Fold hourly 0551 HIGH

[24601]

Fresh PR off latest main after merged #64 (hourly 0445 / notes.md §140). This fold is notes.md §141 / composition items 665–680 / findings batch #123. Does not bump 0.5.0. Do not reopen or amend PR #23–#64. Skip Archer (promised_not_landed).

What was folded

• jaredpalmer/kev Night-2 densify (PRIMARY, §45). HEAD 4f8110a3f862 README SHA d497d4b89427. Dates+unknowable deltas promoted for 0.8B/4B/9B; v7-base tags; locked OOD 0.684/0.837/0.852 theirs; T≈2.0 Brier/ECE already carded. Not a controlled architecture comparison. SHA move is not a replica. Densify §45, not a sibling first sighting.
• bespokelabsai/nimble densify (§35). Publish original 2676 training examples and frozen 324 holdout. 90.1% vs Jev 93.2% vs base 66.4% theirs. Did not distill from Jev.
• Hub first cards. hf:gump2049/APUS-OpenJev-v1 Frozen80 n=80, 9B 85.0% vs Jev API 82.5% theirs not Harbor. Candidate probabilities are not calibrated confidence. hf:SeanLiu/Jev-Vision POPE/MME/NLVR2 theirs; JevBench hard 52% at 91% mean confidence theirs. Qwen3.5/Qwen3-VL-8B ≠ Archer. LoRA ≠ RLCD replica.
• Catalogs / namesakes. AbdelStark/awesome-typesafe-jev IS AbdelStark/awesome-typesafe (same GitHub id 1374058281). Independent community project; catalog ≠ endorsement. cookiespiggy/agentic-rl ch.25 Jev vs RL. DevMortimer/pi-typesafe ≠ twilwa/pi-typesafe. Named remainder (JARVIS, containment, clinic 98.6% theirs, compaction can-only-score, hermes 82%/89% holdout theirs). Skip-thin empty SHA.
• Description-rewrite revisits. evoke-build/evoke densify §139: Jev is the first adapter, the design is bound to no engine. 47thtechcorner GLiNER: Zero Hallucinations marketing. Locate ≠ decide.

Why

Augustus owns class placement, not vendor how-tos. Night-2 OOD, Frozen80, Nimble holdout, and GLiNER locate stay sensors. Soft Noul ≠ hard safety. Formal compose: a Noul is a SENSOR.

Checks

• uniqueness_gate.py ok (UNIQ_0551, 21 overlays; CHANGELOG has no dump wall)
• evaluate_decisions.py --self-test ok (Night-2 OOD / Nimble distill / Frozen80 / Vision overconfidence / Locate ≠ decide)
• revisit_fingerprints.py --self-test ok

  

[24601]

ADV_PASS. Adversarial review + testing of #65 (HEAD c0bb05249935f4a6c24a809de70c4850e8aac932).

Honesty checked: kev Night-2 densify §45 HEAD 4f8110a3f862 README SHA d497d4b89427; locked OOD 0.684/0.837/0.852 theirs; v7-base tags; dates+unknowable deltas; SHA move is not a replica; densify §45 not a sibling first sighting (fingerprints notes_section 45). nimble densify §35 HEAD f136b3f75721; 2676 training + frozen 324 holdout; 90.1% vs Jev 93.2% theirs; did not distill from Jev (notes_section 35). hf:gump2049/APUS-OpenJev-v1 Frozen80 85.0% vs Jev API 82.5% theirs not Harbor; candidate probs ≠ calibrated confidence. hf:SeanLiu/Jev-Vision hard 52% at 91% mean confidence theirs; LoRA ≠ RLCD; Qwen ≠ Archer; ≠ sseanliu/Jev-Vision. AbdelStark/awesome-typesafe-jev IS AbdelStark/awesome-typesafe (same GitHub id 1374058281). evoke densify §139: Jev is the first adapter, bound to no engine (notes_section 139). GLiNER first card §141: Locate ≠ decide. Named remainder: DevMortimer/pi-typesafe ≠ twilwa; cookiespiggy agentic-rl ch.25; hermes 82%/89% holdout theirs. Skip-thin empty SHA. invented_signal false. No version bump. Do not reopen #23–#64.

Tests (python3, all exit 0):

• .agents/skills/augustus/scripts/uniqueness_gate.py → uniqueness-gate ok (0551 chars=3365, 21 overlays, 5 revisit overlays)
• .agents/skills/augustus/scripts/evaluate_decisions.py --self-test → self-test ok
• research/revisit_fingerprints.py --self-test → revisit-fingerprints self-test ok

IDs unique vs main (524c2d2 / merged #64): §141 / items 665–680 / batch #123 (main owns §140 / 649–664 / #122). Occupancy steal FAIL-as-expected for §141, 665, 680, #123; this branch only. SKILL 0.5.0 frozen. Archer still promised_not_landed. CHANGELOG Unreleased has no uniqueness dump wall. CI build success. READY_TO_MERGE. Do not merge from this comment.