I’m curious about that too. Claude Design feels (and from the blog post) sounds like the typical IC passion project that got just polished enough for a “beta” (in the sense of gauging customer interest, not commitment to release).
If someone with clout at Anthropic takes it up as a champion and pushes it into the portfolio of apps at the importance of Claude code, it could be an amazing product, eventually a threat to Adobe, Lovable, Canva market share. It needs a better direct editing layer like 2-way figma sync, so you’re not stuck only prompting for every small adjustment. It could really use smart routing of image generation requests to a nano-banana level image generation model for the images that get poured into the ui mockups (so you don’t need to round trip to external sources by hand). And of course, like all modern LLMs doing visual work, it lacks “taste” if not given extensive examples during prompting.
Right now it feels like a fun proof of concept that could be killed at any time ala Google’s many “I got my PM promoted at launch and then was promptly abandoned” projects.
I’m curious about that too. Claude Design feels (and from the blog post) sounds like the typical IC passion project that got just polished enough for a “beta” (in the sense of gauging customer interest, not commitment to release).
If someone with clout at Anthropic takes it up as a champion and pushes it into the portfolio of apps at the importance of Claude code, it could be an amazing product, eventually a threat to Adobe, Lovable, Canva market share. It needs a better direct editing layer like 2-way figma sync, so you’re not stuck only prompting for every small adjustment. It could really use smart routing of image generation requests to a nano-banana level image generation model for the images that get poured into the ui mockups (so you don’t need to round trip to external sources by hand). And of course, like all modern LLMs doing visual work, it lacks “taste” if not given extensive examples during prompting.
Right now it feels like a fun proof of concept that could be killed at any time ala Google’s many “I got my PM promoted at launch and then was promptly abandoned” projects.