zarazhangrui · Lyrazhouqy · Jul 26, 2026
diff --git a/.github/workflows/generate-feed.yml b/.github/workflows/generate-feed.yml
@@ -1,9 +1,12 @@
 name: Generate Feeds
 
 on:
+  push:
+    paths:
+      - "config/default-sources.json"
   schedule:
-    # All feeds (tweets, podcasts, blogs): once daily at 6:17am UTC
-    - cron: '17 6 * * *'
+    # All feeds: 07:17 Asia/Shanghai (23:17 UTC on the previous day).
+    - cron: '17 23 * * *'
   workflow_dispatch:
     inputs:
       mode:

diff --git a/README.md b/README.md
@@ -15,7 +15,7 @@ A daily or weekly digest delivered to your preferred messaging app (Telegram, Di
 WhatsApp, etc.) with:
 
 - Summaries of new podcast episodes from top AI podcasts
-- Key posts and insights from 26 curated AI builders on X/Twitter
+- Key posts and insights from 36 curated AI builders on X/Twitter
 - Full articles from official AI company blogs (Anthropic Engineering, Claude Blog)
 - Links to all original content
 - Available in English, Chinese, or bilingual
@@ -61,8 +61,18 @@ Edit the files in the `prompts/` folder:
 - `summarize-tweets.md` — how X/Twitter posts are summarized
 - `summarize-blogs.md` — how blog posts are summarized
 - `digest-intro.md` — the overall digest format and tone
+- `digest-intro-expanded.md` — expanded digest rules that retain medium-density items
 - `translate.md` — how English content is translated to Chinese
 
+To generate the expanded edition explicitly:
+
+```bash
+node scripts/prepare-digest.js --expanded
+node scripts/deliver.js --file /tmp/fb-digest-expanded.md --expanded
+```
+
+The expanded edition is exported as `data/YYYY-MM-DD/digest-expanded.md` by default.
+
 These are plain English instructions, not code. Changes take effect on the next digest.
 
 ## Default Sources
@@ -75,8 +85,8 @@ These are plain English instructions, not code. Changes take effect on the next
 - [The MAD Podcast with Matt Turck](https://www.youtube.com/@DataDrivenNYC)
 - [AI & I by Every](https://www.youtube.com/playlist?list=PLuMcoKK9mKgHtW_o9h5sGO2vXrffKHwJL)
 
-### AI Builders on X (26)
-[Andrej Karpathy](https://x.com/karpathy), [Swyx](https://x.com/swyx), [Josh Woodward](https://x.com/joshwoodward), [Boris Cherny](https://x.com/bcherny), [Thibault Sottiaux](https://x.com/thsottiaux), [Peter Yang](https://x.com/petergyang), [Nan Yu](https://x.com/thenanyu), [Madhu Guru](https://x.com/realmadhuguru), [Amanda Askell](https://x.com/AmandaAskell), [Cat Wu](https://x.com/_catwu), [Thariq](https://x.com/trq212), [Google Labs](https://x.com/GoogleLabs), [Amjad Masad](https://x.com/amasad), [Guillermo Rauch](https://x.com/rauchg), [Alex Albert](https://x.com/alexalbert__), [Aaron Levie](https://x.com/levie), [Ryo Lu](https://x.com/ryolu_), [Garry Tan](https://x.com/garrytan), [Matt Turck](https://x.com/mattturck), [Zara Zhang](https://x.com/zarazhangrui), [Nikunj Kothari](https://x.com/nikunj), [Peter Steinberger](https://x.com/steipete), [Dan Shipper](https://x.com/danshipper), [Aditya Agarwal](https://x.com/adityaag), [Sam Altman](https://x.com/sama), [Claude](https://x.com/claudeai)
+### AI Builders on X (36)
+[Andrej Karpathy](https://x.com/karpathy), [Swyx](https://x.com/swyx), [Josh Woodward](https://x.com/joshwoodward), [Boris Cherny](https://x.com/bcherny), [Thibault Sottiaux](https://x.com/thsottiaux), [Kevin Weil](https://x.com/kevinweil), [OpenAI Developers](https://x.com/OpenAIDevs), [Anthropic](https://x.com/AnthropicAI), [Peter Yang](https://x.com/petergyang), [Nan Yu](https://x.com/thenanyu), [Madhu Guru](https://x.com/realmadhuguru), [Amanda Askell](https://x.com/AmandaAskell), [Cat Wu](https://x.com/_catwu), [Thariq](https://x.com/trq212), [Google Labs](https://x.com/GoogleLabs), [Amjad Masad](https://x.com/amasad), [Guillermo Rauch](https://x.com/rauchg), [Alex Albert](https://x.com/alexalbert__), [Aaron Levie](https://x.com/levie), [Ryo Lu](https://x.com/ryolu_), [Garry Tan](https://x.com/garrytan), [Matt Turck](https://x.com/mattturck), [Zara Zhang](https://x.com/zarazhangrui), [Nikunj Kothari](https://x.com/nikunj), [Peter Steinberger](https://x.com/steipete), [Dan Shipper](https://x.com/danshipper), [Aditya Agarwal](https://x.com/adityaag), [Sam Altman](https://x.com/sama), [Claude](https://x.com/claudeai), [Matt Van Horn](https://x.com/mvanhorn), [Simon Willison](https://x.com/simonw), [Harrison Chase](https://x.com/hwchase17), [Cognition](https://x.com/cognition_labs), [Cursor](https://x.com/cursor_ai), [LangChain](https://x.com/LangChainAI), [LlamaIndex](https://x.com/llama_index)
 
