Soleio interviews Joel Lewenstein, Head of Design at Anthropic. Their roadmap comes from internal traction on half-baked vibe-coded prototypes rather than top-down planning, and new designers figure that out by pattern matching on who gets respect. Claude Design is a wrapper over Claude Code with a friendlier front end, and his own designers keep going back to the terminal, because the production codebase is the source of truth for the design system and the rules of how things work while Claude Design has to rummage through Google docs that aren’t.

ChatGPT now has Sketch (type “@Sketch”, draw a rough layout, get it rendered), templates for formats like posters and merch, comments pinned directly onto an image for focused edits, and shareable prompts so someone else can run your idea on their own photos. That’s a design tool’s vocabulary showing up inside a chat. Generation latency is reduced by up to 50% compared with Images 2.0, it’s live on every ChatGPT tier including free, and developers get two new API models: Flare as the default, Sunburst for precision work with longer generation times.

Adam Wathan and Steve Schoger are taking the Tailwind Labs to Shopify. Tailwind CSS stays MIT-licensed and the same people keep leading it, but the business around it ends – both Tailwind Plus and ui​.sh are closed to new signups.

The framework is installed over 110 million times a week, but in January Adam had to lay off three of his four engineers with revenue down close to 80%. As AI agents read the docs developers no longer do, and the docs were selling the templates that funded the framework. Building a business around the open-source framework is incredibly hard, so I’m glad Shopify is ready to foot the bill.

Marek Minor, founder of Amsterdam studio Minor Adventures, spent about a year drawing Cursor’s 600+ icons in two sizes and two styles, every exploration by hand. The rule I keep thinking about: ​“Cursor’s pointer runs from bottom left to top right, and so does every icon that could go either way.” Slashes run the opposite way, because a slash cancels a direction and should cut against it.

Everything above is agents arriving in Figma, but this one runs the other way. Nate Parrott, the Anthropic designer behind Claude Design, put its artboard canvas inside Claude Code, so options get drawn where the code already lives. One brief, several editable artboards, and then Claude implements the one you pick. /design-sync can load your design system from a repo or your local codebase first. 

“The failure is not in the components; it is in the composition.” Matt Rothenberg, Principal Engineer at Cloudflare, handed three isolated agents the same shadcn/​ui library, the same tokens, the same app shell, and three ordinary settings tickets. Back came tabs, cards, and a long form, all three of which a token linter would happily pass. He then gave the next three runs a named vocabulary and a short page of rules, and they finally read as one product. The agents didn’t cause the drift, they just made it impossible to keep pretending it wasn’t there.

Meng To sits down with Ridd at Dive Club for a 35-minute screen-share walkthrough of his current Codex workflow. He’s no longer opening Figma to create anything new — all design exploration, prototyping, and iteration happens in Codex, starting from screenshots and referencing his own past work as context. A few specific tactics stand out: using GPT Image 2 to generate UI concepts that then get turned into HTML, feeding a video of an animation to the AI and asking it to describe the right prompt to recreate it, and using design​.md files as a portable style guide that encodes years of taste into something the model can actually read.

GPT‑5.6 is now available in Figma Make alongside the existing model options. Select it from the model picker in the chatbox. Available on all plans.

Ryo Lu, Head of Design at Cursor, gave the standout talk at Cursor’s first conference, Compile. The title is ​“Closer to the Material,” and the core argument lands hard: as AI makes execution cheaper, the real risk is that humans become approvers rather than authors — people who accept or reject outputs without ever being inside the decision. He distinguishes between output (which ends the loop) and material (which invites you back in), and argues that the future he wants is tools that keep people close enough to the work to still have judgment. A sharp and honest take on the AI-design moment, and easy to connect to what Figma is navigating with its own agent features right now.

Meaghan Choi, lead designer on Claude Code, walks through her actual daily setup at a Dive Club talk in NYC: always in a worktree so parallel Claudes don’t overwrite each other, always in auto mode so the classifier handles permission prompts, and a custom /prototype skill that spins up five HTML variants of any feature and picks one with reasoning before she even looks. She demos it by asking Claude to add autocomplete to Excalidraw with no design spec, then has it open a PR with a screenshot recorded via Claude in Chrome. Her line that lands: she doesn’t review Claude’s terminal output anymore, she reviews the PR.

