Product use — evidence for the remaining meaning choice

This check prepares one meaning choice. It does not establish a completed pipeline audit or authorize data changes.

Prior decisions and the residual

The Moments explanation already separates appearance from product role, and framing from activity and story. The earlier promotion list has three different draft definitions: Active Use describes using the product mid-task; Hands-on Use describes handling and using it directly; Hands-on Interaction describes hands working with it. Their stored assignments do not establish equivalence.

The independent challenge identified a missing boundary: physical handling can happen without operation. The review therefore asks explicitly whether hands-on references include deliberate handling and adjustment or only operation. It preserves overlapping UI Interaction, avoids another required Active Use label and does not approve the separate Product Interaction merge.

Primary examples

The Apple controls guide documents stem presses. The Siri guide documents voice activation while wearing connected AirPods. The ear-tip guide documents removal and attachment. These text instructions support the conceptual distinction between operating with touch, operating without touch and physical adjustment. They do not certify a collected video's actual action.

Sources are linked on the review page. No new media classification was accepted. A browser attempt to inspect a collected ChatGPT interval stopped after the requested end; that frame was excluded as evidence of the interval. Earlier mismatched media descriptions remain evidence gaps.

Bounded live and local checks

A fresh information-schema probe confirmed the selected fields before the query. Selected stored rows show all three proposed labels. This is a sample of assignments, not a validated taxonomy or prevalence estimate.

In the current local classifier, the moment-type and product-role values are resolved separately, carried separately into the built record and written separately to the classification record. The local Product in Use prompt also admits a result or payoff; that wider browse definition must not be silently narrowed by this small decision about actual interaction. Deployed parity, complete consumer behavior and old-record mapping remain unverified.

Technical reference

• Live read scope: public content_segment, content_segment_classifications and selected content_item fields.
• Local writer inspected: backend/routers/pipeline_scene_classify.py; prompt around 185–196, role wording around 319, normalization around 945–1000, classification write around 1179–1207.
• The inspected frontend lab reader selects moment_type and presents it as momentType. This limited reader check does not prove all browsing paths support product-role filtering.
• Literal queries and results, source hashes, browser exclusion and review boundary: docs/reference/2026-10/2026-10-05/product-use-review-preflight.json.
• Existing indexing concern remains in moment-use-boundaries-preflight.json. Its historical causality is unproven; no repair occurred here.

Required later verification

Trace the complete deployed classification and browsing paths. Verify intervals and media identity, preserve provenance and human corrections, and review each old assignment. Test that equivalent browse and tracker rules agree and that unsupported operation is not inferred from product visibility or capability. Implementation and product acceptance remain separate from the meaning decision.

Prepared 5 October 2026.