 ### Official Blogs (2)
 - [Anthropic Engineering](https://www.anthropic.com/engineering) — technical deep-dives from the Anthropic team
@@ -114,7 +124,19 @@ is fetched centrally and updated daily.
    (blog articles via web scraping, YouTube transcripts via Supadata, X/Twitter via official API)
 2. Your agent fetches the feed — one HTTP request, no API keys
 3. Your agent remixes the raw content into a digestible summary using your preferences
-4. The digest is delivered to your messaging app (or shown in-chat)
+4. The digest is saved as Markdown and delivered to your messaging app (or shown in-chat)
+
+Each delivered digest is exported by default to:
+
+```text
+data/YYYY-MM-DD/digest.md
+```
+
+Power users can override the export destination when calling the delivery script:
+
+```bash
+node scripts/deliver.js --file /tmp/fb-digest.md --output data/2026-05-12
+```
 
 See [examples/sample-digest.md](examples/sample-digest.md) for what the output looks like.
 

diff --git a/README.zh-CN.md b/README.zh-CN.md
@@ -11,7 +11,7 @@
 每日或每周推送到你常用的通讯工具（Telegram、Discord、WhatsApp 等），包含：
 
 - 顶级 AI 播客新节目的精华摘要
-- 26 位精选 AI 建造者在 X/Twitter 上的关键观点和洞察
+- 36 位精选 AI 建造者在 X/Twitter 上的关键观点和洞察
 - AI 公司官方博客的完整文章（Anthropic Engineering、Claude Blog）
 - 所有原始内容的链接
 - 支持英文、中文或双语版本
@@ -54,8 +54,18 @@ Skill 使用纯文本 prompt 文件来控制内容的摘要方式。你可以通
 - `summarize-tweets.md` — X/Twitter 帖子的摘要方式
 - `summarize-blogs.md` — 博客文章的摘要方式
 - `digest-intro.md` — 整体摘要的格式和语气
+- `digest-intro-expanded.md` — 扩展版摘要规则，保留更多中等信息密度动态
 - `translate.md` — 英文内容翻译为中文的方式
 
+显式生成扩展版：
+
+```bash
+node scripts/prepare-digest.js --expanded
+node scripts/deliver.js --file /tmp/fb-digest-expanded.md --expanded
+```
+
+扩展版默认保存为 `data/YYYY-MM-DD/digest-expanded.md`。
+
 这些都是纯文本指令，不是代码。修改后下次推送即生效。
 
 ## 默认信息源
@@ -68,8 +78,8 @@ Skill 使用纯文本 prompt 文件来控制内容的摘要方式。你可以通
 - [The MAD Podcast with Matt Turck](https://www.youtube.com/@DataDrivenNYC)
 - [AI & I by Every](https://www.youtube.com/playlist?list=PLuMcoKK9mKgHtW_o9h5sGO2vXrffKHwJL)
 
