Carried out of #28 (its F2), which closes with the loop-engineering PR. Nothing here is applied.
The gap
Base guide §3 defines live as "excluded from the offline run", for criteria that need network. The first clean-room --env=coding-agent-lab regeneration of the Qdrant Edge course (2026-09-14) hit an input a spec cannot author at all: a spoken audio recording. Synthetic silence produces no transcript, so the voice lane's acceptance criterion cannot run on authored fixtures. That run extended the mechanism itself, under a tag of its own invention, and said so.
Every course with an audio, video, or sensor-capture modality hits this. The guide's fixture doctrine ("facts you author") has no answer, so each generation invents one.
Proposed rule (§3, beside the live bar) — not applied
A criterion whose input cannot be authored as a fixture (recorded speech, camera frames, sensor traces) is tagged live-<kind> and excluded from the offline run, exactly as live is. The spec must additionally carry an offline criterion covering the wiring that input feeds — the path is testable even when the payload is not.
Plus a §14 line: every live-<kind> criterion has a sibling offline criterion for its wiring.
Why it did not ship with the loop-engineering pass
That pass (branch course/loop-engineering-with-langchain-deepagents, matrix docs/research/guide-sources-and-tiebreaks-property-matrix.md) validated its guide changes by generating the loop-engineering course, which has no audio, camera, or sensor input — nothing there could show this rule landing. Fix and test travel together, so it waits.
How to validate
Regenerate the Qdrant Edge course (courses/building-on-device-ai-memory-with-qdrant-edge/, voice lane) on the amended guide and check: the voice criterion carries live-audio; a sibling offline criterion exercises the speech→embed→store wiring with an authored transcript; no other criterion changed category.
Carried out of #28 (its F2), which closes with the loop-engineering PR. Nothing here is applied.
The gap
Base guide §3 defines
liveas "excluded from the offline run", for criteria that need network. The first clean-room--env=coding-agent-labregeneration of the Qdrant Edge course (2026-09-14) hit an input a spec cannot author at all: a spoken audio recording. Synthetic silence produces no transcript, so the voice lane's acceptance criterion cannot run on authored fixtures. That run extended the mechanism itself, under a tag of its own invention, and said so.Every course with an audio, video, or sensor-capture modality hits this. The guide's fixture doctrine ("facts you author") has no answer, so each generation invents one.
Proposed rule (§3, beside the
livebar) — not appliedPlus a §14 line: every
live-<kind>criterion has a sibling offline criterion for its wiring.Why it did not ship with the loop-engineering pass
That pass (branch
course/loop-engineering-with-langchain-deepagents, matrixdocs/research/guide-sources-and-tiebreaks-property-matrix.md) validated its guide changes by generating the loop-engineering course, which has no audio, camera, or sensor input — nothing there could show this rule landing. Fix and test travel together, so it waits.How to validate
Regenerate the Qdrant Edge course (
courses/building-on-device-ai-memory-with-qdrant-edge/, voice lane) on the amended guide and check: the voice criterion carrieslive-audio; a sibling offline criterion exercises the speech→embed→store wiring with an authored transcript; no other criterion changed category.