Social Listen Data Pipeline + Registry Expansion
- NewSocial data persistence — all raw social items (YouTube, TikTok, X, Reddit, Blog) are now stored in Supabase for future content discovery and insights.
- NewWeekly query rotation — 3 query variants per platform, rotated weekly to discover different content over time instead of re-fetching the same results.
- NewAdded 190 new programs across 6 batches: schema hardening, PartnerStack delta (+96), AI-native wave 1 (+25), AI wave 2 (+25), YC AI companies (+20), marketplace scans (+24).
- NewSVG badge embed — embeddable Affiliate Score badge with size variants and web component support.
- ImprovedMaximized data collection per API call — YouTube RapidAPI 6→50 results, Apify 20→30, removed slice limits on TikTok and X. Zero cost increase (per-request pricing).
- ImprovedSocial Listen now uses PPR (Partial Prerendering) with dynamic holes — no more API calls during build, fixing the 10-minute build time regression.
- ImprovedYouTube fetch priority swapped: RapidAPI (free) first, Apify ($0.03/call) as fallback. Saves ~$22/day on Apify costs.
- ImprovedAffiliate Score algorithm redesigned (0–100) with better commission parsing and tier-aware scoring.
- FixedBuild time regression (10+ minutes) caused by API calls during static generation for all 446 programs.
- Fixed98 missing logos from Batch 2 PartnerStack delta import.
- FixedVercel deploy failures — inline index build, tolerate missing workspace packages, disable lint during build.