-### X 上的 AI 建造者（26位）
-[Andrej Karpathy](https://x.com/karpathy), [Swyx](https://x.com/swyx), [Josh Woodward](https://x.com/joshwoodward), [Boris Cherny](https://x.com/bcherny), [Thibault Sottiaux](https://x.com/thsottiaux), [Peter Yang](https://x.com/petergyang), [Nan Yu](https://x.com/thenanyu), [Madhu Guru](https://x.com/realmadhuguru), [Amanda Askell](https://x.com/AmandaAskell), [Cat Wu](https://x.com/_catwu), [Thariq](https://x.com/trq212), [Google Labs](https://x.com/GoogleLabs), [Amjad Masad](https://x.com/amasad), [Guillermo Rauch](https://x.com/rauchg), [Alex Albert](https://x.com/alexalbert__), [Aaron Levie](https://x.com/levie), [Ryo Lu](https://x.com/ryolu_), [Garry Tan](https://x.com/garrytan), [Matt Turck](https://x.com/mattturck), [Zara Zhang](https://x.com/zarazhangrui), [Nikunj Kothari](https://x.com/nikunj), [Peter Steinberger](https://x.com/steipete), [Dan Shipper](https://x.com/danshipper), [Aditya Agarwal](https://x.com/adityaag), [Sam Altman](https://x.com/sama), [Claude](https://x.com/claudeai)
+### X 上的 AI 建造者（36位）
+[Andrej Karpathy](https://x.com/karpathy), [Swyx](https://x.com/swyx), [Josh Woodward](https://x.com/joshwoodward), [Boris Cherny](https://x.com/bcherny), [Thibault Sottiaux](https://x.com/thsottiaux), [Kevin Weil](https://x.com/kevinweil), [OpenAI Developers](https://x.com/OpenAIDevs), [Anthropic](https://x.com/AnthropicAI), [Peter Yang](https://x.com/petergyang), [Nan Yu](https://x.com/thenanyu), [Madhu Guru](https://x.com/realmadhuguru), [Amanda Askell](https://x.com/AmandaAskell), [Cat Wu](https://x.com/_catwu), [Thariq](https://x.com/trq212), [Google Labs](https://x.com/GoogleLabs), [Amjad Masad](https://x.com/amasad), [Guillermo Rauch](https://x.com/rauchg), [Alex Albert](https://x.com/alexalbert__), [Aaron Levie](https://x.com/levie), [Ryo Lu](https://x.com/ryolu_), [Garry Tan](https://x.com/garrytan), [Matt Turck](https://x.com/mattturck), [Zara Zhang](https://x.com/zarazhangrui), [Nikunj Kothari](https://x.com/nikunj), [Peter Steinberger](https://x.com/steipete), [Dan Shipper](https://x.com/danshipper), [Aditya Agarwal](https://x.com/adityaag), [Sam Altman](https://x.com/sama), [Claude](https://x.com/claudeai), [Matt Van Horn](https://x.com/mvanhorn), [Simon Willison](https://x.com/simonw), [Harrison Chase](https://x.com/hwchase17), [Cognition](https://x.com/cognition_labs), [Cursor](https://x.com/cursor_ai), [LangChain](https://x.com/LangChainAI), [LlamaIndex](https://x.com/llama_index)
 
 ### 官方博客（2个）
 - [Anthropic Engineering](https://www.anthropic.com/engineering) — Anthropic 团队的技术深度文章
@@ -105,7 +115,19 @@ cd ~/.claude/skills/follow-builders/scripts && npm install
 1. 中心化 feed 每日更新，抓取所有信息源的最新内容（博客文章通过网页抓取，YouTube 字幕通过 Supadata，X/Twitter 通过官方 API）
 2. 你的 agent 获取 feed——一次 HTTP 请求，不需要 API key
 3. 你的 agent 根据你的偏好将原始内容重新混编为易消化的摘要
-4. 摘要推送到你的通讯工具（或直接在聊天中显示）
+4. 摘要保存为 Markdown，并推送到你的通讯工具（或直接在聊天中显示）
+
+每次生成的摘要默认会导出到：
+
+```text
+data/YYYY-MM-DD/digest.md
+```
+
+高级用户也可以在调用投递脚本时指定导出位置：
+
+```bash
+node scripts/deliver.js --file /tmp/fb-digest.md --output data/2026-05-12
+```
 
 查看 [examples/sample-digest.md](examples/sample-digest.md) 了解输出示例。
 

diff --git a/SKILL.md b/SKILL.md
@@ -15,6 +15,10 @@ Philosophy: follow builders with original opinions, not influencers who regurgit
 (X/Twitter posts and YouTube transcripts) is fetched centrally and served via
 a public feed. Users only need API keys if they choose Telegram or email delivery.
 
