AI Fluency Proof-Tail Fixtures
A safe split of the remaining proof/rubric tail into source-backed local fixtures and a repair-first blocked queue.
The prior top-band AI Fluency proof tail was not filled with fuzzy title aliases. It was split into safe source-backed local evaluator fixtures and an explicit repair queue, so the engine gets better proof/rubric context without pretending weak metadata is learner proof.
A New Exact-Match Supplemental Fixture Lane
Added ai-fluency-proof-tail-fixtures. It reads ai_fluency_top_band_capability_coverage_v1, finds rows that still lack local proof/rubric fixtures, and only creates supplemental fixtures when the row is source-backed, non-generic, and has workflow or named-tool context.
uv run aina-data-engine --root /srv/aina/aina-data-engine-room ai-fluency-proof-tail-fixtures
| Artifact | Purpose |
|---|---|
ai_fluency_proof_tail_fixtures_v1.json | Receipt and checks |
ai_fluency_proof_tail_fixtures_v1.jsonl | Source-backed supplemental fixture rows |
ai_fluency_proof_tail_fixtures_v1_blocked_queue.jsonl | Repair-first blocked rows |
ai_fluency_proof_tail_fixtures_v1_semantic_sample_50.jsonl | QA sample |
The Tail Is Smaller And More Honest
| Metric | Value |
|---|---|
| Prior proof/rubric gap rows | 47 |
| Top-500 proof/rubric proxies | 486 |
| Top-1,000 data-boundary policy rows | 1,000 |
| Tool-context gap queue rows | 0 |
| Embedding or batch manifests | 0 |
| Production unlocks | 0 |
It Is Context, Not Learner Proof
This slice did not create learner proof. It created local source-backed evaluator/proof context that can support AI Fluency hardening while keeping measured learner outcome evidence explicitly gapped.
Rows stayed blocked when they were source-less, generic, title-family repair rows, or missing title authority. The remaining 35 rows should be handled by source-ref repair and JD/source-context title resolution, not broad aliases.
All Gates Passed
uv run pytest tests/test_ai_fluency_proof_tail_fixtures.py tests/test_ai_fluency_capability_coverage.py -q uv run ruff check src/aina_data_engine/ai_fluency_proof_tail_fixtures.py src/aina_data_engine/ai_fluency_capability_coverage.py tests/test_ai_fluency_proof_tail_fixtures.py tests/test_ai_fluency_capability_coverage.py src/aina_data_engine/cli.py uv run aina-data-engine --root /srv/aina/aina-data-engine-room ain-506-p0-gate uv run aina-data-engine --root /srv/aina/aina-data-engine-room ain-510-retrieval-promotion-gate uv run aina-data-engine --root /srv/aina/aina-data-engine-room production-runtime-readiness uv run aina-data-engine --root /srv/aina/aina-data-engine-room validate
AIN-510 remains promotion_ready with 6,506 matched vectors, 0 stale vectors, complete top-500/top-1,000 vector coverage, and runtime embedding authority still false.
Proof Posted
| Issue | Comment ID |
|---|---|
| AIN-520 | 283218f8-8ddb-4a6e-9a34-84b9742d6309 |
| AIN-527 | 09c820e2-eeba-4be8-9369-6d96d4acc84b |
| AIN-510 | 4e2b00e2-181c-4537-924d-f93278fcf6b0 |
Exact Commands
cd /srv/aina/aina-data-engine-room
git status --short --branch
git log -5 --oneline
uv run aina-data-engine --root /srv/aina/aina-data-engine-room ai-fluency-proof-tail-fixtures
uv run aina-data-engine --root /srv/aina/aina-data-engine-room ai-fluency-capability-coverage
uv run aina-data-engine --root /srv/aina/aina-data-engine-room ain-510-retrieval-promotion-gate
jq '{valid, status, metrics, checks, failed_checks}' artifacts/validation/ai_fluency_proof_tail_fixtures_v1.json
jq '{valid, status, metrics, quality_gates, failed_checks}' artifacts/validation/ai_fluency_top_band_capability_coverage_v1.json
Resolve the remaining 35 blocked proof-tail rows by repairing source authority first, then add only source-backed local fixtures that pass the same checks.