Record sealed ECP-7 utilization decay batch

[FatosLocos]

Summary

• add a validated late-decay schedule for the existing code-utilization coefficient
• preregister and execute sealed ECP7-B11-I while keeping the complete Batch 10 setup unchanged through step 5,000
• record the valid negative result, integrity evidence, and exact artifact hashes
• return to Batch 10 as the strongest base and narrow ECP7-B12 to one generic decoder-depth change

Result

The unchanged positive control remained at 100%. ECP7-B11-I regressed population train exactness to 79.09%, validation exactness to 72.22%, and translator validation exactness to 72.41%. The translator threshold still passed, but the registered train, validation, and injectivity gates did not. The result falsifies the late-utilization-dominance hypothesis.

The confirmatory ECP-7 test split remains sealed.

Verification

• 79 passed
• ECP7-B11 configuration validates with split sizes 14336/1024/1024
• both sealed analyzers passed with 65/65 artifact hashes and no test metrics
• all eight hashes recorded in batch11-manifest.json match the local run artifacts