Behind the Scenes

Full transparency into how this digest is produced. Every article is AI-synthesized from multiple sources, scored, clustered, and editorially generated — all autonomously.

Processing Pipeline

Init
Initialize
Fetch
Collect articles
AI Analysis
Score & summarize
AI Curation
Group & select
AI Synthesis
Editorial writing
Publish
Generate & publish

← swipe →

Init — Initialize

Resets processing state for the target date and ensures all data sources are synchronized. Makes every run idempotent — safe to re-run anytime.

Fetch — Collect articles

Collects articles from 30+ curated RSS sources with automatic retry and deduplication. Each source contributes up to 5 of its latest articles. Yesterday's content appears in today's digest, like a morning newspaper.

AI Analysis — Score & summarize

AI reads each article and produces a relevance score (0–10), bilingual headlines (EN/ZH), and concise summaries. Articles unrelated to AI are filtered out. All articles get bilingual titles for the full source overview.

AI Curation — Group & select

AI reviews all scored articles and groups related stories into thematic clusters, then ranks them by editorial importance, source diversity, and reader interest. The top 10 stories become the day's digest.

AI Synthesis — Editorial writing

For each selected story, AI reads the full source articles and writes an original bilingual editorial — weaving facts from multiple sources, noting disagreements, and providing context. Original voice, never copy-pasted.

Publish — Generate & publish

The finished digest is assembled into a structured document with full source attribution and automatically published to the website.

Data Sources

Technology Stack

GitHub Actions
Daily Automation
MiniMax M2.5
AI Processing
Turso
Cloud Database
Astro 5
Static Site
0:00