Language /
Lenny's Podcast × Anthropic
設計師必讀・2026 春 Essential for Designers · Spring 2026
AI 時代設計洞察
Jenny Wen · Head of Design, Claude @ Anthropic

設計流程,
已經死了。
The Design Process
Is Dead.

接下來,什麼在取代它? Here's what's replacing it.

Lenny's Podcast 2026 年 3 月March 2026 逐字稿摘要Transcript Summary
完整訪談逐字稿Full Interview Transcript
Jenny Wen × Lenny Rachitsky
Lenny's Podcast · Mar 1, 2026 · 77 min
第一部分 · 開場介紹
LR
我們今天的嘉賓是 Jenny Wen。Jenny 是 Claude 的設計總監,現在領導 Claude Cowork 的設計。在這之前,她是 Figma 的設計總監,領導了 FigJam 和 Slides 背後的設計團隊。她也曾在 Dropbox、Square 和 Shopify 做過設計師。我喜歡這次對話的地方是,Jenny 正活在設計作為一個職業的未來,她在這裡為我們展示那看起來像什麼,以及設計師將會改變多少。這真的很瘋狂也很有趣。非常感謝 Noah Levin 和 Emily Lynn Hasham 為這次對話建議主題和問題。別忘了去 Lenny's Product Pass 點點 com 查看獨家提供給 Lenny 新聞稿訂閱者的優惠。讓我們在我們出色贊助商的短暫廣告後進入正題。Jenny,非常感謝你來,歡迎來到 podcast。
JW
是的,很興奮能在這裡。
LR
我一直期待這次對話,因為我在這個 podcast 上花了很多時間討論軟體工程的未來,這個角色如何改變,產品管理的角色,那個角色如何改變。我甚至花了很多時間討論設計如何改變。顯然,它也在以真的很大的方式改變。而你有著一個如此前排的視角看向事情的走向。我也知道你對事情的走向有著很多非常強烈的意見,所以有很多東西我想談談。我想先從一個很寬泛的問題開始。隨著 AI 的興起,設計流程如何改變?
第二部分 · 設計流程已死
JW
改變了很多。我認為在改變的方式上它還有很長的路要走。我認為我們實際上在過去一小段時間裡看到工程改變得遠比設計多,但我認為工程改變很多的結果是設計某種被迫改變。所以我認為一些背景資訊是,我幾個月前在九月份在柏林的一個會議上做了一個演講,我稱之為「別相信設計流程」,基本上我就是說:「嘿,你知道設計師們一直被教導的那個設計流程嗎,你去發掘和探索,然後你發散,收斂,發散,收斂。」而它就像這個流程,我們某種程度把它當作福音對待,很努力地想要保留,我們就像「相信流程」。那基本上已經死了。我認為它甚至在 AI 時代之前就某種程度在衰亡,但考慮到現在工程師們可以去旋轉他們的七個 Claude,我認為作為設計師,我們真的需要放手那個流程。我認為那是改變的大事情。但我認為即使在我做那個演講以來的過去三到四個月,那個演講實際上開始感覺很...它對我來說有點過時,有點尷尬,但特別是伴隨 Opus 46 的大轉變和一群人在假期裡發現並使用 Claude Code,我認為我們看到改變我們流程的力量發生得更多。我現在看到的方式是基本上有兩種類型的設計工作,設計工作在這個新世界變得真的非常分層。所以第一種,真的就是支持實施和執行。所以這是工程師們使用他們的七個 Claude 創造所有這些功能的地方,任何人都可以提出一個想法,你可以只是談論一個想法,然後某個人,通常實際上是工程師因為他們仍然比我們更擅長實施這些東西,他們會只做一個粗糙的版本讓你試一下。而你作為設計師實際上沒有時間再做這些漂亮的 mock 了,或者以這種方式領導。然後我認為第二種工作也感覺真的很重要,這是創造某種願景或方向。這種感覺像最難撥出時間的,而且它是我們之前仍然做的,但我認為它的形狀非常在改變,因為我認為我們曾經去說,「我們要做這個設計願景。我們要去做這個兩年、五年、不管怎樣,甚至十年的願景,我們要指向什麼。」但技術現在改變的方式,我們不知道兩年內會發生什麼。變化太多了,它通常變成一個三到六個月之外的願景,並且不一定是創造那個被精美講述故事的漂亮牌組。有時候它只是創造指向正確方向的一個原型。我認為這種工作在這個世界上仍然真的很重要,因為在人們可以旋轉他們的七個 Claude、可以製作任何功能用任何方向或實施方式的世界裡,你需要指向他們什麼。而為了確保我們都做著有意義的東西並且也以高效的方式完成。如果我們都朝向有一個更大目標的東西工作,比起只做隨機的東西那樣做會更有效率。而所以那是我看到的大轉變。我認為我現在有意見,但三個月後問我它可能會改變得更多。
LR
所以你在說這裡不是像你或設計領域是,「我們需要改變?」而是工程和你能夠這麼快地構建的事實,只是強制設計師的角色改變,因為如你所說的,工程師們可以只是出貨,出貨,出貨,出貨,出貨。而你發現在這裡的是你最好不要阻擋那個,讓他們馬上做,就像他們說的。然後有這種模式幫助他們一邊出貨,把它聚在一起,確保它全都連接,引導他們一點。
JW
是的,我認為是這樣。是的。我認為沒有一個統一的聲音說,「設計師,我們現在需要改變。」但是是的,有工程工具真的改變的後續效果。我認為我們可能會在明年左右看到設計工具改變,但現在很多東西都落後於那個。我認為這對我們來說也真的很賦能,因為作為設計師,我們現在也有著進入很多這些編碼工具的機會。而我們可以以我們實施東西的方式參與流程。我做了很多最後一英里的東西,我實施所有的打磨,並與工程師緊密合作來取得功能過線,也用實際代碼原型東西而不是依賴工程師再做一次。
LR
你認為這在所有公司有多真實,比如說 AI 公司、非 AI 公司?有人可能聽到這個,好的,Anthropic Claude,好的,他們在最尖端,一個。二,有點是開發的,但我認為人們可能感覺像,「好的,這不會發生在 Salesforce。這不會發生在,我不知道,ServiceNow,不管在哪裡。」所以我猜測,你認為這是所有團隊前進的地方嗎?它大多是 AI、尖端公司嗎?設計流程轉變有多廣泛?
JW
所以我去年做的那個演講真的是我做的最有共鳴的演講。我認為它是人們在整個產業開始感受到的東西,他們就像,「哦是的,我們不能再做舊的設計流程了。我們正在使用像 Claude Code 和 v0 這樣的工具來開始旋轉原型,而且 PM 也正在開始旋轉原型等等。」所以我認為有某些東西在出現。但那個演講有趣的其他觀察也是,實際上有一個相當的反彈。人們清楚地已經投資了他們整個職業生涯來學習、教導、使用這個真的很穩定的設計流程。我認為有很多貶低的聲音就像,「哦是的,我們不能沒有發掘。我們不能沒有這些流程片段。」所以我認為還有產業的一部分在這方面還沒有完全到達,如果這有意義的話。
LR
是的。
JW
是的。
LR
而一個大的事情你可以爭論...問題是什麼導致最好、最成功的產品和公司?而你可以爭論它是花時間做發掘、用戶研究、mock、迭代、beta 測試,或它可能只是工程師出貨的東西,可以,不驚人,夠好。我們學,迭代,構建,迭代。你的感覺是那第二條路,不僅是那只是每個人都在做什麼,但那實際上在這一點帶來更好的產品嗎?
JW
我認為你某種程度必須選擇並使用你的判斷力何時真的要出貨某些東西,但我認為執行的能力,嘗試某個東西並用真實數據試一下它,和一個真實用戶的某種心態在產品裡,我認為那確實導致一個更好的產品,特別是當我們都用這些新開發的非決定論的 AI 模型一起工作,你不能 mock 所有的狀態,而且你不能理論化,你甚至不能用它做一個可點擊的原型。你某種程度必須使用下面的實際模型,而且你必須看人們用他們的用例試著使用它,因為用這些模型,你可以為不同的用例設計它們,但你實際上發現用例當你看人們使用它們時。所以是的。
第三部分 · 在 Anthropic 的一天
LR
另一件我總是聽到的事,並在你剛說的基礎上構建的是就是你不知道人們會用 AI 做什麼。你不知道它在某些事情上有多好,非決定論的片段。所以你可以創造這些關於它可能是什麼的驚人 mock,然後人們用一個完全不同的方式使用它,這是 Cowork 來自的地方,可能甚至 Claude Code 在開始時。而所以對於在 Anthropic 做一個設計師是什麼就像?只是給我們一個在 Anthropic 工作的一天,在風暴中心。
JW
在 Anthropic 的很多時間實際上只是追上公司發生的東西。我認為這是公司,...我在幾個其他這個規模的公司工作過,我認為有大量資訊和很多東西在發生,但我感覺真的被迫去保持與它一致。有模型開發在研究方面的東西。然後在任何給定的時間,有大量不同的團隊原型化和嘗試不同的想法,而且有一堆不同的代碼名稱等等。而很多時間我只是試著導航並弄清楚那些項目是什麼,因為我認為我試著發現並看,嘿,什麼來臨在前面為我?因為有來自研究團隊的東西,但也有一些我們更接近研究的 labs 團隊,試著和原型化東西。然後有我想試著的東西。我們有一堆原型和我們內部可以使用的產品,我非常好奇並想嘗試那些東西。然後我認為也有很多內部的人有很多對行業去向的洞察和意見。而有些那些就是真的有趣讀,因為很多這些都是哲學辯論或公司的方向等等。而且是的,我感覺我只是想保持與這些東西一致。相比之下,我認為在一個正常公司裡,我就像,「沒關係。這是超出我能力的東西。我沒有真的那麼在意。」在這裡,我認為它既是體積也是正在發生的那些事物的種類,我真的對保持與它們一致感興趣。而然後除了那個某種保持與它,那不是我工作的一個巨大部分,但我認為它是真的一個有趣的部分。
LR
好吧,它連接到你早先做的那個點,一個大的設計角色部分現在是幫助工程師和團隊執行,而不只是告訴他們,「這是 mock,這是設計。」它幫助他們保持在軌道上,幫助他們連接想法,當它發生時創造一個凝聚力的體驗。所以那有意義。
JW
是的。是的。我認為部分只是好奇。感覺我有這個前排座位看到這麼多在產業發生的。而且所以很多它就像,是的,我們的 Slack 是一個金礦。我只是對閱讀人們正在工作的東西、他們正在說的東西興奮。
LR
我從未想過作為一個普通人要追蹤已經有這麼多 AI 新聞。然後實際上看到什麼真的在一個實驗室內發生是一個全新的一組 feeds 看。
JW
是的,是的。我認為那最好的 AI 新聞可能在內部,如果你曾經在一個在 Slack 的這些公司之一。
LR
是的。問題持續變得更難。我正在跟蹤什麼在發生。好的。好的。所以那是工作的一部分。還有什麼?
JW
還有一些傳統的,讓我想想什麼將在未來發生,讓我為那做一些設計。那是某些東西,例如,這個星期我已經分配了一些時間,我就像,「好的,很酷。我們一直在為 Cowork 的很多執行模式裡,現在我想留出一些時間來想想,嘿,接下來的三個月看起來像什麼,而那實際上可以去給定市場在哪裡、模型在哪裡、那可能是什麼?」因為我認為它仍然真的幫助可視化那並向團隊展示那,並指向每個人同樣的方向。然後我也花我一天的一堆時間只是與工程師一起即興。很多它只是一個對話,或白板,或通過他們構建的某些東西並給他們關於它的反饋並以那種方式成為一個設計師,我們真的是諮詢。然後我花一部分我的一天在代碼裡,打磨,實施東西。有時候發生的是一個工程師和我已經理出某些東西,他們已經實施了它的第一個版本,我只是進去用他們打磨它。而那是一個真的有趣的工作部分,我認為幾個月前不是那麼存在的。
LR
你還在做傳統設計流程的元素嗎?原型化、用戶研究、panels,我不知道,只是出去和你描述的整個東西?
JW
是的,我認為我們仍然在某種程度上做所有的那些。我們在隊伍上有一個用戶研究者,她正在組合傳統研究和調查,整個團隊正在閱讀那些研究和反饋。我們仍然在原型化東西。我仍然在 mock 東西。我認為它只是我現在有更廣泛的工具集,而且我認為我花在做每件事上的時間比例只是改變了。
LR
明白了。好的。所以那是一個真的有趣的要點。它曾經是那個真的很大的...我猜什麼將是你在那之前的生活的圓形圖,其中它就像傳統思考、計劃、原型化、mocking、研究,然後只是反饋和執行和外出,今天?
JW
是的。我認為作為一個設計師幾年前,我會說也許 60 到 70% 的它是 mocking 和原型化東西,然後花一些最後 20 左右的做某種與工程師的 jamming、諮詢他們,最後可能 10% 做協調會議等等。但現在我感覺 mocking 的部分是 30 到 40%。然後有那個其他 30 到 40% 那現在是 jamming 並直接與工程師配對。然後有一個 slice,我不知道我還有多少,但有一個 slice 現在也是實施。是的。
LR
實際上構建和出貨?
JW
是的。
第四部分 · 設計師的 AI 工具
LR
太棒了。所以有點跟著那個線,你的 AI stack 裡有什麼?什麼你,作為一個設計師,我知道你是一個經理而我想談談你實際上如何也是 IC。你的 AI stack 裡有什麼?什麼工具你在你的角色上使用?
JW
我的 AI stack 裡有什麼?好吧,我們在 Anthropic 工作,所以我們深入 Claude stack。我使用,顯然,chats、Claude Chat,但日益更多的 Claude Cowork。我基本上已經轉移所有我的 chat 用例到 Cowork,因為我發現它某種程度在這些更長的執行任務上更好。而且大多數我問 Claude 的東西都是這些更長的執行任務。然後有 Claude Code,當然。我用它主要用 VS Code 和 IDE 因為我通常調整前端東西,而且它幫助只是能夠看到代碼然後也與 Claude 談話。我一直在嘗試實際上更多遠程使用 Claude Code,通過行動電話和通過 Slack 也一樣。對於某個人說「哦是的,這個圖標有點偏離或東西」然後你只是在提及 Claude 和 Claude 做它,然後你拿起 PR 並它完成了,那真的很有趣。那已經真的、真的很有趣。而且是的,我認為我們在這裡是一個完全的 Claude 之家。所以是的,那基本上是我的 stack。
LR
你還在用 Figma 作為一個設計師嗎?
JW
我仍然使用 Figma,是的,是的。
LR
好的。我正在等著聽。好的。所以 Figma 仍然是你生活的一部分。作為一個前 Figmate,那是你們怎麼稱呼的嗎?
JW
是的,Figmates,是的。
LR
是的。是的。好的。所以我知道 Twitter 上有這個大辯論,就像,代碼是設計的未來嗎?我們需要許多更多,我們需要一個設計?你的感覺是什麼?Figma 仍然重要嗎?
JW
我的意思是,作為一個前 Figmate,也許我以那種方式有偏見,但我認為仍有...當我使用 Figma 時,我就像,「是的,那是我應該使用的。」而且它仍然為我填補一個非常好的缺口。我認為很多那實際上只是,一個是探索很多不同的選項。我認為那是設計流程的一個真的很重要的部分,能夠只是思考 8 到 10 種不同的做某個東西的方式。我認為最好的設計發生在你能夠只是扔一堆想法在牆上,並策展和推自己想出一堆這些不同的方向。現在,編碼,或現在用某些這些編碼工具工作不貸自己那麼超級好,因為它超級線性,你變得超級投資在一個方向,你只是與 Claude 在它們上迭代,例如。所以 Figma 已經真的偉大於只是探索所有這些不同的選項,我認為它仍然某種程度要以那種方式存在。然後我認為有真的精細視覺和互動細節也真的非常偉大要能夠只是在 Figma 試一下。再說,它是很多不同的方向,但它是微方向。它是能夠想想不同的排版或風格。在一個 canvas 擁有那些當你可以只是探索那具體是仍然那麼、那麼有幫助,並不是某些我總是想要直接到代碼。
LR
很有趣你仍然使用一個 IDE,因為在工程,它清楚地轉移到命令行、agents、IDE 某種程度移動到不是酷了。而且它有很多意義。你只是想編輯一些 CSS 東西,一些顏色東西。而且所以我可以看為什麼不只是告訴 agent,「嘿,只是來吧,改變這個一個十六進制值。」只是改變它那麼容易多了。
JW
是的。它真的很煩人像,「你可以改變這個到這個 class 嗎?」當你可以只是進去並改變它到一個不同的 class。
LR
所以它有趣。我想知道如果 IDE 現在變得對設計師和 PM 有用,而且工程師已經移動了?
JW
是的,也許。是的。
第五部分 · 與工程師協作
LR
好的。所以很多你的時間你花與工程師合作,給他們反饋,在正確的方向上輕推他們。有一個感覺,我感覺,的只是你的建議是放手。別感覺像你需要是這個守門人,但有這個片段,好的,幫助他們在一個凝聚的方向移動,並創造產品我們自豪於。很多設計師我認為現在在這個船上,只是像,「哦我的天,我不能追上所有這些工程師整天出貨。」什麼是某些東西你學習了關於只是任一個怎樣幫助你的工程師在設計上變得更好所以它最後只是更好,或只是保持在這個頂上並不發瘋?
JW
不管何時我做工作與工程師在項目上,而且它更多在一個諮詢基礎,我真的只是嘗試解釋為什麼我思考我思考一種方式,幫助他們提取原則。相比之下我只是像,「不,我不認為那應該在這裡。」它就像,「不,我認為我們應該有一個按鈕在這裡因為不是每個人實現你可以提示這個。」而且這是一個例子它來自研究等等。所以我也只是嘗試指向工程師我們的設計系統,以及像那樣的代碼,因為現在 Claude 寫很多代碼而它不總是挑起設計系統裡的東西等等。所以盡可能多我可以用東西裝備他們他們可以在未來沒有我使用,如果那是有幫助的。然後在你的不試著發瘋的點,我認為它是硬。我認為它真的硬現在。**而我看很多這個實際上來自工程師和設計師都,其中它就像現在我們某種程度能夠做這麼多,我們想做更多。而且我認為它不只是設計師們感覺像,哦是的,我們必須追上工程師。**我認為甚至工程師是像,「我們怎麼追上我們自己現在?」所以那是某些東西我聽到很多。
LR
那麼真實。哦老兄,怎樣追上我們所有的 agents、我們的七個 agents 我們不斷執行?
JW
是的。
第六部分 · 速度、品質與信任
LR
好的。所以然後作為一個設計師,在這個職業中手工藝和偉大的體驗和品質和信任是這麼一個核心部分,幫助灌輸那在產品中,因為那理論上導致真的成功的產品和公司,你怎樣只是認為關於保持手工藝品質、信任,當你的產品只是每天出貨 1,000 倍,而且你不能保持在他們頂上,並且沒有設計師涉及?
JW
它不是沒有設計師涉及。它更像就是它幾乎是對一個設計師來說有太多要處理。但我認為藉由這個,我認為特性或產品在哪裡,他們在採用對比早期預覽的循環中在哪裡。例如,我們有時候會啟動東西,而且我們會說,「嘿,這是一個研究預覽。它很早。它會有一堆這些缺陷,」而且我們警告那一堆。我認為 Claude Cowork 實際上是一個好例子,其中我們標籤了它一個研究預覽,而我們把它放在那裡知道那,「嘿,這相似於我們的模型。那是最糟糕它曾經會是,但我們要把它放在那裡因為我們相信,在內部我們已經試了一堆,有某些東西真的強大在這裡某些人會從中受益。它也許還不是最容易使用的。它也許不是最高品質。它也許有某些問題與它,但我們要把它放在那裡因為我們相信好處超過缺點。」我認為那是好的去做,特別是當有某些已經真的有價值於產品,而且它值得把它放在那裡。但我認為承諾你某種程度必須做給你的用戶是像,「嘿,我們要把它放在那裡,但我們要迭代。我們要取得你的反饋,我們要迭代,我們要讓它更好。」而且你必須提交那。你必須向世界展示那,你必須回應人們的反饋,而且你必須展示你不斷出貨和改進它。因為我認為你真的失去關於品質的信任和發佈早期的東西的方式,是如果你早期發佈它,然後沒有什麼曾經發生。那是某些當你發佈早期時降級品牌的東西,這是某些真的損害一個品牌。但不管何時你放某些東西出早期,它是可能做那的並保持你的公司的品牌。而且我認為那是某些我們一直在做相當好的。而且我認為任何聽的人可以帶走從它,它就像,是的,好吧,我們不斷地做那。而且我認為那實際上真的有趣對我作為一個設計師,因為你放某些東西在那裡而且你實際上學並立即得到反饋關於它,而且你知道接下來做什麼。
LR
我聽過你描述這個方式是通過速度構建信任。
JW
是的,確定。是的,它通過速度構建信任,但也只是讓人感覺像他們已經被聽到而且我們根據他們試著使用它為它修復東西,而且他們的反饋實際上被感謝和使用。
LR
是的,它清楚當 labs 啟動東西,而且你全部非常好於這個,團隊上的每個人正在推特和只是回應推文和評論然後出貨,「嘿,我們修了這個昨天和這正在發生。」所以有一個清楚的感覺,「這只是今天,我們知道這破碎,我們將修它。」然後因為 Claude Code 可以非常快地編碼,修復來得非常快。好的。所以另一個大問題人們問那我經常在這個 podcast 上問,是關於只是什麼技能變得有價值?而且另一個方式我一直認為它,Lex 最近放了它那種方式,是其中人腦將繼續有價值當 AI 變得聰慧。所以我們已經經歷過這個進展的 tab 完成...segments 的代碼,到 100% 的代碼現在由 AI 寫,它瘋狂,到現在 AI 正在審核它自己的代碼。Boris 在 podcast 最近說 Claude Code 現在幫助他們想出想法並決定什麼去構建,這就像,好的,哇,看在它。看它去。整個產品工作流程,產品開發流程慢慢被 AI 吃掉。所以問題只是其中人腦仍然會有用,至少直到我們有超智能?你認為 AI 會變得真的、真的好於品味、判斷、設計嗎?
JW
我認為它會變得更好於品味和判斷和設計。是的,我認為我們也許持有太多的而說,「哦是的,一個設計師或某個人將永遠知道最好的東西去出貨或這的最好版本。」但我認為 AI 的品味感會變得更好。在一天結束時,某個人必須決定什麼實際上會被構建而什麼實際上重要。而當我想到人們說,「哦,AI 只是要為我們構建這個軟體,」很多構建軟體的難部分實際上不是構建它。如果你想到你在工作上擁有的最難時間,Lenny,它也許是像,哦,你和某個其他人不同意什麼應該進入這個功能或什麼不應該進入這個功能。而且那些東西仍然感覺像,是的,AI 可以稱重,但它不一定可以解決你和某個其他人之間的爭執。而且所以有某些關於決定什麼實際上進入我們構建的東西,我猜那是某種品味,但也許不是品味按我們想到美學品味或什麼。有某些排序,它是關於接下來做什麼的判斷。
LR
就是看 AI 有多快速地接管了編碼,我想一年前,肯定是兩年前,大多數人會說:「我不認為會這樣。我不認為 AI 會變得這麼好。」而世界上最優秀的工程師信任度如此之高,他們甚至都不看代碼了。這就是我們現在的狀況。這只是讓我重新評估了我之前對許多假設,比如說,AI 永遠不會像真正優秀的 PM、設計師一樣善於判斷什麼是偉大的,以及決定應該構建什麼。但我現在開始認為,我認為它會到達那個境界。即使是你分享的例子,它可以給那兩個試圖做決定的人說:「這是你做決定所需的全部數據,這就是為什麼這是正確答案,你只需按是,按一,我就會去構建它。」所以我認為就你的觀點而言,我認為我們低估了它在這方面會變得多好。好的。所以你的感覺是它會變得更好,但你的感覺是我們仍然會需要很棒的設計師參與其中,還有我們和 PM,幫助做出這些決定,工程師當然也是。
JW
是的。是的。我認為仍然會有人必須決定,哦,我們想要構建這種產品。或者考慮到 AI 呈現給我們的東西,仍然有人需要對決定負責。就像即使 Claude 今天可以為你寫所有代碼,仍然是工程師需要對代碼是否真的有效、這在產品中是否真的有意義負責。所以我認為有那種決策制定/判斷層,這感覺像也許有一天會來臨,我們不必做那樣的事,但它仍然落在我們身上。是的。
LR
這沒有意義。這讓我想起放射學的例子,總是有一種感覺是 AI 會接管放射學領域,告訴你發生了什麼。但人類主要是有用的,用來簽署對決定的批准,因為如果出錯,必須有人負責。這不是世界上最好的工作,但這是一個不同的遊戲,比如說代碼。
JW
是的。
第七部分 · Chat 介面與 UI 的未來
LR
好的。AI 和設計中的另一個持續問題是,感覺聊天機器人和終端就像,我不認為任何人期望這會成為 AI 的持久用戶界面。聊天機器人,好的,不,不,這只是沿著這個旅程的一個臨時停留點,但現在它甚至進一步發展了,只是終端。你對這個有想法嗎,我不知道,你認為會有下一步關於我們如何與 AI 互動,或者你認為聊天機器人和終端主要就是我們最終所在的位置?
JW
可能會有兩者的結合,你正在互動、點擊的 UI 和界面,以及感覺更具觸覺感。我們已經在 Claude 聊天機器人中看到這個並正在使用它。所以我們最近發佈了許多這些小工具,讓 Claude 引出並問你問題,以及以互動的方式向你展示天氣和股票之類的東西。我認為那些獲得了非常好的反響,因為人們仍然喜歡看 UI 並觸摸它們、點擊它們,而且它們比輸入什麼給 Claude 效率高得多。但同時,當我們真的投入這種聊天機器人範式時,我認為這只是為我們打開了一個全新的靈活性世界,我們用那些内建的 UI 無法得到的。所以我的理解是,我不認為 Chat 會消失,因為這打開了一種全新的方式,無限多的方式與模型一起工作,與計算機交談,我們以前根本沒有過。但我認為對於非常具體的事情,在這個 UI 中存在仍然是最直接的。我認為可能發生的是,許多那些 UI 會越來越經常被模型生成,而不是我們手工編碼每個實例。但我認為我們在這個空間中,我不認為聊天......也許甚至與終端交談不會消失。
LR
有趣的是,對於 OpenClaw、Claude、Moltbot,所有這些名稱,一個重大創新是另一種與它聊天的方式,通過 WhatsApp 和 Telegram 和 SMS,就像聊天機器人的另一種形式,但就像,那是一件大事。哦,我只是可以通過 WhatsApp 與它聊天。
JW
是的。就像與某人聊天和交談仍然......我們作為人類正在做它,這是我們以一種真正豐富的方式互動的方式。現在我們只是有了另一個媒介與計算機互動,基本上。
LR
是的。所以 Kevin Wheel,他在另一個我不會提到的 AI 實驗室工作,他在播客中有一個很好的觀點,即交談是處理每個智力水平的一種非常優美的方式。我們可以與非常非常聰明的人和不那麼聰明的人交談,就是交談,它在整個光譜中的擴展性非常好。我們可以與 IQ 200、300 的人交談,就像交談仍然有效。所以這就是為什麼它一直是處理模型不斷增長的智力的一種非常優美的方式,因為它繼續工作。
JW
是的,那完全有意義。是的。
第八部分 · IC 與管理
LR
好的。我想回到這整個管理和 IC 的概念。所以在很多方面,你有點把自己放回了 IC 角色。談談那個。如果你認為這只是設計經理需要做的事情。
JW
是的,我對此有看法。是的。所以在過去的一年在 Anthropic,我最初加入作為一個 IC,然後我在一個某種程度上需要它的組織結構中管理了一個團隊幾個月,現在我實際上回到了全職 IC 工作。我加入 Anthropic 作為一個 IC,因為我對在這裡作為一個 IC 能夠完成的工作類型感到非常興奮,但也因為我感覺我有點想更接近工作,我認為在我上升公司階梯之前,這感覺像一個真正重要的時刻來做。我有這些問題和疑慮,中層管理,這在未來是否安全?我們現在工作的方式實際上是,這會是一個在未來持續的工作嗎?還是我應該嘗試其他東西並讓我的手變髒?為了完全誠實,我實際上喜歡兩者。我喜歡管理人員。我喜歡設立團隊並在那個水平,但我也真的很喜歡 IC 工作。我有點像一個不太願意的經理,當我做的時候,我就像,「好的,我會做。」所以我同樣喜歡兩者。但我認為作為一個 IC 在過去一年所教我的是,它實際上只是給了我許多技能,我認為如果我只是在這一年中管理,我就不會獲得。就像我提到的,設計流程在過去一年中改變得如此多,我感覺我只是學到了許多我如果只是管理一個團隊就不會必然有時間做的硬技能。所以這實際上是它給我的最好的東西。我認為在任何時候如果我再次管理一個團隊,我認為它會給我同情和理解設計流程如何改變的。我認為那實際上是現在真正重要的事情,因為團隊的工作方式非常不同。我認為如果你不以這種方式工作,或者你不總是測試所有工具並嘗試東西,實際上真的很難同情。但是是的,現在是一個有趣的成為設計師的時刻。如果我沒有在這個環境中工作,我不知道我是否會完全理解它,或者知道該怎麼做或如何引導我的團隊。所以這有點像這一年真正給了我什麼。
LR
那麼你之前是 Figma 的設計總監,對吧?
JW
是的。
LR
你的團隊有多大?你的組織有多大,只是為了給人們一個參考?
JW
最多,我可能有,我想,12、15 名設計師左右,我還有幾位經理。
LR
酷。然後你回到了 IC?
JW
這像是一個組織。
LR
是的。好的。所以你有中層管理可能不會持續的感覺。你目前的感受如何?你認為設計管理是一種長期存在的東西,還是你認為每個人都變成 IC?
JW
我認為只要有一群人,就有幫助讓某個人管理一個團隊。我認為經理有真實的價值。這取決於經理的形狀和他們實際做什麼。但我對現在什麼是有幫助的經理的看法是,某個人不只是,我認為純粹的人力管理,比如,哦,只是有人設置你,幫助你在你的職業中,進行一對一,確保你在工作中感到很好。我認為那不再是那麼多的事情了,但我認為有人可以真正地發揮給團隊方向的作用,以及做一些人力管理的東西,結合在一起,我認為那是管理未來看起來的樣子,至少現在是這樣。有人可以真正與團隊就工作進行互動並在那裡提供方向,以及為他們創造做他們最好的工作的環境。
LR
你認為自己長期內會回到管理嗎?
JW
我會。我可能會。我認為我真的只是喜歡幫助一個團隊構建最好的產品。而我的座右銘是,無論如何都要做到。如果是某個人,如果團隊需要某個人給團隊方向和建立團隊等等,那可能是我。如果團隊只是需要某個人去執行它,那也可能是我。
LR
所以我聽到的建議是針對設計中的人,尤其是經理們,你幾乎需要回到 IC,以便真正理解發生了什麼以及它改變了多少,這樣你才能成為一個更好的經理。
JW
我認為是這樣。而且我認為傳統上,至少我所看到的,許多工程學科,當他們聘請 EM 或有時甚至主任時,他們實際上讓 EM 進行幾個月的輪換並撿起一些任務,並在他們成為全職經理之前真正理解技術如何工作。我認為設計可能也需要做類似的事情,我認為在過去,設計更多地面向人力管理。
LR
當你回到 IC 設計師時,你發現自己最生疏的是什麼?
JW
實際上進行評論,並且只是真的......
LR
得到批評。
JW
......是的,得到批評。你就像,哦是的,接受批評反饋並聽到它,並定期聽到它是很困難的,因為這是你作為設計師必須做的事情,就是分享工作並與你的團隊一起呈現它,以及只是得到很多批評反饋並一直接受那樣。是的。
LR
所以目前你在 Cowork 上領導設計/IC 設計。這是對的嗎?
JW
是的。
第九部分 · Cowork 與產品開發
LR
太棒了。所以 Boris 最近在播客上,談論了關於 Cowork 應該是什麼有很多辯論,有所有這些大想法,他就像,「最後,讓我們只讓它像一個終端,基本上,在產品中,只是一個花哨的終端。」有什麼你可以分享關於著陸你著陸的地方的過程,那個 Cowork 體驗?我這裡在我的監視器上,順便說一句,看著它。
JW
具體來說,對於 Cowork,我們在內部對那可能是什麼樣子進行了大量的不同原型設計。它是其中之一,我們嘗試了很多東西,然後我認為我們並不確定它何時實際上準備好運送。然後它有點一下子全部。我們就像,「好的,我們很快就要運送它。」
LR
我認為那是 10 天,10 天的構建。
JW
是的,從我們內部擁有的東西到我們準備外部運送的東西,肯定比那要長。所以我們已經構建了一段時間,但我們對它實際上要採取的形式並不確定。所以它到達那裡的方式實際上是,有許多不同的其他我們在內部進行的探索,在不同的代理工具等之上。我們只是原型化了最終進入 Cowork 的不同互動的小部分。所以當 Claude 給你一個待辦事項列表的東西,我們嘗試了許多不同的形式因素。我們嘗試了許多不同的形式因素來呈現你不同的多項選擇問題的方式。我們嘗試了許多不同的方式來教導人們什麼是使用情況。我不知道我們是否著陸在有史以來最好的形式因素,但本質上它是已經在內部工作的東西,人們喜歡的東西,我們只是認為我們會通過發佈它來獲得更多信號。所以我認為強制自己在我們進行的那 10 天內發佈它,它只是有點像,無論我們有什麼,讓我們把它放出去,然後讓我們出去並從那裡迭代,這就是我們正在做的。
LR
它在你推出時爆炸了網際網路,所以它奏效了。
JW
是的。
LR
Cowork 現在有你最自豪的特性,或者只是迫不及待想要修復和改進的特性嗎?
JW
誠實地說,我認為我最自豪的只是我們實際上只是運送它,並把它放出去。而且是的,我不知道是否有一件特定的東西,因為我認為當你從事某些事情並工作這麼長時間時,特別是作為一個設計師,你就像,我不知道。我所能做的就是看到其中的缺陷,但我認為有很多我興奮的東西。我們一直在迭代,特別是在主頁上,並使那成為一個感覺更像,「嘿,這些是你可以給 Claude 的任務,Claude 正在進行的任務。」所以那實際上應該會推出。到此出來的時候,它可能已經推出了。
LR
我看到這個小隨機化東西,你點擊它,它給你所有這些不同的想法。
JW
是的,是的。然後當你實際上開始與 Claude 一起工作時,它感覺更像一個待辦事項列表。感覺更像這些是 Claude 正在進行的事情,這些是 Claude 需要你注意的事情。我認為這裡有機會使它感覺更像你和 Claude 之間的這個共享待辦事項列表。所以興奮迭代那個。然後我也很興奮思考更多關於這的真實形式因素是什麼?它是一直困在這個屏幕中,還是這如何到達它正在與之工作的不同表面?
LR
我喜歡你分享它不只是 10 天來做這個東西。有這些人拋出的數字,「我們在 10 天內構建了它。」而你的觀點是有時間思考方向應該是什麼,以及原型設計、模擬稿,嘗試東西。然後它就像,好的,現在我們知道我們想要它是什麼。讓我們構建它並運送它。
JW
是的。我認為出於某種原因,那成為了從所有 Cowork 公告中帶走的病毒式東西,它只花了 10 天,但我認為只是有許多不同的探索,以及已經在 Cowork 的不同部分工作的人,那不只是 10 天,許多不同的人都參與了。它是其中之一,想法不斷回來,它從不是正確的時刻或有不同的變化,然後突然間它是正確的時刻,它感覺像,哦,一直這麼明顯,但有一個長、長的旅程才能到達那裡。
LR
順便說一句,對於不太了解 Cowork 的人,我對它的思考方式,它就像有手的 Claude,或在你的計算機上做東西。你會如何在一兩句話中描述它?
JW
這是一個很好的描述。我實際上沒有聽過那個,但我喜歡那個。我可能會更經常使用它,作為有手的 Claude。我還認為它就像 Claude,但 Claude 真的很擅長拿你所有的垃圾,然後把它變成漂亮的東西。我認為我最喜歡的任何我真的喜歡的 Cowork 使用案例,只是給它一個我的東西的文件夾,它實際上不重要那個文件夾中是什麼,但我能夠從中提取一些好東西。
LR
我已經做過很多次了。好的。回到管理和成為一個經理以及設計師的角色,我要談談一點招聘。所以看到設計師角色中發生了多少變化,你現在尋找可能是新的什麼?當你招聘設計師時,你現在尋找什麼,你認為真的對他們在這個新世界中成功很重要?
JW
好吧,我確實認為在我工作的特定環境中工作,只是有一種適應力和隨之而來的感覺,我認為這真的很重要,因為周圍發生了這麼多變化,你必須真的願意適應,嘗試新方法,嘗試新工具並學習東西,而不是陷入舊流程和舊方式。但隨後我也認為,可能有三種類型的人現在對我來說真的很有趣。而且我認為這些人在之前對我來說已經很有趣了,但我感覺在這個時代感到特別重要。所以首先我會稱之為強大的通才。所以不只是他們在很多事情上不錯的常規通才,而是幾乎在 T 形框架中塊形的人,這樣他們就像在幾個核心技能上真的很好,就像第 80 百分位好。我認為這相當罕見並且難以僱用,說實話,但我喜歡這個,因為設計角色我們已經看到正在拉伸和跨越。我們都變得越來越 PM 形,我們都變得工程形。所以如果你已經有幾個不同桶中的強大技能,對你來說很容易靈活並擴展你的角色。所以那對我來說真的很令人興奮。只是有人真的很擅長很多事情,再次,一個巨大的要求。然後另一個對我來說真的很令人興奮的人是在 T 形框架中,一個深層專家,一個 T 形的人,但 T 的尖端可能比大多數其他人走得更遠。所以也許是這個行業前 10% 的人等等。再次,超級難找。我感覺很幸運,在一些這些地方工作,像這樣的人,你可以某種程度上負擔得起僱用他們並實際上讓他們上船。然後我的最後一個可能是我認為我們都在忽視的那個,我稱之為工藝新畢業生。它只是一個早期職業的人,感覺,聰慧和經驗超過他們的年齡,但也只是非常謙虛和非常渴望學習。我認為這個人現在真的很有趣,因為我認為大多數公司只是在僱用高級人才,那些以前做過事情的人,超級有經驗,但鑑於角色改變有多少,我們期望做什麼是改變,我認為有人幾乎有一個白板紙,並且只是一個真正快速的學習者,真的渴望學習新的戰術等等,並且沒有所有這些已經烤制在他們的腦海中的流程和儀式,那是超級有價值的。所以我認為那些我認為我們很多人正在忽視的人,但我對他們真的很興奮。
LR
這很棒。在深 T 形上,有人在設計中有什麼例子......他們真的很擅長什麼技能?
JW
有時有設計師,他們的技術真的很強,他們可能是......他們基本上是一個軟體工程師。那真的很有趣,特別是因為現在很多它是,至少對我們來說,就像你直接與模型一起工作,所以當你有深度工程專業知識時會有幫助。但另一個深層專家 T 只是也許他們只是真的很擅長視覺設計或只是設計圖標或什麼的,那樣的東西,鑑於任何人都可以製作任何東西,有那個深層專家傾斜感覺像,哦是的,他們可以真的幫助區分我們正在構建的東西。
LR
太棒了。好的。然後有塊形。我們在播客上有 Mark Andrews,我們有點稱之為 F 形或 E 形,其中有許多 T 的東西,橫 F,橫 E,我猜。那是你描述的嗎,你真的、真的很擅長許多東西?
JW
是的。是的。基本上,我不知道,如果你幾乎擁有他們的技能組合,它看起來像一個塊而不是 T。
LR
因為有這麼多技能,T 被展開。
JW
是的。
LR
好的。和裂紋新畢業生。所以這只是一個渴望、開放思想、堅韌、我想像非常聰慧的人。
JW
是的。
LR
太棒了。
JW
是的。
LR
如果有人是一個年輕設計師試圖突破,試圖成功,你的建議是什麼來幫助他們有機會加入 Anthropic,例如?
JW
我只會說只是建造很多東西,嘗試很多東西,建造實際的東西。我認為那可以感覺......我真的不知道設計教育或教育的狀態現在是什麼,但至少從我上學的時候回來,一切都非常理論化,這裡,我們會教你一些方法等等。但我認識的最好的裂紋新畢業生人員只是只是使用技術、構建實際東西的人,不要受到他們可能擁有的經驗少的限制。我認為有時他們實際上不受那個負擔,因為在行業中這麼長時間後,我們對自己有期望,但他們實際上沒有,他們有點感覺任何事都是可能的。所以只是構建很多東西並與人分享,並找到也這樣做的人的社區。是的。我認為我在這裡也有一個呼籲,是我上了一所在我畢業後幾年開始 Socratica 的學校,實際上在我畢業後很久。基本上他們的整個事情就是構建東西並幾乎像一個科學項目一樣展示它。我認為只是有一個真的很酷的人的運動,他們只是構建東西和做事。例如,有人構建了這個 Claude 機器人項目,這也是幾年前的事,他們只是組裝了在 Claude 上運行的機器人,然後有人做了她只是想在波士頓的公共汽車上放上 Googly 眼睛或什麼的。並且只是有一種行動感,是的,我們可以只是做東西,但隨後也是這個社區,人們只是嘗試和構建東西並互相分享。所以無論那看起來像什麼,鑑於有人畢業的學校或從,做那種東西是將使人們脫穎而出的東西。
