The Daily

The Daily is something I built for myself to act as a sort of personal daily newspaper. It’s live at jonmoshier.com/daily, and a fresh edition is built every morning. It pulls from the mix of sources I actually want to keep up with:

• arXiv — recent papers, scored for readability
• FRED — Federal Reserve economic charts
• Anthropic news
• Coffee futures & roasting news (for thepress.coffee, the newsletter I publish weekly)
• Top trending projects on GitHub and Hugging Face
• The day’s most-read Wikipedia pages
• Hacker News and Lobsters
• Simon Willison’s blog
• ScienceDaily
• …and more

How it works

Each source has its own scraper. They run every morning, triggered by an EventBridge scheduled rule. New items are deduplicated against a DynamoDB table — so the same story doesn’t resurface day after day — and the result is published as a single edition.json file to S3. The site (built with Astro) reads that JSON and renders the day’s edition.

There’s also an optional curation pass: feed everything into Claude (via the Anthropic API) and let it summarize, enrich, and order the edition. That runs about ~$0.40/day with Sonnet. For now I keep it switched off and let the pipeline curate deterministically, but the LLM path is still there to flip back on whenever I want it.