匠人学院 JR Academy
JR Academy · Blog职业洞察

Anthropic上市募935亿美元,Gemini能替你操作Mac,Signal封锁系统AI读屏

Anthropic以每股83美元在纳斯达克上市,含超额配售共募约935亿美元,首日收97.17美元、次日跌12%;Google给Gemini Mac版加上computer use,暂限美国付费用户;Signal在iOS 27和Android上挡截屏,不让系统级AI助手读聊天内容。

发布日期
阅读时长2 分钟
作者

今天三条:Anthropic上市后的头两个交易日,Google给Mac补上的代操作功能,Signal对系统级AI助手动手。每条都对过公司原文和独立媒体报道,对不上或没查到的地方文里写出来。

没放进来的三件事。The Verge报道Anthropic数学家Levent Alpöge称自己和Claude一起证明了黎曼猜想,但Verge自己写的是「has yet to be verified independently」,Anthropic没有表态,他本人网站上的文档我们也没打开,等数学界有说法再写。CNBC报道Cursor告诉美国众议院委员会,考虑逐步停用中国开源权重模型,可那份报告原文我们在委员会网站上没找到,只有CNBC转述。特朗普周五在记者会上说要对在海外建数据中心的美国公司征「tremendous tax」,目前只有口头表态,没有任何正式文件。

01 Anthropic上市,共募约935亿美元

Anthropic 11月18日定价,公告原文:“Today Anthropic, PBC priced its initial public offering of its Class A common stock at $83.00 per share. The offering consists of 980,000,000 shares of Class A common stock”。TechCrunch补了背景:这个价格高于公司原先给的区间,“which anticipated a price between $70 and $78 per share.”

11月19日开始交易。11月20日Anthropic发了交割公告,承销商把超额配售权全部用掉,多买了1.47亿股。总数是:“The aggregate gross proceeds from the offering, before deducting underwriting discounts and commissions and other offering expenses, were approximately $93.5 billion, of which approximately $91.6 billion was received by Anthropic.” 差额是老股东卖的那部分,钱不进公司。

CNBC的说法是这笔发行“surpassing SpaceX, which held a record IPO in June.”

还有一笔同步的定向增发,容易被忽略。公告写的是向“Advanced Micro Devices, Inc., Broadcom Inc., Google LLC, Micron Technology, Inc., Microsoft Corporation, NVIDIA Corporation, and TSMC Partners, Ltd.”按同样的83美元卖了约3.13亿股Class N股票。IPO加定增合计:“approximately $119.5 billion”。买家名单几乎就是Anthropic的算力供应链:芯片、存储、代工、云。

股价呢?CNBC:“The stock opened at $111.50 on the Nasdaq Thursday after Anthropic raised $81.3 billion in the largest IPO of all time, but shares slid throughout the day and closed at $97.17.” 比发行价高约17%,但比开盘价低了不少。第二天更难看:“Anthropic shares fell 12% on Friday, drifting toward the IPO price of $83, as concerns about artificial intelligence trade continued to drag tech stocks lower.” 周五的具体收盘价我们没查到,这里不写。

我的判断:钱是募到了,而且买定增的全是卖它算力的公司,这种「供应商入股客户」的结构,好的时候互相抬,差的时候一起跌,周五Nvidia、AMD跟着一起跌就是例子。对用Claude做产品的团队,短期最实在的变化是这家公司手里现金非常充足,算力供给和价格的不确定性比上市前小。股价两天的波动说明不了技术路线,别拿它当选型依据。

来源:Anthropic定价公告;Anthropic交割公告;CNBC;CNBC超额配售;TechCrunch。

02 Gemini能替你操作Mac了

Google 11月19日给Gemini的Mac应用加了computer use。博文原话:“Starting today, computer use in the Gemini app for Mac helps you get multi-step tasks done across your local files, desktop apps, and the web.” 你描述要做什么,它去点菜单、打字、填表。

和同类功能比,Google强调的一点是不抢你的屏幕:“Gemini works inside an isolated background window with a live preview right in your chat. Your main screen and cursor remain completely free”。它在后台窗口干活,聊天界面里有实时预览,你可以继续做自己的事。

