Starting from scratch used to be the whole job — maybe not anymore
There’s a specific kind of dread that comes with opening a fresh Figma file. No structure, no direction – just a white screen and the feeling you should already be halfway done.
Claude Design is trying to get rid of that.
Launched this week by Anthropic, it’s building working layouts, prototypes, even rough front-end code.
You prompt it, and it gives you something you can actually click around.
We gave it a simple brief: a rough, psych-rock festival site with decent navigation. About 40 seconds later, it delivered something usable. Not perfect, but real – buttons worked, layout held together, and it didn’t feel like a throwaway mockup.
So after 24 hours, here’s what actually works.
Getting to 80% fast
This is where it shines. First drafts, rough layouts, internal tools — anything that usually takes hours to set up, it can get you most of the way there in minutes. It’s not final, but it’s a strong starting point.
Feeding it context
The more you give it — logos, styles, code — the better it gets. Without that, it defaults to a pretty safe, slightly generic look. With it, things start to feel intentional.
Using it once, then stepping back
Power users are already figuring this out: use the heavy model to build the structure, then stop. Tweak elsewhere. If you stay in the loop too long, you’ll burn through limits fast.
Treating it like a rough tool, not a finisher
The best workflow isn’t “design inside Claude.” It’s generate → export → refine in Figma. It gets you most of the way there, but the last 10–20% still needs a human eye.
But spend five minutes on Reddit and the cracks show up quickly.
The biggest complaint is usage. People are hitting limits fast, sometimes within a couple of hours. There’s already a whole strategy around not getting “cooked” by tokens.
Then there’s the look. Designers are calling out a “Claude aesthetic” – clean, rounded, slightly safe. If you don’t guide it, everything starts to blur together.
There’s also a quieter shift happening. Not full job replacement, but displacement. This thing can handle a lot of junior-level work — quick mockups, first passes, internal builds.
At the same time, it’s surprisingly useful as a second set of eyes. Some users say it gives blunt feedback that actually helps, more like a junior designer than a chatbot.
And that’s really the shift.
It’s not replacing designers — it’s cutting the excess. Smarter prompts in, less wasted output, fewer dead-end iterations.
You spend less time pushing pixels, more time deciding what’s actually worth keeping.
It’s early, and still behind a paywall. But the direction is clear: starting from scratch is starting to feel very optional.
