Ai Daily

Name: Ai Daily
Author: geekjourneyx

geekjourneyx/mind-skills

260 installs
26 repo stars
Updated February 8, 2026
geekjourneyx/mind-skills

Stay current on AI news, model releases, and tooling shifts so you can spot opportunities and inform product or agent design before you commit to a build.

About

ai-daily from mind-skills supports a recurring research ritual: surfacing what changed in AI each day so builders and PMs can connect fresh capabilities to ideas, competitive moves, and agent workflows before scoping or coding.

Daily AI trend digest
Model and tool release tracking
Informs early ideation
Reduces blind spots on new capabilities
Lightweight research habit

Ai Daily by the numbers

260 all-time installs (skills.sh)
Ranked #2,479 of 16,546 AI & Agent Building skills by installs in the Skillselion catalog
Data as of Jul 24, 2026 (Skillselion catalog sync)

npx skills add https://github.com/geekjourneyx/mind-skills --skill ai-daily

Add your badge

Show developers this skill is listed on Skillselion. Paste this into your README.

[![Listed on Skillselion](https://skillselion.com/badge/skills/geekjourneyx/mind-skills/ai-daily.svg)](https://skillselion.com/skills/geekjourneyx/mind-skills/ai-daily)

Installs	260
repo stars	★ 26
Last updated	February 8, 2026
Repository	geekjourneyx/mind-skills ↗

What it does

Stay current on AI news, model releases, and tooling shifts so you can spot opportunities and inform product or agent design before you commit to a build.

Files

SKILL.mdMarkdownGitHub ↗

AI Daily News

Fetch AI industry news from smol.ai RSS feed.

Quick Start

# Basic queries
昨天AI资讯
今天的AI新闻
2026-01-13的资讯

Query Types

Type	Examples	Description
Relative date	`昨天AI资讯` `今天的新闻` `前天`	Yesterday, today, day before
Absolute date	`2026-01-13的新闻`	YYYY-MM-DD format
Date range	`有哪些日期的新闻`	Show available dates

Workflow

- [ ] Step 1: Parse date from user request
- [ ] Step 2: Fetch RSS data
- [ ] Step 3: Check content availability
- [ ] Step 4: Format and display results

---

Step 1: Parse Date

User Input	Target Date	Calculation
`昨天`	Yesterday	today - 1 day
`前天`	Day before	today - 2 days
`今天`	Today	Current date
`2026-01-13`	2026-01-13	Direct parse

Format: Always use YYYY-MM-DD

---

Step 2: Fetch RSS

python skills/ai-daily/scripts/fetch_news.py --date YYYY-MM-DD

Available commands:

# Get specific date
python skills/ai-daily/scripts/fetch_news.py --date 2026-01-13

# Get date range
python skills/ai-daily/scripts/fetch_news.py --date-range

# Relative dates
python skills/ai-daily/scripts/fetch_news.py --relative yesterday

Requirements: pip install feedparser requests

---

Step 3: Check Content

When NOT Found

Sorry, no news available for 2026-01-14

Available date range: 2026-01-10 ~ 2026-01-13

Suggestions:
- View 2026-01-13 news
- View 2026-01-12 news

---

Step 4: Format Results

Example Output:

# AI Daily · 2026年1月13日

> not much happened today

## Content

[News content from smol.ai RSS...]

---
Source: smol.ai

---

Configuration

Variable	Description	Default
RSS_URL	RSS feed URL	`https://news.smol.ai/rss.xml`

No API keys required.

---

Troubleshooting

Issue	Solution
RSS fetch fails	Check network connectivity
Invalid date	Use YYYY-MM-DD format
No content	Check available date range

---

References

Output Format - Markdown template
HTML Themes - Webpage theme specifications

HTML Theme Prompts

This document contains prompt templates for generating themed HTML webpages.

---

Apple Style Theme

Generate a clean, minimalist HTML page inspired by Apple's design language.

