Descriptions:
Claire Valle, product leader and creator of the How I AI channel, offers a hands-on review of two recently released AI design tools: Anthropic’s Claude Design and OpenAI’s GPT Images 2.0 model. Valle tests Claude Design by importing a real-world design system—specifically Lenny’s Newsletter—saving the homepage HTML, logo, and key assets, then feeding them into Claude Design’s design system builder to see how faithfully it can replicate an existing brand when generating prototypes and landing pages.
The strongest differentiator she identifies for Claude Design is that it treats design systems as a first-class citizen, which addresses the top complaint she hears about AI prototyping tools like Lovable, V0, and Bolt: generated output never matches the actual brand. Claude Design’s approach shows genuine promise on this dimension. The central limitation, however, is speed: individual generation cycles take 5 to 10 minutes, which fundamentally undermines the tight feedback loops that make Figma so effective for iterative design work. Valle argues this is a structural constraint of LLM-in-the-loop tools, not merely a performance issue, and that Figma will continue to win for rapid iteration for this reason.
The episode also covers GPT Images 2.0 use cases and includes a broader discussion of where each tool fits in a real product development workflow. Valle’s framing is explicitly practical—aimed at product managers, designers, and founders deciding whether these generative design tools are ready to replace or complement existing parts of their pipeline.
📺 Source: How I AI · Published April 22, 2026
🏷️ Format: Review
