Claude just got another superpower...

Anthropic has launched Claude Design, a new platform powered by their Opus 4.7 model, which aims to generate prototypes, pitch decks, and production-ready UIs from Figma designs without requiring users to open design tools. This announcement has had significant repercussions, with Figma's stock dropping and AI design startups facing funding challenges, while UI designers are reportedly updating their job titles to prompt engineers. Claude Design is built on Opus 4.7, described as more tasteful and creative than its predecessor, capable of understanding images up to 2576 pixels. While Opus 4.7 shows improvements in programming benchmarks, some online users claim it's a regression from Opus 4.6, even suggesting a conspiracy of intentional nerfing. The platform's capabilities are demonstrated through interactive demos featuring working animations and the ability to generate variations. Claude Design can produce complex animations, including those involving shaders, and even full-length video animations. A notable feature is its ability to integrate with existing design systems by linking GitHub repositories or uploading Figma files directly, theoretically allowing designers to create a few screens and have Claude generate the rest. In a practical test, Claude Design was prompted to create an iOS onboarding flow for "Horse Tinder" using a PDF design system. While it produced five screens, it failed to adhere to the provided design system, resulting in a "Claudy" rather than a "fire shippy" aesthetic. The generated logo was also problematic. Attempts to fix the logo by drawing on the canvas and adding a prompt were unsuccessful, with the AI only managing to slightly alter the background color. This suggests that Opus 4.7 may still require further development for such tasks. The video also features a sponsorship from Google Cloud Run, a fully managed serverless platform for running code on Google's infrastructure. It supports various languages and frameworks, batch jobs, and hosting LLMs with GPUs, offering deployment from the terminal or Cloud Code. Google AI Studio is highlighted as a full-stack development environment for building and deploying applications to Cloud Run. Cloud Run scales to zero for idle projects, autoscales for traffic, and offers 2 million free requests per month.