LR
對於目前的設計師......職業,......高級的,你認為你需要變得技術性和學習編碼,至少構建,或者你認為你可以真的成功並且只是不靠那個而變得更好在其他東西上?
JW
我認為肯定有幫助也許不要那麼學習如何編碼以至於你從頭開始構建東西,但它確實感覺像現在設計師詞彙越來越多的是實現一些東西。我想知道,隨著模型,模型和產品都變得更好,我是說,我們可能會繼續上移抽象層,你不必知道代碼的每一行會如何工作。但我認為我會說的是開始將那個帶入你的工具包,編碼工具,無論或不是你實際變成技術,我認為任何設計師都應該真的意識到,並且知道如何使用手頭的工具,而不是也許學習並去學習 React 或等等等等。
LR
Claude 作為設計師有多好,你會說,或 Claude Code,無論如何你想描述......你會僱用 Claude 作為設計師,還是它還沒有到達那裡?
JW
我不認為 Claude 現在到達了那裡。我不認為 Claude 現在在你會僱用的設計師方面到達了那裡。我認為它不是強大的通才或深層專家或裂紋新畢業生。我認為它在首次通過時相當不錯,並為你呈現很多不同的想法,但那裡沒有什麼感覺像特殊和可聘用的。
LR
這對設計師現在來說是好消息。它現在在這方面很爛。我非常好奇它在這方面能有多好。這是一個很大的開放問題,它是否可以泵出驚人的新穎、獨特的創意體驗,或者它永遠不會像人類設計師那樣好?
JW
我是說,即使在過去一年左右,它也好多了。所以是的。
第十部分 · 框架、建議與快問快答
LR
有幾個管理方面的,我不知道,儀式或看法,我從與你共事的人那裡聽說過,我想探討一下。其中之一是你有個熱點想法,就是低槓桿時間對經理來說根本不存在,你可以從人們認為是低槓桿的事情中獲得很多好處。你能談談這個嗎?
JW
是的。是的。我記得我剛開始當經理時,我想我從某個課程、書籍或其他地方得到的建議之一是,「現在你是經理了,你必須真正優先考慮你的時間並對你的工作進行分類。」有個二乘二的矩陣,我不記得具體內容了,但本質上是說,「這些是只有我能做的事情。這些是任何人都能做的事情,其他的,這些是低槓桿的,你不應該再做了。」很多低槓桿的事情就是那些非常瑣碎的、細節的東西,或者說,是的,可能確實有人能做這些任務。但當我想到我最尊敬的領導和經理時,我其實認為他們最好的特點之一就是他們選擇那些看起來低槓桿的任務自己去做,結果實際上變成了非常高槓桿的事情,因為是他們在做。一個例子是,當高級領導徹底測試產品時,他們非常了解它,他們 dog food 它,他們重現 bug,他們花很多時間與工程師分享日誌、挑剔細節之類的。我認為這最終變成了超級高槓桿的,即使它花費了很多時間,花費了很多細節工作時間,因為它創造了對產品的熟悉感,我認為這真的很好。它也創造了一種氛圍,就像,哦是的,這位高級領導真的很在乎,他們實際上知道產品的來龍去脈,他們捲起袖子,他們給出這個反饋並與團隊合作。我想類似地,我見過的是當高級領導能夠現在修復 bug 時。我實際上見過 Mike Krieger 自己提交 PR,這真的很好,因為就像,好吧,我們都在這個團隊裡,沒有什麼是這個人做不了的。還有一件我喜歡的稍微更文化性的事情,就是當有人費力為某人製作週年紀念卡或什麼的,並給他們寫一些超級漂亮的東西,或為他們製作一張超級漂亮的卡片,因為我認為它只是表明就像,一個 EA 或什麼人可以製作卡片,但這位領導就是一個如此關心他們團隊的人,他們付出了努力。所以這是我試圖體現的東西,就是選擇看似低槓桿但值得我花時間的任務。
LR
是的,這太有趣了。你在說的是,從某種意義上講,低槓桿的東西通常是影響最大的東西,因為你的下屬不會期望你花時間在這件事上,低槓桿最終變成了高槓桿。
JW
是的。我認為這就是讓你的領導風格脫穎而出的原因,或者對某個人來說感到特殊。
LR
太棒了。另一個我從你那裡聽說的,我不知道,儀式和運營團隊的方式,是你鼓勵團隊成員互相調侃,從表面上看這聽起來不像一個很棒的環境,但另一方面,我經常聽到你建立的團隊就是最快樂、表現最好的團隊。談談這個調侃的想法和鼓勵它,以及你關於構建很棒團隊學到的東西。
JW
是的,我認為不是說我像,「哦,你們應該互相調侃。」我不是那樣強制執行的或什麼,但當我想到心理安全、團隊和彼此相處融洽的人時,當你想到你的朋友時,你總是願意推行一點邊界並調侃他們。你經常在調侃你的朋友,但你可能實際上沒有在調侃你的同事很多,因為這都是關於舒適和安全的。所以不是說我像,「哦,我希望我的團隊互相調侃。」但我認為這當人們在你的團隊中感到舒適,只是彼此互相開玩笑時,這可能是一個非常好的信號。我認為這也可能是一個很好的信號,當人們對你作為領導的感受也是一樣的時,就像,有一種成分,他們不太害怕你,他們感到有一種安全感,如果他們說什麼,他們不會被解僱。所以一個例子是,在我的上一個團隊中,我感覺他們會嘲笑我在 crit 時說的一些東西,我說的某些短語。
LR
那有什麼例子嗎?
JW
哦,我總是會說,「好的,下一步是什麼,我們如何跟進呢?」然後他們會說,「好的,下一步是什麼?」他們會那樣模仿我。是的,我只是認為這表明了一定程度的,好的,這些人不一定害怕我,他們知道他們可以信任我,他們可以信任我,然後他們對彼此了解足夠,並且對我個人和我們的個人生活了解足夠,能夠知道那些邊界在哪裡。但同時,我認為你在那個領地犯的錯誤是,作為經理,你是和你的下屬做朋友嗎?這是我認為人們告訴你不要做的事情。所以我對這個的想法方式是你必須創造這個心理安全的基線,人們對彼此和你都感到舒適,但你也必須確保他們知道你有非常高的標準。我認為這兩件事可能感覺像是在衝突,但我認為它們真的很好地協作,因為就像一旦你有了那種心理安全,你有了彼此相信的人和你,應用高標準實際上我認為可能變得更容易,因為你可以在沒有恐懼的情況下做到,我認為。我從某種程度上像嚴格父母的方式來思考這個。就像,「哦是的,我的團隊,我與他們的工作方式是他們知道我總是會在那裡,我不會突然解僱他們或什麼。但同時,他們也知道我希望他們最好,我有高標準,我正在與他們合作以做最好的工作。」所以是的,我認為你需要達到的平衡就像,你能創造一個環境,其中你的團隊感到舒適地調侃你,但同時他們知道他們必須做很好的工作,他們會與你一起做很好的工作。
LR
這太棒的建議了。有趣的是,這種管理風格經常回歸,或管理,好的管理,它回歸,提醒我,那是什麼,激進坦誠,就是這種關心和直接挑戰的結合。這就是我在這裡聽到的,就是確保人們知道你深深關心他們,但也要非常直接並有高標準。
JW
是的。是的。
LR
那太有趣了。好的。也許最後一個問題。我一直在尋找有趣的框架、方法和流程,人們在他們的工作中發現有用。我聽說你是一個叫做可讀性框架的東西的大粉絲。談談這個,談談你如何使用它,為什麼它如此寶貴。
JW
是的。這個框架,我想我去年在 Twitter 上看到的什麼時候,是 Evan Tana,他是 SPC 的合夥人。他是一個風投。所以基本上它是這個二乘二的矩陣。我認為它沒有得到那麼多關注,但一旦我開始看到它,我實際上無法停止思考它。所以在二乘二上,他基本上有創始人。創始人要麼不可讀,要麼可讀,然後想法要麼不可讀,要麼可讀。基本上他說的是,「好的,如果創始人和想法都非常可讀,這個想法可能不是那麼新穎,有人已經像,他們已經要實施它或做它,你實際上沒有找到新的東西。」但是後來變得真正有趣的地方是想法本身不可讀的地方。他說的不可讀是什麼意思,他的意思是,「哦,它在某種程度上真的在前沿,人們可能還沒有理解,」或者說它的方式,它不是以對人們最有意義的方式被講述的。我認為這顯然對風投來說是一個好方法,因為你試圖尋找人們看不到的機會並將它們呈現給世界。但我也認為設計師的角色的一部分,至少,至少在 Anthropic 的前沿實驗室,是發現那些不可讀的想法,並試圖理解那裡有什麼,以及如何接受它,無論是通過故事講述,還是通過實際的 UX 和形式因素,並將其呈現給世界。我認為當我提到 Slack 並查看人們正在製作的所有東西時,這就是我正在做的。我試著看,哦是的,有什麼想法,有一些能量在周圍,但可能還沒有意義,值得我在我的工作中更多地思考?有一個很好的例子,實際上,與 cowork 相關,其中有一個我們稱為 Claude Studio 的內部原型,我認為有人去年中途建造的,它本質上只是這個真的很密集、強大的介面,建立在一些 agentic 工具上。那時它可能也是 Claude Code。它有所有這些顯示,顯示你所有的知識和 Claude 正在做的所有技能以及預覽它的輸出。我認為,作為設計師,我看著它,我想,「我不知道發生了什麼。我真的不懂。」但是後來我開始看到研究人員、構建它的人以及內部的人,周圍就是很多能量。我想,「很酷,我認為這裡有什麼東西,但我只是還沒有理解它。」我認為這真的是一個不可讀想法的例子。最後從它出來的是技能框架和 markdown 文件,它指導 Claude 如何做某事。這不是我直接參與的東西,這更多是,處理這個原型的人從中提取的。但是當我確實開始在 Claude cowork 上工作並思考,哦是的,這些東西的形式因素是什麼時?看到那個原型,看到人們發現真的有幫助的信息類型,看到 Claude 的計劃和待辦事項,看到 Claude 的上下文以及它正在進行的文件,這些是我最終從那個原型中提取到 Claude cowork 中的東西。所以是的,我考慮設計師如何幾乎可以在看原型時更像風投,內部地。
LR
這超級有趣。我最近與這傢伙 Terrance Rohan,一個風投,進行了一個研究項目,我們看了什麼是在公司早期加入的人中的模式,最終成為了巨大的成功,像 Palantir、Stripe、Linear 和 Notion,所有這些公司。在許多這些公司早期加入的人,他們在尋找什麼?其中一個因素是這個想法是如此瘋狂,每個人都在嘲笑。「這是不可能的。你永遠不會做到這一點。這是最瘋狂的事情。你為什麼甚至會想……」OpenAI 實際上是其中之一,只是一些研究實驗室在做一些事情。所以有趣的是……它不像每次這都會奏效,也不像每個沒有意義的瘋狂想法都會很好。但我認為你在說的是注意對你有趣的東西,但不是完全清楚。也許你可以成為幫助將其匯集在一起的人。
JW
是的。是的。那個,但也如果有一些能量在周圍,但我不太總是理解能量是什麼,更深入地挖掘並試著理解那是什麼。
LR
明白了。
JW
是的。因為我認為經常傾向於這些早期想法的人,他們不能總是清楚地表達為什麼。這在某種程度上是你要更深入地挖掘並理解的。
LR
所以有三種模式我們發現。其中另一個是,只是,注意人們對這件事非常興奮,即使你不懂。聽起來很瘋狂。這太有趣了。好吧,另一個是什麼?哦,創始人就像是百分之一,是另一個部分,每個人都有 Anthropic 已經。所以你得到了那個。哦天哪,Jenny,我們生活的時代有多瘋狂。多麼不可思議的世界。
JW
多麼瘋狂的時代。
LR
哦,有太多變化。好的。在我們開始非常令人興奮的快速輪之前,還有其他我應該問你的嗎,你想留給聽眾什麼,你想再次強調什麼?
JW
我認為我只是想呼籲 Anthropic 設計團隊並為他們喝彩,只是因為這是一個人真的很謙虛的團隊,他們不總是在 social 上,但他們正在做很多真正偉大的工作。特別是在這個時代,我們的工作變化如此之大,團隊非常有彈性,他們跨越了整個人員光譜,從真的很技術性和原型性的人,一直到真的非常高工藝和交付東西的人,那些東西正在出門並且是棒極了。我們全年都在招聘,所以我只想指出這一點。如果我沒有用我們在 Anthropic 內部工作的方式嚇到你,如果這聽起來比可怕更令人興奮,很樂意聯繫。
LR
聽起來令人興奮的是獲得進入那些正在構建未來的 Slack 的訪問權限。
JW
是的,那是核心好處。
LR
現在讓我們與超級智能交談,告訴我我應該在哪裡投資。我只是在開玩笑。然後在招聘方面,有什麼具體的人應該思考,如果他們想申請?
JW
想一想我提到的原型,特別是強大的通才和深層次的專家。那些人是我們真的很興奮的。但通常那些人——
LR
方塊和深層 T。
JW
方塊和深層 T。我們在談什麼?是的,感覺這些原型與他們共鳴的人,但也人們真的對技術很興奮,一直在構建很多,並且有點想在前沿。
LR
太棒了。嗯,有了那個,我們已經到達了我們非常令人興奮的快速輪。我為你準備了五個問題。Jenny,你準備好了嗎?
JW
好的。我準備好了。
LR
我們開始吧。你最經常向其他人推薦的兩三本書是什麼?
JW
第一本是《權力掮客》由 Robert Caro 著,這是一個相當激進的推薦,考慮到它有 1100 頁。但我認為,在這個時代,當我們的注意力跨度如此之短時,我認為這實際上值得從頭到尾閱讀,因為我認為很少有收藏或回憶錄跨越某人的整個生活,你有點看到某人如何在那些十年中改變。它也是一個真的很有爭議的人。讀一個真的很微妙的關於某人的看法,Robert Moses,這很好。我認為我們只是因為我們這麼多地思考現在而失去了一些這個長期的想法。所以這只是一個重要的提醒,職業生涯很長,並且對於理解某人如何真的做得很好也很有用。所以《權力掮客》。我向很多人推薦的第二本是一本叫《失眠城》的書,由 Bill Hayes 著,他是科學家 Oliver Sacks 的伴侶,在 Oliver Sacks 去世的時候。它只是這個真的很漂亮、飄飄然的回憶錄,關於 Oliver Sacks 的最後日子,以及他們的愛情故事。我認為這與你的播客 Lenny 上的東西關係很小,但這只是我真的很喜歡的一本書,只是讓你思考死亡,還有愛和生活和這樣的東西。所以那是我最喜歡的書之一。
LR
我這裡的目標是,我們試圖創造文藝復興時期的人類,所以所有這些其他東西都是很好的。
JW
很好。
LR
有趣的是,我看到 Julie Zhuo,著名的設計領導者,最近正在讀《權力掮客》。我不知道這裡發生了什麼。在設計界傳播。好的。你最近看過最喜歡的電影或電視節目嗎?
JW
我最近看了《感傷價值》。我在飛機上看的,這是導演希望你看他們電影的方式,但這是一部挪威電影,由同一個導演製作的,他做了《世界上最差的人》。我認為只是節奏、寫作、角色之間的關係,就是真的微妙和漂亮。它基本上是關於這個家庭,有點像家庭戲劇,但也是關於他們一生中住過的這所房子。它很漂亮,因為房子有點像一個角色。所以我不知道還能說什麼關於那個,但那是一部真的很好的電影。然後我也會顯然推薦《The Pitt》第二季。我們在那上面。我認為每個人都喜歡看真正擅長他們工作的人做什麼。所以是的。
LR
想像一下成為那個節目的演員,就像你必須學習和記憶所有這些術語的程度有多高。
JW
是的。是的。這似乎也真的快節奏。他們在一個鏡頭中做了這麼多東西,有這麼多動作和這樣的東西。在上面成為演員似乎真的真的難。
LR
我最近才意識到 Noah Wiley 作為年輕人曾經在《ER》中,現在他是這個的負責人。是的。是的。哦,天哪。好的。你最近發現的最喜歡的產品,你真的很喜歡,不能是 Cowork?
JW
不是我最近發現的,但 Retro,我一直在使用它,自從它推出已經大約兩年了。我認為我最近發現了它的新好處。所以對於不知道 Retro 的人,它有點像這個小社區照片分享應用程式,你只能分享來自每週的照片,來自給定的週,而不是所有時間。它基本上沒有任何社交媒體的東西。你無法真的看到計數,沒有廣告等。但一個真的很好的東西是現在我已經使用了兩年,我現在可以回頭看看每一年,看到,哦是的,這週,兩年前,我在做這個,它已經成為了這個真的特殊的方式來活過我生活的每一週,基本上。
LR
哇。它也是一個設計精美製作的應用程式,如果你想要構建自己的設計品味。
JW
是的。設計師喜歡 Retro。
LR
我可以看到那個。好的。你有最喜歡的人生座右銘,你在工作或生活中經常回到嗎?
JW
是的。不確定它是否是我最喜歡的人生座右銘,但我經常發現自己說的一件事就是,「就是這樣。」
LR
我的爸爸一直都這樣說。我喜歡它。
JW
是的。
LR
是的。
JW
它聽起來超級投降主義者,但我保證它不是。我認為只是鑑於世界上發生了多少東西,特別是對於行業和諸如此類的東西,你無法控制一切。所以有時候就是這樣,只是帶給你前進所需的輕鬆感。
LR
我前段時間進行了一個 10 天的冥想靜修,回來後就像,「爸爸,你一直都是對的。這是所有答案。就是這樣。」不要執著,不要試著改變。就是這樣。
JW
就是這樣。
LR
有這麼多深度那時我就像,好的,比我想的都要聰明。好的。最後一個問題。回到 cowork,有沒有,我不知道,令人印象深刻的用例,什麼只是像,哇,cowork 可以做到這個,這太酷了。要麼是你用它做的事情,要麼你聽說有人用它做的事情?
JW
我真的喜歡的一件事就是內省。所以我有這個文件夾,基本上,是我有的本地筆記,我用 IA writer 來做。我基本上只是寫任何東西,多年來一直在用很多不同的筆記來收集它,他們跨越所有不同的東西,像一對一的筆記、隨意的想法、小備忘錄、採訪筆記等。我最喜歡的,對我來說很酷,只是使用 cowork 來分析它並獲得洞察,並實際上從中創建東西。所以任何時候我都可以學到關於自己的新東西,我喜歡那個。但我認為一個非常實用的事情是我前幾天用它做的是沿著招聘的思路,我想,「哦是的,我想有點表達我在尋找設計工藝時的樣子,」因為我認為實際上很多人都難以表達那個。我只是讓它讀遍我所有的筆記,包括採訪筆記和我關心的其他東西,以及備忘錄和我過去寫的那樣的東西。然後它為我製作了一個評估的標準。所以那種內省,就像,哦,我甚至不會意識到我一直在隱含地提出的這些關於自己的事情。那對我來說真的很酷。
LR
那太酷了。所以只是為了讓人們非常清楚,你有一個文件夾,裡面有你寫的所有東西,一對一、會議,就像你可以做的,我不知道你是否被允許使用 Granola 或類似的東西,會議記錄,並詢問它。我想知道你使用的提示是什麼,但它只是像,「讀我寫的所有這些東西,並幫助我可能只是理解我對什麼是 DesignCraft 的感受。」
JW
是的,基本上我認為它像,「嘿,我在這個文件夾中有一堆採訪筆記和一堆與 DesignCraft 相關的筆記。讀它,然後幫助我製作一個備忘錄/標準,說明我如何在採訪中評估工藝。」
LR
太棒了。
JW
是的。
LR
Jenny,這太棒了。多麼不可思議的時代。
JW
多麼不可思議的時代。
LR
兩個最終問題。人們在哪裡可以找到你在線,如果他們想聯繫,聽眾如何對你有幫助?
JW
是的,我在 Twitter/X 上,這是我們這些天所說的。這是 @jenny_1。這可能是最好的地方,不是真的那麼多在 LinkedIn 上,所以那是最好的地方。人們可以如何幫助我?給我們你的產品反饋。我們正在處理 Cowork,或真的任何 Claude 相關的東西,只是給我們你的反饋,我們很樂意為你改進它。
LR
Jenny,非常感謝你在這裡。
JW
是的,當然。與 Lenny 聊天很棒。
LR
這很棒。Jenny,謝謝。再見大家。非常感謝你的收聽。如果你覺得這很有價值,你可以在 Apple Podcasts、Spotify 或你最喜歡的播客應用程式上訂閱該節目。另外,請考慮給我們評分或留下評論,因為這真的幫助其他聽眾找到播客。你可以在 lennyspodcast.com 找到所有過去的劇集或了解更多關於該節目的信息。下一集見。
✦ ✦ ✦
完整訪談逐字稿 · Lenny's Podcast · 2026.03.01
Part 1 · Introduction
LR
Today's guest is Jenny Wen. Jenny was head of design for Claude, is now leading design for Claude Cowork. Prior to that, she was director of design at Figma, where she led the design teams behind FigJam and Slides. She was also a designer at Dropbox and Square and Shopify. And what I love about this conversation is that Jenny is living in the future of where design as a profession is heading, and she's here to give us a glimpse into what that looks like, and how much things are going to be changing for designers. It is pretty wild and extremely interesting. A huge thank you to Noah Levin and Emily Lynn Hasham for suggesting topics and questions for this conversation. Don't forget to check out Lennysproductpass.com for an incredible set of deals available exclusively to Lenny's newsletter subscribers. Let's get into it after a short word from our wonderful sponsors. Jenny, thank you so much for being here, and welcome to the podcast.
JW
Yeah, excited to be here.
LR
I've been looking forward to this conversation because I spend a lot of time on this podcast talking about the future of software engineering, how much that role is changing, the role of product management, how much that role is changing. I even spent a lot of time on how design is changing. Clearly, it is also changing in a really big way. And you have such a front row seat to where things are heading. I also know you have a lot of very strong opinions about where things are heading, so there's a lot of stuff I want to talk about. I want to just start with just this broad question. How is the design process changing with the rise of AI?
Part 2 · The Design Process Is Dead
JW
It's changing a lot. I think it's still also got a long way to go in terms of the way it's changing. I think we've actually seen engineering change a lot more in the past little while than design actually has, but I think the result of engineering changing a bunch is that design is sort of forced to change. And so I think some context around this, is I did a talk at a conference in Berlin a few months ago in September, and I called it Don't Trust the Design Process, where I basically just said, "Hey, you know this design process that designers have been taught, where you go and you go off and you do a bunch of research and discovery, and then you diverge, you converge, diverge, converge." And it's like this process that we sort of treated as gospel and tried so hard to preserve and we were like, "Trust the process." That's basically dead. I think it was sort of dying before the age of AI, but given now that engineers can go off and spin off their seven Claudes, I think as designers, we really have to let go of that process. And I think that's the big thing that's changing. But I think even in the past three to four months since I did that talk, that talk actually starts to feel pretty … It kind of feels outdated to me, which is a little embarrassing, but especially with the big shift of Opus 46 and a bunch of folks just really discovering and using Claude Code over the holiday break, I think we're seeing this force to change our process happen even more. The way I see it now is there are basically two types of design work, and design work is becoming really stratified in this new world. So there's the first one, which is really just supporting the implementation and execution. So this is the one where engineers are using their seven Claudes to create all these features, and anybody can put an idea out there, and you can just talk about an idea and somebody, usually actually an engineer because they're still better at implementing this stuff than we are, they will just make a scrappy version of it and you can try it out. And you as a designer actually do not have the time to make these beautiful mocks anymore, or to lead in this way. And then I think there's the second kind of work that feels also really important, which is creating the sort of vision or direction for things. This one feels like the hardest to make time for, and it's one that we still did before, but I think the shape of it's very much changing, because I think we used to go off and say, "We're going to do this design vision. We're going to go off and make this two-year, five-year, whatever, 10-year vision even, and we're going to point us towards something." But the way that the technology is changing now, we don't know what's going to happen in two years. There's too much changing, and it usually becomes a vision that's three to six months out, and isn't necessarily something that is creating this beautiful deck that's beautifully story told. It's sometimes just creating a prototype that points people in the right direction. And I think this kind of work is still really important in this world because in a world where people can spin off their seven Claudes, make whatever features they want in any kind of direction or in implementation, you need to point them towards something. And in order to make sure that we're all making something that makes sense together and is also done in a way where it's efficient. If we're all working towards something that has one greater cause, it's much more efficient to do that than just random things. And so that's the big shift that I'm seeing. And I think I have opinions about it now, but ask me in three months and it might actually change even more.
LR
So what you're saying here is it's not like you or the design field is like, "We need to change?" It's engineering and the fact that you can build so quickly, just forces the role of a designer to change because as you said, engineers can just ship, ship, ship, ship, ship. And what you're finding here is you're better off not blocking that, letting them cook, as they say. And then there's this mode of helping them along as they ship, bring it together, make sure it all connects, guide them a little bit.
JW
Yeah, I think so. Yeah. I don't think there is one unifying voice that's like, "Designers, we need to change right now." But yeah, there is the follow-on effects of engineering tooling really changing. I think we'll probably see design tooling change in this next year or so as well, but a lot of it right now is trailing that. And I think it's also really empowering for us too, because as designers, we also now have access to a lot of these coding tools. And we can be a part of the process in a way where we're implementing stuff. I'm doing a lot of last mile stuff where I'm implementing all the polish, and working with engineers really closely to get the feature across the line, and also prototype stuff in actual code as opposed to relying on engineers to do that again.
LR
How true do you think this is at all companies, at say AI companies, non-AI companies? Someone may be hearing this, okay, Anthropic Claude, okay, they're at the bleeding edge, for one. Two, it's developery a little bit, but I think people might be feeling like, "Okay, this is not going to happen at Salesforce. This is not going to happen at, I don't know, ServiceNow, wherever." So I guess, do you feel like this is where all teams are heading? Is it mostly AI, bleeding edge companies? How widespread do you think the design process shift is going to be?
JW
So the talk that I did last year has really been the most resonant talk that I've done. And so I think it's something that people are starting to feel across the industry, where they're like, "Oh yeah, we can't do the old design process anymore. We are using tools like Claude Code and v0 and whatnot, to start to spin up prototypes, and PMs are starting to spin up prototypes and stuff like that as well." So I think there's something there emerging. But the other interesting observation with that talk too, was there was actually also a decent amount of backlash. People clearly have invested their entire careers in learning, teaching, using this really stable design process. And I think there was a lot of discrediting like, "Oh yeah, we can't do without discovery. We can't do without these pieces of process." So I think there is still a piece of the industry that is not quite there yet in terms of this way of working, if that makes sense.
LR
Yeah.
JW
Yeah.
LR