+The central feed is generated from `config/default-sources.json`. GitHub Actions
+refreshes it daily at 07:17 Asia/Shanghai and also refreshes it when the central
+source registry changes.
+
 ## Detecting Platform
 
 Before doing anything, detect which platform you're running on by running:
@@ -173,6 +177,7 @@ cat > ~/.follow-builders/config.json << 'CFGEOF'
   "language": "<en, zh, or bilingual>",
   "timezone": "<IANA timezone>",
   "frequency": "<daily or weekly>",
+  "digestMode": "<standard or expanded>",
   "deliveryTime": "<HH:MM>",
   "weeklyDay": "<day of week, only if weekly>",
   "delivery": {
@@ -321,6 +326,11 @@ You do NOT fetch anything yourself.
 cd ${CLAUDE_SKILL_DIR}/scripts && node prepare-digest.js 2>/dev/null
 ```
 
+Use `node prepare-digest.js --expanded` for the expanded edition, which keeps
+more medium-density product, workflow, tool, and early-signal updates. The same
+edition can be selected persistently with `"digestMode": "expanded"` in
+`~/.follow-builders/config.json`; the default remains `standard`.
+
 The script outputs a single JSON blob with everything you need:
 - `config` — user's language and delivery preferences
 - `podcasts` — podcast episodes with full transcripts
@@ -359,6 +369,10 @@ Read the prompts from the `prompts` field in the JSON:
 
 Assemble the digest following `prompts.digest_intro`.
 
+The prepare script has already selected `prompts.digest_intro` for the requested
+edition. Follow `config.digestMode` and do not silently switch between standard
+and expanded output.
+
 **ABSOLUTE RULES:**
 - NEVER invent or fabricate content. Only use what's in the JSON.
 - Every piece of content MUST have its URL. No URL = do not include.
@@ -397,15 +411,31 @@ Read `config.language` from the JSON:
 
 Read `config.delivery.method` from the JSON:
 
+First write the final digest text to `/tmp/fb-digest.md`, then call
+`deliver.js`. The delivery script always exports a Markdown copy before sending.
+Default export path:
+
+```text
+data/YYYY-MM-DD/digest.md
+```
+
+Expanded output defaults to `data/YYYY-MM-DD/digest-expanded.md`.
+
+Use `--output <path>` only when the user explicitly asks for a custom export
+file or directory. If `<path>` has no file extension, treat it as a directory and
+save `digest.md` inside it.
+
 **If "telegram" or "email":**
 ```bash
-echo '<your digest text>' > /tmp/fb-digest.txt
-cd ${CLAUDE_SKILL_DIR}/scripts && node deliver.js --file /tmp/fb-digest.txt 2>/dev/null
+cd ${CLAUDE_SKILL_DIR}/scripts && node deliver.js --file /tmp/fb-digest.md 2>/dev/null
 ```
 If delivery fails, show the digest in the terminal as fallback.
 
 **If "stdout" (default):**
-Just output the digest directly.
+```bash
+cd ${CLAUDE_SKILL_DIR}/scripts && node deliver.js --file /tmp/fb-digest.md
+```
+Then show the digest text to the user and mention the saved Markdown path.
 
 ---
 
@@ -414,16 +444,26 @@ Just output the digest directly.
 When the user says something that sounds like a settings change, handle it:
 
 ### Source Changes
-The source list is managed centrally and cannot be modified by users.
-If a user asks to add or remove sources, tell them: "The source list is curated
-centrally and updates automatically. If you'd like to suggest a source, you can
-open an issue at https://github.com/zarazhangrui/follow-builders."
+The central source registry is `config/default-sources.json`. When maintaining
+this repository, add or remove sources there, validate the JSON, and commit and
+push that file. The GitHub Action then regenerates the central feed; editing only
+`~/.follow-builders/config.json` or generated `feed-*.json` does not publish a
+source change. For a suggestion without repository access, open an issue at
+https://github.com/zarazhangrui/follow-builders.
+
+The registry currently contains 36 X builders, 6 podcasts, and 4 official blogs.
 
 ### Schedule Changes
 - "Switch to weekly/daily" → Update `frequency` in config.json
 - "Change time to X" → Update `deliveryTime` in config.json
 - "Change timezone to X" → Update `timezone` in config.json, also update the cron job
 
+### Digest Edition Changes
+- "Show the expanded digest" → Run `node prepare-digest.js --expanded` and pass
+  `--expanded` to `deliver.js` when exporting it
+- "Use expanded digest by default" → Set `digestMode` to `expanded` in config.json
+- "Use the concise digest" → Set `digestMode` to `standard` or pass `--standard`
+
 ### Language Changes
 - "Switch to Chinese/English/bilingual" → Update `language` in config.json
 

diff --git a/config/config-schema.json b/config/config-schema.json
@@ -25,6 +25,12 @@
       "default": "daily",
       "description": "How often to deliver the digest"
     },
+    "digestMode": {
+      "type": "string",
+      "enum": ["standard", "expanded"],
+      "default": "standard",
+      "description": "Digest breadth: standard concise edition or expanded edition with medium-density items"
+    },
     "deliveryTime": {
       "type": "string",
       "default": "08:00",