Synthetic Feed and Provenance Research Note

Synthetic Feed and Provenance Research Note

Editorial conclusion

"AI slop feed" is useful as a cultural phrase but weak as a technical definition. The E084 explainer should define the system beneath the insult: a recommendation feed where generated and remixed media can expand content supply with fewer human production constraints.

The term does not prove that every generated item is low quality or that every human-created item is valuable.

Vibes product record

Meta's September 2025 launch record describes Vibes as a personalized feed of short AI-generated videos where a user can discover, create, remix, and share. A user can start from a prompt or existing media, add music, change style, post to Vibes, send directly, or cross-post to other Meta surfaces. Meta Vibes launch record

The official description establishes the product loop. It does not establish the amount of content produced, user satisfaction, moderation quality, or effects on attention.

What changes in the supply loop

A conventional creator feed is constrained by recording, editing, creator time, permissions, and audience building. Generative tools can reduce some production constraints and make variants or remixes cheaper. Recommendation, generation, and remixing can then form a tighter loop.

That does not make supply literally infinite. Compute, model access, policy, rights, moderation, attention, distribution, and user effort still constrain the system. "Near-unlimited" should be understood as a relative product condition.

Provenance

C2PA develops an open standard for cryptographically bound Content Credentials that can record an asset's origin and edit history. C2PA also says provenance is not always complete and cannot by itself prove that content is true or accurate. C2PA specifications and C2PA explainer

Meta says it uses "AI info" labels when it detects industry signals or receives creator disclosure on supported surfaces. A label is useful context, not a complete rights record or quality judgment. Meta labeling approach

Public-analysis questions

The public explainer should ask whether the viewer can tell what is generated, identify the source and remix chain, understand why the item was recommended, correct attribution, report a rights or safety issue, reduce synthetic recommendations, and leave the feed.

It should separately ask whether the product helps the user create or merely produces another item to continue scrolling.