And a big part of this is you could argue … the question is what leads to the best, most successful products in companies? And you could argue it's spending time doing discovery, user research, mocks, iterating, beta testing, or it could be just engineer shipper stuff that's okay, not amazing, good enough. We learn, iterate, build, iterate. Is your sense that that second path, not only is that just what everyone's doing, but that actually leads to the better product at this point?
JW
I think you sort of have to choose and use your discretion as to when to actually ship something, but I think the ability to execute, try something out and try it with real data, and a real user's kind of mindset in the product, I think that does lead to a better product, especially as we're all working with these new developing AI models that are non-deterministic, you can't mock up all the states, and you can't theorize and you can't even make a clickable prototype with it. You sort of have to use the actual models underneath, and you have to see people try it out with their use cases, because with these models, you can design them for different use cases, but you actually discover use cases as you see people using them. So yes.
Part 3 · A Day at Anthropic
LR
The other thing I always hear, and building on what you just said is just you don't know what people will do with AI. You don't know how good it'll be at certain things, the non-deterministic piece of it. So you can create these amazing mocks of what it might be, and then people use it in a completely different way, which is where Cowork came from, and probably even Claude Code at the beginning. And so what's it just like to be a designer at Anthropic? Just give us a day in the life of working at Anthropic, at the center of the storm.
JW
A good amount of time at Anthropic is actually just catching up on what's happening at the company. I think this is the company where … I've worked at a few other companies around this size where I think there's just a lot of information and a lot of things going on, but I feel really compelled to keep up with it. There's stuff that is model developments on the research side. And then at any given time, there are just so many different teams prototyping and trying different ideas out, and there's a bunch of different code names and stuff like that. And a lot of time I'm just trying to navigate and figure out what those projects are, because I think I'm just trying to spot and see, hey, what's coming up ahead for me? Because there's stuff from both the research team, but also some of our labs teams that are closer to research, and trying out and prototyping stuff. And then there's just stuff I want to try out. We have a bunch of prototypes and products internally that we can use, and I am just curious and I want to try those things out. And then I think there's also a lot of folks who internally have a lot of insights and opinions on where the industry's going. And some of those are just really interesting to read, because a lot of these are philosophical debates or directions of the company and stuff like that. And yeah, I feel like I just want to keep up with these things. Whereas I think in a normal company, I'm like, "It's fine. This is stuff that's happening outside of my reach. I don't really care as much." Where here, I think that it's both the volume and the kinds of things that are happening that I'm really interested in keeping up with. And then aside from that sort of keep up, that's not a huge part of my job, but I do think it's a really interesting part of it.
LR
Well, it connects to the point you made earlier, where a big part of the design role now is helping engineers and teams execute, not just telling them, "Here's the mock, here's the design." It's helping them stay on track, helping them connect ideas, create a cohesive experience as it's happening. So that makes sense.
JW
Yeah. Yeah. And I think part of it is just curiosity. It feels like I have this front row seat to so much that's happening in the industry. And so a lot of it is like, yeah, our Slack is a goldmine. I'm just excited to read through the things that people are working on, they're saying.
LR
I never thought about how there's already so much AI news to keep track of as a regular person. And then actually seeing what's actually happening inside a lab is a whole new set of feeds to watch.
JW
Yeah, yeah. I think that is the best AI news is probably internally, if you're ever at one of these companies in the Slack.
LR
Yeah. Just problem keeps getting harder. I'm just keeping track of what's going on. Okay. Okay. So that's part of the job. What else?
JW
There is still some of the traditional, let me think about what's happening in the future and let me make some designs for that. That's something that, for example, this week I've allotted some time to, where I'm like, "Okay, cool. We have been in a lot of execution mode for Cowork, and now I want to set aside some time to think about, hey, what does the next three months look like, and where could that actually go given where the market's at, where the models are at, and what could that be?" Because I think it still really helps to visualize that and show that to the team, and point everyone in the same direction. And then I also spend a bunch of my day just jamming on stuff with engineers. A lot of it is just a conversation, or white boarding, or going through something that they built and giving them feedback on it and being a designer in that kind of way, we're really consulting. And then I spend a part of my day in code, polishing, implementing stuff. Sometimes what happens is an engineer and I have worked through something and they've implemented a first version of it, and I just go in and polish it with them. And that's a really fun part of my job that I think didn't exist as much a few months ago.
LR
Are you still doing elements of the traditional design process? Prototyping, user research, panels, I don't know, just going out and the whole thing you described?
JW
Yeah, I think we're still doing all of that to some extent. We have a user researcher on the team who is putting together both traditional studies as well as surveys, and the whole team is reading those studies and that feedback. We are still prototyping stuff. I'm still mocking stuff up. I think it's just I have a wider set of tools now, and I think the proportion of time I spend doing each thing just has changed.
LR
Got it. Okay. So that's a really interesting takeaway. It used to be that was a huge … I guess what would be the pie chart of what your life was before, where it's like traditional thinking, planning, prototyping, mocking, research, and then just feedback and execution and out, today?
JW
Yeah. I think as a designer a few years ago, I would say maybe 60 to 70% of it was mocking and prototyping stuff up, and then spending some of the last 20 or so doing the sort of jamming with engineers, consulting with them, and the last 10% maybe doing coordination meetings, et cetera. But now I feel like the mocking up part of it is 30 to 40%. And then there's that other 30 to 40% there that is now jamming and pairing directly with engineers. And then there's a slice, I don't know how much I have left, but there's a slice of it that is now implementation as well. Yeah.
LR
Actually building and shipping?
JW
Yeah.
Part 4 · The Designer's AI Stack
LR
Amazing. So kind of following that thread, what's in your AI stack? What do you, as a designer, I know you're a manager and I want to talk about how you actually are IC also. What's in your AI stack? What tools are you using in your role?
JW
What is in my AI stack? Well, we work at Anthropic, so we're going deep on the Claude stack. I'm using, obviously, chats, Claude Chat, but increasingly more and more Claude Cowork. I've basically shifted all of my chat use cases over to Cowork, because I've been finding that it sort of is better at these longer running tasks. And most of the things I was asking Claude for are these longer running tasks. And then there's Claude Code, of course. I use it mostly with VS Code and the IDE because I'm usually tweaking front-end stuff, and it helps to just be able to see the code and then talk to Claude as well. I've been trying to actually use Claude Code more remotely, through both mobile and through Slack as well. It's really fun for somebody to say, "Oh yeah, this one icon's off or something," and you just at-mention Claude and Claude does it, and then you pick up the PR and it's done. That's been really, really fun too. And yeah, I think we're a fully Claude house here. So yes, that's basically my stack.
LR
Are you still using Figma as a designer?
JW
I am still using Figma, yes, yes.
LR
Okay. I was waiting to hear. Okay. So Figma is still part of your life. Being a former Figmate, is that what y'all are called?
JW
Yeah, Figmates, yeah.
LR
Yeah. Yeah. Okay. So I know there's this big debate on Twitter, just like, is code the future of design? Do we need many more, do we need a design? What's your sense? Figma's still important?
JW
I mean, as a former Figmate, maybe I'm biased in that way, but I think there is still … When I use Figma, I'm like, "Yes, this is what I should be using." And it still fills a very good gap for me. I think a lot of that is actually just, one is exploring a lot of different options. I think that's a really important part of the design process, to be able to just think about 8 to 10 different ways to do something. I think the best design happens when you're able to just throw a bunch of ideas at the wall, and curate and push yourself to come up with a bunch of these different directions. Right now, coding, or right now working with some of these coding tools doesn't lend itself super well to that, because it's super linear, you get super invested in one direction and you just iterate with Claude on them, for example. So Figma has been really great at just exploring all these different options, and I think it's still going to exist that way to some extent. And then I think there's really fine visual and interaction details that are also really great to be able to just try out in Figma. Again, it's a lot of different directions, but it's micro directions. It's being able to think about different typography or styles. Having those in a canvas where you can just explore that specifically is still so, so helpful, and is not something that I always want to go directly to code in.
LR
It's interesting you still use an IDE, because in engineering, it's clearly shifting to command lines, agents, IDEs are kind of moving to not be cool anymore. And it makes a lot of sense. You just want to edit some CSS things, some color stuff. And so I could see why not just telling the agent, "Hey, just come on, change this one hex value." Just changing it is so much easier.
JW
Yeah. It's really annoying to be like, "Can you change this to this class?" When you can just go in and change it to a different class.
LR
So it's interesting. I wonder if IDEs now become the useful for designers and PMs, and engineers have moved on?
JW
Yeah, maybe. Yeah.
Part 5 · Working with Engineers
LR
Okay. So a lot of your time you spend working with engineers, giving feedback, nudging them in the right direction. There's a sense, I feel, of just your advice is let go. Don't feel like you need to be this gatekeeper, but there's this piece of, okay, help them move in a direction that is cohesive and is creating products we're proud of. A lot of designers I think are in this boat right now, just like, "Oh my God, I can't keep up with all these engineers shipping stuff all day." What's something you learned about just either how to help your engineers get better at design so that it just ends up being better, or just keeping on top of this and not going crazy?
JW
Whenever I do work with engineers on projects, and it's more on a consulting basis, I do just try to explain why I'm thinking a way that I'm thinking, to help them extract principles. As opposed to me just being like, "No, I don't think this should go here." It's like, "No, I think we should have a button here because not everybody realizes you can prompt this." And here's an example where it comes from research and whatnot. So I also just try to point engineers to our design system and stuff like that in code, because right now Claude is writing a lot of the code and it's not always picking up stuff in the design system and whatnot. So as much as I can equip them with stuff that they can use in the future without me, if that's helpful. And then on your point of trying not to go crazy, I think it's hard. I think it's really hard right now. **And I see this a lot from actually both engineers and designers, where it's like now that we're sort of capable of doing so much, we want to do more. And so I think it's not just designers who are feeling like, oh yeah, we have to keep up with engineers.** I think even engineers are like, "How do we keep up with ourselves right now?" So that's something I'm hearing a lot.
LR
So true. Oh man, how to keep up with all our agents, our seven agents we're constantly running?
JW
Yeah.
Part 6 · Speed, Quality & Trust
LR
Okay. So then as a designer where in this profession, craft and great experience and quality and trust are such a core part of the job, to help instill that in the products, because that in theory leads to really successful products in companies, how do you just think about maintaining craft quality, trust, as your products are just shipping 1,000 times a day, and you're not able to stay on top of them and there's no designer involved?
JW
It's not that there's no designer involved. It's more just like it's almost that there's too much for one designer to handle. But I think with this, I think about where the features or products are, where they are in the cycle of adoption versus early preview. So for example, we sometimes will launch things and we will say, "Hey, this is a research preview. It's early. It's going to have a bunch of these flaws," and we caveat that a bunch. I think Claude Cowork is actually a good example of this, where we labeled it a research preview and we put it out there knowing that, "Hey, this is similar to our models. This is the worst it's ever going to be, but we're going to put it out there because we believe, internally we've tried it a bunch, and there's something really powerful here that some people will benefit from. It might not yet be the easiest to work with. It might not be the highest quality. It might have some issues with it, but we're going to put it out there because we believe the benefits outweigh the cons." I think that is okay to do, especially when there is something really valuable with the product already, and it's worth putting it out there. But I think the promise you sort of have to make your users is like, "Hey, we're going to put it out there, but we're going to iterate. We're going to take your feedback and we're going to iterate and we're going to make it better." And you have to commit that. You have to show that to the world, you have to respond to people's feedback, and you have to show that you are continuously shipping and improving it. Because I think the way that you really lose trust around quality and releasing something early, is if you release it early and then nothing ever happens. That is something that degrades a brand. But whenever you put something out early, it's possible to do that and maintain the brand of your company. And I think that's something that we've been doing pretty well. And I think anyone who's listening can take away from it, it's like, yeah, well, we're continuing to do that. And I think that is actually really fun for me as a designer, because you put something out there and you actually learn and you get feedback about it immediately, and you know what to do next.
LR
The way I've heard you describe this is building trust through speed.
JW
Yeah, for sure. Yeah, it's building trust through speed, but also just making people feel like they've been heard and that we're fixing things based on what they're trying to use it for, and their feedback is actually appreciated and used.
LR