Design Principles

Minimalist philosophy: Every pixel serves a purpose
Generous white space: Content density ≤ 40%
Typography-first: SF Pro Display for headings, SF Pro Text for body
Subtle visual effects: Gentle glows, smooth transitions
Rim lighting: Studio-quality soft lighting

Color Palette

--bg-color: #000000              /* Pure black background */
--glow-start: #0A1929           /* Deep blue */
--glow-end: #1A3A52             /* Midnight blue */
--title-color: #FFFFFF           /* Pure white */
--text-color: #E3F2FD            /* High brightness blue-white */
--accent-color: #42A5F5          /* Sky blue for highlights */
--secondary-color: #B0BEC5       /* Medium gray for secondary info */

Structure

<body>
  <div class="background-glow"></div>
  <div class="geometric-lines"></div>

  <div class="container">
    <header>
      <div class="logo-icon">🤖</div>
      <h1>AI Daily</h1>
      <div class="date-badge">2026年1月13日 星期一</div>
    </header>

    <main>
      <section class="summary-card">
        <h2>核心摘要</h2>
        <!-- Summary items -->
      </section>

      <section class="category-section">
        <h2>模型发布</h2>
        <!-- News cards -->
      </section>

      <!-- More categories -->

      <footer class="keywords-footer">
        #Anthropic #Google #MedGemma
      </footer>
    </main>
  </div>
</body>

Key CSS Effects

/* Background glow - bottom-right corner */
.background-glow {
  position: fixed;
  bottom: -20%;
  right: -20%;
  width: 70%;
  height: 70%;
  background: radial-gradient(
    circle at center,
    var(--glow-end) 0%,
    var(--glow-start) 40%,
    transparent 80%
  );
  opacity: 0.6;
  filter: blur(80px);
  z-index: -2;
}

/* Geometric lines - 15% opacity */
.geometric-lines {
  position: fixed;
  background-image:
    linear-gradient(90deg, transparent 49%, var(--accent-color) 50%, transparent 51%),
    linear-gradient(0deg, transparent 49%, var(--accent-color) 50%, transparent 51%);
  background-size: 200px 200px;
  opacity: 0.08;
  z-index: -1;
}

/* Glowing logo icon */
.logo-icon {
  animation: glow-pulse 3s ease-in-out infinite;
}

@keyframes glow-pulse {
  0%, 100% {
    box-shadow: 0 0 40px var(--accent-color), 0 0 80px var(--accent-color);
  }
  50% {
    box-shadow: 0 0 50px var(--accent-color), 0 0 100px var(--accent-color);
  }
}

Full Prompt

Generate an HTML webpage with Apple-style minimalist design for AI news summary.

## Content
{markdown_content}

## Design Requirements

**Colors** (Black background with blue glow):
- Background: #000000
- Title: #FFFFFF
- Body text: #E3F2FD
- Accent: #42A5F5 (for highlights and icons)
- Secondary: #B0BEC5

**Layout**:
- Centered container, max-width 900px
- Content density ≤ 40% (generous white space)
- Padding: 40px 20px

**Typography**:
- Headings: SF Pro Display, 24-48px, pure white
- Body: SF Pro Text, 16px, #E3F2FD
- Line height: 1.6

**Visual Effects**:
- Background glow from bottom-right corner (radial gradient, blur 80px)
- Subtle geometric lines pattern (15% opacity)
- Glowing logo icon with pulse animation
- Smooth hover transitions on cards

**Structure**:

<header>: Logo + Date badge <main>:

Summary card (rounded, subtle border)
Category sections with cards
Keywords footer

<footer>: Copyright


**Card Style**:
- Background: rgba(255,255,255,0.05) with backdrop-filter blur
- Border: 1px solid rgba(255,255,255,0.1)
- Border radius: 12px
- Padding: 20px
- Hover: border color changes to accent color

**Output**: Complete single HTML file with inline CSS, no external dependencies.

---

Ocean Calm Theme (深海蓝)

Business professional style for product announcements.

Color Palette

--bg-color: #000000
--glow-start: #0F1C3F           /* Deep indigo */
--glow-end: #1A2F5A             /* Royal blue */
--title-color: #FFFFFF
--text-color: #E3F2FD
--accent-color: #5C9FE5          /* Bright blue */
--secondary-color: #BBDEFB       /* Light blue */

Prompt

Generate a professional business-style HTML webpage for AI news.

## Content
{markdown_content}

## Design: Ocean Calm Theme

Professional blue tones suitable for business updates and product launches.

