Trading & coding daily https://xitter.cf/t.co/lAp2QWwQjq for agent mails https://xitter.cf/t.co/aUK2VwjOTS for easy trading https://xitter.cf/t.co/04T0tl5AXc for agent web search
Beijing
Joined January 2014
- Tweets3.4K
- Following1.7K
- Followers6.1K
- Likes2.1K
随着模型能力的提升，网络安全只会越来越重要，这次 PHP 的事件只是个开始
$BUG +3.8%. PHP2Shell was exploited before anyone published a fix. That's the new normal: stronger models mean more bugs found, faster, by more people. Patching stays human speed. The spend moves to whatever stops an attack before the patch lands.
Search1API 针对知乎的文章和回答，做了专门的优化
Zhihu articles and answers, as clean Markdown.
Send a Zhihu link to Crawl:
> column article: headings, lists, code and images kept
> answer link: that exact answer, never an excerpt or a different one
> author and publish time when the page has them
s1.dev/zhihu-scraper-api?utm…
ALT Search1API Zhihu scraper landing page: a real Zhihu answer page on the left, the clean Markdown output on the right
长鑫快跌回 IPO 当天价格了，还有人一直持有着吗🤔
688825 CXMT: record profit, then -9% intraday. Q3 revenue +15–20% QoQ, net profit just +1–4%. Micron's last quarter: profit +33%, carried by HBM. CXMT sells mostly commodity DRAM, and that price curve is flattening. Without HBM, this is what a plateau looks like.
ALT Pulse menu bar app showing CXMT (688825) daily chart
Search1API 的 crawl 接口，不仅能获取网页内容，还能获取 markdown 格式完整的 YouTube 字幕
YouTube videos, as timestamped transcripts.
Send a YouTube URL to Crawl:
> video: the full transcript, every line timestamped
> playlist: each video's title, channel and duration
> no captions: an honest partial, gaps named
No Google account needed.
s1.dev/youtube-scraper-api?u…
ALT Left: a YouTube video page. Right: the Search1API Crawl output, a Markdown transcript with a timestamp on every line.
越来越多知名开发者做了 macOS 的垃圾清理、活动监视类软件，我自己也有需求，经常清一清，没觉得有什么问题。
不过，我突然想起最开始选择 macOS 而不是 Windows 的一个重要原因，不是不用做这些事情吗 😂
Search1API 的 DeepCrawl 支持一次爬取整个网站为 Markdown 格式，打包成压缩包供下载
Crawl a whole docs site from the browser, no code. DeepCrawl is now in the Search1API Playground.
> Paste a URL, crawl the sitemap or all links
> Watch progress, then download a ZIP
> cURL, TypeScript and Python snippets for when you move it to code
s1.dev/deepcrawl-api?utm_sou…
ALT Search1API Playground DeepCrawl page: crawling s1.dev from its sitemap completed, with a Download ZIP button and code snippets
Search1API & LangChain
Search1API has an official LangChain toolkit. @LangChain
> pip install search1api-langchain
> Search1APIToolkit().get_tools() gives your agent web search, news search and page crawling
> Built on our Python SDK: auth, retries and timeouts handled
s1.dev/integrations/langchai…
ALT Code card: pip install search1api-langchain and using Search1APIToolkit with a LangChain agent
重构了 Pulse 的股票搜索功能，比之前快了 300 倍 😂 pulseticker.app
Pulse 0.15.14: ~300x faster search, in any script: tesla, AAPL, 腾讯, 台積電, トヨタ, 삼성전자, btc.
pulseticker.app/changelog#v0…
会飞的汽车，被太空公司收购，不是很自然的事情吗 🚀
$TSLA A Roadster on SpaceX thrusters, revealed at SpaceX's test site. It ships in 2028, so the car isn't the story yet. The shared stage is: Barclays sees an all-stock SpaceX–Tesla deal possibly before December's shareholder meeting. +5.3% pre-market.
ALT Pulse daily chart of Tesla (TSLA) with position
放一个 mailly.site 的实际使用截图，一个巴基斯坦的视频博主联系到我聊合作，Mailly 接收邮件，Agent 总结，并直接输出回复草稿。
因为我已经提前给这个 Agent 配置了 skills 和 memory，所以其实他完全知道我的项目是怎么一回事，草稿可以直接发送。如果我有更多要讨论的，也有在对话框里继续跟 Agent 聊。
小范围发布一个拖了半年的新产品， Mailly，简单理解就是 Resend+Agentmail，我自己在实际业务中已经用了半年度。
收发邮件支持 Web、API、MCP、CLI，也内置了 Agent，支持 skills、memory，可以自动总结、生成回复草稿。
Free 用户可以有 3 个子域名，每个月 3000 封邮件，欢迎体验：mailly.site
Great works of @SwamiMalode
Your agent found the sources. Now it shows them.
render_results in our MCP server turns results it already has into expandable source cards in ChatGPT and Claude.
Cards adapted from @rareuiHQ's Web Search component.
s1.dev/mcp-server?utm_source…