Yeah, it's clear when the labs launch stuff, and you all are very good at this, everyone on the team is tweeting and just responding to tweets and comments and then shipping, "Hey, we fixed this yesterday and this is happening." So there's a clear sense of, "This is just today and we know this is broke and we will fix it." And then because Claude Code can code very quickly, the fixes come very fast. Okay. So another big question that people are asking that I ask a lot on this podcast, is around just what skills become valuable? And another way I've been thinking about it, Lex put it this way recently, is where will human brains continue to be valuable as AI gets smarter? So we've gone through this progression of tab completing [inaudible 00:27:56] segments of code, to 100% of code is written by AI now, it's crazy, to now AI is reviewing its own code. Boris on the podcast recently was saying Claude Code is now helping them come up with ideas and decide what to build, which is like, okay, wow, look at that. Look at it go. The whole product workflows, the product development process slowly get eaten up by AI. So the question is just where will human brains still be useful, at least until we have super intelligence? Do you think AI is going to get very, very good at taste, judgment, design?
JW
I think it will get better at taste and judgment and design. Yeah, I think we might be holding onto that a little bit too much and saying, "Oh yeah, a designer or somebody will always know the best thing to ship or the best version of this." But I do think AI's sense of taste will get better. At the end of the day, someone has to decide what is actually going to get built and what actually matters. And when I think about people saying, "Oh, AI is just going to build this software for us," a lot of the hard parts of building software are actually not building it. If you think about the hardest times that you've had at work, Lenny, it's probably things like, oh, you and some other person disagreeing about what should go into this feature or what shouldn't go into this feature. And those things still feel like, yes, AI can weigh in, but it can't necessarily solve this dispute between you and somebody else. And so there is something about deciding what actually goes into the things we build, which I guess is taste in some way, but maybe not taste in the way we think about aesthetic taste or whatnot. There's some sort of, it's judgment around what to do next.
LR
Just watching how quickly AI took over coding, which, I think a year ago, definitely two years ago, most people are like, "I don't think so. I don't think AI will get this good." And that the best engineers in the world trusted so much, they're not even looking at the code anymore. That's where we've gone. It just made me reevaluate all these assumptions I've had about, okay, AI will never be as good as really good PMs, designers, at judging what is great and deciding what to build. But I'm just starting to think, I think it will get there. Even an example you've shared, it could give these two people trying to make a decision, "Here's all of the data you need to make a decision and here's why this is the right answer and just press yes, press one, and I'll go ahead and build this." So I think to your point, I think we undervalue just how good it'll get at this stuff. Okay. So your sense is it'll get better, but your sense is we'll still need awesome designers to be involved, us and PMs, to help make these decisions, engineers, of course?
JW
Yeah. Yeah. I think someone will still have to decide, oh, we want to build this kind of product. Or given what the AI is presenting us, someone still needs to be accountable for the decision. The same way that even though Claude can write all this code for you today, it is still an engineer who's accountable for, does that code actually work? Does this actually make sense in the product? So I think there's that decision making/judgment layer, which feels like maybe one day will come when we won't have to do that, but it still falls on us. Yeah.
LR
It doesn't make sense. It makes me think about the radiology example, where there's always the sense that AI is going to take over that field of radiology and tell you what is going on. But the human is mostly useful for signing off on the decision, because someone needs to be liable if they're wrong. Which isn't the best job in the world, but that's a different game as [inaudible 00:31:38] code.
JW
Yeah.
Part 7 · Chat Interfaces & the Future of UI
LR
Okay. Another ongoing question in AI and design is just, it feels like chatbots and terminals are just like, I don't think anyone expected this to be the lasting user interface to AI. Chatbots, okay, no, no, this is just a temporary stop along the journey, but now it's gone even further, and just terminals. Do you have thoughts on just, I don't know, do you think there will be a next step of how we interface with AI, or do you think chatbots and terminals are mostly where we end up?
JW
There will likely be a combination of both, both UIs and interfaces that you are interacting with, clicking with, and that feel more tactile., We are already seeing this and playing with this within Claude, the chatbot. So we recently released a bunch of these widgets that let Claude elicit and ask you questions, and also show you things like the weather and stocks and whatnot in interactive ways. And I think those have had a really good reception, because people still like to see UIs and touch them and click them, and they are much more efficient than typing something to Claude. But at the same time, when we really leaned into this chatbot paradigm, I think that just gave us this whole world of flexibility that we didn't get with these sort of baked-in UIs. So my read here is I don't think Chat is ever going away because this opened up this new way of infinite ways to work with the model, and to talk to the computer, that we just didn't have before. But I think that it will still be most direct for very specific things to exist in this UI. And I think that what will probably happen here, is that a lot of those UIs will be generated more and more often by the models, as opposed to something that we're hand coding each instance. But I think we're in this space where I don't think chat … And maybe even talking to the terminal is going to go away.
LR
It's interesting that with OpenClaw, Claude, Moltbot, all the names, one of the big innovations is another way to chat with it through WhatsApp and Telegram and SMS, just like another form of chatbot, but just like, that was a big [inaudible 00:33:54]. Oh, I could just chat with it through WhatsApp.
JW
Yeah. And it's like chatting and talking to someone is still … We as humans are doing it, and it's a way for us to interact in a really rich way. And now we just have this other medium to interact with a computer, basically.
LR
Yeah. So Kevin Wheel, who works at another AI lab I won't mention, he had this great point on the podcast that talking is such a beautiful way to handle every level of intelligence. We can talk to people that are very, very smart and not so smart and it's talking, and it scales so well across the spectrum. We can talk to people at 200 IQ, 300, like its talking still works. So that's why it's been this beautiful way to deal with the growing intelligence of models as it continues to work.
JW
Yeah, that totally makes sense. Yeah.
Part 8 · IC vs. Management
LR
Okay. I want to come back to this whole idea of management and IC. So you've kind of put yourself back into the IC role in a lot of ways. Talk about that. And if you think that's just a thing design managers need to be doing.
JW
Yeah, I have takes on this. Yeah. So this past year at Anthropic, I joined as an IC at first, and then I managed a team for a few months in an org structure that sort of needed it, and now I'm actually back to doing full-time IC work. And I joined Anthropic as an IC because I was just really excited about the kind of work that there was to be done as an IC here, but also because I was feeling like I sort of want to be closer to the work, and I think this feels like a really important time to do it, before I ascend the corporate ranks. And I was having these questions and doubts about, is middle management, is that safe in the future? Is the way that we're working actually, is this going to be a job that persists in the future? Or should I try something else and get my hands dirty kind of thing? And to be totally fair, I actually love both sides of the coin. I love managing people. I love setting up teams and being at that level, but I also just really love IC work. I was sort of a reluctant manager when I did it, and I was like, "Okay, I'll do it." So I love both sides of the coin pretty equally. But I think actually what being an IC across this past year has taught me, is that it actually just gave me a lot of skills that I don't think I would've gained if I was just managing throughout this year. Like I mentioned, the design process has changed so, so much in this past year, and I feel like I've just picked up so many hard skills that I wouldn't have necessarily had the time to do if I was just managing a team. So that's actually the best thing it's afforded me. And I think at any point if I'm managing a team again, I think it will give me the empathy and understanding of how the design process has changed. And I think that's actually a really important thing right now because the teams are working so differently. And I think it's actually pretty hard to empathize if you are not working in that way, or you're not always testing all the tools and trying stuff. But yeah, it's an interesting time to be a designer. And if I had not worked in this environment, I don't know if I would've totally understood it or knew what to do or how to guide my teams. So that's sort of what this year really gave me.
LR
And so you were previously a director of design at Figma, right?
JW
Yeah.
LR
How big was your team? How large was your org, just to give people a reference?
JW
At the max, I probably had, I think, 12, 15 designers or so, and I had a few managers as well.
LR
Cool. And then you went back to an IC?
JW
It was like an org.
LR
Yeah. Okay. So you had the sense that middle management might not last. What's your current feeling? Do you think design management is a thing that persists long-term, or do you think everyone turns into IC?
JW
I think as long as there is a team of people, it helps to have somebody who is managing a team. I think there's real value in managers. It depends what the shape of the manager is and what they actually do. But the way I think about what a helpful manager is these days, is somebody who is not just, I think pure people management like, oh, just somebody to set you up, help you in your career, have one-on-ones, make sure you're feeling good at work. I think that that is not a thing as much anymore, but I think somebody who can really function as giving the team direction, as well as doing some of the people management stuff, that tied together, I think is the future of what managing looks like, at least for now. Somebody who can really engage with the team in terms of the work and giving direction there, as well as creating the environment for them to do their best work.
LR
And do you see yourself going back into management long-term?
JW
I will. I probably will. I think I really just love helping a team build the best product possible. And my motto there is, whatever it takes. If it's somebody that, if the team needs somebody to give the team direction and set up the team and whatnot, that could be me. If the team just needs somebody to execute on it, that could be me as well.
LR
So the advice I'm hearing for people in design that are especially managers, is you almost need to move back into IC in order to truly understand what is happening and how much it's changing, so that you can be a better manager.
JW
I think so. And I think traditionally, at least what I've seen, a lot of the engineering disciplines, when they hire EMs or even sometimes directors there, they actually make the EMs take a rotation for a few months and pick up a few tasks, and really understand how the technology works before they become a full-time manager. And I think design probably needs to do something similar too, where I think in the past design has been much more people management oriented.
LR
What did you find yourself most rusty in when you went back to IC designer?
JW
Actually doing crits, and just really-
LR
Getting criticized.
JW
… Yeah, getting criticized. You're like, oh yeah, it is hard to get critical feedback and to hear it, and to hear it on such a regular basis, because that's the thing you have to do as a designer, is it's a pretty vulnerable exercise to share work and present it with your team, and then also just get a lot of critical feedback and take that all the time. Yeah.
LR
So currently you're leading design/IC designing on cowork. Is that right?
JW
Yeah.
Part 9 · Cowork & Building Products
LR
Awesome. So Boris, he was on the pod recently, talked about how there's a lot of debate about what cowork should be, and there's all these big ideas and he's like, "In the end, let's just make it like a terminal, basically, in the product, and just kind of a fancy terminal." Is there anything you could share about just the process of landing on where you landed for that experience of cowork? I have it here on my monitor, by the way, looking at it.
JW
With Cowork specifically, we have had a bunch of different prototypes internally of what that could look like. And it's one of those things where we tried a lot of things, and then I think we weren't really sure when it was actually going to be ready to ship. And then it was sort of everything all at once. We were like, "Okay, we're going to ship it soon."
LR
I think it was 10 days, 10 days of building.
JW
Yeah, it was definitely longer than that overall. It was 10 days to get it from what we had internally to something that we were ready to ship externally. So we'd been building it for a while, but we weren't really sure about the actual form it was going to take. And so the way it got there is actually, there was a lot of different other explorations that we had internally on top of different agent harnesses and whatnot. And we just had prototyped little parts of the different interactions that ended up in cowork. So things like when Claude gives you a to do list, we tried a bunch of different form factors for that. We tried a bunch of different form factors for the way it presents you different multiple choice questions. We tried a bunch of different ways to teach people what the use cases are and whatnot. And I don't know if we landed on the best form factor ever, but essentially it was stuff that was already working internally that people liked, that we just thought we were going to get some more signal on by releasing it. So I think forcing ourselves to release it within that 10 days that we did, it was just sort of like, whatever we had, let's put it out there and then let's go out there and iterate from there, which is what we're doing.
LR
And it blew up the internet when you launched it, so it worked out.
JW
Yeah.
LR
Is there a feature of Cowork today that you're either most proud of, or just can't wait to fix and improve?
JW
Honestly, I think I'm just most proud of us actually just shipping it, to be honest, and putting it out there. And yeah, I don't know if there's one specific thing yet, because I think when you work on something and you work at so long, especially as a designer, you're like, I don't know. All I can do is see flaws in it, but I think there's a lot of stuff that I'm excited about. We have been iterating, especially on the homepage, and to make that something where it feels more like, "Hey, these are tasks you can give Claude and the tasks that Claude are working on. " And so that actually should be rolling out. It might already be rolled out by the time this comes out.
LR
I see this little randomizer thing, where you click it and gives you all these different ideas.
JW
Yeah, yeah. And then so when you actually start to work with Claude on stuff, it feels more like a to do list. It feels more like these are things Claude's working on, these are things that Claude needs your attention on. And I think there's an opportunity here to make it feel much more like this shared to do list between you and Claude. So excited to iterate on that. And then I'm also excited to think more about what is the actual true form factor of this? Is it stuck in this screen always, or how does this reach out to the different surfaces that it's working with?
LR