**Colors**:
- Background: #000000
- Glow: Deep indigo (#0F1C3F) to royal blue (#1A2F5A)
- Title: #FFFFFF
- Body: #E3F2FD
- Accent: #5C9FE5 (bright blue)
- Secondary: #BBDEFB

**Style**: Corporate, trustworthy, data-driven.

---

Autumn Warm Theme (秋日暖阳)

Warm vibrant style for community updates and lively discussions.

Color Palette

--bg-color: #000000
--glow-start: #1F1410           /* Deep brown */
--glow-end: #3D2415             /* Warm brown */
--title-color: #FFFFFF
--text-color: #FFF3E0            /* Warm white */
--accent-color: #FFA726          /* Bright orange */
--secondary-color: #FFCCBC       /* Light orange */

Prompt

Generate a warm, vibrant HTML webpage for AI news.

## Content
{markdown_content}

## Design: Autumn Warm Theme

Warm orange tones for community-focused content.

**Colors**:
- Background: #000000
- Glow: Deep brown (#1F1410) to warm brown (#3D2415)
- Title: #FFFFFF
- Body: #FFF3E0 (warm white)
- Accent: #FFA726 (bright orange)
- Secondary: #FFCCBC

**Style**: Energetic, friendly, community-focused.

---

Theme Comparison

Theme	Style	Best For
Apple Style	Minimalist, professional	Technical content, product reviews
Ocean Calm	Corporate, business	Business updates, announcements
Autumn Warm	Energetic, friendly	Community news, discussions

---

Usage in Skill

When user selects a theme, apply the corresponding prompt template and generate the HTML.

Example workflow:

1. User: "昨天AI资讯，生成网页" 2. Claude: "可选主题: 苹果风 / 深海蓝 / 秋日暖阳" 3. User: "苹果风" 4. Claude: Uses Apple Style Theme prompt to generate HTML 5. Save to docs/{date}.html

Markdown Output Format

This document defines the standard markdown output format for AI Daily news.

---

Template

# AI Daily · {年}年{月}月{日}日

> {一句话核心摘要}

## 核心摘要

{3-5条核心要点，每条一句话}

## {分类名称}

### [{标题}](链接)

{详细摘要}

**关键信息**: {相关标签}

---

数据来源: smol.ai
生成时间: {YYYY-MM-DD HH:MM}

---

Complete Example

# AI Daily · 2026年1月13日

> Anthropic 整合 Agent 产品线，Google 发布医疗 AI 模型，LangChain 推出 Agent 构建工具

## 核心摘要

- Anthropic 发布 Cowork 统一 Agent 平台，整合 Claude Code 和 MCP
- Google 开源 MedGemma 1.5 医疗多模态模型，支持 3D 影像分析
- LangChain Agent Builder 正式发布，支持内存和人工审核循环
- 社区讨论"Vibe Coding"定义，强调工程师验证的重要性
- 开源项目快速复制 Cowork 功能，Agent 技术商品化加速

## 模型发布

### [MedGemma 1.5](https://news.smol.ai/issues/26-01-13-not-much/)

Google 发布 4B 参数医疗多模态模型 MedGemma 1.5，专为离线医疗场景设计。支持 3D 体积（CT/MRI）处理、纵向对比和解剖定位。声称在 EHR 理解上达到 89.6% 准确率（+22%），X 光定位达到 38% IoU。

**关键信息**: Google, MedGemma, 医疗AI, 多模态, 3D影像

### [Open 复现 Cowork](https://news.smol.ai/issues/26-01-13-not-much/)

开发者使用 QEMU + bubblewrap + seccomp 构建了跨平台类 Cowork VM 环境。这表明 Agent shell 技术正在快速商品化，成为基础设施而非产品护城河。

**关键信息**: 开源, QEMU, Agent, 虚拟化

## 产品动态

### [Cowork 品牌整合](https://news.smol.ai/issues/26-01-13-not-much/)

Anthropic 将其 AI agent 产品统一到 Cowork 品牌，整合了之前的 Claude Code、Claude for Chrome 等工具。使用 Apple 的虚拟化技术和 bubblewrap 实现安全沙箱。

**关键信息**: Anthropic, Cowork, Claude Code, MCP

### [LangChain Agent Builder GA](https://news.smol.ai/issues/26-01-13-not-much/)

LangChain 宣布 Agent Builder 正式发布（GA），提供无代码但强大的 agent 编排功能：内存、触发器、人工审核循环和 agent 收件箱。

