Intelligence and context scout evidence

This public-safe evidence snapshot retains the completed 2026-09-13–14 qualification, promotion, bounded client acceptance, and post-promotion context evidence. Native artifacts are sanitized; private originals remain outside this bundle.

2026-09-17 editorial note

Earlier derivative prose called this bundle a draft or incomplete and said restoration and client receipts were missing. That was stale publication text. The retained receipts establish the selected GLM no-spec text lane as admitted, with bounded Pi, Hermes, and OpenClaw acceptance and 9/9 post-promotion context after a managed cold reload. The original native scores and failures are unchanged. A fresh systemd start or reboot was not tested, and the declared Qwen FP8 rollback was not exercised after promotion.

Retained derivative records

• summary.json: bounded decision summary and original raw counts.
• configuration-identity.json: public-safe model/runtime pins and source digests.
• redaction-provenance.json: source digests and field-redaction audit.
• native/: 69 sanitized native evidence files and one derived environment diagnosis for r7 quality/strict120/context plus matched no-spec strict120/context/agentic/SWE/comparison and Next quality, context, agentic, SWE, image, and strict120.
• source-registry.json: primary model and runtime sources selected from the campaign research landscape.
• artifact-manifest-source.json: ten-role publication ledger, with its current hashed artifact-manifest.json derivative.

Publication boundary

The manifest is a completed, hashable ledger for the retained public evidence. The dated finding and maintained portal describe the recorded decision; this bundle does not alter live serving state or expand the retained limitations.