
Anatomy of AI Input
Layer-by-layer breakdown of modern AI chat inputs: field behavior, context bar, tool triggers, attachments, assist suggestions, and send/stop states that stay usable while streaming.
UI 靈感、工具、文章與 Figma 資源——為工作室設計師整理、索引、隨取隨用。
每週二精選推送 · 絕不濫發 · 隨時退訂
248 designers subscribed across the beta — resources, articles, tools & more.
從全球設計社群手挑選的工具、文章與設計靈感。

Layer-by-layer breakdown of modern AI chat inputs: field behavior, context bar, tool triggers, attachments, assist suggestions, and send/stop states that stay usable while streaming.

Agent skill for naïve folk-craft narrative illustration: multi-panel story scenes with flat color, decorative borders, and hand-drawn character storytelling (XXD Panel 100).

Agent skill that turns datasets into editorial HTML charts and reports: Lupi for slow-reading data stories, Glance for ranking in seconds, plus 12 bilingual full-page templates.

One-ink and two-ink editorial image skill for posters, zines, and packaging: halftone photography, visible paper, and restrained type. Not a monochrome filter.

Codex skills that bind everyday photos into paper zine pages, either keeping the photograph as an anchor or distilling it into a new illustrated artwork.

Codex skill that turns a theme, photo, or sentence into a sparse paper-texture zine poster: lots of negative space, one small subject, one loud color accent.

Codex skill that keeps the original photo and adds a lower abstract memory panel plus a short poetic English title. Not a filter or style transfer.

Codex skill for 3:4 bitmap editorial posters: fine dot-matrix subjects, quiet type, and print texture. Not retro game pixel art.

Gemini prompt that animates a still papercraft city — tiny vehicles, water, clouds — without leaving the paper-art look.

Threads post by @ai_traveler_k with a landmark-led geometric travel poster prompt, shown on Seoul, Hong Kong, and Shanghai.

Beautifully designed, copy-paste UI components for AI agent conversations: thinking states, tool calls, streaming text, citations, tables and more. Ships in React, Vue and Svelte.

A collection of animated thinking orb components for React, including states like 'Working...', 'Searching...', 'Thinking...', and 'Listening...'. These orbs provide visual feedback during loading sta

Ultramock is a tool that turns product screens into premium visuals with professional mockups. It offers various templates and effects to enhance product presentations.

A comprehensive checklist resource for designers with Figma plugin integration

The words designers use when they know what they're looking at. Covers vocabulary across typography, color, iconography, layout, interaction, motion, accessibility, IA, copywriting, tools, analysis, a

A glossary of 91 common animation patterns. Use these names to describe what you want when prompting an AI — and tweak each one live to feel how it moves.

Fluid text transitions powered by Framer Motion. A React component library for animating typography with style.

Stunning 3D mockup generator with 30+ battle-tested devices. Create studio-grade product shots in minutes with no 3D experience needed.

Figma AI-powered design workflows platform that combines multiple AI models (Google, OpenAI, Kling, Runway, etc.) with professional design tools. Build custom AI workflows and apps.

AI-powered vector graphics generation tool that generates, edits, and animates SVGs from text or images using the Arrow model. Raised $8.3M seed from a16z.
色彩、工具、流程與觀點——為工作室設計師整理的長文與指南。
上週寫完 Airtable 交割、Astra 出現在哪些選擇器裡,這週第一個後果不是模型,是所有權。9 月 10 日,Bending Spoons 簽下 Miro:企業價值 $1.355B$1.79B,預計第四季交割。同一家公司,兩個星期內收走資料庫跟白板。 同一週 OpenAI 讓你在 ChatGPT 裡畫一筆當參考圖;Adobe 讓 Premiere 的時間軸自己生影片、音效、配樂;OpenAI 工程部落格則叫你把去年堆起來的 skill 跟 `AGENTS.md` 砍掉一截。Jira 開始掃描待辦、自己派工、自己開 PR。 設計師這週該拿走的不是「又一個生圖模型」,而是地圖更新:工具可以換老闆,工作流也可以在你沒按送出的時候自己開始。 你還能設計的,是可帶走的白板、短到不會互相打架的指令,以及一條「什麼時候人必須按合併」的閘門。 <!-- weekly-toc --> 9 月 10
Apple 第一支摺疊 iPhone,要的不是為每種硬體狀態另做一套「摺疊模式」,而是讓同一個 App 在姿態與兩塊螢幕之間自適應。尺寸、長寬比、哪一面在作用都在變時,設計目標仍是:體驗像同一個產品。 先看官方材料:iPhone Duo for developers、HIG:Designing for iPhone Duo,以及技術講座 Design for iPhone Duo、Raise the bar、Multiple displays and scenes。 把 Duo 當成同一產品的多種姿態。優先用系統的 size class(compactregular)、layout margins 與 safe area insets,不要為每個姿態各做一套自訂版面。若某個姿態必須特製,也要維持階層與導覽語言一致,讓它仍像同一個 App。 控件往外緣靠,好保住垂直內容高度。裝置摺起時,內容
上週寫完那句「Astra 從一開始就不會出現在 Cursor 裡」,這週它真的出了。9 月 3 日 OpenAI 把 GPT-6 Astra 送進 ChatGPT、API、Azure、Bedrock;隔天 GitHub Copilot 跟 Vercel AI Gateway 也接上。官方可用性清單裡,沒有 Cursor。 同一週 Anthropic 把 Fable 5.1 的 cache read 砍了 75%,長跑 agent 終於比較負擔得起;Figma 讓 shader 跟 generative plugin 可以發布、下載程式碼、交給 MCP;Cursor 則把 agent 的電腦操作搬進你自己的機器。Google 把 Pics 塞進 Docs 跟 Slides。 設計師這週該拿走的不是「又一個更聰明的模型」,而是地圖更新:模型可以出、可以貴、可以便宜,但出現在哪一個選擇器裡,仍是