json
{
  "evolutionary_verdict": "evolves_discriminator",
  "progress_score": 1.0,
  "preservation_ratio": 1.0,
  "intermediate_promotion_ratio": 0.0,
  "poles_ok": true
}
```

W20 is preserved. W16/W16.5 remain unpromoted.

### Round 2 - stronger perturbation

```json
{
  "evolutionary_verdict": "needs_more_physical_null",
  "progress_score": 0.9625,
  "preservation_ratio": 1.0,
  "intermediate_promotion_ratio": 0.125,
  "poles_ok": true
}
```

W20 is still preserved, but `tilted_disorder_strong` closes W16.5 as
`local_global`.

### Round 3 - wider / size 8

```json
{
  "evolutionary_verdict": "needs_more_physical_null",
  "progress_score": 0.9625,
  "preservation_ratio": 1.0,
  "intermediate_promotion_ratio": 0.125,
  "poles_ok": true
}
```

W20 is still preserved, but `mixed_opposed` closes W16.5 as `local_global`;
`anisotropic_y_strong` moves W16.5 to `global_only`.

## Interpretation

This is not a failure. It is a sharper boundary.

The closed reference W20 is robust across all perturbation rounds. The
intermediate band W16/W16.5 is not rigidly non-promoted: under stronger
perturbations, W16.5 can close sporadically. Therefore the landing is not a
stable final claim; it is a regenerative discriminator with a sensitive
intermediate channel.

## Boundary

- No cron reactivation.
- No public sync.
- No physics-law promotion.
- No claim that Anderson graph/localization entity is established.

## Productive Next Move

The next useful move is to isolate the W16.5 closing condition:

1. sweep `onsite_tilt` around `0.04 -> 0.10`;
2. hold hopping fixed, then combine with anisotropy;
3. measure whether W16.5 closure is a robust physical channel or a small-size
   artifact;
4. keep W20 as reference and W16/W16.5 as landing-sensitive transition zone.

This should be a focused W16.5 closure-map, not another broad metric.