I love that you shared that it wasn't just 10 days to do this thing. There's these numbers that people throw out there, "We built it in 10 days." And your point is there was time spent thinking about what direction it should go, and prototyping, mocking, trying stuff. And then it's like, okay, now we know what we want it to be. Let's build it and ship it.
JW
Yeah. I think for some reason that became the viral thing that got taken away from all of the cowork announcements, is that it only took 10 days, but I think there have just been so many different explorations, and people that have worked on different pieces of cowork, that it was not just 10 days, and there was a lot of different people involved. It's one of those things where it's like the idea kept coming back, and it's never the right moment or there's different variations of it, and then all of a sudden it's the right moment and it feels like, oh, so obvious all along, but there was a long, long journey to get there.
LR
And by the way, for people that don't know much about cowork, the way I think about it, it's like Claude with hands, or to do stuff on your computer. How would you describe it, just in a sentence or two?
JW
That's a good description. I actually haven't heard that, but I like that. I might use it more often as Claude with hands. I also think about it as it's like Claude, but Claude is really good at taking all your garbage and then turning it into something nice. I think one of my favorite, any sort of use case that I really like out of cowork, is just giving it a folder of my stuff, and it doesn't really matter what's in that folder, but I'm able to extract something good out of it.
LR
I've done that many times. Okay. Coming back to managing and being a manager and the role of a designer, I'm going to talk about hiring for a little bit. So seeing how much is changing in the role of a designer, what do you look for that's maybe new? What do you now look for when you're hiring designers, that you think is really important for them to be successful in this new world?
JW
Well, I do think working specifically in the kind of environment that I do, there's just a sense of resilience and roll with it, kind of thing, that I think is really important because so much is changing around us and you have to be really willing to adapt, to try out new methods, to try out new tools and learn stuff, as opposed to just be stuck in the old processes and the old ways. But then I think about also, there's probably three archetypes of folks that are really interesting to me right now. And I think these folks were already interesting to me before, but I feel like in this era feels especially important. So the first one I would call is strong generalists. So not just regular generalists where they're kind of good at a lot of things, but people that are almost block shaped, in that T-shaped framework, where it's like they're really good at a few core skills, like 80th percentile good. I think this is pretty rare and hard to hire for, to be honest, but I like this because the design role we've already seen is stretching and spanning. We're all becoming more PM shape, we're all becoming engineering shaped. And so if you already have strong skills in a few different buckets, it's really easy for you to flex around and expand your role. So that's really exciting to me. It's just somebody who is really good at a bunch of things, again, a huge ask. And then the other person that's really exciting to me, is in that T-shaped framework, a deep specialist, someone who is T-shaped, but the tip of the T probably goes down farther than most other people. So folks that are maybe the top 10% in the industry and whatnot. Again, super hard to find. And I feel very lucky that working at some of these places, folks like these, you could sort of afford to hire them and actually bring them on board. And then my last one is probably the one that I think we're all overlooking, which is what I call the craft new grad. It's just somebody who's early career and feels, like, wise and experienced beyond their years, but is also just very humble and very eager to learn. And I think this person is really interesting right now because I think most companies are just hiring senior talent, folks that have done things before, are super experienced, but given how much the roles are changing and what we're expected to do is changing, I think having somebody who almost has a blank slate, and is just a really quick learner and is really eager to learn new tactics and stuff like that, and doesn't have all these baked in processes and rituals in their mind, that's super valuable. So I think those are the folks that I think a lot of us are just overlooking, but I'm really excited about.
LR
This is awesome. On the deep T shape, what's an example of someone in design that has a … What's a skill that they're really good at?
JW
Sometimes there's designers who are just really technical in a way where they could be 50% of … they're basically a software engineer. That's really interesting, especially because right now a lot of it is, at least for us, it's like you're working directly with the model, so it helps when you have just deep engineering expertise. But another deep specialist T is just maybe they're just really good at visual design or just designing icons or something, where things like that, given that anybody can make anything, having that deep specialist slant feels like, oh yeah, they can really help differentiate the things that we're building.
LR
Awesome. Okay. And then there's block shape. We had Mark Andrews on the podcast, we kind of called it the F shape or E shape, where there's multiple T things, sideways F, sideways E, I guess. Is that what you're describing, where there's many things you're really, really good at?
JW
Yeah. Yeah. And basically, I don't know, if you almost had their skillset, it would look like a block instead of a T.
LR
Because there's so many skills that the T is spread out.
JW
Yeah.
LR
Okay. And the crack new grad. So this is just someone that is eager, open-minded, gritty, very smart, I imagine is a big part of it.
JW
Yeah.
LR
Awesome.
JW
Yeah.
LR
If someone's a young designer trying to break in, trying to be successful, what would your advice be to them to help them have a shot at joining Anthropic, for example?
JW
I would just say just build a bunch of stuff, try a bunch of stuff out, build actual things. I think that can feel … I don't really know what the state of design education or education is these days, but at least from back when I was in school, everything was very theoretical, and here, we're going to teach you some approaches and whatnot. But the best crack new grad folks I know are just people who just use the technology, build actual things, don't feel limited by how little experience they might have. I think that sometimes they're actually unburdened by that, because we have expectations of ourselves after being in the industry for so long, but they actually don't and they sort of feel like anything is possible. And so just building a bunch of stuff and sharing it with people, and finding a community of folks that also do that. Yeah. I think my one call it here too, is I went to a school that started something called Socratica a few years after, actually a while after I graduated. And basically their whole thing is building stuff and showcasing it almost like a science project. And I think there's just been a really cool movement there of folks who just build things and do things. For example, somebody built this Claude robot project, this was a few years ago too, where they were just assembling robots that were running on Claude, and then somebody else did something where she just wanted to put Googly eyes on a bus in Boston or something. And there's just a sense of both agency in terms of, yeah, we can just do stuff, but then also this community where people were just trying and building things and sharing things with each other. So whatever that looks like, given the school that someone's from or graduating from, doing that kind of stuff is the stuff that will make people stand out.
LR
For current designers that are [inaudible 00:52:54] career, [inaudible 00:52:55] senior, do you think that you need to get technical and learn to code, at least build, or do you think you could be really successful and just not lean into that and get better at other stuff?
JW
I think it definitely helps to maybe not learn how to code so much that you're building something from scratch, but it does feel like more and more of the designer's vocabulary right now is implementing some stuff. I wonder though, as the models, both the models and the products get better, I mean, we probably will continue to move up the abstraction layers and you won't have to actually know how each single line of code will work. But I think what I would say is start to bring that into your toolkit, the coding tools, whether or not it's like you're actually becoming technical, I think any designer should just be really aware, and know how to use the tools that are at hand, as opposed to maybe learning and going off and learning React or et cetera, et cetera.
LR
How good of a designer is Claude, would you say, or Claude Code, however you want to describe … Would you hire Claude as a designer or is it not there yet?
JW
I don't think Claude is there yet. I don't think Claude is there yet in terms of a designer you would hire. I think it is not yet the strong generalist or the deep specialist or the crack new grad. I think it's pretty good at a first pass, and at presenting a bunch of different ideas to you, but nothing there quite feels like special and hireable yet.
LR
Which is good news for designers for now. It sucks at this for now. And I'm so curious to see how good it could get at this. That's such a big open question, is can it pump out amazing novel, unique creative experiences, or it's just never going to be that good as a human designer?
JW
I mean, it's gotten a lot better in the last year or so, even. So yeah.
Part 10 · Frameworks, Advice & Lightning Round
LR
There's a couple management, I don't know, rituals or takes you have that I've heard from folks that you work with, that I want to touch on. One is that you have this hot take that low leverage time for managers is just not a thing, that there's a lot of benefit you can get out of things that people consider low leverage. Can you talk about that?
JW
Yeah. Yeah. I remember first becoming a manager, and I think one of the pieces of advice that I either got from a course or a book or something, is like, yeah, "Now that you're a manager, you have to really prioritize your time and categorize your work." And there was this two by two of … I don't remember what it was in it, but you essentially say, "Oh, these are the things that only I can do. These are things anybody else can do, and everything else, it's low leverage and you shouldn't do that anymore." And a lot of the low leverage things were just things that are really nitpicky in the weeds, or just literally, yes, probably somebody else could do those tasks. But when I think about leaders and managers that I respect the most, I actually think some of their best traits is that they choose low leverage tasks that they take on themselves, and that actually ends up being actually a very high leverage thing, because it's them who's doing it. So one example is whenever you have senior leaders who just test the shit out of the product, and they're just so in tune with it, and they dog food it, they repro the bugs, they spend a bunch of time with engineers sharing the logs and nitpicking and stuff like that. And I think that ends up being super actually high leverage, even though it's a lot of time, of nitty-gritty time, because it creates this familiarity with the product, which I think is really good. It also creates this vibe where it's like, oh yeah, this senior leader really cares deeply and they actually know the ins and out of the product, and they're rolling up their sleeves and they're giving this feedback and working with the team on it. And I think similarly to what I've seen is when a senior leader is able to fix a bug now. I think I've actually seen Mike Krieger before put in PRs himself, and it's really nice because it's like, okay, cool, we're all on this team together and nothing is below this person. And I think another thing that I love that's a little bit more cultural, is when somebody goes out of the way to make somebody's anniversary card or something, and vibe code them something super nice, or make them something, a super nice card, because I think it just shows that it's like an EA or somebody can put together the card, but this leader is just somebody who cares so much about their team that they put in the effort. So that's something I try to embody, is choosing the seemingly low leverage tasks that are worth my time.
LR
Yeah, that is so interesting. What you're saying there is, in a sense, the low leverage stuff is the stuff that often has the most impact because your reports wouldn't expect you to spend time on this thing, and the low leverage ends up being high leverage.
JW
Yeah. And I think it's what makes your style of leadership stand out, or feel special to a certain person.
LR
Amazing. Another, I don't know, ritual and way of running teams that I heard about you, is you encourage team members roasting each other, which on the surface doesn't sound like a wonderful environment, but on the other hand, I hear constantly that the teams that you've built are just the happiest, the highest performing teams. Talk about, I guess, this idea of roasting and encouraging that, and just what you've learned about building awesome teams.
JW
Yeah, I think it's not that I'm like, "Yo, you should roast each other." I'm not forcing it in that way or anything, but when I think about the psychological safety, and teams and people that just get along with each other, when you think about your friends, you're always willing to push the boundaries a little bit and roast them. You're roasting your friends a lot, but you actually might not be roasting your coworkers a lot, because it's all just about comfort and safety. So it's not that I'm like, "Oh, I want my teams to roast each other." But I think it can be a really good sign when the people on your team feel comfortable just poking fun at each other a little bit. And I think that also can be a good sign when folks also feel the same way about you as a leader, where it's like there's just an element of they don't fear you as much, and they feel like there's a sense of safety where if they say something, they're not going to get fired. So an example of this is, with my last team, I feel like they would make fun of things that I would say at crits sometimes, certain phrases I would say.
LR
What's an example of that?
JW
Oh, I would always be like, "Okay, what are next steps, and how do we follow up on this?" And then they'd be like, "Okay, what are next steps?" And they would sort of channel me in that way. Yeah, I just think it shows a level of like, okay, these people are not necessarily afraid of me, they know that they trust me, they can trust me, and then they sort of know enough about each other, and me personally and our personal lives, to be able to know where those boundaries are. But at the same time, I think the thing that you err into in that territory, is as a manager, are you friends with your reports? Which is I think a thing people tell you not to do. And so the way I think about balancing this out is you have to create this baseline of psychological safety, and people feel comfortable both with each other and with you, but you also have to make sure that they know that you have really high standards. And I think these two things can feel like they're at tension, but I think they work really well together, because it's like once you have that psychological safety, you have people trusting each other and you, applying the high standards actually I think becomes potentially easier because you can do it without fear, I think. And I sort of think about this from the approach of being a tough parent a little bit. It's like, "Oh yeah, my team, I work with them in a way where they know I'm always going to be there and I'm not just going to fire them on a whim or something. But at the same time, they also know that I want the best for them, and that I have high standards, and that I'm working with them to make the best work possible." And so yeah, that's the balance I think you just strike is like, can you create this environment of one where your team feels comfortable roasting you, but at the same time they know they have to be doing great work, and they will do great work with you.