控制权方面:“you choose which apps Gemini can access, watch every step live, pause at any time, and approve key actions before they happen.” 能选它可以碰哪些应用,随时暂停,关键动作要你批准。

Google举的用法都很具体:从本地PDF里提取重量、尺寸、地址去填多箱运费报价;从几个区域的在线看板拉数据和截图,做成Google Slides;给一批照片统一调色。全是重复、多步骤、跨应用的活。

限制要看清。“Computer use is rolling out in English in the U.S. in the Gemini app for Mac.” 脚注写的是:“Requires an Apple Silicon Mac running macOS Sequoia 15.0+. Available for personal Google accounts 18+, available to Google AI Plus, Pro and Ultra users.” 也就是美国、英文、Apple芯片、付费订阅、个人账号。澳洲用户现在用不上。TechCrunch确认:“The company didn’t say when it would expand to other markets or platforms.”

TechCrunch对这次发布的评价不客气:“The tech giant is a little late to the game with this launch, as other leading AI model makers like OpenAI, Anthropic, and Meta have already been offering agentic features in their desktop applications for some time.”

我的判断:同一周内,Gemini先在手机上开放了代办任务,现在又到了桌面,Google在补agent这块短板的速度很快。对在澳洲的同学,这条暂时是「知道就行」。但如果你的工作里有大量跨应用的重复操作,现在就可以把这些流程列出来,等功能开放或者用已有的同类工具试,哪一步必须人工确认要提前想好。

来源:Google博客;TechCrunch。

03 Signal封锁系统AI助手读屏

上一条是AI替你看屏幕、操作电脑,这一条是有人不想被看。

Signal发了一篇博文,宣布对手机系统自带的AI助手动手:“So, we are blocking such capture by preventing screenshots in the app on iOS 27 and when AI assistants attempt to capture screen data on Android.” 两个平台做法不同:iOS 27上是应用内直接禁截屏,Android上是检测到AI助手要抓屏幕数据时拦截。

它针对的是谁,博文点了名:“Google and Apple are integrating agentic capabilities like Gemini Intelligence with Google Gemini and Onscreen Awareness with Siri AI deep into their operating systems.” 这些功能靠扫描屏幕内容来理解你在干什么。聊天内容一旦被扫走,存在哪、谁能读,就不归Signal管了。

这不是Signal第一次这么干。博文提到去年它用同样的截屏拦截挡过Windows 11上的Microsoft Recall。

你还是可以截屏,但要自己去关保护,Signal把后果写得很直白:“People who want the ability to take screenshots can disable this protection within the Signal app’s settings. However, doing so will potentially expose your (and your contacts’) Signal content to these AI agents, and break our privacy guarantees.”

Signal也承认这招不彻底:“restricting screenshots does not address other more sneaky ways AI agents could potentially access application data, such as retrieving content already present on a user’s device via the filesystem.” 而且桌面端有缺口:“This is also the reason we are currently unable to implement these protections on macOS.”

它向Apple、Google、微软提的要求之一:“Developers should have an easy and well-documented way to opt their app out of AI data access and processing”。The Verge的报道和博文一致,还提到Signal认为需要更多工具来防第三方agent,原话是“like Meta’s Muse, from using deep accessibility settings in apps like a “skeleton key” to access application data.”

三家系统厂商对这篇博文的回应,我们在这两个来源里都没看到。

我的判断:系统级agent要好用就得看得见屏幕,应用要守隐私就得不让看,这个矛盾现在只能靠「禁截屏」这种粗办法解决,说明操作系统还没给出正经的权限模型。做App的团队值得现在就查一遍:自己的应用里哪些页面有敏感数据(聊天、支付、健康、身份信息),在iOS和Android上有没有标记为禁止截屏或录屏。这是眼下开发者手里唯一确定管用的开关。

来源:Signal博客;The Verge。

作者
一键分享或复制链接

相关文章推荐

查看全部文章 →

看完了,下一步