Design Thursday #104
A weekly recap of everything you need to know about tools, events, guidelines and design in general.
Figma Updates
Custom skills in Make
Skills are markdown files that capture the conventions and workflows you want to reuse across prototypes. Instead of re-explaining your standards every time, you define them once and call them with a slash command. Use a skill like `/insert-sample-data` to drop in company-approved test data, or set up `/build-from-prd` paired with a Notion or Confluence connector to turn any PRD into a prototype that already meets your requirements. Right now, skills are personal. Sharing them across your team and org is coming soon.
The Figma agent is here
Figma's own agent is now rolling out in beta. It lives where design already happens, generates and remixes designs, handles repetitive work, and respects your design system out of the box. During the beta period it won't consume credits, so you can try our and experiment with it. You can sign up for early access here.
Sketch Edinburgh Beta
The Edinburgh beta is out with a some great new improvements. Multi-paste now works across multiple selected frames, groups, or layers in a single action, and paste behavior overall is a lot more predictable: pasting into a group centers the content, and pasting on the canvas places it near where you're working instead of somewhere off-screen. Two new gradient interpolation methods have been added: Perceptual (Oklab) for smooth, natural-looking transitions and Vibrant (Oklch) for punchier, more saturated results. Both are based on human color perception, unlike the classic RGB mode that can produce muddy midpoints. Copy Properties (⌥⌘C) is also expanded, now covering fills, borders, shadows, effects, size, position, and prototyping interactions.
Rive Updates
Two Rive editor got quite some updates:
- Semantics for Accessibility in Early Access, letting you add semantic meaning directly in Rive and have runtimes translate it into native accessibility trees for screen readers.
- Stateful Components: a new component type that exposes properties without needing view model instances, so you can quickly override values in the inspector for faster iteration.
- Bring-your-own Agent API keys (Anthropic now, OpenAI and Gemini coming soon)
- An experimental Flash-to-Rive converter that lets you drag and drop a Flash file to convert it. Wow the early 2000s called.
Webflow Updates
Use component props in custom code
Code Embeds can now reference component props, locale settings, and page settings directly in custom code, just like CMS fields. You can reference a Text prop in custom CSS to allow per-instance styling, or build a JSON-LD block that pulls from props to generate structured data. A new Code prop type is also available, letting you swap entire SVG blocks, third-party embeds, or other custom code across instances.
Single-page publishing
Admins can now restrict full-site publishing for users with page-specific access, so collaborators only publish what they're allowed to edit. The publish modal has also been updated to make shipping individual pages to production faster.
Jitter Updates
Jitter AI
Jitter AI lets you describe something in plain language and it builds it for you inside the interface. Ask for a fluted glass background, a glitch transition, a ripple effect, or a 3D card flip, and Jitter generates it as a reusable, editable tool. The tools you create can be shared with your team, so everyone can produce consistent animated assets without starting from scratch each time.
Glass effect
A glass effect is now available as both a static design effect and an animation. You can dial in refraction, depth, dispersion, frost, and lighting to get anything from subtle frosted UI to rich refractive glass, then animate each of those properties for fluid transitions.
LottieFiles for VS Code 2.0
LottieFiles rebuilt their VS Code extension from scratch. The new version includes a dotLottie player with scrubbing, speed, and live reload controls, inline state machine debugging, an explore panel with 100,000+ animations, a workspace for team assets and CDN hosting, and a Composer for bundling animations, themes, and state machines. It also adds Motion Tokens with live preview, code snippets for React, Vue, Angular, Svelte, and vanilla JS, TypeScript type generation, and CI/CD integrations for GitHub Actions.
Google Flow: Tools
Google Flow added a Tools feature that lets you build custom creative workflows from scratch. You can browse a gallery of premade Tools built by other people, remix them to fit your project, or describe one from scratch and let Flow generate it.
Google Stitch Updates
A big batch of updates dropped for Google Stitch at I/O. The most notable is streaming: Stitch now shows its work in real time on the canvas so you can steer it before the final screen is done. In-place AI edits let you ask for targeted changes without regenerating the whole screen. The canvas now renders native HTML by default, with support for scripts, SVGs, and shaders, so animations and hover states are immediately interactive. On the import/export side there is now support for .fig files, one-click export to Netlify, Lovable, and Bolt, and new MCP skills to import screens from your codebase and sync changes back.
Recraft V4.1
Recraft updated their image generation model to V4.1. The main improvements are in photorealism, illustration, and short-prompt accuracy. Backgrounds are less randomly busy, group shots look more candid, and vector and typography outputs are cleaner overall. There's also a new Utility variant that focuses on flat lighting and front-facing composition, which is the better pick for mockups and product shots over the standard model.
AI in Design Report 2026
A new state-of-the-industry report is out, based on responses from over 900 designers at startups, enterprises, and agencies, plus 20+ leadership interviews. It covers how AI is changing tools, workflows, and team structures in tech design. Case studies from companies like Anthropic, Stripe, Notion, Shopify, Linear, and Framer are coming throughout the year.
Mobbin MCP
Mobbin launched an MCP server that connects your AI agents to 600,000+ real product screens. The idea is similar to Refero MCP: give your agent real design references before it starts building so the output is grounded in how actual products solve real problems instead of generating something that looks generic.
Dive Club: Tommy Geoco on the state of the design industry
Tommy Geoco spent the last few months visiting top design teams at companies like Vercel, Perplexity, Metalab, and Ramp, and this episode is a deep dive into what he found. Some highlights: why 59% of designers are building their own tools, how design workflows are starting to resemble creative director roles, what signals hiring teams look for to gauge AI fluency, and why internal tool building is becoming a real career path. Worth watching if you want a grounded picture of where the industry is right now.
Dive Club: Katarina Batina on making big bets with design
Design Director Katarina Batina from Shopify talks through how the best designers shape product strategy. The conversation covers redesigning the shopping cart UX, building the Shop super feed, how product strategy has shifted with AI, and why it matters to know when to stop adding to an app. A good one for anyone thinking about design leadership.