LR
That is awesome advice. It's interesting how often this just management style comes back, or management, good management, it comes back, reminds me of, what was it, radical candor, just this combination of caring deeply and challenging directly. And that's kind of what I hear here, is just make sure people know that you care deeply about them, but also be very direct and have high standards.
JW
Yeah. Yeah.
LR
That's so interesting. Okay. Maybe a final question. I'm always looking for interesting frameworks and methods and processes that people have found useful in their work. And I hear you're a big fan of something called the legibility framework. Talk about this, talk about how you use it, why it's so valuable.
JW
Yeah. This framework, I think I saw it on Twitter, maybe last year or something, and it was Evan Tana, who's a partner at SPC. He's a VC. So it basically is this two by two. I don't think it got so much attention, but once I started seeing it, I actually couldn't stop thinking about it. So on the two by two, he basically has founders. Founders can be either illegible or legible, and then ideas can either be illegible or legible. And basically he was saying that, "Okay, if both the founder and idea is super legible, the idea is probably not that novel, and somebody's already like, they're already going to implement it or do it and you're actually not finding something new." But then where it gets really interesting is where the idea itself is illegible. And by illegible, he means, "Oh, it's sort of really on the frontier, people might not get it yet," or the way it's being told, it's not being told in the way that makes the most sense to people. And I think this is obviously a good way for a VC to operate, because you're trying to look for the opportunities that people don't see and put them out there in the world. But I also think that part of the role of the designer, at least, at least at a frontier lab at Anthropic, is spotting the ideas that are illegible, and trying to understand what's there, and how to take that and transform it, whether it's through storytelling, or whether it's through the actual UX and the form factor, and put it out there. And I think when I mentioned going through Slack and looking at all the stuff that people are making, that's kind of what I'm doing. I'm trying to see, oh yeah, what are the ideas that there's some energy there around, but might not make sense yet, that are worth me thinking about more in my work? There's one good example, actually, that ties to cowork where there was this internal prototype that we called Claude Studio that I think somebody built partway through last year, and it essentially was just this really kind of dense, powerful interface that was built on top of some agentic harness. It might've been Claude Code at that point in time too. And it had all these displays where it was showing you all this knowledge and all these skills and things Claude was doing, and previewing its outputs. And I think, to a designer, I looked at it and I was like, "I don't know what's going on. I don't really get it." But then I sort of saw the folks in research, the folks building it and just folks internally, there's just a lot of energy around it. And I was like, "Cool, I think there's something here, but I just don't understand it yet." And I think that really was an example of an illegible idea. And ultimately what came from it was the skills framework and the markdown files that instruct Claude on how to do something. So that came out of that specific prototype. That was not something I was directly involved in, and that was more of, the folks working this prototype extracted that out of it. But when it did come to work on Claude cowork and I was thinking about, oh yeah, what is the form factor for those things? Seeing that prototype and seeing the kinds of information that people found really helpful, seeing Claude's plan and to-dos, seeing Claude's context and the files that it was going through, those kinds of things are things that I ended up pulling out of that prototype into Claude cowork. So yeah, I think about how can designers almost be more like VCs in this way, internally when we're looking at prototypes.
LR
This is super interesting. I did a research project recently with this guy Terrance Rohan, a VC, actually, we looked at what are patterns across people that have joined companies very early that ended up being massive successes, like Palantir and Stripe and Linear and Notion, all these companies. People that have joined many of these companies early, what were they looking for? And one of the factors was the idea is so crazy that everyone's laughing at this. "This is impossible. You're never going to do this. This is the craziest thing. Why would you even think … " OpenAI actually was one of them, just some research lab doing some stuff. So it's interesting that … And it's not like every time this will work, and it's not like every crazy idea that makes no sense will be good. But I think what you're saying is pay attention to stuff that's interesting to you and isn't totally clear. Maybe you can be the person that helps pull it together.
JW
Yeah. Yeah. That, but also if there is some energy around it, but I don't quite always understand what the energy is, it's to dive deeper and try to understand what that is.
LR
Got it.
JW
Yeah. Because I think people who often gravitate towards these early ideas, they can't always articulate why. And it's sort of up to you to dive deeper and understand that.
LR
So there's three patterns we found. One of the other ones was, there's just, pay attention to people getting very excited about this thing, even if you don't get it. And it sounds crazy. That's so interesting. Okay. And then what was the other one? Oh, the founders are just like top 1% was the other piece there, which everyone had Anthropic already. So you got that one. Oh man, Jenny, what a crazy time we're living through. What a world.
JW
What a crazy time.
LR
Oh, so much change. Okay. Before we get to our very exciting lightning round, is there anything else that I should have asked you that you want to leave listeners with, that you want to double down on?
JW
I think I just want to call out the Anthropic design team and shout them out, just because it's a team of folks that are just really humble, and they're not always [inaudible 01:07:49] on social, but they're doing a lot of really great work. And especially through this time our jobs are changing so much, the team is so resilient, and they sort of span this whole spectrum of people who are really technical and prototypey, to all the way to folks that are really high craft and delivering stuff that's going out the door and is fabulous. And we are hiring throughout the year, so I just wanted to call that out. If I didn't scare you with the way that we work internally at Anthropic, if it sounds more exciting than terrifying, would love to connect.
LR
What sounds exciting is getting access to these Slacks where all the future is being built.
JW
Yeah, that's the core benefit.
LR
Let's talk to super intelligence right now and tell me where I should invest. I'm just joking. And then in terms of hiring, is there anything specific that people should think about, if they want to apply?
JW
Think a little bit about the archetypes that I mentioned, especially the strong generalists and the deep specialists. Those folks we're really excited about. But generally folks who are-
LR
The block and the deep T.
JW
The block and the deep T. What could we be talking about? Yeah, folks who feel like those archetypes resonate with them, but also folks that are just really excited about the technology, have been building a lot, and sort of want to be on the frontier.
LR
Amazing. Well, with that, we've reached our very exciting lightning round. I've got five questions for you. Jenny, are you ready.
JW
All right. I'm ready.
LR
Here we go. What are two or three books that you find yourself recommending most to other people?
JW
The first one is The Power Broker by Robert Caro, which is an incredibly aggressive recommendation, given that it's like 1100 pages. But I think, in this era, when our attention spans are so short, I think this is actually worth reading end to end, because I think there are very few collections or memoirs where it spans through someone's entire life, and you sort of see how somebody changes throughout those decades. And it's also somebody who's really controversial too. And it's nice to read a really nuanced view of somebody, Robert Moses. And I think we just lose out on some of this long arc thinking, because we're thinking so much about right now. So it's just an important reminder that careers are long, and is also really good for understanding how somebody just gets things done really well. So Power Broker. The second one that I recommend to a lot of people is a book called Insomniac City, which is written by Bill Hayes, who was the partner of the scientist, Oliver Sacks, around the time that Oliver Sacks died. And it's just this really beautiful, ethereal memoir of Oliver Sacks' last days, and their sort of love story. I think this has very little to do with the stuff on your podcast, Lenny, but it's just a book that I really love, and just makes you think about mortality, but also love and life and stuff like that. So that's one of my favorite books.
LR
My goal here is, we're trying to create Renaissance humans, so all of this other stuff is excellent.
JW
Cool.
LR
Interestingly, I saw Julie Zhuo, famed design leader, was reading The Power Broker recently. I don't know what's going on over here. Spreading in design land. Okay. Do you have a favorite recent movie or TV show that you really enjoyed?
JW
I watched A Sentimental Value recently. I watched it on a plane, which is how directors want you to watch their films, but it's a Norwegian film by the same director who did The Worst Person in the World. I think just the pacing, the writing, the relationship between the characters is just really subtle and beautiful. It's basically about this family, sort of a family drama, but also about this house that they lived in their entire lives. It's beautiful because the house is sort of a character. So I don't know what else to say about that, but that was a really good movie. And then I would also recommend obviously The Pitt, season two. We're on that. I think everybody just likes to watch people who are really competent at their jobs do something. So yeah.
LR
Imagine being an actor on that show, just like how much you have to learn and memorize all these terms.
JW
Yeah. Yeah. It just also seems really fast paced too. They do so much stuff in one shot, and there's just so much movement and stuff like that. It seems really, really hard to be an actor on.
LR
And I only recently realized Noah Wiley was in ER as a younger person, and now he's the head of this. Yeah. Yeah. Oh, man. Okay. Favorite product you recently discovered that you really love, cannot be Cowork?
JW
Not one that I've actually discovered recently, but Retro, I've been using it for basically two years now since it came out. And I think I discovered new benefits of it recently. So for folks who don't know about Retro, it's sort of this small community photo sharing app, in which you can only share photos from each week, from a given week as opposed to all time. And it basically has none of the social media stuff. You can't really see counts, there's no ads, et cetera. But one really nice thing is now that I've been using it for two years, I can now look back at each year and see, oh yeah, this week, two years ago, I was doing this, and it's become this really special way to live through each week of my life, basically.
LR
Wow. And it's also a beautifully crafted app if you're looking for building your own taste in design.
JW
Yeah. Designers love Retro.
LR
I could see that. Okay. Do you have a favorite life motto that you often come back to in work or in life?
JW
Yeah. Not sure if it's my favorite life motto, but one thing I do catch myself saying a lot is just, "It is what it is."
LR
My dad says that all the time. I love it.
JW
Yeah.
LR
Yeah.
JW
It sounds super defeatist, but I promise it's not. I think just given how much stuff is going on in the world right now, and especially with the industry and whatnot, you can't control everything. And so sometimes it is what it is, just brings the levity you need to move forward.
LR
I did a 10-day meditation retreat a while ago, and I came back from it and it's like, "Dad, you've been right all along. This is the answer to it all. It is what it is." Don't cling, don't try to change. Just it is what it is.
JW
It is what it is.
LR
There's so much depth then. I was like, okay, smarter than I even thought. Okay. Final question. Coming back to cowork, is there, I don't know, mind-blowing use case, something just like, wow, that's so cool that cowork can do that. Either something you use it for, or you've heard somebody using it for?
JW
One thing I really like is just introspection. And so I have this folder, basically, of local notes that I have that I use IA writer for. And I basically just write whatever, and over the years have collected it with a bunch of different notes, and they span all different things like one-on-one notes, random thoughts, tiny memos, interview notes, et cetera. And my favorite, it's cool to me, is just using cowork to analyze that and have insights out of it, and actually create things out of it. So anytime I can learn something new about myself, I love that. But I think a very practical thing I did with it the other day was along the lines of hiring, I was like, "Oh yeah, I want to sort of articulate what it is that I look for when I look for in design craft," because I think actually a lot of people struggle to articulate that. And I just had it read through all of my notes, both interview notes and other things that I cared about, and memos and stuff like that I've written in the past. And then it made me this rubric for evaluating that. So that kind of introspection where it's like, oh, I wouldn't have realized even these things about myself that I'd been putting out there implicitly. That's been really cool for me.
LR
That is so cool. So just to make this very clear for people, you have a folder with all these things you've written, one-on-ones, meetings, like you could do, I don't know if y'all are allowed to use Granola or something like that, meeting transcripts, and ask it. And I was going to ask what prompt you used, but it's just like, "Read all these things I've written, and help me probably just understand how I feel about what is DesignCraft."
JW
Yeah, basically I think it was like, "Hey, I have a bunch of interview notes and a bunch of notes related to DesignCraft in this folder. Read it and then help me craft a memo/rubric for how I assess craft in interviews."
LR
So cool.
JW
Yeah.
LR
Jenny, this was awesome. What a time to be alive.
JW
What a time.
LR
Two final questions. Where can folks find you online if they want to reach out and how can listeners be useful to you?
JW
Yeah, I'm on Twitter/X is what we're calling it these days. It's \@jenny\_1. That's probably the best place, not really on LinkedIn as much, so that's the best place. And how can folks be helpful to me? Send us your product feedback. We're working on Cowork, or anything Claude related really, just send us your feedback, we'd love to make it better for you.
LR
Jenny, thank you so much for being here.
JW
Yeah, of course. It was great chatting Lenny.
LR
It was wonderful. Jenny, thank you. Bye everyone. Thank you so much for listening. If you found this valuable, you can subscribe to the show on Apple Podcasts, Spotify, or your favorite podcast app. Also, please consider giving us a rating or leaving a review, as that really helps other listeners find the podcast. You can find all past episodes or learn more about the show at lennyspodcast.com. See you in the next episode.
✦ ✦ ✦
Full Interview Transcript · Lenny's Podcast · Mar 1, 2026
「那套設計師奉為圭臬的流程——已經死了。」
"This design process that we as designers have been taught, we treat it as gospel — that's basically dead."
— Jenny Wen, Head of Design, Claude @ Anthropic
第一章Chapter 1

