diff --git a/domains/entertainment/ai-video-adoption-is-demand-constrained.md b/domains/entertainment/ai-video-adoption-is-demand-constrained.md
new file mode 100644
index 00000000..572de7da
--- /dev/null
+++ b/domains/entertainment/ai-video-adoption-is-demand-constrained.md
@@ -0,0 +1,36 @@
+---
+type: claim
+domain: entertainment
+description: "Sora's 12M downloads with <8% D30 retention demonstrates AI video tools face adoption barriers on the demand side, not supply side, even among early adopters"
+confidence: likely
+source: "AI Journal / Evolink AI / Lantaai benchmark review, 2026-02-01"
+created: 2026-03-10
+---
+
+# AI video generation adoption is demand-constrained despite sufficient supply-side capability
+
+The 2026 benchmark data reveals a striking disconnect between technological capability and consumer adoption in AI video generation. Seedance 2.0 achieves near-perfect hand anatomy scores (the primary visual tell of AI-generated video since 2024), 2K native resolution, and 15-second duration capabilities that clear the technical threshold for live-action substitution in many production contexts. Yet Sora's standalone app, despite achieving 12 million downloads, retains fewer than 8% of users by day 30—well below the 30%+ benchmark for top consumer applications.
+
+This retention gap suggests that even among technology enthusiasts and early adopters, AI video generation has not yet created a compelling consumer habit. The supply side has cleared capability thresholds; the constraint now lies in demand-side adoption. This aligns with the existing claim that GenAI adoption in entertainment will be gated by consumer acceptance rather than technology capability.
+
+## Evidence
+- Seedance 2.0 achieves near-perfect hand anatomy scores with complex finger movements (magician shuffling cards, pianist playing) showing zero visible hallucinations
+- Native 2K resolution (2048x1080 landscape) represents a 2x improvement over Seedance 1.5 Pro's 1080p maximum
+- Dynamic duration extends to 15 seconds per generation, the longest in the flagship category
+- Sora standalone app: 12 million downloads but retention below 8% at day 30, versus 30%+ benchmark for top consumer apps
+
+## Challenges
+- The benchmark data uses synthetic test prompts (50+ generations per model, identical prompt set of 15 categories), not real production scenarios. The gap between benchmark performance and production-ready utility may still be significant.
+- Retention data reflects Sora specifically, which may have unique product-market fit issues unrelated to the broader category
+
+---
+
+Relevant Notes:
+- [[GenAI adoption in entertainment will be gated by consumer acceptance not technology capability]] — this claim is directly supported by the Sora retention evidence
+- [[consumer definition of quality is fluid and revealed through preference not fixed by production value]] — quality thresholds being cleared shifts the moat from capability to consumer preference
+
+Topics:
+- [[entertainment]]
+- [[ai-video-generation]]
+- [[adoption-curves]]
+- [[demand-constraints]]
diff --git a/domains/entertainment/ai-video-benchmarks-2026-capability-milestone.md b/domains/entertainment/ai-video-benchmarks-2026-capability-milestone.md
new file mode 100644
index 00000000..796dd734
--- /dev/null
+++ b/domains/entertainment/ai-video-benchmarks-2026-capability-milestone.md
@@ -0,0 +1,40 @@
+---
+type: claim
+domain: entertainment
+description: "2026 AI video benchmarks show Seedance 2.0 leading in creative control while Kling 3.0 leads in ease of use, with capability gaps to traditional video narrowing across resolution, duration, and realism dimensions"
+confidence: likely
+source: "AI Journal / Evolink AI / Lantaai benchmark review, 2026-02-01"
+created: 2026-03-10
+---
+
+# 2026 AI video benchmarks show capability convergence with traditional production across key quality dimensions
+
+The 2026 benchmark data reveals a maturing competitive landscape in AI video generation where leading models have cleared multiple capability thresholds simultaneously. Seedance 2.0 (ByteDance) ranks #1 globally on the Artificial Analysis benchmark, achieving native 2K resolution (2048x1080 landscape / 1080x2048 portrait), dynamic duration from 4 to 15 seconds, and 30% faster throughput than its predecessor. Kling 3.0 edges ahead for straightforward video generation on ease-of-use metrics, while Seedance 2.0 wins for precise creative control. Google Veo 3 adds audio generation capability, representing multimodal integration.
+
+The competitive differentiation has shifted from raw capability (all flagship models now clear the quality threshold) to use-case fit: ease of use versus creative control versus multimodal integration. This mirrors patterns in other technology categories where capability commoditization precedes market consolidation.
