A daily AI briefing that translates yesterday's most useful AI developments into plain-English summaries, importance flags, source links and searchable history.
AI moves too fast for normal people to track manually. AI Yesterday collects high-signal RSS feeds from trusted sources every day, ranks stories by likely importance, and converts technical headlines into plain-English explanations:
| Feature | What It Means |
|---|---|
| Plain-English summaries | Every story is translated into clear, jargon-free language |
| Importance flags | Critical, High, Medium or Low - so you know what's worth your attention |
| Source links | Every item links back to the original article |
| Searchable history | All past digests archived by date, searchable from the page |
| Weekly trends | See which categories are heating up over the past 7 days |
| Category pages | Browse by Models, Products, Research, Business, or Policy |
| Source quality scores | Each RSS feed is scored so noisy sources are visible |
Three premium social preview variants are included:
preview-linear.png- dark Linear-style command centre (active aspreview.png)preview-vercel.png- clean white Vercel-style launch cardpreview-superhuman.png- luxury purple Superhuman-style editorial card
Building on the same idea as my Year 4 Prep project - information you need, organised automatically, so you don't have to hunt for it yourself. AI news is the same problem: too many sources, too little signal, and nobody curating it with a non-technical reader in mind.
The site runs as a static GitHub Pages page with no backend. A scheduled GitHub Action collects, scores, enriches (via OpenRouter), and archives the digest daily.
| File | Purpose |
|---|---|
index.html |
Single-file premium UI (dark/light theme, search, archive) |
preview.png |
Active 1280x640 social preview image (Linear variant) |
preview-linear.png |
Dark Linear-style preview variant |
preview-vercel.png |
Clean white Vercel-style preview variant |
preview-superhuman.png |
Purple Superhuman-style preview variant |
scripts/collect_ai_news.py |
Collector, scorer, history archiver, OpenRouter enricher |
scripts/generate_preview.py |
Preview image generator using Pillow |
data/digest.json |
Latest daily digest |
data/history/*.json |
Archived daily snapshots |
python3 scripts/collect_ai_news.py
python3 -m http.server 8777Then open http://localhost:8777.
Set environment variables before running the collector:
export OPENROUTER_API_KEY="..."
export OPENROUTER_MODEL="openrouter/free"
python3 scripts/collect_ai_news.pyFor GitHub Actions, add OPENROUTER_API_KEY as a repository secret. The scheduled workflow enriches top items and stops on rate limits so it does not burn through the free quota. No key = falls back to rules-based summaries.
pip install pillow
python3 scripts/generate_preview.pyFeeds monitored daily for new signals. Source quality is scored and visible on the site.
Hacker News AI, MIT Technology Review AI, Ars Technica AI, VentureBeat AI, The Verge AI, OpenAI, Hugging Face, Google DeepMind, arXiv AI
Powered by GitHub Actions and OpenRouter. Built because busy humans deserve to understand AI without drowning in AI news.