Global AI News Aggregation and Summary

The Problem

Tracking AI progress often means switching between arXiv, TechCrunch, OpenAI Blog, and Chinese tech media. Signals are scattered, duplicates are common, and marketing filler makes manual digesting slow. This skill turns source collection, filtering, summarization, and layout into a repeatable pipeline for daily or weekly reports.

How It Works

• Multi-source collection: Covers arXiv, TechCrunch, Jiqizhixin, QbitAI, OpenAI, Google DeepMind, 36Kr, The Verge, Meta AI, and Zhidongxi, with time filtering, title-source deduplication, and ad filtering.
• Processing: Removes low-quality content, tags items into research, industry, engineering, and policy categories, and prioritizes top-conference papers and major-vendor updates.
• Output: Produces 80–150-character Chinese summaries and 50–120-word English summaries in consistent Markdown, suitable for docs or channels.
• Archiving: If MCP auth is configured, it can append dated rows to Tencent Docs or Feishu sheets, preserving the full Markdown digest and trend note.

Limits

It is aimed at recent AI industry news, typically within the last 30 days. Daily and weekly item caps apply, and the skill skips requests focused only on model theory, code, or deployment tutorials. It produces structured Markdown reports, not images, table attachments, formal due diligence, or legal compliance review.