AI Yesterday: A Daily AI News Briefing Tool for Everyone

An open-source project that transforms fast-evolving AI news into accessible English summaries, featuring importance ranking, searchable history, and categorized browsing—all running as a static GitHub Pages site.

• Original Author/Maintainer: SahirVhora
• Source Platform: GitHub
• Original Title: ai-yesterday
• Original Link: https://github.com/SahirVhora/ai-yesterday
• Publication Date: 2026-06-04

The AI field is evolving at a breath-taking pace. Dozens of papers are published, multiple models updated, and countless products launched every day. For average developers, product managers, or tech enthusiasts, keeping up with this rhythm is nearly an impossible task.

Worse still, most AI news sites are either full of marketing jargon, overly technical, or just list links without context. Readers spend a lot of time browsing but struggle to figure out which content is truly important.

AI Yesterday was created to address this pain point. It's not another AI news aggregator; it's a "calm daily briefing"—fewer links, better context, and clear impact labels.

AI Yesterday's design philosophy can be summed up with a few key terms:

Curated, Not Piled: Only collects "yesterday's" AI news daily, auto-scraped via RSS feeds but intelligently filtered. It's not about quantity; every entry must have value.

Accessible, Not Technical: Translates technical jargon into "English that everyone can understand". It's not dumbed down; it's clearly expressed, allowing non-technical readers to grasp the AI development trajectory.

Ranked, Not Equal: Each news item has an importance label—Critical, High, Medium, Low. Readers can instantly identify content that needs priority attention.

Historical, Not Fleeting: All content is archived in JSON format with date history. This means you can trace trends and see how a technical topic has evolved.

Static, Not Dynamic: The entire site is a static GitHub Pages site—no backend server, no database, fast loading, low cost, and easy to maintain.

The core script is scripts/collect_ai_news.py, which is responsible for:

• Collect AI news from high-signal RSS feeds (including arXiv, tech blogs, official announcements, etc.)
• Score news importance based on rules
• Convert technical titles into accessible summaries
• Optional: Use OpenRouter API to enhance summary quality
• Save results in JSON format for version control and historical tracing

The OpenRouter integration is a highlight. If OPENROUTER_API_KEY is set, the script calls large language models to generate higher-quality summaries; if no API key is present, it falls back to rule-based summary generation. This design provides advanced features while maintaining the possibility of zero-cost operation.

index.html is a single-file advanced UI that includes:

• Scrolling weekly trend view showing recent developments by category
• Categorized pages: Models, Products, Research, Business, Policy
• Source quality scoring, marking noisy or unavailable RSS feeds
• Elegant visual design, referencing the styles of modern SaaS products like Linear, Vercel, and Superhuman

The project even offers three social preview image variants:

• preview-linear.png: Dark Linear-style command center (currently in use)
• preview-vercel.png: Clean white Vercel-style launch card
• preview-superhuman.png: Luxurious purple Superhuman-style edit card

This attention to detail reflects the developer's focus on product experience.

.github/workflows/daily-digest.yml is a scheduled GitHub Actions workflow that automatically runs the collection script and updates the site daily. This means:

• Zero server maintenance cost
• Automated content updates
• Complete version history (since each update is a Git commit)

Every morning, the team lead opens AI Yesterday and quickly skims through Critical and High-level news. In 5 minutes, they get up to speed on the most important AI developments of the past 24 hours, providing context for the team's daily work planning.