+
+## Evidence
+- Seedance 2.0 ranked #1 globally on Artificial Analysis benchmark
+- Native 2K resolution (2048x1080 landscape / 1080x2048 portrait), up from 1080p max in Seedance 1.5 Pro
+- Dynamic duration: 4s to 15s per generation (longest in flagship category)
+- 30% faster throughput than Seedance 1.5 Pro at equivalent complexity
+- Kling 3.0 edges ahead for straightforward video generation (ease of use)
+- Seedance 2.0 wins for precise creative control
+- Google Veo 3 combines visual and audio generation
+- Benchmark methodology: 50+ generations per model, identical prompt set of 15 categories, 4 seconds at 720p/24fps, rated on 6 dimensions by 2 independent reviewers, normalized to 0-100
+
+## Challenges
+- Synthetic benchmark prompts may not reflect real production complexity
+- The benchmark-to-production gap remains unquantified
+
+---
+
+Relevant Notes:
+- [[non ATL production costs will converge with the cost of compute as AI replaces labor across the production chain]] — capability improvements support cost convergence
+- [[media disruption follows two sequential phases as distribution moats fall first and creation moats fall second]] — creation moat erosion accelerating
+
+Topics:
+- [[entertainment]]
+- [[ai-video-generation]]
+- [[benchmarking]]
+- [[competitive-landscape]]
diff --git a/domains/entertainment/hand-anatomy-capability-threshold-crossed.md b/domains/entertainment/hand-anatomy-capability-threshold-crossed.md
new file mode 100644
index 00000000..657d2636
--- /dev/null
+++ b/domains/entertainment/hand-anatomy-capability-threshold-crossed.md
@@ -0,0 +1,36 @@
+---
+type: claim
+domain: entertainment
+description: "Near-perfect hand anatomy scores in 2026 benchmarks signal that AI video has cleared the primary visual quality threshold that distinguished synthetic from real footage"
+confidence: likely
+source: "AI Journal / Evolink AI / Lantaai benchmark review, 2026-02-01"
+created: 2026-03-10
+---
+
+# Hand anatomy capability threshold has been crossed in AI video generation
+
+The 2026 benchmark data demonstrates that hand generation—the most visible "tell" of AI-generated video since 2024—has achieved near-perfect scores. Seedance 2.0 produces complex finger movements (magician shuffling cards, pianist playing) with zero visible hallucinations or warped limbs. This represents a capability threshold crossing that fundamentally changes the quality landscape for AI video.
+
+When hands were consistently distorted in AI-generated video, viewers could reliably distinguish synthetic from real footage. With this barrier removed, the remaining differentiators shift to creative direction, narrative coherence, and stylistic preference—areas where human judgment remains central. The technical moat in video generation has narrowed significantly.
+
+## Evidence
+- Seedance 2.0 achieves near-perfect hand anatomy score on Artificial Analysis benchmark
+- Complex finger movements (magician shuffling cards, pianist playing) render with zero visible hallucinations or warped limbs
+- Hand anatomy was identified as the most visible "tell" of AI-generated video in 2024
+- Supports 8+ languages for phoneme-level lip-sync, further reducing visual tells
+
+## Challenges
+- Benchmark methodology uses synthetic test prompts (50+ generations, 15 categories, 4 seconds at 720p/24fps) rather than real production scenarios
+- The gap between benchmark performance and production-ready utility may still be significant
+
+---
+
+Relevant Notes:
+- [[consumer definition of quality is fluid and revealed through preference not fixed by production value]] — if quality can no longer be visually distinguished, production value as a moat claim collapses
+- [[non ATL production costs will converge with the cost of compute as AI replaces labor across the production chain]] — capability improvements support cost convergence thesis
+
+Topics:
+- [[entertainment]]
+- [[ai-video-generation]]
+- [[quality-thresholds]]
+- [[capability-milestones]]
diff --git a/inbox/archive/2026-02-01-seedance-2-ai-video-benchmark.md b/inbox/archive/2026-02-01-seedance-2-ai-video-benchmark.md
index b0c317b2..f50b2467 100644
--- a/inbox/archive/2026-02-01-seedance-2-ai-video-benchmark.md
+++ b/inbox/archive/2026-02-01-seedance-2-ai-video-benchmark.md
@@ -7,9 +7,14 @@ date: 2026-02-01
 domain: entertainment
 secondary_domains: []
 format: report
-status: unprocessed
+status: processed
 priority: medium
 tags: [ai-video-generation, seedance, production-costs, quality-threshold, capability]
+processed_by: clay
+processed_date: 2026-03-10
+claims_extracted: ["ai-video-adoption-is-demand-constrained.md", "hand-anatomy-capability-threshold-crossed.md", "ai-video-benchmarks-2026-capability-milestone.md"]
+extraction_model: "minimax/minimax-m2.5"
+extraction_notes: "Extracted three claims from the AI video benchmark source. Primary insight is the Sora retention data (<8% D30 despite 12M downloads) demonstrating demand-side constraint, which directly supports the existing claim about GenAI adoption being gated by consumer acceptance. Also extracted claims about hand anatomy capability threshold crossing (quality moat erosion) and the broader 2026 capability milestone showing convergence with traditional production. All claims cite specific benchmark data from the source. Checked against existing claims - no duplicates, though these claims enrich several existing claims about production costs, quality definition, and adoption constraints."
 ---
 
 ## Content