**关键信息**: LangChain, Agent Builder, 编排, 工具链

## 研究方向

### [MemRL: 记忆即强化学习](https://news.smol.ai/issues/26-01-13-not-much/)

DAIR AI 强调 MemRL 方法，将记忆检索视为强化学习问题。保持基础模型冻结，学习情节记忆的 Q 值（意图-经验-效用），两阶段检索：语义过滤 + 效用排序。

**关键信息**: MemRL, 强化学习, 记忆检索, DAIR AI

### [递归语言模型 (RLMs)](/root/ai-daily-skill/docs/2026-01-12.html)

Omar Khattab 等人指出大多数"子 agent"实现错过了核心思想：需要类似指针的符号访问提示词来递归遍历。这可以实现超过 1000 万 tokens 的上下文而无需重新训练。

**关键信息**: RLMs, 递归模型, 长上下文, 符号访问

## 工具框架

### [Diffusers 统一注意力后端](https://news.smol.ai/issues/26-01-13-not-much/)

Hugging Face Diffusers 发布统一注意力后端，结合了 Ring 和 Ulysses 的属性。这是持续推动注意力内核/后端可互换和性能可移植化的一部分。

**关键信息**: Hugging Face, Diffusers, 注意力机制

### [量化的注意事项](https://news.smol.ai/issues/26-01-13-not-much/)

TensorPro 报告称 MXFP4 量化的注意力可能破坏因果建模，发布了诊断和修复"泄漏量化"行为的方法。对于从业者："以 FP8/4 位训练"日益可行，但数值边缘情况仍是活跃的研究/运维问题。

**关键信息**: 量化, FP8, MXFP4, 数值稳定性

## 关键词

#Anthropic #Google #MedGemma #LangChain #Agent #MemRL #RLMs #Diffusers #量化

---

数据来源: smol.ai
生成时间: 2026-01-15 10:30

---

Category Names

Use these Chinese category names:

Category	Chinese Name	Icon
Model Releases	模型发布	🤖
Product Updates	产品动态	💼
Research	研究论文	📚
Tools & Frameworks	工具框架	🛠️
Funding & M&A	融资并购	💰
Industry Events	行业事件	🏆

---

Output Guidelines

1. Title format: # AI Daily · {年}年{月}月{日}日 2. Summary: 3-5 bullet points, one sentence each 3. Categories: Use category sections above 4. Links: Include original smol.ai links 5. Keywords: 5-10 hashtags, comma separated 6. Footer: Source and generation time

#!/usr/bin/env python3
"""
AI Daily News Fetcher
Fetches AI news from smol.ai RSS and returns structured data.
"""
import sys
import json
import argparse
from datetime import datetime, timezone, timedelta
from pathlib import Path

try:
    import feedparser
    import requests
except ImportError:
    print("Error: Required packages not installed.")
    print("Run: pip install feedparser requests")
    sys.exit(1)

# RSS URL
RSS_URL = "https://news.smol.ai/rss.xml"
REQUEST_TIMEOUT = 30


def fetch_rss():
    """Download and parse RSS from smol.ai"""
    try:
        response = requests.get(RSS_URL, timeout=REQUEST_TIMEOUT)
        response.raise_for_status()
        return feedparser.parse(response.content)
    except requests.RequestException as e:
        print(json.dumps({"error": f"Failed to fetch RSS: {e}"}))
        sys.exit(1)


def get_date_range(feed):
    """Get available date range from RSS entries

    Returns:
        tuple: (min_date, max_date) in YYYY-MM-DD format, or (None, None)
    """
    dates = []
    for entry in feed.entries:
        # Method 1: Parse from link (format: .../issues/YY-MM-DD-...)
        if hasattr(entry, 'link'):
            date_from_link = extract_date_from_link(entry.link)
            if date_from_link:
                dates.append(date_from_link)

        # Method 2: Parse from pubDate
        elif hasattr(entry, 'published_parsed') and entry.published_parsed:
            dt = datetime(*entry.published_parsed[:6], tzinfo=timezone.utc)
            dates.append(dt.strftime("%Y-%m-%d"))

    if not dates:
        return None, None

    return min(dates), max(dates)


def extract_date_from_link(link):
    """Extract date from RSS link

    Args:
        link: URL string like https://news.smol.ai/issues/26-01-13-not-much/

