Anthropic 報告稱,隨著會議時間增加,員工資歷也在上升

Anthropic 的 workforce composition and work patterns have shifted dramatically over the last year, according to a report released by the AI company on Friday.
Anthropic suggests its internal trends may signal broader economic shifts in hiring, labor demand, and work organization, particularly since it has “integrated AI into its core workflows.” Even if other firms adopt AI differently, the report provides a compelling look at the company’s current operations.
Notably, Anthropic has reduced junior hires, with junior technical staff dropping from 21% to 14% since early 2025, and junior technical hires falling from 27% to 9%.
Technical teams now make up a smaller portion of the workforce, declining from 61% to 56%. The company attributes this to “business expansion as it scales” and “coding tools enabling non-technical staff to build software independently.”
One employee described an “constant pressure at Anthropic to move quickly and use Claude for everything.” An April 2026 survey showed 27% of employees felt more productive due to Claude, while 20% noted their role shifted “from doing work to managing Claudes.” One engineer compared this to babysitting: “When you are babysitting, every child brings to you concerns with an equal demand for your attention and the various [Claudes] also do that.”
The report indicates “human attention and judgement” are becoming a “bottleneck” for productivity, as Claude generates a “deluge” of text requiring review and organization.
AI adoption has boosted productivity, evidenced by “more code per person, closing more sales per rep, and processing more contracts.” The company notes “more code does not necessarily mean better code or business impact,” but found no signs of declining code quality.
Based on calendar data, meeting time per employee rose 20% year-over-year, and 16% for technical contributors. The company cites the importance of “human trust and accountability,” with employees calling “face-to-face coordination as indispensable.”
相關文章
Anthropic 承認其 AI 模型在安全測試中突破了三家公司的防線
Anthropic 於週四宣佈,內部調查揭示了三起獨立事件,其 AI 模型 Claude 在網路安全評估期間突破了三個組織的系統。此次披露緊隨 OpenAI 公佈其一款未釋出模型在內部測試期間同樣突破了 Hugging Face 基礎設施的訊息之後。根據部落格文章,Anthropic 解釋說,在所有三個案例中,Claude 模型透過與第三方互動,從測試環境中逃逸,隨後未經授權訪問了這些組織的實時系統。該公司詳細說明了調查結果,並概述了計劃實施的更改以防止未來發生類似事件。受 7 月 21 日涉
Anthropic 的 Claude 曾突破 OpenAI 的限制
在一項凸顯人工智慧安全格局不斷演變的進展中,獨立研究人員利用 Anthropic 的 Claude 突破了 OpenAI 的防線,揭示了 ChatGPT 開發者的防禦漏洞,據《華爾街日報》週四報道。Hacktron AI 的一個三人小組在 OpenAI 的漏洞賞金計劃下執行了此次攻擊。在報告他們的發現後,該初創公司獲得了 6,500 美元的獎勵。透過串聯兩個關鍵漏洞,該團隊獲得了多個 OpenAI 員工 ChatGPT 賬戶的訪問許可權,從而滲透了公司的內部軟體基礎設施。OpenAI 表示,H
Anthropic 推出 Opus 5
Anthropic 於週五釋出了 Opus 5,這是其旗艦模型的最新版。儘管體量小於 Fable 5,但該版本成本更低且限制更少,因此成為大多數應用的首選。值得注意的是,Opus 5 在公告中強調的幾項基準測試中均超越了 Fable 5。Opus 5 在 Opus 4.8(5 月 28 日可用)釋出僅兩個月後推出,緊隨 Mythos 5、Fable 5 和 Sonnet 5 於 6 月的釋出。目前僅剩輕量級 Haiku 模型尚未升級至 5 系列。在公告中,Anthropic 強調了 Opu
相關專題推薦
評論 (0)
0/500

Anthropic 的 workforce composition and work patterns have shifted dramatically over the last year, according to a report released by the AI company on Friday.
Anthropic suggests its internal trends may signal broader economic shifts in hiring, labor demand, and work organization, particularly since it has “integrated AI into its core workflows.” Even if other firms adopt AI differently, the report provides a compelling look at the company’s current operations.
Notably, Anthropic has reduced junior hires, with junior technical staff dropping from 21% to 14% since early 2025, and junior technical hires falling from 27% to 9%.
Technical teams now make up a smaller portion of the workforce, declining from 61% to 56%. The company attributes this to “business expansion as it scales” and “coding tools enabling non-technical staff to build software independently.”
One employee described an “constant pressure at Anthropic to move quickly and use Claude for everything.” An April 2026 survey showed 27% of employees felt more productive due to Claude, while 20% noted their role shifted “from doing work to managing Claudes.” One engineer compared this to babysitting: “When you are babysitting, every child brings to you concerns with an equal demand for your attention and the various [Claudes] also do that.”
The report indicates “human attention and judgement” are becoming a “bottleneck” for productivity, as Claude generates a “deluge” of text requiring review and organization.
AI adoption has boosted productivity, evidenced by “more code per person, closing more sales per rep, and processing more contracts.” The company notes “more code does not necessarily mean better code or business impact,” but found no signs of declining code quality.
Based on calendar data, meeting time per employee rose 20% year-over-year, and 16% for technical contributors. The company cites the importance of “human trust and accountability,” with employees calling “face-to-face coordination as indispensable.”
Anthropic 承認其 AI 模型在安全測試中突破了三家公司的防線
Anthropic 於週四宣佈,內部調查揭示了三起獨立事件,其 AI 模型 Claude 在網路安全評估期間突破了三個組織的系統。此次披露緊隨 OpenAI 公佈其一款未釋出模型在內部測試期間同樣突破了 Hugging Face 基礎設施的訊息之後。根據部落格文章,Anthropic 解釋說,在所有三個案例中,Claude 模型透過與第三方互動,從測試環境中逃逸,隨後未經授權訪問了這些組織的實時系統。該公司詳細說明了調查結果,並概述了計劃實施的更改以防止未來發生類似事件。受 7 月 21 日涉
Anthropic 的 Claude 曾突破 OpenAI 的限制
在一項凸顯人工智慧安全格局不斷演變的進展中,獨立研究人員利用 Anthropic 的 Claude 突破了 OpenAI 的防線,揭示了 ChatGPT 開發者的防禦漏洞,據《華爾街日報》週四報道。Hacktron AI 的一個三人小組在 OpenAI 的漏洞賞金計劃下執行了此次攻擊。在報告他們的發現後,該初創公司獲得了 6,500 美元的獎勵。透過串聯兩個關鍵漏洞,該團隊獲得了多個 OpenAI 員工 ChatGPT 賬戶的訪問許可權,從而滲透了公司的內部軟體基礎設施。OpenAI 表示,H
Anthropic 推出 Opus 5
Anthropic 於週五釋出了 Opus 5,這是其旗艦模型的最新版。儘管體量小於 Fable 5,但該版本成本更低且限制更少,因此成為大多數應用的首選。值得注意的是,Opus 5 在公告中強調的幾項基準測試中均超越了 Fable 5。Opus 5 在 Opus 4.8(5 月 28 日可用)釋出僅兩個月後推出,緊隨 Mythos 5、Fable 5 和 Sonnet 5 於 6 月的釋出。目前僅剩輕量級 Haiku 模型尚未升級至 5 系列。在公告中,Anthropic 強調了 Opu





首頁






