The Problem
I kept opening news apps and feeling worse. Too much noise, too many opinions, not enough signal.
I wanted one place that told me the 10 most important things happening in the world today — ranked by actual coverage reach, not by engagement bait.
So I built it.
What Dawnly Does
Every day at 6AM EST, an automated pipeline:
- Pulls headlines from 50+ RSS feeds across Tier A/B/C sources
- Clusters similar stories using sentence-transformers and DBSCAN
- Ranks clusters by mention count × source tier weight × geographic diversity
- Summarizes the top 10 using Claude AI
- Publishes to GitHub Pages as a clean static site
The result resets every 24 hours. No archive, no scroll. Just today.
What I Learned
Building Dawnly was the first time I owned a full product end-to-end — not just an analysis or a dashboard, but something that runs on its own every day without me touching it.
The hardest part wasn't the NLP. It was the pipeline reliability — making sure it degrades gracefully when a feed is down, when Claude returns unexpected output, or when GitHub Actions has a bad day.
That's the real engineering work nobody talks about.
Try It
dawnly.news — resets daily at 6AM EST.