Daily Word

Overview

Daily Word was built to make a fragmented content journey feel effortless. Instead of forcing readers to jump between websites, YouTube, and newsletters, the product assembles the day's Catholic readings, commentary, reflection, AI summary, and illustration into one focused destination in Polish.

Key Highlights

• Daily content pipeline: readings and reflections are scraped from external sources, YouTube commentary is matched by date, and the full page is assembled automatically.
• AI where it matters: summaries, commentary timestamps, hero copy, and gospel illustrations are generated with a multi-step fallback strategy instead of a single brittle prompt.
• Built for reliability: cache hits still trigger missing-field backfills, so the experience improves even when upstream content appears late.
• Distribution included: a daily newsletter is sent only when the right content is ready, using an idempotent send log to avoid duplicates.

Technical Notes

• Next.js App Router with async server components, SSR, and route-level revalidation for both today's page and recent history.
• Supabase used as the content cache, subscriber store, send log, and image storage layer.
• Summary generation cascades through transcript mode, Gemini video mode, and gospel-text fallback, with deterministic transcript parsing for commentary timestamps.
• Vercel Cron pre-fills the cache every morning, but page visits can also safely complete any missing parts of the pipeline.

Results

• Turned a manual daily routine into a repeatable digital product with automation across content ingestion, AI enrichment, and delivery.
• Created a full-stack content system that stays useful even when upstream sources are incomplete or late.
• Extended the product beyond text by adding AI-generated visuals and newsletter distribution without breaking the calm reading experience.