Stephen Haney announces QuiverAI’s SVG generation inside Paper – vector output from text or references, aimed at logos, illustrations, and animations. QuiverAI​’s models have been on my radar since Dann Petty joined the team as a Founding Product Designer.

Hannah Hearth runs a tooling Show and Tell with her team at Vercel and writes up the results: Codex + Claude pair programming, Conductor for parallel agent threads, UI Fork for in-browser variant exploration, and Cleanshot’s Pin tool still earning its place.

Kris Puckett, Design Manager at Stripe, spent months building Epilogue, a real iOS app with 14,000 lines of Swift, entirely through conversation with Claude. This essay is a specific and honest account of what the designer-building-with-AI experience actually looks like: what broke, what he learned about asking precise questions, what ​“vague frustration keeps you stuck, specific confusion gets you answers” actually looks like in practice. ​“I realized the bottleneck was never coding ability. It was articulation. The ability to describe what I wanted clearly enough that something else could build it.”

“Designers have always (and will always) answer the question ​‘What’s worth making?’ ” Joel Lewenstein, Head of Design at Anthropic, argues that as the cost of software drops, the most important decisions shift from ​“can we make this?” to ​“should we?” Design, in his framing, is what narrows the possibility space fast enough to keep up with the speed of delivery. He describes Claude Design as a tool for getting ideas ​“good enough to move discussion forward,” cutting idea-to-internal-feedback time from days to hours.

“A state-of-the-art image model that can take on complex visual tasks and produce precise, immediately usable visuals, with sharper editing, richer layouts, and thinking-level intelligence.” OpenAI’s second-generation image model promises a step change in instruction following, precise object placement, dense text rendering, and cross-aspect-ratio generation.

Soleio (designer, angel investor) draws a three-way taxonomy to cut through the noise: production design gets automated on economic grounds; strategic design expands its medium set to include live production experiments iterated by agents; and exploratory design actually grows in importance because humans and machines together can traverse a far larger combinatorial space of novel forms. The real competitive edge, he argues, shifts to ​“problem-finders” who build new systems for channeling desire: ​“The inputs and systems we create to find opportunities will reward the most intrepid problem-finders. Design stops being a method of sitting and ruminating on possible forms or solution spaces. Design becomes active, research-based, and built around speed of discovery and expression.”

Ben Blumenrose offers a useful counterpoint to the critical takes: ​“What if it’s not a Claude Design vs. Canva vs. Figma vs. xxx but a ​‘yes and’ as millions more people are exposed to design and use this accessible tool to begin a journey into being better at type, color, composition, effective communication, building great experiences, etc?”

Sam Henri Gold argues that Figma’s decade of components, variables, and tokens has become an overgrown, pre‑agentic system that no one would design from scratch now that LLMs work directly with code. He frames Claude Design as ​“truth to materials”: HTML/JS-first, tightly coupled to Claude Code, and better aligned with a future where the source of truth moves back into repositories instead of opaque design files. The interesting provocation is his prediction that tooling will split in two—agentic, code-native tools like Claude Design on one side and pure, system-free exploration playgrounds on the other—leaving Figma facing its own ​“Sketch moment” unless it radically rethinks what it’s for.

Paul Stamatiou with the first impression most experienced designers will relate to: Claude Design didn’t land on first try, but the trajectory outlined by Sam Henri Gold is unmistakable. Lately, I’ve been working in pretty much the same way: ​“So much of my iteration happens directly in the code, Figma is where the direction gets sorted out, then it gets a million iterations and refinements in code. The last thing a designer on a team moving at hyperspeed wants to do is backport designs in code back to Figma to maintain a single source of truth.”

Jacob Miller left the PM role at Figma to build Diffui, a new way to approach the design to dev process with AI. In this thread, he compares designs generated by his tool and Claude Design. One is more expressive and diverse, another is more polished and organized. The point is not that one is better than another, but how can we combine different approaches and tools to get to the most interesting result.