FavCollege
College comparison platform (kulisama81/favcollege) built with Next.js 15 and Supabase. 74% Claude commits across Jan-Feb 2026. First project with 16 “DO NOT” rules in CLAUDE.md and first use of Opus 4.6 alongside Sonnet 4.5.
Architecture
Two data pipelines feed the platform:
-
College Scorecard API — US Dept of Education’s free API providing 40+ fields per college across ~7,000 institutions. Imported via a mapper that auto-generates tags from Carnegie classifications and normalizes data to application types. See source-college-scorecard-api.
-
Student Review Scraping — Twitter and Reddit scraping with sentiment-analysis pipeline (keyword-based, -1 to +1 normalized to 1-5 stars), topic categorization (7 categories), and a denormalized cache pattern for fast page loads. See source-review-scraping.
Both use the provider-pattern — a USE_MOCK_DATA toggle switches between mock data and Supabase-backed real data.
Notable Patterns
- GitHub Actions integration (
create-ticket.yml) for mobile ticket creation from Issues - 16 “DO NOT” rules in CLAUDE.md — the most extensive at the time
- Denormalized caching: aggregate review stats cached on colleges table, synced by separate script
See also: source-college-scorecard-api, source-review-scraping, source-claude-usage-patterns