# AI Daily Feed — Full Guide

## Overview

Daily AI content feed from 40+ bilingual sources (English tech blogs + Chinese WeChat AI accounts). Every article scored 0-100, assigned a verdict, and broken down into structured fields: key takeaways, CEI reasoning, practical relevance.

Updated most weekdays. 427+ articles in the cumulative dataset. Data since 2026-01-26.

## Use Cases

### "What's worth reading today?"

Fetch today's picks grouped by verdict:

  GET https://raw.githubusercontent.com/makinotes/ai-daily-feed/master/lists/daily-picks.json

```json
{
  "date": "2026-02-27",
  "must_read": [
    {
      "title": "Google API Keys Weren't Secrets. But then Gemini Changed the Rules.",
      "link": "https://simonwillison.net/...",
      "source": "Simon Willison",
      "score": 100,
      "why_matters": "API keys silently gained Gemini access — thousands of leaked keys now exploitable"
    }
  ],
  "worth_reading": [...],
  "neutral": [...],
  "noise": [],
  "overhyped": []
}
```

### "What's happening in a specific domain?"

30-day category index:

  GET https://raw.githubusercontent.com/makinotes/ai-daily-feed/master/indexes/by-category.json

```json
{
  "updated": "2026-02-27",
  "categories": {
    "AI/Tech": [
      {
        "title": "...",
        "source": "PaperWeekly",
        "score": 95,
        "verdict": "must_read",
        "pub_date": "2026-02-27",
        "why_matters": "New training optimization may reduce LLM costs by 40%"
      }
    ],
    "Builder 实践": [...],
    "AI 使用": [...]
  }
}
```

### "What's trending this week?"

  GET https://raw.githubusercontent.com/makinotes/ai-daily-feed/master/indexes/trending.json

```json
{
  "updated": "2026-02-27",
  "window": "7d",
  "article_count": 107,
  "trending": [
    {"keyword": "Anthropic", "count": 9},
    {"keyword": "Agent", "count": 8}
  ]
}
```

### "Full article data for deep analysis"

  GET https://raw.githubusercontent.com/makinotes/ai-daily-feed/master/api/{YYYY-MM-DD}.json

```json
{
  "date": "2026-02-27",
  "total": 39,
  "verdict_counts": {"must_read": 5, "worth_reading": 14, "neutral": 20},
  "articles": [
    {
      "title": "Google API Keys Weren't Secrets. But then Gemini Changed the Rules.",
      "link": "https://simonwillison.net/...",
      "source": "Simon Willison",
      "source_channel": "overseas",
      "category": "AI/Tech",
      "pub_date": "2026-02-26",
      "summary": "Google silently expanded API key permissions to include Gemini model access...",
      "core_point": "Google API keys now grant Gemini access by default → Thousands of leaked keys become attack vectors → Developers must audit existing key permissions",
      "highlights": [
        "API keys that were previously low-risk now grant Gemini access",
        "Google made this change without notifying developers"
      ],
      "why_matters": "If you have Google API keys in public repos or client-side code, they may now be exploitable",
      "score": 100,
      "level": "精读",
      "verdict": "must_read"
    }
  ]
}
```

### "Training data for a content scoring model"

  GET https://raw.githubusercontent.com/makinotes/ai-daily-feed/master/datasets/scored-articles.jsonl

Each line is one article with all fields. 427+ articles, growing daily. Deduplicated by (pub_date, title). See datasets/README.md for methodology.

## Article Schema

### Core Fields
- title: Article headline (string)
- link: Original URL (string)
- source: Publisher or author name (string)
- source_channel: "overseas" (English) or "wechat-ai" (Chinese WeChat)
- category: "AI/Tech", "Builder 实践", "AI 使用", etc.
- pub_date: Publication date, YYYY-MM-DD

### Analysis Fields
- summary: One-sentence overview
- core_point: CEI reasoning — Claim → Evidence → Implication
- highlights: 2-3 key takeaways (array of strings)
- why_matters: Practical relevance — when and how to apply

### Scoring Fields
- score: Quality score 0-100, algorithm-driven
- level: Reading priority — "精读" (deep read), "收藏" (bookmark), "速览" (skim)
- verdict: must_read | worth_reading | neutral | overhyped

## Verdict Definitions

- must_read: High quality, novel insights — prioritize
- worth_reading: Good depth, novelty, or actionability — worth your time
- neutral: Acceptable, nothing remarkable
- overhyped: Appears important but lacks substance

## Integration Patterns

**RAG Pipeline** — Index highlights + why_matters. Use verdict to weight relevance. Filter by category for domain-specific indices.

**Daily Briefing** — Fetch lists/daily-picks.json. Summarize must_read using summary + highlights.

**Knowledge Base** — Ingest api/{date}.json. Use core_point for reasoning chains. Track categories for emerging themes.

**MCP / Tool Use** — Point your agent at the raw GitHub URLs. Static files, no auth needed.

## Data Availability

- Coverage: 2026-01-26 to present
- Frequency: Most weekdays
- Volume: 25-40 articles/day after filtering
- Sources: 40+ curated (English + Chinese)
- Dataset: 427+ cumulative articles
- Note: articles before mid-February 2026 may have empty why_matters and core_point fields (added to pipeline later, not backfilled)
- Gaps: Some dates may be missing (holidays, maintenance)

## License

CC BY-NC 4.0. Non-commercial, personal use only. Article metadata and AI-generated analysis only — original content belongs to respective authors. Content owners can request removal via GitHub issues.