工程變了,設計被迫跟著變 Engineering changed. Design had no choice.

「不是設計師主動想改,是工程工具逼著大家改。」 "It's not because designers chose to change — it's the follow-on effect of engineering tooling changing."

「工程師現在可以同時跑七個 Claude,隨時把任何想法直接做出來。設計師再也沒有時間慢慢做那些漂亮的 mock 了。」

"Engineers can spin off their seven Claudes, make whatever features they want. You as a designer actually do not have the time to make these beautiful mocks anymore."

— Jenny Wen

傳統設計流程(已死) The Old Process (Dead)

  • 發散 → 收斂 → 再發散:奉為聖典的雙鑽石流程
  • Diverge → Converge → Diverge: the double-diamond treated as gospel
  • 2–10 年願景 deck:精美簡報指向遠方
  • 2–10 year vision decks: beautiful story-told presentations pointing to a far-off future
  • 60–70% 時間做 mock 和 prototype
  • 60–70% of time spent mocking and prototyping
  • 設計師是「守門員」,先設計才交給工程
  • Designer as gatekeeper: design first, then hand off to engineering

現在的設計工作 Design Work Today

  • 3–6 個月短期願景,以可跑的 prototype 呈現
  • 3–6 month vision horizon, expressed as a runnable prototype
  • 30–40% 時間做 mock,其餘在配合執行
  • 30–40% of time on mocks; rest on pairing and real-time support
  • 與工程師即時協作,親自下 code 做最後潤飾
  • Real-time pairing with engineers; designer writes code for last-mile polish
  • 放手讓工程師衝,設計師提供方向與品質把關
  • Let engineers ship; designer provides direction and quality guidance
70% 過去花在 MockPreviously on Mock
35% 現在花在 MockNow on Mock
35% 現在花在與工程師 PairNow on Eng Pairing
3–6mo 願景視野縮短至此Vision Horizon
第二章Chapter 2

設計工作正在「層級分化」 Design Work Is Stratifying Into Two Types

「不是所有設計工作都消失,而是分裂成兩種截然不同的形態。」 "Design work is becoming really stratified in this new world."

類型一Type 1
支援執行Supporting Implementation
工程師用 Claude Code / v0 快速打出功能原型,設計師即時給意見、做最後一哩路潤飾。重點是:不擋、配合、快速反應。 Engineers use Claude Code / v0 to spin up features. Designer responds in real-time — give feedback, polish the last mile. The key: don't block, stay flexible, react fast.

Code ReviewLast-mile PolishDesign System
類型二Type 2
創造方向Creating Direction
在眾人各自衝刺的混亂中,設計師必須提出短期願景 prototype,讓整個團隊對齊方向,避免七個 AI 做出七個不兼容的東西。最難留出時間做,也最重要。 Amid everyone shipping independently, the designer must produce a short-term vision prototype to align the team and prevent seven AIs building seven incompatible things. Hardest to make time for — and most important.

Vision PrototypeDirection SettingTeam Alignment
第三章Chapter 3

在 AI 颱風眼工作是什麼感覺? What It's Like Working at the Center of the Storm

「我們的 Slack 是金礦——最好的 AI 新聞就在公司內部。」 "Our internal Slack is a goldmine. The best AI news is probably internal."

第四章Chapter 4

速度與品質,可以並存 Speed and Quality Can Coexist

「用速度建立信任。」 "Build trust through speed of iteration."

「你可以先把不完美的東西推出去,但你必須承諾:我們會修,我們在聽。讓使用者知道我們回應了他們的反饋,而且是持續在改的。這才是維持品牌信任的方式。」

"You can put something out early, but the promise you have to make your users is: we're going to iterate, we're going to take your feedback. The way you really lose trust is if you release something early and then nothing ever happens."

