Paste everything below the line, unchanged, into each model. Do not adapt it per tier.
Do not send a follow-up message.
Build me a live world news globe. Working app, real data, one shot.
A user lands on a topic screen and picks one or more subjects — Health, Politics, Science, Tech, Business, Sport, Climate, Conflict. They confirm, and the app pulls the 50 most recent news items for those topics, live.
Each item is pinned onto a 3D globe at the place in the world it is about. The user spins the globe, finds a pin, clicks it, reads the summary, and can open the original source.
All news comes from the SocialCrawl API. Do not use a news API, do not scrape sites, do not seed with sample data.
Base URL: <https://www.socialcrawl.dev>
Auth: x-api-key: $SOCIALCRAWL_API_KEY (header, every request)
Envelope: { success, platform, endpoint, data, credits_used, credits_remaining, request_id, cached }
Primary endpoint:
GET /v1/search/everywhere?query=<topic query>&lookback_days=2
It fans one query out across 12 sources (reddit, twitter-ai-search, youtube, tiktok, instagram, hackernews, polymarket, github, threads, pinterest, perplexity, tavily) and returns ranked, clustered candidates with engagement and freshness signals. It costs 20 credits per call.
sources, exclude, from_date and to_date are also accepted.
Sending Accept: text/event-stream streams typed SSE chunks (meta, source_started, items, source_failed, ranked_final, clusters, done) instead of returning one JSON body.
Narrower endpoints exist if you want them:
/v1/reddit/subreddit?subreddit=worldnews/v1/hackernews/search?query=/v1/twitter/ai-search?query=