    Returns:
        Date string in YYYY-MM-DD format, or None
    """
    import re

    patterns = [
        r'/issues/(\d{2})-(\d{2})-(\d{2})-',  # YY-MM-DD
        r'/issues/(\d{4})-(\d{2})-(\d{2})-',  # YYYY-MM-DD
    ]

    for pattern in patterns:
        match = re.search(pattern, link)
        if match:
            year, month, day = match.groups()
            if len(year) == 2:
                year = "20" + year
            return f"{year}-{month}-{day}"

    return None


def get_content_by_date(feed, target_date):
    """Extract content for a specific date

    Args:
        feed: Feedparser parsed feed
        target_date: Date string in YYYY-MM-DD format

    Returns:
        dict with keys: title, link, content, pubDate, or None if not found
    """
    for entry in feed.entries:
        # Check by link date
        if hasattr(entry, 'link'):
            date_from_link = extract_date_from_link(entry.link)
            if date_from_link and date_from_link == target_date:
                return extract_entry_content(entry)

        # Check by pubDate
        if hasattr(entry, 'published_parsed') and entry.published_parsed:
            dt = datetime(*entry.published_parsed[:6], tzinfo=timezone.utc)
            entry_date = dt.strftime("%Y-%m-%d")
            if entry_date == target_date:
                return extract_entry_content(entry)

    return None


def extract_entry_content(entry):
    """Extract content from an RSS entry

    Returns:
        dict with keys: title, link, content, pubDate
    """
    content = {
        "title": entry.get("title", ""),
        "link": entry.get("link", ""),
        "pubDate": ""
    }

    # Get published date
    if hasattr(entry, 'published'):
        content["pubDate"] = entry.published

    # Get full content
    if hasattr(entry, 'content') and entry.content:
        content["content"] = entry.content[0].get('value', '')
    elif hasattr(entry, 'summary'):
        content["content"] = entry.summary
    else:
        content["content"] = content.get("title", "")

    # Clean HTML entities
    content["content"] = content["content"].replace('&lt;', '<').replace('&gt;', '>').replace('&amp;', '&')

    return content


def main():
    parser = argparse.ArgumentParser(description='Fetch AI news from smol.ai')
    parser.add_argument('--date-range', action='store_true', help='Show available date range')
    parser.add_argument('--date', type=str, help='Get content for specific date (YYYY-MM-DD)')
    parser.add_argument('--relative', type=str, choices=['yesterday', 'today', 'day-before'],
                       help='Relative date: yesterday, today, day-before')

    args = parser.parse_args()

    # Fetch RSS
    feed = fetch_rss()

    # Date range mode
    if args.date_range:
        min_date, max_date = get_date_range(feed)
        print(json.dumps({
            "min_date": min_date,
            "max_date": max_date,
            "total_entries": len(feed.entries)
        }, indent=2))
        return

    # Calculate target date
    if args.relative:
        if args.relative == 'yesterday':
            target_date = (datetime.now(timezone.utc) - timedelta(days=1)).strftime("%Y-%m-%d")
        elif args.relative == 'day-before':
            target_date = (datetime.now(timezone.utc) - timedelta(days=2)).strftime("%Y-%m-%d")
        else:  # today
            target_date = datetime.now(timezone.utc).strftime("%Y-%m-%d")
        date_arg = target_date
    elif args.date:
        target_date = args.date
        date_arg = args.date
    else:
        # Default: yesterday
        target_date = (datetime.now(timezone.utc) - timedelta(days=1)).strftime("%Y-%m-%d")
        date_arg = target_date

    # Get content
    content = get_content_by_date(feed, target_date)

    if content:
        print(json.dumps(content, indent=2, ensure_ascii=False))
    else:
        # Return empty result with available range
        min_date, max_date = get_date_range(feed)
        print(json.dumps({
            "error": "not_found",
            "message": f"No content found for {target_date}",
            "target_date": target_date,
            "available_range": {
                "min": min_date,
                "max": max_date
            }
        }, indent=2))


if __name__ == "__main__":
    main()