— Jenny Wen
原則一Principle 1
標示「Research Preview」Label It "Research Preview"
誠實告訴用戶這是早期版本,設定正確預期,讓你得以在真實環境中學習。 Be honest that it's early. Set correct expectations and earn the right to learn from real usage.
原則二Principle 2
持續公開迭代Iterate Publicly
推出後不能靜默,必須公開修復進度,讓社群看見你真的在改——Twitter、Changelog、公告都是武器。 Silence after launch destroys trust. Show your fixes publicly — Twitter, changelogs, announcements are all instruments of trust-building.
原則三Principle 3
AI 模型不能只靠 MockAI Can't Be Fully Mocked
非確定性模型的行為無法在靜態 prototype 中呈現,必須讓真實用戶在真實環境中試用,才能發現真正的使用情境。 Non-deterministic behavior can't be captured in static prototypes. Real users in real contexts are the only way to discover actual use cases.
第五章Chapter 5

AI 世代,人還剩什麼? Where Will Human Designers Stay Valuable?

「品味會進步,但決策責任永遠是人的。」 "AI will get better at taste — but someone still has to be accountable for the decision."

「軟體最難的部分根本不是建造它。想想你工作中最難的時刻,大多是兩個人對同一個功能意見不合。AI 可以給建議,但解決不了人與人之間的那場爭論。到最後,永遠需要一個人站出來說:這個,我們要做。」

"A lot of the hard parts of building software are not building it. Think about the hardest times you've had at work — it's probably two people disagreeing about a feature. AI can weigh in, but it can't necessarily solve that dispute. Someone still needs to be accountable for the decision."

— Jenny Wen
第六章Chapter 6

Jenny 現在招什麼樣的設計師? Three Designer Archetypes Jenny Hires For Now

「這個時代最難找、也最珍貴的三種人。」 "These archetypes were always interesting to me — but now they feel especially valuable."

方塊型通才The Block Shape
Strong Generalist · Block-Shape
在多個領域都達到 80 分以上的人。不是樣樣通樣樣鬆,而是技能側寫近似一個方塊。當設計師的角色持續向 PM、向工程延伸,這種人能最快適應。非常稀有,非常珍貴。 80th percentile good across multiple skill buckets. Not "okay at everything" — nearly block-shaped. As the design role keeps stretching toward PM and engineering, this person adapts fastest. Very rare. Very valuable.
最難找Rarest
T
深度 T 型專家The Deep T
Deep Specialist · Deep T-Shape
T 字的那一豎比所有人都深——業界前 10% 的單一技能。在「人人都能做好設計」的時代,極度深入的專業能力是最難被取代的差異化。 The stem of the T goes deeper than anyone else's — top 10% in a single skill. In an era when "anyone can design," radical depth is the hardest thing to replicate.
最有護城河Strongest Moat
閃耀應屆生The Craft New Grad
Craft New Grad
早期職涯,但展現出超齡的判斷力與謙遜的學習渴望。沒有被舊流程制約的包袱,對新工具、新工作方式完全開放。最被低估的族群。 Early career, but wise beyond their years — and genuinely humble. Unburdened by old processes, completely open to new tools and methods. The most overlooked archetype right now.
最被低估Most Underrated
第七章Chapter 7

Chat 介面會消失嗎? Will Chat Interfaces Disappear?

「對話,是跨越所有智識層次最美的互動方式。」 "Talking scales beautifully across every level of intelligence."

  • Chat 不會消失:它開啟了無限彈性的互動方式,是過去 baked-in UI 給不了的自由度
  • Chat isn't going away: it opened up infinite flexibility that baked-in UIs simply can't provide
  • UI 元件會增加:天氣、股票、互動 widget——特定場景 UI 比打字更快
  • UI elements will grow: weather, stocks, interactive widgets — specific UIs are faster than typing for certain tasks
  • UI 將更多由模型動態生成,而非全部手工撰寫
  • UIs will increasingly be generated by models rather than all hand-coded

「說話,是一種能跨越所有智識層次的互動方式。你可以跟 IQ 300 的人說話,跟 IQ 80 的人說話,形式都是說話。這就是為什麼對話界面能隨著模型能力無限擴展。」

"Talking is such a beautiful way to handle every level of intelligence. You can talk to people who are very smart and not so smart — and it's still just talking. It scales so well across the spectrum. That's why it's been such a natural fit for growing model intelligence."

— Lenny Rachitsky, citing Kevin Weil
第八章Chapter 8

設計管理者,現在該怎麼辦? What Should Design Managers Do Right Now?

「回去當 IC 的那一年,是我職涯最值得的一年。」 "Going back to IC gave me things I could not have gotten from just managing."

Jenny 的選擇Jenny's Choice
從 Figma 總監,回到 ICFrom Figma Director Back to IC
從管理 12–15 人的設計總監,Jenny 主動選擇回到 Anthropic 當個人貢獻者。理由是:在 AI 轉型最劇烈的時刻,待在離工作最近的地方,學到了任何管理角色都給不了的東西。 From directing a team of 12–15 at Figma, Jenny chose to return to IC work at Anthropic. The reason: at the most disruptive moment in AI, being closest to the actual work gave her skills no management role could have provided.
未來的好主管The Future Manager
方向給予者,而非人事管理者Direction-Giver, Not Just People Manager
純粹的「關心、one-on-one、讓你感覺良好」的人事管理角色正在萎縮。未來好主管的定義是:既能給團隊清晰方向、又能真正投入工作本身的人。 Pure people management — the "care about you, 1:1s, feel good at work" role — is shrinking. The future manager is someone who can give clear direction AND genuinely engage with the work itself.
第九章Chapter 9

Claude Cowork 是怎麼做出來的? How Was Claude Cowork Actually Built?

「10 天只是最後衝刺,前面那段沒有人算過。」 "The 10 days was just the sprint to go from internal to external. The journey was much longer."

「我最喜歡的 Cowork 使用情境,是把一個裝滿隨機檔案的資料夾丟給它,然後它能從那堆垃圾裡提煉出有用的東西。Claude with hands——Claude 有了手。」

"One of my favorite Cowork use cases is just giving it a folder of my stuff — it doesn't really matter what's in that folder — and being able to extract something good out of it. Claude with hands. Claude has hands now."

— Jenny Wen
第十章Chapter 10

設計師要像 VC 一樣看世界 Designers Should Think Like VCs

「看那些你不懂但感覺有能量的東西——那裡往往藏著金礦。」 "Dive deeper into what that energy is around something you don't quite understand yet."

「Legibility Matrix:創辦人和想法都一目瞭然?那機會太明顯了,別人早就做了。真正值得投注的,是那些本身模糊、說不清楚——但有能量圍繞著的想法。」

"If both the founder and the idea are super legible, the idea is probably not that novel. Where it gets really interesting is where the idea itself is illegible — on the frontier, people might not quite get it yet. That's where the value often is."

— Evan Tana (SPC), referenced by Jenny Wen
框架說明The Framework
Legibility Matrix
一個 2x2:橫軸是「想法是否清晰」,縱軸是「創辦人是否清晰」。兩者都清晰 = 太顯而易見。想法模糊但有能量圍繞 = 這裡有寶。設計師的工作是發現它、詮釋它,讓它變得可被理解和執行。 A 2×2: legibility of founder vs. legibility of idea. Both legible = already being done. Idea illegible but energy present = treasure. The designer's job is to find it, interpret it, and make it understandable enough to execute.
對設計師的啟示Implication for Designers
成為內部的「機會偵測器」Become the Internal Opportunity Detector
Jenny 在公司 Slack 裡刷到各種早期 prototype 時,她在做的事正是這個:找那些「我不太懂,但感覺有能量」的東西。設計師最獨特的價值,是把一個 illegible idea 透過 UX 和 storytelling 轉化成可以被組織理解、執行的方向。 When Jenny scrolls through early prototypes in Slack, this is exactly what she's doing: finding things she doesn't fully understand but that have energy around them. The designer's most unique value is translating an illegible idea — through UX and storytelling — into a direction the whole organization can act on.
✦ ✦ ✦
外部視角章節 Outsider Perspective
Lenny 的追問:局外人的好奇心 Lenny's Questions: What Outsiders Actually Want to Know
Lenny Rachitsky 不是設計師。他是 PM 出身的創業者、podcast 主持人,代表的是所有在設計圈外、但每天需要和設計師協作的人。他的追問不預設立場、不保護行業自尊——往往比圈內人更犀利,也更接近真正的焦慮所在。 Lenny Rachitsky isn't a designer. He's a PM-turned-founder-turned-podcaster who represents everyone who works alongside designers without being one. His follow-up questions carry no industry self-preservation instinct — they're often sharper than what insiders ask, and closer to where the real anxiety lives.
LR
Lenny Rachitsky
這個轉變,只發生在 Anthropic 這種 AI 前沿公司?還是所有公司的設計師都要面對?
"How true is this at all companies? Someone may be hearing this thinking — this isn't going to happen at Salesforce, at ServiceNow. Is this where all teams are heading, or mostly AI-first companies?"
JW
Jenny Wen
我在柏林做的那場演講是我做過共鳴最強的一場——人們普遍感受到「舊的設計流程走不下去了」。但同時也有很大的反彈,因為有人把整個職涯都押注在那套流程上了。
"The talk I gave last year was the most resonant talk I've ever done — people are feeling this across the industry. But there was also a decent amount of backlash. People have invested their entire careers in learning and teaching this stable design process."
外部視角Outside View 設計師在害怕的,不只是 AI 工具取代自己——還有整套被視為專業核心的「設計流程」正在失去公信力。這對設計師的身份認同是雙重衝擊。Lenny 的問題讓 Jenny 不得不承認:這個焦慮不只在前沿公司,而是整個行業都在感受的。 What designers fear isn't just AI tools replacing them — it's that the design process that defined their professional identity is losing credibility. A double hit to the self. Lenny's question forced Jenny to admit: this anxiety isn't only at frontier companies. It's industry-wide.
LR
Lenny Rachitsky
快速出貨、邊做邊學,真的會做出比傳統「發現 → 設計 → 驗證」流程更好的產品嗎?
"Is it your sense that shipping fast and learning in the wild — not only is that just what everyone's doing — but that actually leads to a better product at this point?"
JW
Jenny Wen
是的,尤其對 AI 產品而言。你沒辦法 mock 出所有的狀態——你必須讓真實用戶在真實環境中試用,才能發現真正的使用情境。因為 AI 模型是非確定性的,你根本無法預測使用者會用它做什麼。
"The ability to try something with real data and real users does lead to a better product. Especially with AI — you can't mock up all the states. You have to use the actual models and see people try it with their actual use cases, because you discover use cases as you see people using them."
外部視角Outside View 這是一個關鍵洞察:AI 產品不只是「功能複雜」,而是本質上不確定。傳統 UX 研究是在「已知的問題空間」裡找答案,AI 產品的問題空間本身就是未知的。Lenny 用一個商業問題逼出了設計方法論的根本轉變。 Key insight: AI products aren't just "complex features" — they're fundamentally non-deterministic. Traditional UX research finds answers in a known problem space. AI products have an unknown problem space. Lenny's simple business question surfaced a fundamental shift in design methodology.
LR
Lenny Rachitsky
AI 最終會在「品味」和「判斷力」上超越人類設計師嗎?看 AI 征服程式碼的速度,我開始重新思考所有我對 AI 能力的假設。
"Just watching how quickly AI took over coding — a year ago people were saying 'I don't think AI will get this good.' Now the best engineers aren't even looking at the code. It made me reevaluate all my assumptions about what AI can't do in design."
JW
Jenny Wen
我認為 AI 的品味和判斷力會持續進步,我們可能太執著於認為這些是人類專利了。但最終,還是需要有人決定「這個,我們要做」,以及「這個,我來負責」。就像放射科的例子——不是因為 AI 不夠準確,而是需要有人簽名負責。
"It will get better at taste and judgment — we might be holding onto that too much. But at the end of the day, someone has to decide what's actually going to get built and someone needs to be accountable. Like the radiology example — AI might be just as accurate, but someone still needs to be liable."
外部視角Outside View Lenny 的這個追問最有力——他從「AI 已征服寫程式」出發,逼 Jenny 正視「設計師可能也會被超越」。這是圈內人通常不願意直接說出口的問題。外部視角反而有打破禁忌的效果,而 Jenny 的回答出乎意料地誠實。 Lenny's most powerful push — starting from "AI already conquered coding" to force Jenny to honestly face "designers might be next." This is the question insiders usually avoid saying out loud. The outsider breaks the taboo. And Jenny's answer was more honest than expected.
LR
Lenny Rachitsky
設計主管是不是應該回去當 IC?這對已經做到管理職的人,是一個建議嗎?
"So the advice I'm hearing for designers who are managers: you almost need to move back to IC in order to understand what's happening — so you can be a better manager?"
JW
Jenny Wen
我認為是的。工程領域很久以前就這樣了——好的 EM 要先輪調 IC,真正理解技術是怎麼運作的,才去做全職的管理。設計也應該學這一套。如果我沒有在這一年親身在這個環境裡工作,我不知道是否能真正理解,或知道怎麼引導我的團隊。
"I think so. Engineering disciplines have been doing this for a long time — good EMs do a rotation in IC to really understand how the technology works before becoming full-time managers. Design probably needs to do something similar. If I had not worked in this environment myself this year, I don't know if I would've totally understood it or known how to guide my teams."
外部視角Outside View Lenny 的問題揭示了一個管理悖論:在變化最快的時代,最有效的領導方式可能是「暫時放棄領導位置」。這對習慣「升職就是升管理」的設計職涯觀是直接挑戰。而 Jenny 從 PM 出身的 Lenny 口中得到確認,反而讓這個建議更有說服力。 Lenny's question surfaces a management paradox: in the fastest-changing era, the most effective leadership may require stepping away from leadership temporarily to recalibrate. A direct challenge to the "promotion = management" career model in design. Getting that confirmed by a PM-turned-CEO makes the advice land harder.
LR
Lenny Rachitsky
你面試的設計師,現在需要會寫程式嗎?這是加分,還是基本要求?
"For designers who are mid-career or senior — do they actually need to get technical and learn to code? Or could someone be really successful staying non-technical?"
JW
Jenny Wen
不必變成真正的工程師,但設計師的詞彙表正在越來越包含「實作」這件事。我的建議是:先把 coding 工具放進你的工具箱——不一定要學 React,但要知道怎麼用這些工具。隨著模型越來越強,我們會繼續往更高層次的抽象移動。
"I think it definitely helps — but maybe not to build something from scratch. More and more of a designer's vocabulary right now includes implementation. I'd say: start bringing coding tools into your toolkit. As models get better, we'll continue moving up abstraction layers and you won't need to know how every single line of code works."
外部視角Outside View Lenny 問的是「要會寫程式嗎」,Jenny 的回答卻是「把工具先帶進來」——這個重新框架很重要。設計師的技術門檻不是「能不能寫程式」,而是「能不能用工具創造可跑的東西」。作為局外人,Lenny 把這個問題問得非常直接,讓 Jenny 沒有辦法給一個政治正確的模糊答案。 Lenny asked "can you code?" Jenny reframed to "get the tools in first" — an important distinction. The technical bar for designers isn't "can you write code?" but "can you use tools to create runnable things?" As an outsider, Lenny asked directly enough that Jenny couldn't give a politically careful vague answer.
結語Conclusion

給設計團隊的五個行動啟示 5 Action Takeaways for Design Teams

從這場對話轉化為可立即執行的方向 Translated from this conversation into immediately actionable directions

01
角色重塑Role Shift
從守門員到嚮導From Gatekeeper to Guide
不再把持交付物,而是讓工程師衝,設計師走在旁邊提供方向、原則和品質反饋。學會說「先做再說,我幫你修」。 Stop owning deliverables as the prerequisite. Let engineers ship, walk alongside them with direction, principles, and quality feedback. Learn to say "go ahead, I'll help you fix it."
02
技能擴展Skill Expansion
把 Code 加入工具組Add Code to Your Toolkit
不必變成工程師,但要能用 Claude Code 打 prototype、改 CSS、做最後潤飾。這是設計師現在最快產生影響力的技能槓桿。 You don't need to become an engineer. But be able to use Claude Code to prototype, edit CSS, do last-mile polish. This is the fastest leverage point for design impact right now.
03
願景輸出Vision Output
用 Prototype 取代簡報Replace Decks with Prototypes
3–6 個月的短期方向,不用做精美 deck,做一個可以跑的 prototype,遠比 100 頁投影片有說服力。 For 3–6 month direction, skip the polished deck. Build a runnable prototype that people can feel. Far more persuasive than 100 slides.
04
品質策略Quality Strategy
用速度建立信任Build Trust Through Speed
不完美的東西可以先出去,但必須快速回應反饋、公開持續改進。讓用戶看見你在聽、在動,這才是長期品牌信任的基礎。 Ship imperfect things, but respond fast and improve visibly. Let users see that you heard them and you're moving. That's the foundation of long-term brand trust.
05
職涯選擇Career Choice
主動讓自己變得「Illegible」Make Yourself "Illegible"
在你的領域深到前 10%,或橫向擴展到接近方塊形——成為那種讓人說「這個人做的東西我說不清楚為什麼好,但就是不一樣」的設計師。 Go deep to the top 10% in your domain, or expand horizontally to near-block-shape — become the kind of designer people say "I can't explain why their work is great, it just is."
參考原始資料Source Material
收聽 YouTube PodcastWatch on YouTube 閱讀完整 TranscriptRead Full Transcript