# Apeiron AP1 – G32C log-H positive-branch recovery preflight

**Updated UTC:** 2026-09-03T06:23:47Z  
**Classification:** `M1_LOGH_POSITIVE_BRANCH_RECOVERY_PREFLIGHT_PASS_CONSUMED_DOMAIN_PILOT_OPEN_NEW_HELDOUT_LOCKED_PHYSICAL_KERNEL_BLOCKED`  
**AP1 status:** ORANGE

The registered G32B heldout was attempted exactly once.  It stopped in the
first coarse background integration with `background left finite positive-H branch` before a level,
mode/operator audit, or frozen-tolerance comparison completed.  No heldout
report, checkpoint, trajectory, G33 promotion, or seed was written.  The
interval through relative `N=0.10` is consumed and can never be claimed as a
blind validation result.

## Pure numerical-coordinate recovery

The new background chart is `ell_H=log(H/H_reference), H=H_reference*exp(ell_H)`.  On `H>0`,
`d ell_H/dN = (dH/dN)/H`, so reconstructing `H` recovers the unchanged direct
equation.  The physical `H` absolute tolerance remains
`2e-17` and is mapped to the dimensionless chart as
`H_atol/H_reference`.  Equations, matter content, parameters, counterterms,
PV basis, moving split, mode transport, fixed-point settings, and every
existing threshold remain unchanged.

- Maximum H round-trip defect: `0.0`
- Maximum RHS reconstruction defect: `0.0`
- Positive-H analytic toy defect: `0.0`
- Machine-scaled cap: `1.4210854715202004e-14`
- Structural gates: `16/16` PASS

## Chronology

- Non-blind recovery pilot: `N=0.085..0.1` on `1281/2561/5121` nodes.
- New prospective heldout: `N=0.105..0.12` on `1537/3073/6145` nodes.
- The new heldout remains locked until the recovery pilot passes completely and
  new method tolerances are frozen separately.

This checkpoint contains zero background runs and no physical response-kernel
run.  AP1 remains ORANGE; G03, G11/G12, AP2/AP3, production curves, fits,
observables, and significances remain locked.