Keep pulling the thread on Felix Lee.
Figma recently launched a feature that converts code from a running application back into an editable Figma design canvas.
The majority of designers are not using advanced AI coding tools like Cloud Code or Cursor and are still primarily using Figma.
Cloud Code produces significantly higher quality and faster output compared to other coding agents like Gemini AI Studio and Cursor's Codex.
The visual recognition capability of Anthropic's 'Opus LIM API' is not very effective, particularly for identifying precise 'hotspot' locations on an image.
Google's Gemini Pro model is superior for visual recognition tasks, such as accurately placing hotspots on a webpage screenshot, compared to Anthropic's Opus model.
Based on data from over 55,000 monthly mentorship bookings on ADP List, designer conversations about advanced AI tools like Cloud Code and Cursor have only increased by 10-20% over the past six months.
Figma's new code-to-canvas feature generates editable vector components (SVGs) within Figma, not just static screenshots.
Felix Lee believes designers are currently over-reliant on Figma as both a company and a tool.
The daunting, code-centric interfaces of tools like Cloud Code and Cursor, which feature terminals and IDEs, present a high barrier to entry for many designers.
Felix Lee predicts that for designers who adopt AI coding tools, Figma's primary role will shift to collaboration and final review, as using it for initial design will become a bottleneck.
The Figma MCP automatically extracts and includes design assets like icons and images when converting a design to code, which is an advantage over using screenshots.
Cloud Code was able to generate a 3D globe asset for a web project based on a screenshot of the Shopify Globe.