Back to Details

twitter-alpha

Compare original and translation side by side

🇺🇸

Original

English
🇨🇳

Translation

Chinese

twitter-alpha

Twitter-alpha

Turn a company or person into a short list of the highest-signal ideas their circle is posting right now.
Read the 
deepapi
skill before any search or scrape. All web, LinkedIn, and X work goes through DeepAPI. Never use built-in search, fetch, or a browser.
将某企业或个人转化为一份其圈子当前发布的最高价值观点的简短列表。
使用本技能前，请先阅读
deepapi
技能。所有网页、LinkedIn及X平台的操作均需通过DeepAPI完成。禁止使用内置搜索、抓取或浏览器工具。

Seed

初始信息

Need a company or a person. If missing, ask once.
If the seed is a person, their current company is the company.
需要一个企业或个人作为初始信息。若缺失，询问用户一次。
若初始信息为个人，则其当前任职的企业即为目标企业。

Pick 7 people

挑选7人

Always 3 insiders + 4 outsiders. Name them before scraping tweets.
3 insiders: the most impactful people at the company (founders, CEO, president/CTO, CRO, VP eng, VP product). Prefer people who actually post.
4 outsiders: people the seed actually talks to on X. Not competitors you guessed. Not investors unless they show up in the timeline.
How to find outsiders:
  1. Resolve the seed's X handle (company account plus founder handles).
  2. POST /v1/scrape/twitter/search
    with those handles, 
    sort: "latest"
    , 
    maxItems
    50+.
  3. Count replies, retweets, quote-tweets, and @mentions. Rank other people by how often the seed engages them.
  4. Take the top 4 who are not employees of the company.
If the seed has no X, use founder handles. If you still cannot build an engagement graph, say so and stop. Do not invent a peer set.
固定选择3名内部人员 + 4名外部人员。在抓取推文前需列出他们的姓名。
3名内部人员：企业中最具影响力的人员（创始人、CEO、总裁/CTO、首席营收官、工程副总裁、产品副总裁）。优先选择实际发布内容的人员。
4名外部人员：初始信息对象在X平台上实际互动过的人员。不可自行猜测竞争对手，除非投资者出现在初始对象的时间线中，否则也不选择投资者。
寻找外部人员的方法：
  1. 解析初始对象的X平台账号（企业账号及创始人账号）。
  2. 调用
    POST /v1/scrape/twitter/search
    接口，传入这些账号，设置
    sort: "latest"
    ，
    maxItems
    不少于50。
  3. 统计回复、转发、引用推文及@提及的次数。根据初始对象与其他用户的互动频率进行排名。
  4. 选取排名前4的非企业员工。
若初始对象没有X账号，则使用创始人账号。若仍无法构建互动关系图，请告知用户并停止操作。不得虚构同行群体。

Handles and fallback

账号及备选方案

Find each person's X handle with DeepAPI (
/v1/scrape/twitter/user
, 
/v1/search/web
only as a last resort). Verify the bio matches the person.
No X, private X, or almost no posts → scrape LinkedIn posts instead (
POST /v1/scrape/linkedin/posts
). Keep them in the 7. Do not silently drop them.
Need 7 people with some posts (X or LinkedIn). If you cannot fill 7, say who is missing and why.
通过DeepAPI查找每个人的X平台账号（优先使用
/v1/scrape/twitter/user
接口，仅在最后不得已时使用
/v1/search/web
）。验证账号简介是否匹配对应人员。
若无X账号、账号设为私有或几乎无发布内容 → 改为抓取LinkedIn动态（调用
POST /v1/scrape/linkedin/posts
接口）。仍将该人员计入7人名单，不得悄悄剔除。
需确保7人均有一定量的发布内容（X或LinkedIn动态）。若无法凑齐7人，请告知用户缺失人员及原因。

Scrape

抓取操作

For each of the 7:
  • X: 
    POST /v1/scrape/twitter/search
    with their handle, 
    sort: "latest"
    , 
    maxItems: 50
    .
  • No X: LinkedIn posts, cap at 50.
Follow DeepAPI polling. Skip empty RTs with no added text when scoring alpha. Keep originals, quote-tweets, and threads.
Also pull 
POST /v1/scrape/twitter/user
for the handles you will scrape (follower counts help ranking, not the output).
针对7人中的每一位：
  • X平台：调用
    POST /v1/scrape/twitter/search
    接口，传入其账号，设置
    sort: "latest"
    ，
    maxItems: 50
    。
  • 无X账号：抓取LinkedIn动态，上限为50条。
遵循DeepAPI的轮询规则。评分高价值信息时，跳过无附加文字的空转发。保留原创推文、引用推文及推文线程。
同时调用
POST /v1/scrape/twitter/user
接口获取待抓取账号的信息（粉丝数有助于排名，而非直接使用输出内容）。

What to extract

提取内容

Read all ~350 posts. Rank for alpha: a concrete claim, a non-obvious take, a number, a product/move, a hiring or market tell. Ignore congrats, logo spam, podcast plugs, and vague inspiration.
Prefer ideas that show up across more than one person. A single post can still make the list if it is clearly the strongest item.
阅读全部约350条内容。按“高价值（alpha）”标准排序：具体主张、非显而易见的观点、数据、产品动态、招聘或市场信号。忽略祝贺内容、品牌垃圾信息、播客推广及模糊的励志内容。
优先选择多人提及的观点。若某单条内容明显是最有价值的，也可纳入列表。

Output

输出要求

Short. Plain English. No preamble dump. No "here's what I found."
Numbered list, 5–8 items max. Same shape every time:
markdown
1. [Title of the idea](https://x.com/...) (~808k views)
   One sentence with the actual insight.
Rules:
  • Title is the idea, not the person's name.
  • Link the source post (or LinkedIn post).
  • Put view counts in parentheses when X reports them. Omit if unknown (LinkedIn).
  • One sentence under each title. No second paragraph.
  • Do not paste tweet dumps, bios, or a recap of who the 7 people are unless the user asks.
If nothing is high-alpha, say that in two sentences. Do not pad the list.
简洁明了，使用通俗易懂的中文。不要开场白，不要“以下是我的发现”之类的表述。
采用编号列表，最多5-8项。格式统一如下：
markdown
1. [观点标题](https://x.com/...) (~808k次浏览)
   一句话说明核心洞察。
规则：
  • 标题为观点内容，而非发布者姓名。
  • 链接至源推文（或LinkedIn动态）。
  • 若X平台显示浏览量，需放在括号内。若未知（如LinkedIn）则省略。
  • 每条标题下仅用一句话说明，不得添加第二段。
  • 除非用户要求，否则不得粘贴推文原文、账号简介或7人名单的概述。
若未找到高价值信息，用两句话告知用户，不得凑数填充列表。